## THE ROLE OF MODERN NEURO-ONCOLOGY IN THE TREATMENT OF PRIMARY CNS TUMORS, AND BRAIN AND SPINAL METASTASES

EDITED BY : Marcos Vinicius Calfat Maldaun, Sujit Prabhu, Claudio Tatsui and Luis Souhami PUBLISHED IN : Frontiers in Oncology and Frontiers in Neurology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-852-9 DOI 10.3389/978-2-88963-852-9

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

## THE ROLE OF MODERN NEURO-ONCOLOGY IN THE TREATMENT OF PRIMARY CNS TUMORS, AND BRAIN AND SPINAL METASTASES

Topic Editors:

Marcos Vinicius Calfat Maldaun, Hospital Sírio-Libanês, Brazil Sujit Prabhu, University of Texas MD Anderson Cancer Center, United States Claudio Tatsui, University of Texas MD Anderson Cancer Center, United States Luis Souhami, McGill University, Canada

Citation: Maldaun, M. V. C., Prabhu, S., Tatsui, C., Souhami, L., eds. (2020). The Role of Modern Neuro-oncology in the Treatment of Primary CNS Tumors, and Brain and Spinal Metastases. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-852-9

# Table of Contents


Leomar Y. Ballester, Erin Dunbar, Nandita Guha-Thakurta, John W. Henson, Howard Chandler, Jeremiah Watkins and Gregory N. Fuller

*11 The Role of Intraoperative MRI in Awake Neurosurgical Procedures: A Systematic Review*

Tumul Chowdhury, Frederick A. Zeiler, Gyaninder P. Singh, Abseret Hailu, Hal Loewen, Bernhard Schaller, Ronald B. Cappellani and Michael West

*21 Molecular Evolution of a Glioblastoma Controlled With Tumor Treating Fields and Concomitant Temozolomide* H. Ian Robins, HuyTram N. Nguyen, Aaron Field, Steven Howard,

Shahriar Salamat and Dustin A. Deming

*26 Somatic SMARCB1 Mutation in Sporadic Multiple Meningiomas: Case Report*

Alice S. Wang, Ali O. Jamshidi, Nathan Oh, Ronald Sahyouni, Behdokht Nowroozizadeh, Ronald Kim, Frank P. K. Hsu and Daniela Bota

*32 Identification of Novel Genes Involved in the Pathogenesis of an ACTH-Secreting Pituitary Carcinoma: A Case Report and Literature Review*

Fuyou Guo, Guoqing Wang, Fang Wang, DingKang Xu and Xianzhi Liu

*37 The Comparative Treatment of Intraventricular Chemotherapy by Ommaya Reservoir vs. Lumbar Puncture in Patients With Leptomeningeal Carcinomatosis*

Mariano Montes de Oca Delgado, Bernardo Cacho Díaz, José Santos Zambrano, Vicente Guerrero Juárez, Manuel Salvador López Martínez, Elvira Castro Martínez, Javier Avendaño Méndez-Padilla, Sonia Mejía Pérez, Ignacio Reyes Moreno, Axayacatl Gutiérrez Aceves and Alberto González Aguilar

*44 Advanced Ultrasound Imaging in Glioma Surgery: Beyond Gray-Scale B-mode*

Massimiliano Del Bene, Alessandro Perin, Cecilia Casali, Federico Legnani, Andrea Saladino, Luca Mattei, Ignazio Gaspare Vetrano, Marco Saini, Francesco DiMeco and Francesco Prada


Bernardo Cacho-Díaz, Nydia A. Lorenzana-Mendoza and Luis F. Oñate-Ocaña

*79 Temozolomide Induced Hypermutation in Glioma: Evolutionary Mechanisms and Therapeutic Opportunities* Paul Daniel, Siham Sabri, Ahmad Chaddad, Brian Meehan,

Bertrand Jean-Claude, Janusz Rak and Bassam S. Abdulkarim

#### *86 Health Related Quality of Life in Adult Low and High-Grade Glioma Patients Using the National Institutes of Health Patient Reported Outcomes Measurement Information System (PROMIS) and Neuro-QOL Assessments*

Nicolette Gabel, David B. Altshuler, Amanda Brezzell, Emily M. Briceño, Nicholas R. Boileau, Zachary Miklja, Karen Kluin, Thomas Ferguson, Kaitlin McMurray, Lin Wang, Sean R. Smith, Noelle E. Carlozzi and Shawn L. Hervey-Jumper

*94 Impacts of Surgery on Symptom Burden and Quality of Life in Pituitary Tumor Patients in the Subacute Post-operative Period*

Mark R. Waddle, Mollie D. Oudenhoven, Casey V. Farin, Allison M. Deal, Riane Hoffman, Hojin Yang, Jennifer Peterson, Terri S. Armstrong, Matthew G. Ewend and Jing Wu

*101 Radiomics in Glioblastoma: Current Status and Challenges Facing Clinical Implementation*

Ahmad Chaddad, Michael Jonathan Kucharczyk, Paul Daniel, Siham Sabri, Bertrand J. Jean-Claude, Tamim Niazi and Bassam Abdulkarim

# Editorial: The Role of Modern Neuro-oncology in the Treatment of Primary CNS Tumors, and Brain and Spinal Metastases

Marcos Maldaun<sup>1</sup> \*, Sujit Prabhu<sup>2</sup> , Claudio Tatsui <sup>2</sup> and Luis Souhami <sup>3</sup>

1 IEP Hospital Sírio-Libanês, São Paulo, Brazil, <sup>2</sup> University of Texas MD Anderson Cancer Center, Houston, TX, United States, <sup>3</sup> McGill University Health Centre, Montreal, QC, Canada

Keywords: neuro-oncology, CNS tumors, brain metastases, spinal metsatases, gliomas, meningeomas

#### **Editorial on the Research Topic**

#### **The Role of Modern Neuro-oncology in the Treatment of Primary CNS Tumors, and Brain and Spinal Metastases**

This special multi-disciplinary article collection presents an exciting summary of what is new, promising, and controversial in the neuro-oncology field. Despite major advances in brain and spine tumor research as well as important developments and progress made in treatment delivery, unlike some other malignancies, we are unfortunately still far away from achieving a cure or even significantly improving survival rates in most pathologies in the neuro-oncology domain. On the other hand, many new developments in several areas in neuro-oncology have lately been reported, including new imaging technology, a better molecular and genetic pathology understanding, newer and safer surgical techniques, and promising target-oriented immunotherapy and chemotherapy strategies. In this special edition of Frontiers Neuro-Oncology, several of these Research Topics are presented, bringing new hope for improving outcomes in many areas.

Over the last 20 years, advances in surgery, radiation oncology, and systemic therapy associated with a better understanding of molecular and genetics profile in glioblastoma has lead to an improvement in median survival from 9 to 24 months. Although this is a significant change, it is clear that further improvement is necessary. Chaddad et al., in a comprehensive review, discuss how radiomics may aid in building predictive models for diagnosis, prognosis, and therapeutic response. Daniel et al. also review potential reasons for treatment resistance post-temozolomide treatment and describe the hypermutant genomic outcome (temozolomide-induced) and its possible impact on disease recurrence and newer therapeutic approaches. Del Bene et al. describe how useful advanced intra-operative ultrasound can be by guiding the surgeon to achieve a maximal safe resection in real time. On the same line, Chowdhury et al. report on a systematic review of the role of intra-operative MRI in awake neurosurgical procedures, pointing out that it is an imaging technique feasible and safe to perform.

In the pituitary field, Waddle et al., using validated instruments, for the first time characterized, in a prospective fashion, usual symptoms, complications, and quality of life before and after the subacute period of post-surgical intervention. A disturbed sleep pattern was identified as the culprit for worsened quality of life in such patients. Of interest, Guo et al. describe novel mutant genes in a patient with pituitary carcinoma. The authors were able to identify an uncommon P53 mutation in addition to novel mutations in ATRX and PTEN genes. This discovery can lead to important therapeutic changes by targeting very specific mutations. It may also help our understanding of the innate biology of these tumors.

Edited and reviewed by: David D. Eisenstat, University of Alberta, Canada

\*Correspondence: Marcos Maldaun marcosmaldaun@gmail.com

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology

> Received: 31 October 2019 Accepted: 21 April 2020 Published: 03 June 2020

#### Citation:

Maldaun M, Prabhu S, Tatsui C and Souhami L (2020) Editorial: The Role of Modern Neuro-oncology in the Treatment of Primary CNS Tumors, and Brain and Spinal Metastases. Front. Oncol. 10:759. doi: 10.3389/fonc.2020.00759

**5**

Health-related quality of life (HRQOL) metrics are unequivocally a very important aspect of treatment outcomes for all neuro-oncological disorders and a growing concern among investigators. There is now an increasing interest toward incorporating measurements of quality of life as endpoints in most studies. Two reports in this collection deal with this aspect of patients' management. Gabel et al. report the first use of PROMIS and NEURO-QOL, two validated questionnaires, to assess functional domain that could alter HRQOL in adult patients with low- and high-grade gliomas. The authors show that low-grade gliomas patients experience more pain intensity and greater distress compared to high-grade patients, while the latter do develop more significant neuro cognitive dysfunction. They also demonstrate that these two HRQOL tools are valuable questionnaires. From another front, Cacho-Díaz et al. from Mexico used a Mexican-Spanish version of the QLQ-BN20 instrument in patients with primary or metastatic brain tumors and showed its validity and reliability. Of importance is that the authors studied a bivariate association of HRQOL and overall survival, and they report that several domains affecting HRQOL were associated with a poorer overall survival. These findings challenge us to utilize patient reported outcomes metrics as measures to facilitate specific interventions.

Lastly, Suki et al., from the M.D. Anderson Hospital, describe a structured database from their institution that provides extremely useful guidelines for other investigators who are initiating such process in their centers. This large repository of clinical information has allowed them to obtain significant resources over time leading to hundreds of publications including landmark papers and practical and helpful information for the scientific community all over the world. Big data may further elucidate unknown mechanisms of chemo-resistance, for example, and will hopefully help in developing opportunities for novel treatments.

We hope the papers reported in this special collection (Ballester et al.;Cacho-Díaz et al.; Chaddad et al.; Chowdhury et al.; Daniel et al.; Del Bene et al.; de Oca Delgado et al.; Gabel et al.; Guo et al.; Ian Robins et al.; Suki et al.; Waddle et al.; Wang et al.) will aid investigators in their daily battle to improve outcomes in patients harboring nervous system tumors. We all appreciate how heterogeneous disease of the nervous system can be, and we agree that individualizing treatment based on clinical and pathological parameters remains a valid strategy in many situations. The modern neuro-oncology must integrate findings from the bench, including development of new systemic agents, to newer technological surgical and radiotherapy advances. We firmly believe this collection describes novel treatment paradigms and highlights the need for a multi-disciplinary neuro-oncological approach that emphasizes good perspectives for better and solid results in the management of nervous system tumors.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Maldaun, Prabhu, Tatsui and Souhami. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Primary Leptomeningeal Oligodendroglioma, IDH-Mutant, 1p/19q-Codeleted

Leomar Y. Ballester 1,2 \*, Erin Dunbar <sup>3</sup> , Nandita Guha-Thakurta<sup>4</sup> , John W. Henson<sup>3</sup> , Howard Chandler <sup>3</sup> , Jeremiah Watkins <sup>3</sup> and Gregory N. Fuller <sup>5</sup>

*<sup>1</sup> Department of Pathology and Laboratory Medicine, Houston, TX, United States, <sup>2</sup> Neurosurgery, University of Texas Health Science Center at Houston, Houston, TX, United States, <sup>3</sup> Piedmont Brain Tumor Center, Piedmont Cancer, Atlanta, GA, United States, <sup>4</sup> Department of Diagnostic Radiology, University of Texas MD Anderson Cancer Center, Houston, TX, United States, <sup>5</sup> Department of Pathology, University of Texas MD Anderson Cancer Center, Houston, TX, United States*

We present a case of a 43-year-old woman with a history of headaches and blurry vision. Ophthalmologic examination identified papilledema. MR imaging demonstrated a right parietal region mass with patchy areas of contrast enhancement and focal calcifications. Intraoperative examination and exploration revealed an extra-axial mass with no apparent parenchymal involvement. Microscopic examination revealed solid sheets of tumor cells with clear cell cytologic features and no discernable intra-parenchymal tumor component. Molecular studies demonstrated the presence of *IDH1 IDH1* c.395G>A p.R132H and *CIC* c.601C>T p.R281W mutations and 1p/19q codeletion. The radiographic features, gross appearance, and microscopic and molecular characteristics of the mass support the diagnosis of primary leptomeningeal oligodendroglioma, IDH-mutant, 1p/19-codeleted. This case represents one of a very few reported instances of molecularly-defined solitary, primary, intracranial oligodendroglioma, without definitive involvement of the brain parenchyma.

#### Edited by:

*Luis Souhami, McGill University, Canada*

#### Reviewed by:

*Sunit Das, St. Michael's Hospital, Canada David D. Eisenstat, University of Alberta, Canada*

#### \*Correspondence:

*Leomar Y. Ballester Leomar.y.ballester@uth.tmc.edu*

#### Specialty section:

*This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Neurology*

> Received: *13 June 2018* Accepted: *02 August 2018* Published: *27 August 2018*

#### Citation:

*Ballester LY, Dunbar E, Guha-Thakurta N, Henson JW, Chandler H, Watkins J and Fuller GN (2018) Primary Leptomeningeal Oligodendroglioma, IDH-Mutant, 1p/19q-Codeleted. Front. Neurol. 9:700. doi: 10.3389/fneur.2018.00700* Keywords: leptomeningeal oligodendroglioma, IDH1, 1p/19q-codeletion, diffuse glioma, CIC, ATRX, FUBP1

### CLINICAL PRESENTATION AND IMAGING RESULTS

A 43-year-old woman presented with several months of blurry vision and headaches. Ophthalmologic examination revealed papilledema. She had no personal or family history of malignancy, but had lived near the Chernobyl, Ukraine nuclear disaster site from birth until her late 20s. Brain magnetic resonance imaging (MRI) demonstrated a right paramedian parietal region mass, ∼5 × 6cm in perpendicular dimensions, with patchy areas of contrast enhancement and coarse calcification (**Figure 1D**). The mass exerted substantial local mass-effect but with only minimal vasogenic edema. Pre-operatively, it was difficult to distinguish whether the mass was intra- or extra-axial. Gadolinium-enhanced MRI of the entire spine showed no evidence of additional lesions. No other masses were identified CT imaging of the body. A right parietal craniotomy for maximal safe surgical resection of the mass was performed urgently. Intraoperatively, all visible tumor was removed. The tumor appeared to be entirely extra-axial, without a defined site of origin in the brain parenchyma. There was no evidence of brain invasion and no distal dural deposits were observed. No residual tumor was identified on post-operative MR imaging studies. After extensive multidisciplinary discussion, treatment with concurrent radiation and temozolomide, as per the STUPP protocol, was initiated (1).

## HISTOLOGY AND MOLECULAR RESULTS

H&E-stained sections showed diffuse sheets of a tumor composed of relatively small cells with round-to-oval nuclei and scantto-cleared cytoplasm (**Figure 2**). There were several small areas of stromal sclerosis with cell dropout, and collections of hemosiderin-laden macrophages were identified; in one of these foci, endothelial cell hypertrophy bordering on early hyperplasia was noted. No brain parenchyma was identified in the sections. Molecular signature and immunophenotype determination studies were performed. The tumor cells were immunopositive for GFAP, S100 protein, and mutant IDH1 p.R132H. ATRX immunostaining showed retained wildtype expression, with weak expression of p53 protein in a minor subset of tumor cells. Additional immunostains for synaptophysin, SMA, desmin, EMA, and keratins were negative. Mitotic figures were rare on H&E-stained sections, and were quantified at a maximum of 2 mitoses per 10 high-power fields using the phosphohistone H3 (PHH3) antibody. Computer-assisted quantitation yielded a correspondingly low Ki67 antigen (MIB1) labeling index of 4.8% (6,857 nuclei counted). Fluorescence in situ hybridization (FISH) studies showed a 1p/1q ratio of 0.58 and a 19q/19p ratio of 0.58, indicating the presence of 1p/19q codeletion in the tumor cells. Additional molecular testing (CARIS Molecular Intelligence, please visit https:// www.carismolecularintelligence.com for a complete list of the genes evaluated) confirmed the presence of the IDH1 c.395G>A p.R132H mutation and revealed a CIC c.601C>T p.R201W mutation (**Table 1**). The MGMT promoter (analyzed by pyrosequencing) was methylated. ATRX or FUBP1 mutations and BRAF-KIAA1549 fusion/tandem duplication at 7q34 were not detected. No mutations in the BRAF gene were identified. A final diagnosis of "Oligodendroglioma, IDH-mutant, 1p/19qcodeleted (leptomeningeal)" was rendered.

### DISCUSSION

In this case, the preoperative imaging studies showed an extraaxial mass with a small focus that appears inseparable from cortex (**Figures 1A,B**), raising the possibilities of either tumor pushing against the brain or a potential connection of the tumor to the brain parenchyma. The intraoperative observations of the neurosurgeon (solid tumor without an identifiable connection to brain parenchyma) and the results of the assessment of the resected tissue (as detailed above) indicate that this is an example of primary leptomeningeal oligodendroglioma (2–4). The other entity in the differential diagnosis is diffuse leptomeningeal glioneuronal tumor (DLGT), which very rarely can show combined 1p/19q codeletion (isolated 1p deletion is more common). IDH mutations have not been described in DLGT (2, 5–7). In contrast, DLGT or disseminated oligodendrogliomalike leptomeningeal neoplasms (DOLN) have been shown to frequently carry the BRAF-KIAA1549 fusion/tandem duplication at 7q34 (6).

Review of the preoperative imaging studies demonstrated the presence of multiple prominent and unequivocal foci of ring-like contrast enhancement (**Figure 1**). A ring enhancement pattern is traditionally indicative of anaplastic changes (8). In this case, the tissue available for examination did not show frank vascular proliferation or necrosis; thus, the presence of unsampled WHO grade III tumor (anaplastic oligodendroglioma) is possible. However, recent studies indicate only a very modest, if any, prognostic impact of traditional histologic criteria-based grading for WHO grade II-III IDH-mutant diffuse gliomas, including oligodendrogliomas, with the most important prognostic factors being IDH1/IDH2 mutation status and 1p/19q codeletion status (9, 10). CT imaging showed focal calcification (**Figure 1D**), supporting the clinical suspicion of the tumor's protracted natural history.

Several cases of primary leptomeningeal oligodendroglioma have been reported in the literature (4). An origin from meningeal glial heteroptopia has been postulated (3). However, many of the reported cases were not evaluated for 1p/19q codeletion or IDH1/IDH2 mutation status, raising the question of whether the tumors represent true oligodendrogliomas, as defined in the 2016 WHO classification system. In contrast to primary leptomeningeal oligodendroglial tumors, involvement of the leptomeninges by parenchymal oligodendroglial tumors

scant-to-cleared cytoplasm. (B) GFAP was positive in a subset of tumor cells. (C) IDH1 p.R132H mutant protein immunohistochemistry was strongly positive. (D) ATRX protein wildtype expression was retained. (E) Weak expression of p53 protein. (F) Low Ki67 labeling index of 4.8%. (G) Mitotic activity (arrow) was quantified at a maximum of 2 per 10 high-power fields using phosphohistone H3 (pHH3) immunostaining.

TABLE 1 | Summary of genetic alterations.


occurs only in a minority of patients (11). In this case, no parenchymal component was identified by imaging, intraoperative observation, or histologic studies, thus making this possibility unlikely.

In addition to MGMT promoter methylation, IDH1 c.395G>A p.R132H mutation and 1p/19q codeletion, the tumor showed a CIC c.601C>T p.R201W mutation. Mutations in CIC are a frequent finding in oligodendroglioma (10). This case thus represents one of only a very few reports of molecularly-characterized, primary leptomeningeal oligodendroglioma in an adult patient. Primary leptomeningeal oligodendrogliomas are molecularly distinct from DLGT/DOLN, which are tumors that present in pediatric patients and frequently exhibit BRAF-KIAA1549 fusion and 1p deletions. Although rare, primary leptomeningeal oligodendroglioma should be considered in the differential diagnosis of an extra-axial tumor with clear cell cytology. Testing for the critical molecular alterations (i.e., IDH1/IDH2 mutations and 1p/19q codeletion) is essential for accurate diagnosis of this rare presentation of oligodendroglioma.

There is no standard evaluation and management for primary leptomeningeal oligodendroglioma. The authors suggest baseline staging of the CNS axis by imaging, and, if possible, by cerebrospinal fluid examination. There are no studies regarding long-term outcomes of adjuvant management in primary leptomeningeal oligodendroglioma comparing observation vs. adjuvant radiation with or without chemotherapy. This patient's age, health, preferences, and the possibility of microscopic disease involving the meninges and CSF, factored into the recommendation for adjuvant treatment. Given the presence of MGMT promoter methylation, as well as the accumulation of long-term results of several international prospective randomized clinical trials demonstrating improved outcomes with the addition of chemotherapy to radiation, the patient was commenced on radiation and temozolomide chemotherapy, as per the STUPP protocol (1). An alternative treatment strategy would have been the use of radiation followed by PCV (12). Long-term clinical-radiographic surveillance of the CNS is warranted.

### REFERENCES


### ETHICS STATEMENT

This study was performed with approval of the institutional ethics committee (IRB) and with the patient's written informed consent for publication.

This study was carried out in accordance with the recommendations of the UT-MDACC IRB with written informed consent from all subjects.

### AUTHOR CONTRIBUTIONS

LB: Manuscript writing and figure preparation; ED: Clinical information and manuscript editing; NG-T: Figure preparation and manuscript editing; JH: Clinical information and manuscript editing; HC: Clinical information, specimen; JW: Specimen processing, slide review, manuscript editing; GF: Pathology and manuscript editing.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Ballester, Dunbar, Guha-Thakurta, Henson, Chandler, Watkins and Fuller. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Role of Intraoperative MRI in Awake Neurosurgical Procedures: A Systematic Review

Tumul Chowdhury <sup>1</sup> \*, Frederick A. Zeiler 2,3,4, Gyaninder P. Singh<sup>5</sup> , Abseret Hailu<sup>6</sup> , Hal Loewen<sup>7</sup> , Bernhard Schaller <sup>8</sup> , Ronald B. Cappellani <sup>1</sup> and Michael West <sup>2</sup>

<sup>1</sup> Department of Anesthesiology, Perioperative and Pain Medicine, University of Manitoba, Winnipeg, MB, Canada, <sup>2</sup> Section of Neurosurgery, Department of Surgery, University of Manitoba, Winnipeg, MB, Canada, <sup>3</sup> Clincian Investigator Program, University of Manitoba, Winnipeg, MB, Canada, <sup>4</sup> Division of Anaesthesia, Department of Medicine, Addenbrooke's Hospital, University of Cambridge, Cambridge, United Kingdom, <sup>5</sup> Department of Neuroanaesthesiology & Critical Care, Neurosciences Centre, All India Institute of Medical Sciences, New Delhi, India, <sup>6</sup> Max Rady College of Medicine, University of Manitoba, Winnipeg, MB, Canada, <sup>7</sup> College of Rehabilitation Sciences Librarian, Neil John Maclean Health Science Library, University of Manitoba, Winnipeg, MB, Canada, <sup>8</sup> Department of Primary Care, University of Zurich, Zurich, Switzerland

#### Edited by:

Marcos Vinicius Calfat Maldaun, Hospital Sírio-Libanês, Brazil

#### Reviewed by:

Sunit Das, St. Michael's Hospital, Canada Seunggu Jude Han, Oregon Health & Science University, United States

> \*Correspondence: Tumul Chowdhury tumulthunder@gmail.com

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology

Received: 06 July 2018 Accepted: 17 September 2018 Published: 10 October 2018

#### Citation:

Chowdhury T, Zeiler FA, Singh GP, Hailu A, Loewen H, Schaller B, Cappellani RB and West M (2018) The Role of Intraoperative MRI in Awake Neurosurgical Procedures: A Systematic Review. Front. Oncol. 8:434. doi: 10.3389/fonc.2018.00434 Background: Awake craniotomy for brain tumors remains an important tool in the arsenal of the treating neurosurgeon working in eloquent areas of the brain. Furthermore, with the implementation of intraoperative magnetic resonance imaging (I-MRI), one can afford the luxury of imaging to assess surgical resection of the underlying gross imaging defined neuropathology and the surrounding eloquent areas. Ideally, the combination of I-MRI and awake craniotomy could provide the maximal lesion resection with the least morbidity and mortality. However, more resection with the aid of real time imaging and awake craniotomy techniques might give opposite outcome results. The goal of this systematic review.is to identify the available literature on combined I-MRI and awake craniotomy techniques, to better understand the potential morbidity and mortality associated.

Methods: MEDLINE, EMBASE, and CENTRAL were searched from inception up to December 2016. A total of 10 articles met inclusion in to the review, with a total of 324 adult patients.

Results: All studies showed transient neurological deficits between 2.9 to 76.4%. In regards to persistent morbidity, the mean was ∼10% (ranges from zero to 35.3%) with a follow up period between 5 days and 6 months.

Conclusion: The preliminary results of this review also suggest this combined technique may impose acceptable post-operative complication profiles and morbidity. However, this is based on low quality evidence, and is therefore questionable. Further, well-designed future trials with the long-term follow-up are needed to provide various aspects of feasibility and outcome data for this approach.

Keywords: intraoperative magnetic resonance imaging (iMRI), awake craniotomies, outcome, complications, brain tumors

**11**

## INTRODUCTION

The role of maximal surgical resection in the case of brain tumors especially, intrinsic gliomas has been widely debated in neurosurgery, neuro-oncology, and radiation oncology with the underlying principle of this technique focused of extensive surgical cytoreduction prior to aggressive chemotherapeutic and radiation therapies. Current literature suggests the increasingly strong link between maximal safe surgical resection of highgrade glial neoplasms and both progression free and overall survival (1–8). The role of aggressive surgical approaches for low grade gliomas remains unclear especially in the case of asymptomatic incidental presentations, though some circles argue for similar aggressive resection in younger patient cohorts prior to neuropathologic transformation to high grade lesions (7, 9, 10). In addition, across the spectrum of intrinsic glial tumors, as we begin to better understand the molecular signatures associated with these lesions, it is becoming clearer that certain subtypes of gliomas may benefit from aggressive resection (11–15). However, one must acknowledge the decision to pursue aggressive operative intervention is one that is made weighing the risk and benefit profile, allowing the individual patient to decide what are the "acceptable" risks and potential morbidities.

In order to improve the safety profile associated with extensive surgical resections, awake craniotomy techniques have been implemented, particularly in those lesions located in or near eloquent structures (16). Awake craniotomy is a commonly performed neurosurgical procedure for the resection of brain lesions near to an eloquent area (17–21). This technique increases the safety profile and potentially improves the overall neurological outcome of the patient (17, 18, 20). Importantly, it has become a standard of care in many centers in the world. Such techniques require specialized neurosurgical, neuroanesthesia, and intra-operative neurophysiologic monitoring and nursing personal. The premise relies on the fact that the patient is under "light" sedation throughout the procedure, comfortable enough to tolerate a craniotomy, and tumor resection, but able to awaken and participate in intra-operative clinical examination during resection and direct electrical stimulation of neural structures. This allows the treating team to identify eloquent territories, avoiding aggressive resection in these areas and reducing the risk of permanent post-operative morbidity (22).

In line with the premise of awake craniotomy, intraoperative magnetic resonance imaging (I-MRI) provides real-time imaging and can potentially increase the degree of resection of brain tumors, either identifying missed areas or residual disease not grossly apparent to the neurosurgeon by direct inspection of the surgical field (1, 23). I-MRI has been used in many neurosurgical procedures including primary brain tumor resection, pituitary tumors surgeries and deep brain stimulation for various movement disorders (1, 23–25). In glioma surgery, I-MRI has been employed in patients under general anesthesia in order to optimize surgical results (26–28). To date, literature supports improved resection of gross imaging based T1 and T2 weighted MRI abnormalities. However, under such anesthetic conditions, some argue that with use of I-MRI may increase post-operative transient and permanent morbidity, particularly language and motor deficits (26, 29, 30).

Combining awake craniotomy techniques and I-MRI may provide optimal safe conditions for aggressive surgical resection of intrinsic glial neoplasms. It is plausible that the combination of I-MRI and awake craniotomy can provide maximum tumor resection with less post-operative morbidity and mortality. Therefore, the addition of these two techniques should produce favorable neurological outcomes (31). However, I-MRI assisted maximum resection can also lead to more language deterioration and new neurological deficits (26, 29, 30). In addition, there can be many complications including surgical, anesthetic or radiological during I-MRI use (32–36). Theoretically, combing two techniques may sometimes act as a double-edge sword, and it remains currently unknown the risk profile associated with using both techniques, and it may improve extent of resection at the cost of functional outcomes. Therefore, the goal of this systematic review is to identify the available literature on combined I-MRI and awake craniotomy techniques, to better understand the potential morbidity and mortality associated.

### METHODS

#### Protocol

This systematic review was registered with PROSPERO, International prospective register of systematic reviews (CRD42016052733). This review involves various steps including preliminary searches, piloting of the study selection process, formal screening of search results against eligibility criteria, data extraction, risk of bias (quality) assessment and data analysis. Though, statistical analysis was not carried out due to heterogeneity.

The protocol is developed on the basis of PICOS [Patient Population or Problem, Intervention (treatment/test), Comparison (group or treatment), Outcomes, and Setting question]. Whether or not, the inclusion of I-MRI with awake craniotomy imposes additional benefit or harm is the basis of this research. This review is reported in keeping with the systematic review guidelines in the preferred reporting in systematic reviews and meta-analysis (PRISMA) statement.

### Search Criteria

The search strategy was developed by the primary investigator (TC) in consultation with a professional librarian at Neil John Maclean Health Science Library, Winnipeg, Canada (HL). A search was conducted in the databases: MEDLINE, from 1946 to December 1, 2016 EMBASE, from 1996 to December 2, 2016, and the Cochrane Central Register of Controlled Trials (CENTRAL), issue 11 of 12 (December 1, 2016). The search strategy included appropriate subject headings and keywords for the concepts terms of awake neurosurgical procedure, awake craniotomy, and intraoperative magnetic resonance imaging. There were no language restrictions on the search. The detailed search terms are given in Appendix A in **Supplementary Material**. The study population of interest included adult patients undergoing awake neurosurgical procedures under I-MRI for brain tumors. Pediatric patients (aged < 18 years), and pregnant patients undergoing the above mentioned procedures were excluded. Retrospective as well as prospective observational studies, randomized clinical trials, and case series involving more than four cases were included for this systematic review.

#### Data Collection and Quality Assessment

On the basis of above defined terms, Initial titles and abstracts were provided by HL (librarian). All data (titles, abstract, exclusion criteria) were recorded in Microsoft Excel 15.0 version (password protected). Three separate sheets were created. The first sheet was for the titles and abstracts, second for the screened titles and abstracts (on the basis of inclusion/exclusion criteria) and third one for the final articles (on the basis of full texts). This part of data collection was done by two independent investigators (AH and GP) and any discrepancy was sorted out by the third (TC). In case, if primary or secondary outcomes defined for the project were not mentioned in the articles, corresponding authors were contacted to provide the data or clarification by the principal investigator (TC). The quality assessment was done by two reviewers (AH and GP). We used the Cochrane Collaboration' tool to assess the risk of selection, performance, detection, attrition, and reporting biases. For reducing selection bias, the fourth reviewer (FZ) reviewed all the data provided on sheet 2 and sheet 3 as well as cross-references. All studies were also categorized as direct, if mentioned awake craniotomies as the primary study subjects, and indirect, if mentioned awake craniotomies as one of the parts of total study subjects.

### Outcome(S)

#### Primary Outcome(s)

The primary objective of this study is to note the effect of I-MRI on overall morbidity in patients undergoing awake neurosurgical procedures. Morbidity is defined as any new neurological deficit or worsening of pre-existing neurological deficits. This is further divided into two: transient (short term) and persistent (long term). Transient deficits were defined as any morbidity that improved during the study period whereas persistent is defined as any morbidity that persisted through out the study period.

#### Secondary Outcome(s)

We noted the effect of I-MRI on various other parameters including extent of resection of brain tumor, intraoperative surgical complications, intraoperative anesthetic complications, intraoperative radiological complications, total duration of procedure and overall mortality.

### Data Synthesis

A descriptive data summary is presented as events numbers/proportions/percentages. To explain the data further, various tabulated aspects are presented in **Tables 1**–**5**. No formal statistical analysis was done. Meta-analysis was not carried out, as we did not have sufficient homogenous data, and there were lack of randomized controlled trials.

### RESULTS

### Study Selection and Characteristics

Our search strategy retrieved 438 titles and abstracts, and the subsequent filtering process is presented as a PRISMA flow chart [**Figure 1**]. After the deletion of duplicate results, 320 titles and abstracts were selected. Out of these, 280 articles were excluded on the basis of the pre-defined inclusion and exclusion criteria, 40 were screened further. After going through full texts for all 40 articles, only 10 articles met the criteria, and selected for final inclusion (37–46). All studies were conducted in a single center except one that involved 6 German centers (45). Seven were retrospective and 3 were prospective studies


T, tesla, I-MRI, intraoperative magnetic resonance imaging; n, number; R, retrospective; P, prospective, CC, case control, MR, multicenter retrospective, D, direct, In, indirect, EOR, extent of resection, NA, not available, PFS, progression free survival, N, No, Y, Yes, \*Number of patients were 34 (number of procedures-38), \*\*Number of patients 41 (number of procedures-42), <sup>a</sup>Average follow up period was 24. 8 months but criteria to divide transient to persistent was one month.


TABLE 2 | Primary outcome (s) in patients undergoing awake craniotomies under I-MRI.

T, tesla; I-MRI, intraoperative magnetic resonance imaging; n, number; (M; F), (Male; Female); P, propofol; R, remifentanil; NA, not available; D, dexmedetomidine; A, alfentanil; AWA, asleep wake asleep; M, midazolam; F, fentanyl; <sup>a</sup> right arm weakness; <sup>b</sup> speech problems; <sup>c</sup>All patients had preoperative deficits; <sup>d</sup>one of the deficits (motor; speech or sensory); <sup>e</sup>4 patients had word-finding difficulties; one had left sided inattention; <sup>f</sup> one left-sided weakness and two had word-finding difficulties; <sup>g</sup>one patient developed both aphasia and hemiparesis and other had hemiparesis; \*Either speech problems or motor deficits or both.

TABLE 3 | Secondary outcome(s) including resection of tumor, intraoperative complications and mortality in patients undergoing awake craniotomies under I-MRI.


T, tesla; I-MRI, intraoperative magnetic resonance imaging; n, number; GTR, gross total resection; NA, not available; S, seizure; PH, postoperative hematoma; \*three patients had seizures during cortical stimulation (out of these; one developed post-ictal right arm weakness); <sup>+</sup>two patients had seizures during cortical stimulation; one had intracranial hemorrhage; one had infarct and one had brain swelling; !one patient had seizure during stimulation; other had seizure but not during stimulation; #one patient had intracranial hemorrhage; ∧one patient had ischemia; three patients developed neurological deficits; and one had intracerebral hemorrhage.

(**Table 1**). Only one study included a control group [craniotomy under general anesthesia]. Three articles discussed 3 Tesla (T) I-MRI, five articles 1.5 T, two articles 0.2 T and one article mentioned both 0.23 T and 1.5 T. All articles showed level IV evidence except one that had a level III evidence. All studies were published between 2008 and 2016. These included 324 patients. Most of the studies primarily aimed at exploring the feasibility, functional outcome, and extent of resection. Few highlighted the progress free survival, adverse events and patient perception. Only four articles conducted the volumetric assessment for tumors.

#### Outcome Results

We included all those studies that mentioned morbidity data (**Table 2**). All studies had mentioned transient and persistent morbidity. All studies showed transient neurological deficits (speech disturbances, and/or motor weakness, and/or sensory deficits) between 2.9 and 76.4% with a mean of 35.6%. In regards to persistent morbidity, the mean was ∼10% (ranges from zero to 35.3%) with a follow up period between 5 days and 6 months. Two studies failed to disclose the exact follow-up duration.

For the secondary outcomes, 9 studies reported percentage of patients with gross total resection (15–70%) on final scans, however, only four included the GTR information (5– 36.7%) after the first scan (**Table 3**). Among intraoperative complications, eight studies noted surgical complications whereas anesthetic as well as radiological problems were mentioned in a single study (**Table 3**). Majority of the surgical complications included seizures during cortical


T, tesla, I-MRI, intraoperative magnetic resonance imaging; n, number, NA, not available; min, minutes; h, hours; Y, yes, (−1, 0), 1 day prior and same day; C; contrast, T2 (i), T2 sequence for the initial scan.

stimulations. Only, one study reported the mortality in one patient (44). Imaging, operative, and tumor information are also presented; however, these data are quite variable (depend upon institutional and I-MRI characteristics) and preclude any relevant interpretation (**Tables 4**, **5**). Along with I-MRI, all studies have utilized multi-modal monitoring techniques to further localize tumors (**Table 5**).

#### DISCUSSION

Our review of the literature on awake craniotomy plus I-MRI in the resection of intrinsic brain tumors has yielded important results, which deserve highlighting.

First, regarding the primary outcome of patient morbidity, the cumulative results are in keeping with literature on those patients undergoing resection of eloquently located glial neoplasms in the absence of awake craniotomy or I-MRI (17, 18, 47–49). Thus, from the 10 studies included within this review, the combined use of awake craniotomy techniques with I-MRI may not increase the post-operative transient and persistent neurological morbidity, with the range identified from 2.9 to 76.4% and 0 to 35.3%, for transient and persistent morbidity respectively. Though, it should be acknowledged that the overall patient numbers for all included studies are low, given the complexities of such techniques and the need for costly equipment. In addition, the studies suffered from a global lack of controls for comparison, in the setting of heterogeneous pathology, location, surgical teams/techniques, and I-MRI types/field strength. Thus, one must be reserved in implying that the combination of awake craniotomy and I-MRI is equivalent in safety to resection in the absence of such techniques, or in the presence of either only awake craniotomy or I-MRI. Therefore, the results of this systematic review provide preliminary evidence only to support safety, with much further investigation required to demonstrate equivalence or superiority. Furthermore, it must be emphasized that the use of these techniques, awake craniotomy and I-MRI, are typically reserved for those patients with eloquently located intrinsic tumors, as was the case for all studies included in this review. As such, the expected post-operative morbidity for resections carried out in such territories is high, and not necessarily a reflection of the combined technique, but the risk of operating in such cortical areas. Furthermore, as we've demonstrated, despite relatively high transient post-operative morbidity, these deficits typically resolve quickly during follow-up.

Second, with the application of this combined technique, the extent of GTR appears to be in keeping with standard I-MRI studies, where patients were under general anesthesia (17, 20, 21) This result provides preliminary evidence to support the notion that awake craniotomy techniques during I-MRI cases do not limit the ability to obtain acceptable operative resections for intrinsic tumors. With appropriate anesthetic techniques, one can perform similar resections to patients under general anesthetics. Though based on the small patient numbers in the included studies, these comments should be considered preliminary, with further investigation required.

Third, the surgical complication profile for these studies is in keeping with that described in other glioma surgical series and I-MRI series (17, 36, 50). This patient population classically carries a high pre-operative rate of epilepsy, and intra-operative rate of seizures. Our review demonstrated almost all intra-operative surgical complications were seizures, a welldescribed complication of cranial surgery, especially in cortically located intrinsic tumors. Thus, the combined technique of awake craniotomy and I-MRI does not appear to increase the intra-operative surgical complication profile. Though one must acknowledge, the use of I-MRI requires extensive surgical team training prior to implementation. Furthermore, the use of awake craniotomy techniques is also a specialized skill set, requiring collaborative efforts between the neurosurgical and neuroanesthesia teams. The appropriate awake craniotomy techniques are acquired through specialized training and require both knowledgeable and attentive teams to carry out successfully for extended duration cases, such as the resection of eloquently located intrinsic tumors while using I-MRI. Thus, the low surgical complication profiles seen in the studies included in this review are likely a reflection of the highly trained teams involved in these operative cases. This is also emphasized by the lack of operative mortality within the described studies.

TABLE 5 | Tumor characteristics, number of patients with pre-operative deficits or symptoms, number of patients with redo-operations and intraoperative localization techniques during awake craniotomies under I-MRI.


T, tesla; I-MRI, intraoperative magnetic resonance imaging; n, number; GTR, gross total resection; NA, not available; L, left; R, right; MEPs, motor evoked potentials; USG, ultrasound; DTI, diffuse tensor imaging; WHO, World Health Organization; F-MRI, functional magnetic resonance imaging; \*number of patients were 34 (number of procedures-38), \*\*number of patients 41 (number of procedures-42).

Fourth, the overall operative durations, when reported, ranged from 2.7 to 13.9 h. This time is including the additional time required for I-MRI scan acquisition. As every tumor is a different entity, it can be difficult to provide hard guidelines on the expected duration for the resection of such lesions. In general, for the resection of eloquently located intrinsic tumors, this operative range is in keeping with other series where the combined awake craniotomy/I-MRI technique is not utilized (36). Thus, based on the small cohorts described in the parent studies included in this review, it appears that the overall operative times are not dramatically increased secondary to the application of this combined approach.

Fifth, one potential concern regarding I-MRI remains various radiologic complications including dye induced adverse reactions and anaphylaxis, image distortions, burn injury, interference with anesthetic monitors, and failure to complete the scan. Our review demonstrated only one complication. This complication was a technical one, precluding scanning, resulting in no direct patient related consequences (38). As such, with the appropriate training and safety precautions, I-MRI in the presence of awake craniotomy techniques, can be safely conducted.

Finally, meticulous anesthetic techniques and medications have provided a safe environment for carrying out these prolonged and complex neurosurgical cases under IMRI. Majority of centers have utilized a combined approach of nerve blocks, local anesthetic infiltration and sedation (37–39, 41, 42, 44–46). Two centers have used general anesthesia (deep sedation) with supra-glottic airway device, laryngeal mask airway to protect the airway during initial and later phases of the procedure, and patients were subsequently awaken during the stimulation and tumor excision phase (40, 43). Only one study had reported an anesthetic complication intraoperatively (39). Notably, very few patients showed agitation, fatigue and noncompliant with the procedure in this study; however, there was no robust study designed exclusively for these parameters (39). Therefore, it is apparent that the present anesthetic techniques with standard monitoring make this challenging procedure safe and comfortable to the patients.

#### Limitations

Despite the interesting results generated from this systematic review, there are some important limitations that deserve highlighting.

First, the overall number of studies where awake craniotomy techniques in combination with I-MRI were used is quite small, at 10 studies identified. Furthermore, most studies focused on small patient populations with heterogeneous patient characteristics, tumor locations and histopathology. As a result, the overall conclusions regarding this combined operative technique for the resection of eloquently located intrinsic brain tumors are limited. Subsequently, the results of this review should be considered preliminary; supporting the need for properly designed prospective studies into the use of such techniques in glioma surgery.

Second, patient morbidity post operatively is influenced by various factors. Such factors include tumor location, preoperative deficits, extent of resection, tumor biology, duration of follow-up and also, surgical experiences. The studies

included were all focused on eloquently located lesions, however, the location and extent of such lesions varied significantly. In addition, the extent of pre-operative deficits was also heterogeneous. Extent of resection is influence by numerous factors, which will be discussed below. With that said, post operative morbidity is intimately linked with the extent of resection for eloquently located intrinsic tumors. Furthermore, tumor biology is important to acknowledge. The tumor histopathologic grade carries important implications for post-operative clinical course and the use of adjunctive chemotherapeutic and radiation techniques. Higher grade lesions tend to have a more complicated post operative and follow-up course, impeding the ability to determine if persistent deficits are related to surgical resection, inherent tumor biology or secondary effects of chemotherapeutic and radiation therapies. Finally, the duration of follow-up is important. The overall follow-up duration in the included studies ranged from 5 days to 6 months. Thus, any deficits seen during these periods may be permanent or in the process of ongoing evolution. It is difficult to comment on operative morbidity accurately with such heterogeneous and short follow-up periods.

Third, the GTR rates described within the included studies is subject to numerous factors. These factors include pre-operative expectations for resectability, patient/surgeon threshold for "satisfactory" and "acceptable" outcome, type of I-MRI used, and the use of various other intra-operative surgical adjuncts. Based on tumor location, size and extension, there is usually a pre-operative notion of how resectable an intrinsic lesion will be. These views based on pre-operative imaging likely continue to influence an individual surgeon's willingness to continue aggressive resection, and the pre-determined goal of a given operation (i.e., "GTR" or subtotal resection). Further, based on pre-operative clinical phenotype of the patient and both the surgeon/patient's view on what is an "acceptable" outcome, the extent of surgical resection of intrinsic brain tumors is dictated by such notions. What is deemed "acceptable" for outcome and morbidity varies significantly from patient to patient, and from surgeon to surgeon. As such, the GTR rates in this review are also likely a reflection of this. In addition, the type of I-MRI utilized can influence the ability to obtain GTR. Low field strength I-MRI was demonstrated to be inferior to high field (i.e., 1.5 or 2T) in the ability to obtain GTR in one study (45). Thus, comparing the resection rates for low and high field I-MRI is controversial, given the information provided by such low field units is inferior. Finally, many of the studies describe the application of various other intra-operative surgical adjuncts to aid with resection, including: MRI neuronavigation, DTI tractography, preoperative fMRI, electrophysiology including cortical mapping, and intra-operative ultrasound. All of these adjuncts aid with localization of tumor and eloquent cortex. Thus, the GTR rates, patient morbidity and operative complication profiles described within this review are likely influenced by all of these factors, making the exact impact of awake craniotomy/I-MRI on these outcomes difficult to discern.

Finally, and arguably the most important, is to re-emphasize that the comments and conclusions of this review should be considered preliminary. Based on the individual limitations highlighted above and the small patient numbers, one should be cautioned into considering the combination of awake craniotomy and I-MRI to be equivalent to standard glioma resection techniques, in the presence or absence of awake craniotomy or I-MRI alone. The significant heterogeneity in patients, pathology, lesion location, surgical teams, resection techniques, equipment, field strength, and follow-up information makes the results presented here preliminary for the combined efforts of awake craniotomy and I-MRI for glioma surgery. This is despite

#### REFERENCES


the data suggesting safety and comparable extent of resection and peri-operative complication profiles, in comparison to the existing literature. Much further work is required to investigate this combined technique, employing multi-center studies with control subjects and standardized surgical techniques, I-MRI technology and clinical follow-up principles.

### CONCLUSION

This systematic review suggests that the awake craniotomy combined with intraoperative MRI is feasible and safe to conduct. The preliminary results of this review also suggest this combined technique may impose acceptable post-operative complication profiles and morbidity. However, this is based on low quality evidence, and is therefore questionable. Further, welldesigned future trials with the long-term follow-up are needed to provide various aspects of feasibility and outcome data for this approach.

### AUTHOR CONTRIBUTIONS

TC developed the hypothesis, assisted substantially in data collection, screening, reviewing, analyzing, contacting the authors, compiling, and writing the manuscript. FZ reviewed and screened the data, and helped substantially in writing the manuscript. GS collected and screened the data and helped writing the manuscript. AH assisted in collecting the data. HL provided the initial data, abstracts, full-texts and formulated the search tables and prisma flow chart. BS assisted in writing, reviewing and editing the manuscript. RC assisted in developing the hypothesis, reviewing and editing the manuscript. MW assisted in writing and editing the manuscript.

### ACKNOWLEDGMENTS

We sincerely appreciate Dongxiao Zhuang and colleagues (Department of neurosurgery, Huashan Hospital, Shanghai Medical College, Fudan University, Shanghai, Neurosurgical Center, Shanghai, China) and Jan Coburger and colleagues (Department of Neurosurgery, University of Ulm, Germany) for providing clarification and data pertaining to their studies.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc. 2018.00434/full#supplementary-material

patients with malignant brain astrocytoma. J Neurosurg. (2009) 110:156–62. doi: 10.3171/2008.4.17536


neurosurgery: safety and utility. J Neurosurg Pediatr. (2017) 19:77–84. doi: 10.3171/2016.8.PEDS15708


50. Eseonu CI, Rincon-Torroella J, Lee YM, ReFaey K, Tripathi P, Quinones-Hinojosa A. Intraoperative seizures in awake craniotomy for perirolandic glioma resections that undergo cortical mapping. J Neurol Surg A Cent Eur Neurosurg. (2018) 79:239–46. doi: 10.1055/s-0037-1617759

**Conflict of Interest Statement:** TC served as a guest associate editor for Frontiers in Neurosciences and is presently serving as an associate editor for Intensive Care Medicine and Anesthesiology section in Frontiers in Medicine. FZ has received salary support for dedicated research time, during which this manuscript was completed. Such salary support came from: the Cambridge Commonwealth Trust Scholarship, the Royal College of Surgeons of Canada—Harry S. Morton Travelling Fellowship in Surgery and the University of Manitoba Clinician Investigator Program. FZ's research is also supported through the Thorlakson Chair in Surgical Research Establishment Grant. BS served as a guest associate editor for Frontiers in Neurosciences.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Chowdhury, Zeiler, Singh, Hailu, Loewen, Schaller, Cappellani and West. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Molecular Evolution of a Glioblastoma Controlled With Tumor Treating Fields and Concomitant Temozolomide

#### H. Ian Robins 1,2,3 \*, HuyTram N. Nguyen<sup>4</sup> , Aaron Field<sup>5</sup> , Steven Howard<sup>3</sup> , Shahriar Salamat <sup>6</sup> and Dustin A. Deming1,7

<sup>1</sup> Department of Medicine, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States, <sup>2</sup> Department of Neurology, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States, <sup>3</sup> Department of Human Oncology, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States, <sup>4</sup> University of Wisconsin School of Medicine and Public Health, Madison, WI, United States, <sup>5</sup> Department of Radiology, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States, <sup>6</sup> Department of Pathology, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States, <sup>7</sup> Department of Oncology, McArdle Laboratory for Cancer Research, University of Wisconsin School of Medicine and Public Health, Madison, WI, United States

#### Edited by:

Marcos Vinicius Calfat Maldaun, Hospital Sírio-Libanês, Brazil

#### Reviewed by:

Christoph Straube, Technische Universität München, Germany Kamalakannan Palanichamy, The Ohio State University, United States

> \*Correspondence: H. Ian Robins hirobins@wisc.edu

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology

Received: 27 July 2018 Accepted: 26 September 2018 Published: 15 October 2018

#### Citation:

Robins HI, Nguyen HN, Field A, Howard S, Salamat S and Deming DA (2018) Molecular Evolution of a Glioblastoma Controlled With Tumor Treating Fields and Concomitant Temozolomide. Front. Oncol. 8:451. doi: 10.3389/fonc.2018.00451

Tumor Treating Field (TTFields) therapy has demonstrated efficacy in a Phase 3 study of newly diagnosed glioblastoma (GB) following radiation (RT) and temozolomide (TMZ). We report the appearance of an isolated satellite anterior temporal lobe lesion, 2 months post primary RT/TMZ directed at the primary GB (MGMT methylated) parietal lobe lesion and one adjuvant cycle of TMZ and TTFields. The mean RT dose delivered to the temporal lobe lesion was negligible, i.e., 4.53 ± 0.95 Gy. Mapping of the generated TTFields demonstrated that both lesions were encompassed by a field intensity in a therapeutic range. The temporal lobe lesion remained under the control of TTFields up to 12 months, at which point progression on a T1 contrast MRI resulted in surgery and a definitive diagnosis of GB without MGMT methylation. The primary parietal lobe at this time was in remission. Molecular sequencing on the GB tissue from multiple time points demonstrates clonal evolution of the cancer over time and in response to treatment.

#### Keywords: glioblastoma, tumor treating fields, optune®, genomics, temozolomide

## INTRODUCTION

Tumor Treating Field (TTFields) therapy has demonstrated efficacy in a Phase 3 study of newly diagnosed glioblastoma (GB) following radiation (RT) and temozolomide (TMZ) (1), as well as in the recurrent setting (2). Interestingly, there have been no reports of TTFields therapy in GB patients who have not received prior RT. In addition, the potential mechanisms by which resistance to TTFields therapy develops has been understudied.

In the report to follow, an analysis of a satellite lesion that developed after standard RT and TMZ therapy in a newly diagnosed GB patient is presented. Therapy with TTFields had been initiated 1 month prior to the appearance of the satellite lesion. The patient was followed longitudinally with MRIs every 2 months; additional analysis of the radiation dose exposure, as well as the TTFields intensity, was performed. The differential diagnosis at the time included an MRI artifact or lesion induced by TTFields, vs. progressive disease. After 12 months, the aforementioned lesion was resected. Molecular alterations from baseline, postprogression on TTFields and following a further recurrence were assayed. The results below summarize these collective findings.

### CLINICAL DETAILS AND DATA ANALYSIS

In March 2016, a 51-year old male presented with left-sided numbness and weakness. A MRI demonstrated a 35 × 25 × 29 mm partially cystic or necrotic, enhancing mass with internal hemorrhage in the right parietal lobe. Subtotal resection was accomplished in March 2016 confirming a grade 4 astrocytoma with IDH1/2 wild type, MGMT methylated, and negative 1p19q co-deletion.

Standard radio-chemotherapy was completed in June 2016 (3), including daily TMZ with a total of 60 Gy radiation give in 30 fractions; adjuvant TMZ began in July 2016. TTFields therapy (1) was initiated in July 2016 and continued until August 2017. A post-radiation MRI was done in August 2016, showing increased thickness of the residual enhancing region in the right parietal lobe in addition to a new lesion in the right middle temporal gyrus (**Figure 1**).

Changes of residual tumor in the right parietal lobe was presumed to be progression vs. pseudo-progression, and the patient continued with six cycles of adjuvant TMZ, which was completed in December 2016. The changes in the parietal lobe lesion resolved over time, confirming pseudo-progression. In spite of the appearance of the temporal lobe lesion, it was decided to continue therapy with both TMZ and TTFields (with frequent monitoring), as the possibility of an artifact of TTFields therapy and/or an unusual form pseudo-progression was raised.

On a series of follow-up MRIs from August 2016 to August 2017, the initial parietal lobe lesion regressed with adjuvant TMZ and appeared stable on both T1+contrast and T2/FLAIR MRIs. The new enhancing lesion in the temporal lobe (during adjuvant TTFields/TMZ therapy) decreased from 9 to 7.7 mm in diameter with decreasing enhancement from August 2016 to November 2016 (**Figure 1**), and stayed stable on bi-monthly follow up MRIs until August 2017. At this time the temporal lobe lesion was at 17.9 mm in diameter (on T1+contrast); the parietal lobe lesion was essentially resolved, confirming pseudoprogression of this tumor (**Figure 1**). T2/FLAIR images showed abnormality with an area of restricted diffusion and peripheral rim enhancement in the region of the right temporal lobe lesion. A gross total resection of the temporal lesion was achieved in August 2017, confirming a grade 4 astrocytoma, with wild type IDH1/2, unmethylated MGMT, and negative 1p19q codeletion (**Figure 2**). The mean prior radiation dose for this temporal lesion was determined to be 4.53 Gy ± 0.95 Gy (5.7 Gy max; 3.5 Gy min; volume 0.1 mL). An isodose cloud is depicted (**Figure 3**). The lesion was 2.5 cm away from the edge of the planning target volume treated to full dose (46 Gy; center lesion dose 60 Gy).

The patient was then treated (September 2017–November 2017) with radiation (60 Gy in 30 fractions), targeting the temporal lobe resection cavity. An MRI in January 2018 demonstrated a possible new nodule (0.7 × 0.7 cm) on the edge of the resection cavity. A subsequent MRI in February 2018 confirmed progression with an increase in the aforementioned nodule to 1.4 × 1.8 cm. In March 2018, the patient underwent reoperation with a gross total resection as part of the TOCA 5 Tocagen Inc. clinical trial and was randomized to the control arm post-operatively. He started bevacizumab therapy in April 2018 which maintained his surgically obtained complete remission until relapse in August 2018.

Molecular analyses demonstrate that at resection of the primary parietal lobe lesion this cancer possessed mutations in BRAF (V600E), PTEN (319fs), and the TERT promoter (C228T). Following progression on TTFields, the separate anterior temporal lesion was resected. This lesion possessed these identical BRAF, PTEN, and TERT alterations, and was also found to possess a deep deletion of CDK2NA and an activating mutation in mTOR (V2006I). The lesion in the anterior temporal lobe that recurred following radiation was also sequenced following resection. This lesion was hypocellular, and similar to the prior resection exhibited mitosis, nuclear atypia and no necrosis; the same BRAF, mTOR, and TERT abnormalities were still able to be observed. No new alterations were detectable in this sample.

Additionally, a retrospective analysis demonstrated the TTFields intensity was in a therapeutic range for both the parietal lobe and temporal lobe lesion, i.e., 1 V/cm (**Figure 4**).

## METHODS

### Bio-Marker Testing

MGMT testing performed by LabCorp, NC; IDH testing done by PCR, UW Health Clinical labs, WI, 1p19q testing by FISH, Wisconsin State Laboratory of Hygiene.

### Determination of Radiation Dose

Using image registration software (Mim Vistag Cleveland, OH) that imports the radiation dose, the axial contrast-enhanced 3D T1-weighted images (T1 3D-SPGR Bravo, GE Healthcare, Waukesha, WI) were fused into a coordinate system of the treatment planning CT. A region of interest was drawn around the anterior temporal lobe lesion, and dosimetric analysis revealed the prior RT dose to the lesion.

### Mapping of TTFields Intensity

In order to estimate field intensity distributions within the lesions, numerical simulations were performed using finite differences Time Domain (FDTD) calculations and a realistic head model as described in Wenger et al. (4). Briefly, a realistic head computational model of a healthy male was created and scaled to match the dimensions the patient's head. Transducer arrays for the delivery of TTFields were positioned on the head model to mimic the personalized transducer array layout that was placed on the patient. In order to establish whether or not TTFields were delivered at therapeutic levels to the tumors, ellipsoidal regions approximately encompassing the lesions were manually marked on the field intensity maps. The field was considered to deliver TTFields at therapeutic levels to the lesion if the median field intensity within the respective ellipsoid exceeded 1 V/cm (5).

mutations in BRAF (V600E), PTEN (319fs), and the TERT promoter (C228T). Following progression on TTFields, the separate anterior temporal lesion was resected and demonstrated BRAF, PTEN, and TERT alterations, and the acquisition of a deep deletion of CDK2NA and an activating mutation in mTOR (V2006I). No other pathologic alterations were identified in the remaining 47 genes of the 88 genes assessed.

### Strata Oncology Hot Spot Sequencing

Patient samples were sequenced through STRATA Oncology CLIA-certified laboratory using the StrataNGS platform. This panel covers 88 genes and examines predefined variants including single and multinucleotide alterations, small insertions/deletions, fusions, exon skipping mutations, copy number variation, and microsatellite instability (www. strataoncology.com).

### DISCUSSION

In this report, we present the first instance of a grade 4 astrocytoma controlled by systemic TMZ and TTFields, with negligible radiation exposure. The patient's initial parietal lobe lesion was MGMT methylated, not IDH mutated, and not 1p19q deleted; the resected temporal lobe lesion was similar histologically, but was not MGMT methylated. Based on the

FIGURE 3 | Demonstration of isodose cloud for temporal lobe lesion (see arrow). Purple denotes 5Gy isodose; green denotes 8.57 Gy isodose.

MRIs between June 28th to August 14th, 2016 (**Figure 1**), the volume doubling time was calculated (6) as 14 days for the temporal tumor.

As the temporal lobe tumor appeared after initial concurrent radiation and TMZ treatment, this tumor may have been TMZ resistant, which is consistent with the absence of methylation on MGMT promoter. Alternatively, a resistant TMZ clone may have evolved over time. The initial radiation field was reconstructed, showing that the temporal lobe was exposed to minimal radiation at the time, 4.53 ± 0.95 Gy. This region, however, was within the TTFields effective region, suggesting that the suppression of tumor growth from August 2016 to 2017 was under the control of adjuvant TMZ and/or TTFields.

The original plan for the placement of OptuneTM arrays using the NovoTALTM methodology (7) targeted the right parietal lesion. It was not intuitively obvious that the field distribution in the temporal lobe region would be sufficiently high to have a therapeutic effect. Hence, numerical simulations (**Figure 4**) were performed; the simulations demonstrate that the field intensity delivered to both lesions was at therapeutic levels (>1 V/cm). Taken collectively, these data support the efficacy of TTFields in a newly diagnosed GB regarding a lesion that received a negligible dose of ionizing radiation. The contribution of adjuvant TMZ in controlling this lesion is indeterminate as discussed above.

Based on the molecular sequencing we can see that the cells within the anterior temporal lobe lesion developed from cells in the original primary parietal lobe lesions as the exact alterations were identified in both instances. The additional alterations identified presumably arose through clonal selection. While many factors could have potentially played into this selection process, we propose that it is quite plausible that the activating mutation in mTOR and/or the deep loss of CDKN2A could be inducing the resistance to TTFields therapy.

Over the last few years, new mechanistic insights have been gained into the anti-cancer effects of TTFields. These potential mechanisms of action include disruption of key

FIGURE 4 | Demonstration of the field intensity distribution in axial slices through the centers of the (A) primary right parietal lobe lesion and (B) the right temporal lobe secondary lesion. The median field intensity in the region of the primary lesion was 1.7 V/cm (mean of 1.66 V/cm). In the region of the secondary lesion the median intensity was 1.48 V/cm (mean of 1.56 V/cm). This suggests that TTFields intensities around both lesions exceeded the therapeutic threshold of 1 V/cm.

cellular functions, such as mitosis, DNA repair, mitochondrial function, and the folded protein response, leading to the induction of cellular stress, autophagy and apoptosis (5, 8–10). TTFields has also been implicated in enhancing the immune response through the induction of immunogenic cell death and modulation of antigen presentation (11). Loss of CDKN2A could lead to cell cycle dysregulation and mTOR activation could lead to inhibition of autophagy, apoptosis, and enhance cell proliferation overcoming some of the potential mediators of response to TTFields (12, 13). In addition, activation of the PI3K/AKT/mTOR signaling pathway has been associated with immune suppressive properties, including the up-regulation of the PD-L1 immune checkpoint.

In summary, this report provides evidence that TTFields may offer prolonged therapeutic benefit for some patients with recurrent GB. The molecular analysis of this patient's cancer over time provides potential insight to mechanisms by which resistance to TTFields might occur. This work also raises several interesting questions about how clonal evolution and spread through the central nervous system occurs, whether targeting therapies, such as mTOR or BRAF inhibitors, could be used in settings like this, and whether more routine molecular profiling should be obtained for patients with GB. Clearly, as we learn more about the biology of individual patients with GB this will lend itself to more precision-based treatment strategies for patients.

#### REFERENCES


### ETHICS STATEMENT

This is a case report regarding a patient treated with FDA approved standard of care treatment. Informed consent was obtained.

### AUTHOR CONTRIBUTIONS

HN contributed to the clinical history review, literature review, and manuscript preparation. SH evaluated radiation dosing and reviewed the manuscript. SS reviewed pathology. AF reviewed the manuscript and MRI scans. HR identified the patient, reviewed the clinical history and the literature, and prepared the manuscript. DD performed the genomic analyses and edited the manuscript.

### FUNDING

This research was supported in part by the: National Institutes of Health (NIH) (grant number P30 CA014520), UW Carbone Cancer pilot grant; the Decker Neuro-Oncology Research Fund.

### ACKNOWLEDGMENTS

We would like to thank Ariel Naveh and Ze'ev Bomzon of Novocure Ltd., for performing the numerical simulations.

In: Proceedings of the 109th Annual Meeting of the American Association of Cancer Research (2018) Apr 14-18. Chicago, IL; Philadelphia, PA: AACR (2018).


**Conflict of Interest Statement:** HR and DD have support for laboratory research from Novocure, Ltd.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Robins, Nguyen, Field, Howard, Salamat and Deming. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Somatic SMARCB1 Mutation in Sporadic Multiple Meningiomas: Case Report

Alice S. Wang<sup>1</sup> , Ali O. Jamshidi <sup>1</sup> \*, Nathan Oh<sup>1</sup> , Ronald Sahyouni <sup>1</sup> , Behdokht Nowroozizadeh<sup>2</sup> , Ronald Kim2,3, Frank P. K. Hsu1,3 and Daniela Bota2,3,4

<sup>1</sup> Department of Neurological Surgery, University of California-Irvine Medical Center, Orange, CA, United States, <sup>2</sup> Department of Pathology and Laboratory Medicine, University of California-Irvine Medical Center, Orange, CA, United States, <sup>3</sup> Chao Comprehensive Cancer Center, University of California-Irvine Medical Center, Orange, CA, United States, <sup>4</sup> Department of Neurology, University of California-Irvine Medical Center, Orange, CA, United States

Background: Multiple intracranial meningiomas account for <10% of all meningiomas. Familial multiple meningiomas have been linked to germline mutations in two genes: neurofibromatosis type 2 (NF2) and SWIch/Sucrose Non-Fermentable (SWI/SNF)-related matrix-associated actin-dependent regulator of chromatin subfamily B member 1 (SMARCB1). Sporadic multiple meningiomas have been associated with somatic NF2 mutations and, to date, there has been no case related to somatic SMARCB1 mutations. Here, we describe the first case.

#### Edited by:

Marcos Vinicius Calfat Maldaun, Hospital Sírio-Libanês, Brazil

#### Reviewed by:

Maria Caffo, Università degli Studi di Messina, Italy Mariko D. DeWire, Cincinnati Children's Hospital Medical Center, United States

> \*Correspondence: Ali O. Jamshidi aliojam@gmail.com

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Neurology

> Received: 04 August 2018 Accepted: 10 October 2018 Published: 26 October 2018

#### Citation:

Wang AS, Jamshidi AO, Oh N, Sahyouni R, Nowroozizadeh B, Kim R, Hsu FPK and Bota D (2018) Somatic SMARCB1 Mutation in Sporadic Multiple Meningiomas: Case Report. Front. Neurol. 9:919. doi: 10.3389/fneur.2018.00919 Case Report: A 45-year-old female suffered a head trauma while snowboarding. Subsequent to her injury, she experienced persistent headache, nausea, vomiting, dizziness, and flashing lights in the right eye. Magnetic resonance imaging (MRI) of her brain revealed multiple intracranial meningiomas. She underwent a two-staged craniotomy to remove frontal/parietal/temporal and occipital extra-axial tumors. Pathology confirmed the masses as meningiomas, WHO Grade I. Tumor genetic testing was positive for SMARCB1 mutation but blood genetic testing was negative for SMARCB1 mutation.

Conclusion: In sporadic multiple meningiomas, somatic NF2 mutations are usually the suspected genetic alternations. Our case illustrates that somatic SMARCB1 mutation is another genetic risk factor for sporadic multiple meningiomas, albeit rare.

Keywords: sporadic multiple meningiomas, familial multiple meningiomas, SMARCB1, NF2, somatic mutation, germline mutation

### INTRODUCTION

The phenomenon of multiple meningiomas is rare, accounting for <10% of all meningiomas (1). Multiple meningiomas can be categorized as familial or sporadic. Familial multiple meningiomas can manifest due to neurofibromatosis type 2 (NF2) disease, in which the NF2 gene on chromosome 22 mutates the growth inhibitory function of Merlin. Alternatively, familial multiple meningiomas can be inherited in an autosomal dominant fashion without the involvement of the NF2 gene (2). Sporadic multiple meningiomas are usually associated with somatic NF2 mutations (heinrich). Somatic SMARCB1 mutations have been associated with sporadic meningiomas in the literature; however, these were sporadic solitary meningiomas (3–5). To date,

**26**

there is no documented report of somatic SMARCB1 mutation as the underlying genetic alternation for sporadic multiple meningiomas. The authors describe the first case of a somatic SMARCB1 mutation in a patient with sporadic multiple meningiomas who harbors no NF2 mutations.

### CASE REPORT

The authors report the case of a 45-year-old female who suffered a snowboarding accident and presented several days following the event complaining of persistent headache, nausea, vomiting, dizziness, and photic auras in the right eye. She visited an urgent care facility twice and during these visits, no imaging was obtained. A neurologist saw her 18 days after the accident. An MRI scan was ordered that showed multiple masses some of which harbored hemorrhagic components: a left frontal parafalcine, calcified 2.0 × 2.2 × 3.3 cm [anteroposterior (AP), transverse (TV), craniocaudal (CC)] mass with associated vasogenic edema, a 4.5 × 2.9 × 4.1 cm (AP, TV, CC) mass in the left frontotemporal convexity with another mass measuring approximately 2.3 × 2.0 × 1.9 cm (AP, TV, CC) located just superiorly, and a 2.6 × 2.9 × 3.9 cm (AP, TV, CC) mass in the left occipital lobe (**Figure 1**). Additionally, there was an 8 mm left to right midline shift (**Figure 1**). Her findings were most consistent with multiple meningiomas. There was also a possible vestibular schwannoma measuring 0.7 × 1.4 × 0.7 cm (AP, TV, CC) in the left internal auditory canal (figure not shown). Initially, she was thought to have NF2. Pre-surgical tumor embolization and a two-staged surgery were recommended. The patient had successful embolization of the left middle meningeal artery and left posterior meningeal artery.

The first surgical stage involved a left-sided craniotomy for resection of the frontal-parietal-temporal meningiomas; pathology reported WHO Grade I meningiomas with a low/moderate proliferation index (percentages of positive Ki-67 tumor nuclei: left occipital mass: 2–3%; midline frontal mass: 3–4%; and left frontal mass: 1–2%; **Figure 1**). NF-2 blood testing (NEUROFIBROMATOSIS TYPE 2 SEQUENCING AND DELETION/DUPLICATION ANALYSIS IN Blood, UAB), which has a mutation detection rate in leukocytes of 93% was negative. This specific study detects truncating mutations (nonsense, frameshift, splicing mutations including deep intronic splice mutations), missense mutations, multi-exon deletions or duplications, and total gene deletions. Postoperative MRIs showed resection of the meningiomas in the left frontal/parietal/temporal convexity (**Figure 2**). Seven months

FIGURE 1 | Initial imaging showing multiple menigniomas. Preoperative contrasted T1 and T2 MRI scans reveal multiple meningiomas. The left calcified parafalcine lobulated mass (2 × 2.2 × 3.3 cm, AP, TV, CC) was associated with vasogenic edema (red arrows in A,B, D–E). In the left frontal temporal convexity, there was a 4.5 × 2.9 × 4.1 cm (AP, TV, CC) mass with another 2.3 × 2.0 × 1.9 cm (AP, TV, CC) mass located superior to it (black arrows in B–F). In the occipital lobe, the mass was measured to be 2.6 × 2.9 × 3.9 cm (AP, TV, CC) (white arrows in A,B,E). There was an 8 mm rightward midline shift. The images seemed to suggest neurofibromatosis type 2.

**Abbreviations:** SWI/SNF, SWItch/Sucrose Non-Fermentable; SMARCB1, SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily B member 1; NF2, neurofibromatosis type 2; MRI, magnetic resonance imaging; WHO, World Health Organization; AP, anteroposterior; TV, transverse; CC, craniocaudal; q, long arm of the chromosome.

FIGURE 3 | Resection of the occipital mass. Preoperative MRI showed a 2.6 × 2.9 × 3.9 cm (AP, TV, CC) mass in the left occipital convexity (white arrows in A–C). Immediate post-operative MRI showed removal of the mass (white arrows in D–F).

after her first-stage surgery, she underwent the second surgical stage: left occipital craniotomy for resection of a 2.9 × 2.7 × 4.2 cm (AP, TV, CC) mass (**Figure 3**). A specimen from the left occipital mass was sent to pathology which classified it as meningioma, WHO Grade I with a Ki-67 of 2–3% (**Figure 4**). Her most recent post-operative MRI scans at 16 months of follow up show multiple stable enhancing extra-axial masses compared to her immediate post-operative ones with no new lesions observed (**Figure 5**).

In addition to having multiple meningiomas, she also had a soft tissue mass in her right palm and moles in her left axilla. MRI scans showed the mass to be approximately 11 × 10 × 20 mm (**Figure 6**) with unclear pathology, and the patient was referred to plastic surgery. The patient underwent surgery to excise her cystic mass and pathology found it to be a benign nerve sheath tumor, consistent with Schwannoma with positive immunostain S-100. As for the moles on her left armpit, a shaved biopsy was obtained and pathology confirmed pigmented seborrheic keratosis, consistent with an atypical mole.

Because she was NF2 negative, further genetic testing was pursued for both somatic and germline mutations. Foundation Medicine, Inc. (Cambridge, Massachusetts, USA) tested her tumor tissue positive for SMARCB1 mutation (but not for NF2) while her blood genetic testing was negative for SMARCB1 mutation.

FIGURE 5 | Latest imaging. Her most recent MRIs (16 months after her last surgery) show multiple enhancing extra-axial masses, stable compared to her immediate post-operative MRIs. Here is a stable 1.8 cm (superior-inferior) meningioma arising from the left posterior falx, adjacent to the prior resection cavity (red arrows in A–C). No recurrence observed.

### DISCUSSION

Multiple meningiomas occur in <10% of patients with meningiomas (1). Multiple meningiomas, whether sporadic or familial, seem to have a clonal origin, rather than an isolated formation of unrelated tumors (6, 7). Cerebrospinal fluid may be the vehicle of transportation for spreading the clonal tumor cells to various locations in the central nervous system (7).

FIGURE 6 | Imaging of her right palm mass. MRI scans showed an approximately 11 × 10 × 20 mm mass in her right palm. Plastic surgery excised the mass and pathology confirmed a benign nerve sheath tumor, consistent with Schwannoma (A,B).

Neurofibromatosis type 2 disease harbors mutations in the NF2 gene, which are well-known risk factors associated with familial multiple meningiomas and sporadic multiple meningiomas (3, 4, 8). The NF2 gene is located at chromosome 22q12.2 and regulates the production of Merlin (schwannomin), a tumor suppressor protein. Approximately 50% of patients with NF2 disease inherit germline NF2 mutations and develop multiple meningiomas (9). The type and position of mutations in the NF2 gene contribute to a differential risk of developing multiple meningiomas. For example, truncating mutations at the proximal 5′ end of the gene pose a higher risk than nontruncating mutations at proximal 3′ end of the gene (10). The most common genetic risk factor is somatic NF2 mutations in sporadic multiple meningiomas (3–5).

Another gene of particular interest is SMARCB1 (aka INI1 and SNF5), which is located on chromosome 22 but at 22q11.23 (4). SMARCB1 is a tumor suppressor protein that is part of the SWI/SNF complex that remodels chromatin structures for transcription (11). For an extensive discussion of the underlying pathophysiological mechanisms, we refer the reader to Kalimuthu et al. (12). Mutations in the SMARCB1 gene have been associated with malignant rhabdoid tumor, schwannomatosis, and meningiomas (11, 13, 14). Bacci et al. found in a family study that germline SMARCB1 mutations, with no somatic SMARCB1 or NF2 mutations, are associated with familial schwannomatosis and multiple meningiomas (14). In a follow-up study by Hadfield et al., they found no germline SMARCB1 mutations in multiple meningiomas. This study contained both sporadic and familial multiple meningiomas (5:1 ratio in patient proportion); therefore, these sample differences could have contributed to this finding. Additionally, the sample size was small; only 6/47 patients had tumor DNA and blood DNA available for analysis (15). In another family study done by Christiaans et al., germline SMARCB1 mutation and somatic NF2 mutations were found in familial multiple meningiomas. They proposed the four-hit mechanism involving both tumor suppressor genes SMARCB1 and NF2 (13). Interestingly, van den Munckhof et al. found that germline SMARCB1 mutation and somatic NF2 mutations preferentially localized the cranial meningiomas at the falx cerebri (16).

Somatic SMARCB1 mutations have also been studied in meningiomas. Somatic SMARCB1 mutations have been reported in sporadic meningiomas; however, a closer examination of these reports reveals the meningiomas were solitary (17, 18). To date, there is no report of somatic SMARCB1 mutations in sporadic multiple meningiomas. The authors describe the first case in which somatic SMARCB1 mutation contribute to the development of sporadic multiple meningiomas.

Initially, the authors hypothesized that the patient likely had a founder germline NF2 mutation given that she has no family history of NF2. Once it was determined that she had no NF2 mutations in either blood or the tumor tissue, the authors tested for other mutations. Fountain Medicine, Inc. (Cambridge, Massachusetts, USA) found a mutation in the SMARCB1 gene. Her blood test was negative for SMARCB1 mutation, indicating that she did not harbor known or expected germline SMARCB1 mutations. It is unlikely, although still plausible, that the genetic testing yielded a false negative result, however, based on the family history and her clinical presentation, this is less likely. Therefore, she had a somatic mutation in the SMARCB1 gene.

#### REFERENCES


Because she harbored only this mutation and she developed sporadic multiple meningiomas, the authors believe somatic SMARCB1 mutation poses a genetic risk for sporadic multiple meningiomas.

### CONCLUSION

Sporadic multiple meningiomas are rare phenomena and germline or somatic NF2 mutations are usually the culprit. Here, the authors describe the first case in which somatic SMARCB1 mutation is responsible for the development of sporadic multiple meningiomas. Somatic SMARCB1 mutation is a genetic risk factor for sporadic multiple meningiomas and should be considered for testing when markers for NF2 are negative in similar clinical situations as it could be a marker for possible future therapeutics.

#### ETHICS STATEMENT

This study received an exemption from our institutional review board due to the nature of the study and the fact that it is a case report with fewer than 3 subjects.

#### CONSENT

Written, informed consent was obtained from the participant for the publication of this case report.

### AUTHOR CONTRIBUTIONS

AW, AJ, NO, RS, BN, RK, FH, and DB contributed to the design and implementation of the research, to the analysis of the results and to the writing of the manuscript.

the majority of sporadic meningiomas. Nature genetics (1994) 6:180-4. doi: 10.1038/ng0294-180


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Wang, Jamshidi, Oh, Sahyouni, Nowroozizadeh, Kim, Hsu and Bota. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Identification of Novel Genes Involved in the Pathogenesis of an ACTH-Secreting Pituitary Carcinoma: A Case Report and Literature Review

Fuyou Guo\*, Guoqing Wang, Fang Wang, DingKang Xu and Xianzhi Liu

*Department of Neurosurgery, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China*

Pituitary carcinomas (PCs) is considerable uncommon entities with a poor prognosis that represents only 0. 1–0.2% of all pituitary tumors. There are fewer than 150 reported cases up to now. In addition, the molecular pathogenesis leading to malignant pituitary transformation remain unclear due to the rarity of PCs. Here we present an uncommon case of ACTH-secreting PCs and explore the gene mutation following pituitary adenoma transformation. Our detailed clinical, histopathological and molecular detection data suggest that novel genes of ATRX and PTEN were implicated in the pathogenesis of PCs by searching Pubmed and the Web of Science databases as well as Cosmic databank. To the best of our knowledge, this is the first documented rare PCs patient with novel gene mutations that included ATRX and PTEN in addition to TP53. Present finding may therefore provide significant information for targeted therapy of PCs.

#### Edited by:

*Claudio Tatsui, University of Texas MD Anderson Cancer Center, United States*

#### Reviewed by:

*Brad E. Zacharia, Penn State Milton S. Hershey Medical Center, United States Dimitris G. Placantonakis, New York University, United States*

> \*Correspondence: *Fuyou Guo chyou666@hotmail.com*

#### Specialty section:

*This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology*

> Received: *03 August 2018* Accepted: *19 October 2018* Published: *06 November 2018*

#### Citation:

*Guo F, Wang G, Wang F, Xu D and Liu X (2018) Identification of Novel Genes Involved in the Pathogenesis of an ACTH-Secreting Pituitary Carcinoma: A Case Report and Literature Review. Front. Oncol. 8:510. doi: 10.3389/fonc.2018.00510* Keywords: pituitary carcinoma, mutant genes, molecular mechanism, targeted therapy, ATRX, PTEN

## INTRODUCTION

Pituitary carcinomas (PCs) constitute an extremely rare clinical entity that represents only 0.1–0.2% of all pituitary tumors, and there are fewer than 150 reported cases (1). PCs usually develop from progressive atypical pituitary adenomas and predominantly consist of hormone-generating tumors, defined by systemic metastases or the presence of disseminations to the cerebrospinal system. However, the molecular pathogenesis leading to malignant pituitary transformation is largely unidentified. A comprehensive understanding of the molecular mechanisms driving malignant pituitary progression would be beneficial for the treatment of pituitary carcinoma. Herein we described an uncommon PCs and the novel genes involved, providing useful evidence for future targeted therapy.

### CASE DESCRIPTION

A 55-years-old male presented with progressive deterioration of visual acuity and dizziness for 2 months. A preoperative computed tomographic (CT) scan revealed a large-mass lesion of the sellar region with extreme suprasellar extension (**Figure 1A**). Magnetic resonance imaging (MRI) demonstrated a large lesion located in the sellar region with heterogeneous enhancement and invasion to both cavernous sinuses (**Figures 1B–D**). The size of the tumor upon MRI was ∼3.0 × 2.5 × 4.0 cm. Endocrinologic tests showed that the levels of adrenocorticotropic hormone (ACTH) were significantly elevated to 411.3 pg/ml (range, 7.2–63.3) at 8 a.m. and 352.1 pg/ml

**32**

FIGURE 1 | Preoperative computed tomographic (CT) scanning revealed a large mass lesion in the sellar region (A). Magnetic resonance imaging (MRI) also demonstrated a large lesion located in the sellar region, with heterogeneous enhancement and invasion to both cavernous sinuses (B–D). Subtotal resection was obtained after operation (E,F). A recurrent tumor was involved in the saddle fossa and left cavernous sinus (G,H). Computed tomographic angiography (CTA) ruled out a left posterior communicating artery aneurysm (I). The PET-CT showed a residual intracranial tumor in the left cavernous sinus after a second operation (J). Multiple metastatic lesions were found in the lung (K,L; blue arrow indicates the metastatic lesion).

(4–32) at 4 p.m. The cortisol concentrations were 1,123.9 ng/ml (171–536) at 8 a.m. and 912.3 ng/ml (64–327) at 4 p.m. Other hormones, such as prolactin, growth hormone, free thyroxine (T3 and T4), and thyroid-stimulating hormone (TSH) were normal. The patient underwent an endonasal transsphenoidal surgery, and subtotal resection was obtained after the operation (**Figures 1E–H**). The postoperative ACTH levels dropped to 96.8 pg/ml at 8 a.m. and 78.3 pg/ml at 4 p.m., and the level of cortisol was reduced to 321.1 ng/ml at 8 a.m. and 165.2 ng/ml at 4 p.m. The residual tumor at the left cavernous sinus was subsequently treated with gamma knife surgery.

The ACTH and cortisol values remained stable during the 4 years of follow-up. However, 5 years after the first surgery, the patient was readmitted with a history of 20 days of left visual disturbance and 10 days of left eyelid ptosis. A MRI scan revealed a recurrent tumor in the sellar region and invasion of the left cavernous sinus. Computed tomographic angiography (CTA) was adopted to rule out an aneurysm of the left posterior communicating artery, and no aneurysm was found on CTA (**Figure 1I**). Hormonal evaluation showed slightly decreased levels of FT3 and FT4 (3.12 pmol/L [3.28–6.47] and 5.22 pmol/L [7.9–18.4], respectively). The ACTH levels were 41.3 pg/ml (7.0–61.1) at 8 a.m. and 38.6 pg/ml (3.5–30.55) at 4 p.m., and the levels of cortisol were 4.9 ug/dl (7–27) at 8 a.m. and 17.6 ug/dl (3.5–13.5) at 4 p.m. A second surgery was performed. The postoperative FT3, FT4, TSH, and cortisol values were significantly decreased compared with respective preoperative hormone levels. The patient was discharged under hormonal replacement therapy with euthyrox (25 ug/d) and cortisone acetate, and adjustment dosages were administrated based on subsequent endocrinologic tests. Postoperative histopathologic examination showed the presence of a PC. PET-CT was used for further evaluation and a residual intracranial tumor was observed in the left cavernous sinus (**Figure 1J**). Multiple metastatic lesions were also found in the lung (**Figures 1K,L**), and biopsy of these lesions revealed a metastatic neuroendocrine tumor.

Initial postoperative histopathologic examination revealed a pituitary adenoma, and microscopic evaluation showed that the tumor consisted of circular cells of uniform morphology

(**Figure 2A**). Immunohistochemical staining was positive for the expression of ACTH and P53 (**Figures 2B,C**), and Ki-67 expression was essentially negative (**Figure 2D**). A second postoperative histopathologic examination showed the presence of a PC. Hematoxylin and eosin staining also revealed the presence of excessive pleomorphic cells and frequent mitoses (**Figure 2E**). Immunohistochemical staining was positive for the expression of ACTH (**Figure 2F**), with strong positive staining for P53 and Ki-67 (**Figures 2G,H**); in fact, Ki-67 expression was up to 80%. In addition, we conducted a systematic review of the literature by searching Pubmed and Web of Science databases, and the Cosmic databank to ascertain all published studies on alterations in gene expression with respect to pituitary adenomas and pituitary carcinomas. Our literature search identified 44 mutant genes in pituitary adenomas. Their protein-protein interaction (PPI) network is shown in **Figure 2I**. Using these 44 genes, we found enrichment of several GO groups using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (**Figure 2J**). Among these GO groups were signaling pathways involving Ras, mTOR, MAPK, FoxO, ErbB, focal adhesion, and PI3K-Akt. However, we only uncovered limited novel gene expression changes and related clinical features from an isolated case report regarding PC (2–7) (see **Table 1**). In light of the exceeding rareness of PCs, molecular profiles of genes in the current rare case were

derived by the Beijing Pangenomics Technology Co., Ltd. Only 3 gene mutations were found among 509 genes examined in the ACTH-producing PC in our study, including mutations in ATRX (alpha thalassemia/mental retardation syndrome X-linked) and P53. In addition, we uniquely identified the novel mutation in PTEN (Phosphatase and tension homolog deleted on chromosome 10) by comparison analysis with published data on PCs. The patient's residual tumor was well-controlled by temozolomide and general radiotherapy for 3 months, and there have been no new lesions or metastases.

### DISCUSSION

PCs are considered uncommon entities and exhibit a poor prognosis due to their highly aggressive biology. Unlikely pituitary adenomas, PCs have received little attention, and the malignant transformation process from pituitary adenomas remains unclear. The present case is consistent with the criteria governing PCs as follows: (1) our patient experienced pituitary adenoma resection and subsequent adjuvant radiation as previous management; (2) a second postoperative histopathologic examination showed that the tumor was highly malignant, with Ki-67 expression up to 80%; and (3) multiple metastases into the lungs were


TABLE 1 | Summary of the literature regarding mutant genes involved in PCs cases.

*ACTH, adrenocorticotropichormone; ATRX, alpha thalassemia/mental retardation syndrome X-linked; DAXX, death-domain-associated protein; NA, none available; PTEN, phosphatase and tension homolog deleted on chromosome 10; y, years.*

observed by PET-CT, and biopsy of lung lesions revealed a metastatic neuroendocrine tumor. In addition, the most common category of PC was ACTH secreting (34.7%), prolactin secreting (23.6%), and null cell (15.3%). The latency period between the presentation of a sellar pituitary adenoma and the manifestation of metastases occupies a surprisingly wide range, from a few months to 18 years (median, 5 years). All accumulating evidence strongly supports the diagnosis of a PC, although it is an extremely rare tumor.

Temozolomide monotherapy is the first-line chemotherapy for pituitary carcinomas based on clinical practice guidelines from the European Society of Endocrinology (8). However, there is a paucity of randomized controlled trials for large-series studies; moreover, this drug is not sensitive and effective for all PCs. In fact, the estimated response rate to temozolomide is 58% in aggressive pituitary adenomas and 55% in PCs (9). Thus, it is necessary to elucidate the precise molecular mechanisms governing malignant transformation, which would then contribute to developing targeted therapy for PCs. Unfortunately, there have only been a few isolated mechanistic studies with respect to PCs described in the recent literature (**Table 1**). We exploited molecular profiles of PCs as they pertain to the pathogenesis of malignant transformation in our rare case; and conducted a comparative study from available data in the Cosmic databank (http://cancer.sanger.ac.uk/cosmic). Evidence showed that the gene mutation frequency for TP53 and HRAS were 33 and 14%, respectively, in PCs using the latest COSMIC databank. In addition, KEGG pathway enrichment analysis showed that the 44 mutant genes were enriched significantly within several signaling pathways, including Ras, mTOR, MAPK, FoxO, ErbB, focal adhesion, and PI3K-Akt. These enriched pathways provided insights into the molecular mechanisms underlying PC initiation and progression, and can therefore be useful in the development of new therapeutic strategies.

Our findings not only detected an uncommon P53 mutation from a total of 509 known genes, but also novel gene mutations in ATRX and PTEN unique to this PC. Mutations in ATRX and PTEN might, then, play vital roles in the malignant transformation of a pituitary adenoma into a PC. Most importantly, PTEN is a tumor suppressor gene that dominates the PTEN/AKT/PI3K pathway, prolonging progression-free survival to 11.4 months on pancreatic neuroendocrine tumors in a sunitinib group compared with 5.5 months in a placebo group (10). We assume that PTEN may be a crucial treatment target for PCs. Thus, we suggest that PTEN inhibitors, such as everolimus be used as an alternative chemotherapy for PCs once treatment failure occurs with temozolomide.

#### CONCLUSIONS

To our knowledge, this is the first documented PCs patient with novel mutant genes, including ATRX and PTEN. The present findings will therefore contribute to the development of promising targeted therapy based on individual gene assay for uncommon PCs, although further studies of a larger cohort of PC patients are necessary to clarify the precise molecular mechanism(s) underlying the pathogenesis of PCs.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of frontiers in oncology. The protocol was approved by the medical ethical committee of Zhengzhou University. This patient gave written informed consent in accordance with the Declaration of Helsinki. I currently state that written informed consent was obtained from the participant for the publication of this case report.

### AUTHOR CONTRIBUTIONS

FG: study concept and design, acquisition of data and writing paper. GW: study concept, analysis and interpretation of data.

#### REFERENCES


FW: study concept and design, data collection. DX: analysis and interpretation of data. XL: critical revision of manuscript for intellectual content.

#### ACKNOWLEDGMENTS

The authors would like to thank the members of the research group for useful discussions.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Guo, Wang, Wang, Xu and Liu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Comparative Treatment of Intraventricular Chemotherapy by Ommaya Reservoir vs. Lumbar Puncture in Patients With Leptomeningeal Carcinomatosis

Mariano Montes de Oca Delgado<sup>1</sup> , Bernardo Cacho Díaz <sup>2</sup> , José Santos Zambrano<sup>1</sup> , Vicente Guerrero Juárez <sup>1</sup> , Manuel Salvador López Martínez <sup>1</sup> , Elvira Castro Martínez <sup>1</sup> , Javier Avendaño Méndez-Padilla3,4, Sonia Mejía Pérez 3,4, Ignacio Reyes Moreno<sup>5</sup> , Axayacatl Gutiérrez Aceves 5,6 and Alberto González Aguilar 1,3,4,5 \*

#### Edited by:

Luis Souhami, McGill University, Canada

#### Reviewed by:

Yoshua Esquenazi, University of Texas Health Science Center at Houston, United States Maria Caffo, Università degli Studi di Messina, Italy

#### \*Correspondence:

Alberto González Aguilar albertogonzalez@ neurocirugia-innn.com

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology

> Received: 24 July 2018 Accepted: 17 October 2018 Published: 20 November 2018

#### Citation:

Montes de Oca Delgado M, Cacho Díaz B, Santos Zambrano J, Guerrero Juárez V, López Martínez MS, Castro Martínez E, Avendaño Méndez-Padilla J, Mejía Pérez S, Reyes Moreno I, Gutiérrez Aceves A and González Aguilar A (2018) The Comparative Treatment of Intraventricular Chemotherapy by Ommaya Reservoir vs. Lumbar Puncture in Patients With Leptomeningeal Carcinomatosis. Front. Oncol. 8:509. doi: 10.3389/fonc.2018.00509 <sup>1</sup> Emergency Department, National Institute of Neurology and Neurosurgery "Manuel Velasco Suárez" (INNN), Mexico City, Mexico, <sup>2</sup> Neuroscience Unit, National Cancer Institute, Mexico City, Mexico, <sup>3</sup> Neurosurgery Department, National Institute of Neurology and Neurosurgery "Manuel Velasco Suárez" (INNN), Mexico City, Mexico, <sup>4</sup> Neurooncology Department, National Institute of Neurology and Neurosurgery "Manuel Velasco Suárez" (INNN), Mexico City, Mexico, <sup>5</sup> Neurological Service, The American British Cowdray Medical Center, Mexico City, Mexico, <sup>6</sup> Radioneurosurgery Department, National Institute of Neurology and Neurosurgery "Manuel Velasco Suárez" (INNN), Mexico City, Mexico

Object: Leptomeningeal Carcinomatosis (LCM) represents a state of systemic malignant disease with poor prognosis. The purpose of this study is to compare overall survival (OS) between intraventricular chemotherapy through Ommaya reservoir (OR) and chemotherapy through lumbar puncture (LP) in LCM.

Patients and Methods: Forty adult patients with LCM were included. All patients underwent lumbar puncture and Magnetic resonance imaging (MRI). Thirty patients received chemotherapy through LP and 10 undergone colocation of Ommaya reservoir for intraventricular chemotherapy.

Results: The most common symptom was headache (Present in 50%). The cranial nerves most affected were VI and VII. Leptomeningeal enhancement was the most frequent finding in MRI. The OS in the LP group was 4 months and Ommaya group was 9.2 months (p = 0.0006; CI:1.8-3), with statistical differences in favor to Intraventricular treatment. Proportional hazard regression showed that receiving chemotherapy through Ommaya reservoir was a protective factor (Hazard ratio = 0.258, Standard Error = 0.112, p = 0.002 and 95% CI 0.110-0.606). Using KPS as a factor did not affect the hazard ratio of Ommaya reservoir itself.

Conclusions: OS was significantly higher in patients with Ommaya reservoir in spite of Karnofsky Performance Status (KPS) previous to chemotherapy. Therefore, intraventricular chemotherapy should be preferred over lumbar puncture chemotherapy administration if there are resources available.

Keywords: leptomeningeal carcinomatosis, overall survival, chemotherapy, ommaya reservoir, lumbar puncture, karnofsky performance status, intraventricular

**37**

## INTRODUCTION

LCM is a rare complication of advanced cancer, which consists in infiltration of the meninges and Cerebrospinal fluid (CFS) space by malignant cells (1). Any cancer can metastasize to meninges but is mainly detected in association with breast cancer, lung cancer, melanoma, and in fewer occasions with other types of cancer (Gastrointestinal, prostate, lymphoma, leukemia, unknown primary cancer) (2). It has an incidence of ∼5% of the patients with cancer but because of the asymptomatic patients or late-onset symptomatology, it may increase even to 20% as biopsies studies have demonstrated (3, 4). The median survival is around 4–6 weeks when untreated but it may improve as well as neurological status because of chemotherapy regimens (2, 4, 5). Karnofsky Performance Status (KPS) is the most reliable prognostic factor in patients with diagnosis of LCM (3, 4, 6). The gold standard remains the identification of malignant cells in CSF cytological study (7). The treatment goals are to improve the neurological status and to prolong survival. Different treatments are used (Radiotherapy and Neurosurgery) but the chemotherapy is essential in the management of LCM. Traditionally the method of election was the lumbar puncture (Intrathecal), but currently there are other options such as the Ommaya reservoir (Intraventricular) that might have better outcomes for patients (8). There is not a standard route of administration and both are recommended taking into consideration that chemotherapy needs good distribution and penetration; Intraventricular chemotherapy acts directly in CSF and probably it is superior to lumbar administration but there is not a trial that confirms this hypothesis completely. The present article is a retrospective study that compares the Intraventricular vs. the lumbar administration of chemotherapy in LCM.

## PATIENTS AND METHODS

We conducted a retrospective study collecting and analyzing data from patients diagnosed with LCM between 1980 and 2016 at National Institute of Neurology and Neurosurgery. We obtained clinical, imaging, histological, and treatment outcome data from electronic database such as gender, age, Karnofsky Performance Status (KPS), overall survival in months (OS, established with date of decease), symptomatology, primary tumor, localization of lesion by neuroimaging, treatment received, date of histological diagnosis, lumbar puncture (Glucose, proteins, cells, malignant cells), HIV status, and type of treatment (Intraventricular and Intrathecal). Diagnosis was established by presence of malignant cells in CSF and by neuroimaging findings in patients with histological diagnosis of cancer. Statistical analysis was performed using Stata/MP 14.1. In an effort to identify potential bias we stablished mean and t-test for scalar variables. Survival was established by Kaplan-Meier method taking on account impact of primary tumor and KPS. We used log-rank test to establish the statistical significance of difference in overall survival.

### Chemotherapy Protocol

The chemotherapy regimen administered was Methotrexate 15 mg (MTX) monotherapy, and IT triple therapy (ITtriple; 15 mg MTX, 30 mg/m<sup>2</sup> Cytarabine and 15 mg/m<sup>2</sup> Hydrocortisone) or Cytarabine (Ara-C) alone 30 mg/m<sup>2</sup> . The regimen was administered up to twice a week, according to the condition of the patient, until negative cytology (Induction phase), followed by once weekly for 4 weeks (Consolidation phase) and the last maintenance phase was once a month until progression, maximal doses or death.

## RESULTS

We identified 40 patients; ten patients had undergone installation of an intraventricular Ommaya reservoir (Between 2000 and 2014) and received chemotherapy for LCM while 30 patients received intrathecal chemotherapy through lumbar puncture. We obtained the following data: 26 patients were male (65%) and 14 were female (35%) with a ratio of 1.8:1. The median age was 52 years range of 18–76 (**Table 1**).

The KPS range was 40–100, with median of 70. Neurological examination and clinical symptoms were as follows: Headache was the most common symptom, present in 20 patients (50%). The rest of signs and cranial nerves (CN) most affected as well as KPS previous treatment are resumed in **Table 2**.


KPS, Karnofsky Performance Status; LP, Lumbar puncture; Ara-C, Citarabine; Mtx: Methotrexate.



The most affected CN were VI and VII. Neuroimaging findings were: meningeal enhancement (especially in cerebellum 21/40 patients) and nodular lesions 15/40 patients (**Figure 1**), in the 30% of patients the Magnetic Resonance Imaging (MRI) was normal.

The most common site of primary tumor was breast with 10 patients (25%), followed by lung (7 patients, 17.5%) and Leukemia (7 patients, 17.5%), then melanoma with 5 patients (12.5%), ovary with 4 patients (10%), then prostate with 3 patients (7.5%) and at last but not least lymphoma and unknown with 2 patients (5%, respectively) (**Figure 2)**.

The classic pattern of LCM is Hypoglycorrhachia, hyperproteinorrhachia, and malignant cells present. We analyzed the percentage of patients that presented this classic pattern and those who present different pattern. Referent to

FIGURE 1 | Axial MRI with nodular contrast enhancement in VII cranial nerve and enhancement in fourth ventricle, patient with diagnosis of lung cancer.

proteins in CSF, all patients presented values >45 (Range of 53–245), mean of 101.8. Glucose outcomes were divided in: Normal (8 patients, 20%), hypoglycorrhachia (30 patients, 75%) and hyperglycorrhachia (2 patients, 5%). Cellularity was present in all patients, with a range of 5–985, mean of 55.3. Classic pattern was present in 30 patients of the sample (75%).

FIGURE 3 | The Overall Survival by Lumbar chemotherapy vs. Intraventricular with Ommaya reservoir in patients with Leptomeningeal Carcinomatosis.

The Overall survival in the group with chemotherapy by LP was 4 months and Ommaya group was 9.2 months (p = 0.0006; CI:1.8-3), significantly higher in patients who received chemotherapy through Ommaya reservoir (see **Figure 3**).

Cox model for proportional-hazard regression showed that receiving chemotherapy through Ommaya reservoir was a protective factor (Hazard ratio = 0.258, Standard Error = 0.112, p = 0.002 and 95% CI 0.110-0.606). Using KPS as a factor did not affect the hazard ratio of Ommaya reservoir itself.

Overall survival by primary cancer had a median of 1.5 months for lung cancer, 2.1 months for prostate, 2.2 months for leukemia, 2.8 months for breast cancer, 3 months for melanoma, 4.3 months for lymphoma, 4.7 months for unknown primary, and 4.9 months for ovary cancer. Cox regression method showed breast and lung cancer as risk factors for poor prognosis with statistical significance (p = 0.069 and p = 0.012, respectively) (see **Figure 4**).

Regarding complications, two patients were reported with neuroinfectious disease with isolation and identification of Klebsiella pneumoniae in both cases. According to CTR (Common Toxicity Criteria) chemotherapy related complications noted were: Two patients with toxicity grade 2 in platelets count, two patients with toxicity grade 2 in WBC count, and one patient with toxicity grade 4 in platelet count.

We analyzed the prognostic factors (age, KPS, Chemotherapy and Intraventricular chemotherapy) without statistical differences (see **Table 3**).

#### DISCUSSION

LCM is a rare complication of advanced cancer, which consists in infiltration of the meninges and CFS space by malignant cells and with the presence of new treatments that increase survival it is likely that its frequency increases. We have reported and analyzed the outcomes obtained with chemotherapy through lumbar puncture vs. Ommaya reservoir in patients with diagnosis of LCM. There are numerous topics to underline. Some of our results were consistent with those reported on previous clinical trials.

As we mentioned before, headache was the most common symptom, present in 20 patients (50%), which is concordant with studies that described clinical features in LCM (9, 10). Cranial nerves are usually affected in patients with LCM, our patients presented mainly affection of VI and VII nerves alone or in combination with other cranial nerves (35% in total). Some reviews have noted that indeed, VI is the cranial nerve most affected (8, 11).

Previous studies showed abnormalities in CSF in more than 90% of the cases (11–13). It is necessary to have in mind that the most frequent pattern profile in lumbar puncture in patients with LCM began to be well identified since the 50's, nowadays CSF analysis has great importance in the approach of LCM (14–16). All of our patients presented abnormalities in CSF and 75% of them presented the classic pattern that we had discussed.

LCM involves the entire axis of CNS, therefore MRI takes great relevance. Contrast enhancement is necessary when obtaining a neuroaxis image. The principal site of lesion in our study was by far cerebellum, but it is clear that sites affected were variable and did not follow a pattern. However, bulky lesions are not always observable and diffuse pattern can be present as well as multiple lesions (17–19).

There are only three drugs available to administer intra-CSF: Methotrexate (MTX), Cytarabine (Ara-C) and less often Thiotepa. Effectiveness of these drugs is demonstrated in LCM, nevertheless is limited in some solid tumors associated to LCM (Melanoma and lung cancer). There was not significant difference between the three distinct types of chemotherapy employed in our study (Methotrexate, Liposomal Cytarabine TABLE 3 | Favorable prognostic factor in Leptomeningeal carcinomatosis by Cox regression method.


and MTX/Ara-C). Until now, effectivity of these drugs remains similar (3, 20). In addition, combination vs. single agent therapy neither has shown overwhelming superiority so it remains controversial, but it can be associated to less tolerance (14, 21). OS was significantly higher in the Ommaya reservoir group. The patient with the highest OS was a young woman with low KPS, nevertheless the log-rank showed no significance to this point. Concerning to Intraventricular chemotherapy administration, there is adequate drug distribution within the leptomeningeal space. Even when CSF flow is unimpeded, the normal CSF circulation carries fluid preferentially to the ventricles (22, 23). As a result, the delivery of drug administered into the lumbar Intrathecal space is unlikely to achieve clinically relevant drug concentrations within the cerebral ventricles, where malignant cells are known to reside (24–27).This may explain the observation that there is better response in patients who receive Intraventricular chemotherapy, in contrast to Intrathecal chemotherapy.

On the other hand, a recent clinical phase II trial (28), demonstrates that disturbances in the CSF flow makes chemotherapy ineffective as it may hinder the drug distribution and increase intracranial pressure. Same authors have pointed that ventroculolumbar chemotherapy showed improvement of increased intracranial pressure, altered mental status and cauda echina symptoms. It must be noted that this trial only includes Methotrexate in evaluation, however is a reliable study about perfusion rate, adverse effects and toxicity.

About primary tumors, results were similar to the rest of literature. Confirming that breast cancer, lung cancer and melanoma are the solid tumors with major association to LCM (16).

OS was affected by primary cancer as other studies have demonstrated (3, 29, 30).

Data regarding complications related to placement of Ommaya reservoir was limited to those noted on database.

### CONCLUSIONS

In our study, OS was determined by the factors previously mentioned, which is consistent with reports on similar trials. LCM represents an advanced stage of cancer and therefore it is a pathology of poor prognosis. Analysis of CSF and MRI to identify sites of lesions are fundamental to achieve diagnosis and to establish management. Recent research indicates that the future in the treatment of LCM is in the study of molecular targeted therapies such as epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) (31).

Another diagnostic studies like rare cell capture technology should be taken into consideration to the approach of LCM on the future. As well as the detection of CSF malignant cells through CellSearch (32, 33).

Chemotherapy is the cornerstone of the treatment and Intraventricular administration through Ommaya reservoir or other dispositive, have shown similar outcomes and also have demonstrated to be the best option when available.

We should mention that our results stablish intraventricular chemotherapy as a better option of treatment in this group of patients, nevertheless, due to the retrospective design and extended time of study, result should be taken cautiously. Further studies must include bigger sample size with data about complications related to the procedures such as increased intracranial pressure and ventriculitis.

#### REFERENCES


A prospective randomized study would be ideal to set conclusions but we consider it particularly difficult to select a homogeneous sample in patients with leptomeningeal disease.

Is important always having in mind that the objective of chemotherapy is to improve neurological status and quality of life more than prolonging survival. Next trials should be focused on improving diagnostic and therapeutic options that may reduce costs, avoid delayed processing, exempt patients from invasive procedures and allow a more precise diagnosis and prognosis.

### ETHICS STATEMENT

Ethical review process was not required due to the retrospective design of study and anonymized data of patients included, in accordance with the local legislation and institutional requirements.

#### AUTHOR CONTRIBUTIONS

MM and AGo: design of study, data collection, manuscript writing, analysis of results, discussion. BC, JS, VG, ML, EC, JA, SM, IR, and AGu: manuscript writing, analysis of results.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Montes de Oca Delgado, Cacho Díaz, Santos Zambrano, Guerrero Juárez, López Martínez, Castro Martínez, Avendaño Méndez-Padilla, Mejía Pérez, Reyes Moreno, Gutiérrez Aceves and González Aguilar. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Advanced Ultrasound Imaging in Glioma Surgery: Beyond Gray-Scale B-mode

Massimiliano Del Bene1,2, Alessandro Perin<sup>1</sup> , Cecilia Casali <sup>1</sup> , Federico Legnani <sup>1</sup> , Andrea Saladino<sup>1</sup> , Luca Mattei <sup>1</sup> , Ignazio Gaspare Vetrano<sup>1</sup> , Marco Saini <sup>1</sup> , Francesco DiMeco1,3,4 and Francesco Prada1,5,6 \*

<sup>1</sup> Department of Neurosurgery, Fondazione IRCCS Istituto Neurologico Carlo Besta, Milan, Italy, <sup>2</sup> Department of Experimental Oncology, IEO, European Institute of Oncology IRCCS, Milan, Italy, <sup>3</sup> Department of Pathophysiology and Transplantation, University of Milan, Milan, Italy, <sup>4</sup> Department of Neurological Surgery, Johns Hopkins Medical School, Baltimore, MD, United States, <sup>5</sup> Department of Neurological Surgery, University of Virginia, Charlottesville, VA, United States, <sup>6</sup> Focused Ultrasound Foundation, Charlottesville, VA, United States

#### Edited by:

Marcos Vinicius Calfat Maldaun, Hospital Sírio-Libanês, Brazil

#### Reviewed by:

Sunit Das, St. Michael's Hospital, Canada Edjah K. Nduom, National Institutes of Health (NIH), United States

\*Correspondence: Francesco Prada francesco.prada@istituto-besta.it

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology

Received: 15 September 2018 Accepted: 16 November 2018 Published: 03 December 2018

#### Citation:

Del Bene M, Perin A, Casali C, Legnani F, Saladino A, Mattei L, Vetrano IG, Saini M, DiMeco F and Prada F (2018) Advanced Ultrasound Imaging in Glioma Surgery: Beyond Gray-Scale B-mode. Front. Oncol. 8:576. doi: 10.3389/fonc.2018.00576 Introduction: Glioma surgery is aimed at obtaining maximal safe tumor resection while preserving or improving patient's neurological status. For this reason, there is growing interest for intra-operative imaging in neuro-oncological surgery. Intra-operative ultrasound (ioUS) provides the surgeon with real-time, anatomical and functional information. Despite this, in neurosurgery ioUS mainly relies only on gray-scale brightness mode (B-mode). Many other ultrasound imaging modalities, such as Fusion Imaging with pre-operative acquired magnetic resonance imaging (MRI), Doppler modes, Contrast Enhanced Ultrasound (CEUS), and elastosonography have been developed and have been extensively used in other organs. Although these modalities offer valuable real-time intra-operative information, so far their usage during neurosurgical procedures is still limited.

Purpose: To present an US-based multimodal approach for image-guidance in glioma surgery, highlighting the different features of advanced US modalities: fusion imaging with pre-operative acquired MRI for Virtual Navigation, B-mode, Doppler (power-, color-, spectral-), CEUS, and elastosonography.

Methods: We describe, in a step-by-step fashion, the applications of the most relevant advanced US modalities during different stages of surgery and their implications for surgical decision-making. Each US modality is illustrated from a technical standpoint and its application during glioma surgery is discussed.

Results: B-mode offers dynamic morphological information, which can be further implemented with fusion imaging to improve image understanding and orientation. Doppler imaging permits to evaluate anatomy and function of the vascular tree. CEUS allows to perform a real-time angiosonography, providing valuable information in regards of parenchyma and tumor vascularization and perfusion. This facilitates tumor detection and surgical strategy, also allowing to characterize tumor grade and to identify residual tumor. Elastosonography is a promising tool able to better define tumor margins, parenchymal infiltration, tumor consistency and permitting differentiation of high grade and low grade lesions.

**44**

Conclusions: Multimodal ioUS represents a valuable tool for glioma surgery being highly informative, rapid, repeatable, and real-time. It is able to differentiate low grade from high grade tumors and to provide the surgeon with relevant information for surgical decisionmaking. ioUS could be integrated with other intra-operative imaging and functional approaches in a synergistic manner to offer the best image guidance for each patient.

Keywords: Glioma, intra-operative ultrasound, contrast enhanced ultrasound, Doppler, B-mode, elastography, fusion imaging, navigated ultrasound

#### INTRODUCTION

Extent of resection (EOR) together with brain function sparing represent the most critical aspects of glioma surgery. A growing body of literature and evidences firmly supports gross total removal (GTR), instead of subtotal (STR) or biopsy (1, 2), defining GTR as the complete resection of contrast-enhancing regions in high grade glioma (HGG) on T1 weighted Gdenhanced magnetic resonance imaging (MRI), and of hyperintense areas on T2/FLAIR MRI in non-enhancing low grade glioma (LGG).

In diffuse LGG, GTR is able to improve progression free survival (PFS), overall survival (OS) and the time needed for malignant transformation (1–5). In HGG, GTR is nowadays considered the first phase of the standard multimodal therapeutic approach in order to extend PFS and OS (1, 2).

Surgeon's perception of gross total removal in glioma surgery is commonly inaccurate (6): portions of intra-axial tumor may resemble healthy brain parenchyma thus leading to sub-optimal resection and subsequently influencing patient's prognosis.

With these premises a growing interest for intra-operative imaging has led to the development of new technologies to localize tumors in order to ultimately help surgeons achieving GTR (7).

Numerous intra-operative approaches have been proposed: computed tomography (ioCT), magnetic resonance imaging (ioMRI), ultrasound (ioUS), fluorescence guided surgery (FGS) [e.g., 5-ALA, fluorescein, second window idocyanine green (ICG)] and other experimental techniques (e.g., optical coherence tomography and Raman spectroscopy) (7–14).

Among all these techniques ioUS is still one of the most employed, studied and developed, despite being the most dated since its first report was in 1978 with Reid (15).

US application in brain is especially favored by cerebral mechanical properties which allow an excellent US propagation and by the absence of superficial layers such as skin and subcutaneous connective which can distort US waves (16).

The main value of ioUS is the possibility to study the surgical scenario in real-time, every time it is needed, without the interruption of surgical work flow and, in specific condition, permitting to operate under direct guidance (17).

Continuous research and development led to US probes and scanners able to provide images with superb temporal and spatial resolution, comparable, or even superior to volumetric MRI (18).

Numerous studies have also investigated the diagnostic properties of ioUS in terms of sensitivity, specificity and ability to increase EOR and subsequently PFS and OS (9, 19–22).

In general ioUS demonstrated to own high diagnostic value in glioma surgery, in particular in low grade lesions, allowing to maximize the extent of resection and consequently to improve prognosis and quality of life of patients (9, 19–25).

It has to be said that in most of the cases ioUS application is limited to standard brightness mode (B-mode) with or without co-registration to pre-operative MRI. B-mode alone, being an anatomical representation of the echo wave for each point in the space, is a truncated application for ioUS. Indeed one pivotal adjunct of this technology is the possibility to implement different modalities to broaden the amount of different information.

The aim of our work is to review and describe different ioUS modalities in glioma surgery underlying the potential implications of standard b-mode and other advanced techniques such as fusion imaging, Doppler (power-, color-, spectral-), CEUS, and elastosography.

#### INTRAOPERATIVE ULTRASOUND IN GLIOMA SURGERY

#### US Equipment

US scanner should include the predisposition for different US modalities, a tracking system, the possibility to support different probes, the option to modify imaging presets through a complete access to all the US parameters. In general a specific designation for neurosurgery application is not required, as in most of cases a last generation general radiology US scanner with different presets is sufficient (10).

The scanner should provide a tracking system to allow fusion imaging with pre-operative MRI, to correct brain shift, and also to acquire a 3D US scan to obtain an updated neuronavigation volume.

The system should be equipped with different probes: a linear multifrequency (3–11 MHz) probe for deep-seated lesion, an high frequency (10–22 MHz) for small superficial lesions and a mini-convex to study the surgical field from inside the surgical cavity, overcoming the limitations of surgical artifacts in the final stages of tumor resection (10, 26).

Another issue is represented by imaging presets: in most US scanner designed for neurosurgery, imaging preset is standardized in order to provide highly contrasted image with few modifiable parameters. US is a demanding technique, with a steep learning curve and high operator dependency, but the only way to obtain the maximum in every situation is to became accustomed to this imaging modality and consequently being able to master the settings accordingly to the scenario (27).

FIGURE 1 | B-mode representation of different glioma grades. (A) Left temporal pilocytic astrocytoma. (B) Left temporal diffuse astrocytoma. (C) Right parietal anaplastic astrocytoma. (D) Right temporo-parietal glioblastoma. (1) pre-operative volumetric MRI. (2) Intra-operative trans-dural US scan. Note the different lesion appearances and in particular the different degree of margins definition.

The typical workflow at our Institution comprise the use of a last generation US scanner (MyLab, Esaote, Italy) with an integrated magnetic tracking system allowing for virtual navigation (MedCom GmbH, Germany). Typically the patient is registered in the 3D frame to the pre-operative MRI volume, permitting fusion imaging between ioUS and MRI and also to navigate with a pointer (as with a standard navigation system), to plan the surgical strategy and designing craniotomy site and shape (28). After bone flap removal the probes (usually a linear multifrequency and a mini-convex) are wrapped in plastic sterile sheath with coupling sterile US gel (Civco, USA) and a first direct trans-dural insonation is performed. The field is continuously irrigated with saline solution to allow US coupling and to improve imaging, reducing air or blood clots between probe and brain/dura. In every case the first US modality applied is B-mode, usually followed by Doppler, CEUS, elastography, depending on which information is necessary to achieve.

Indeed B-mode provides anatomical information requisite in order to understand and exploit the other modalities. Furthermore, B-mode permits to correct the brain-shift and brain deformation that naturally take place as resection advances (29, 30). Multiple US scans are performed throughout the whole surgery. Once tumor resection is completed, the final multimodal scan is conducted to evaluate potential hidden residual tumor and hypothetical tissue and vessels damages.

### B-mode ioUS

B-mode is the most simple and diffuse modality of US imaging, such that in some cases "US" and "B-mode" are erroneously used as synonyms. Literally, B-mode stands for Brightness mode, a two-dimensional US imaging modality formed by bright dots, which represent the amplitude of each reflected eco-wave in a specific point in the space. B-mode permits to visualize and characterize anatomical structures relying on their capacity to reflect, refract, absorb or transmit US beam (31, 32) (**Figures 1**–**3**). The brightness of a structure of interest (e.g., tumor) is evaluated in comparison to surrounding healthy tissue and consequently a structure can be hyperechoic, hypoechoic, or isoechoic. It has to be said that neurosurgical US semiotics is demanding especially for an un-experienced user. In general, structures defined as hyperechoic are: ependyma, choroid plexus, arachnoid interfaces, dural structures, skull, most tumors and their margins. Cisterns, ventricles, cerebro-spinal fluid, and some tumors tend to be hypoechogenic. Gray matter, white matter (typically gray matter is brighter than white matter) and some tumors appear as isoechoic (33, 34) (**Figures 1**–**3**).

Glioma appearance in ioUS B-mode is dependent on lesion grading and consequently biological behavior (**Figure 1**). HGG own an explosive growth, with high proliferation and areas of cysts, bleedings, necrosis, high-cells-density, and invasive zones. All these features lead to an heterogeneous representation in

which is possible to identify the different areas of the lesion. In general, it is possible to say that HGG appear heterogeneously echogenic with hyperechoic margins and iso-hypoechoic central necrotic areas. In general, the margins are more identifiable than in LGG even if is difficult to differentiate between tumor boundaries and peri-lesional edema (10, 23, 33, 35–39) (**Figure 1**). On the other hand, LGG appear slightly hyperechoic if compared to healthy brain, with homogeneous aspect and blurred margins particularly where they merge with healthy white matter (**Figure 1**). In most of cases B-mode imaging overlap pre-operative FLAIR MRI scan in LGG (**Figure 2**). The main difficulty in these tumors is to identify the margins/areas of invasion from peri-lesional edema (23, 33, 35–38).

Numerous studies have reported on the applications of Bmode in oncological neurosurgery and in particular in glioma resection (9, 10, 19–26). Several reviews and meta-analyses have addressed the value of ioUS in glioma surgery even if it is still not available a randomized controlled trial as pointed out in the last version of Cochrane review on intra-operative imaging technologies to enhance EOR in glioma (7).

In his meta-analysis Guangying Zhang found that B-mode provides an high sensitivity and specificity (0.75 and 0.88) in identifying tumor residual in glioma surgery operation, especially in LGG. In their study the Authors also confirmed the difficulty in distinguishing edema from infiltrating zone in particular for HGG and the ability of B-mode to display tumor presence also in areas with preserved blood brain barrier (21).

Bodil Karoline Ravn Munkvold performed a review on the diagnostic properties of ioUS in glioma surgery and analyzed factors influencing EOR. He found an overall specificity of 85% while sensitivity was 46% even if the residual tumor was small (median 1.05 ml) in cases with false-negative ioUS. Specificity was higher in LGG than in HGG (94 and 77%) and lowest in patients who undergone previous radiotherapy. The Authors conclude their analysis stating that ioUS specificity is high while the sensitivity for small residue is lower than post-operative MRI (22).

Syed Mahboob conducted a meta-analysis of the existing literature on the application of ioUS B-mode in glioma surgery. He analyzed 739 cases of LGG an HGG glioma operated under ioUS B-mode guidance in which gross total resection was achieved in 77% of patients (HGG 71.9% and LGG 78.1%). The Authors also examined, through a multivariate analysis, the factors implied in GTR finding that ioUS image quality is

FIGURE 3 | Navigated intra-operative B-mode US in the final stages of surgery. (A) Left temporal anaplastic astrocytoma. (B) Right fronto-temporal glioblastoma. The comparison between ioUS and corresponding pre-operative MRI aid in identifying residual tumor and understanding ioUS semiotics: (A) side-by-side view, (B) superimposition.

one of those and in turn it is influenced by previous surgery and radiotherapy. The Authors conclude that ioUS is able to improve EOR, especially in conjunction with other technologies to enhance anatomic orientation (9).

Jia Wang in his study investigated the role of ioUS in improving the survival time of patients who underwent resection of cerebral gliomas. He compared the survival rate at 6 months, 1-year, and 2-year in LGG and HGG patients operated with and without ioUS. He observed that in those patients in which ioUS was used survival rate at 1 and 2 year were significantly better than the survival rates of the controls (19). Their results were confirmed by Saether et al. in another retrospective study (40).

In our experience B-mode is extremely helpful in each phase of surgery. Before dural opening, it permits to find the lesion, to study its extension and if necessary to modify the craniotomy accordingly (**Figure 1**). Once dura is opened, brain shift take place, and consequently anatomy could be importantly modified. In this context B-mode allows to find the lesion, neighbor anatomical landmarks and vital structures (**Figure 1**). Furthermore, B-mode provides information on which gyrus is infiltrated and which is spared thus allowing to tailor the corticectomy according to lesion extension. Notably, being ioUS real-time, in case of discordance between US and neuronavigation, B-mode provide the most reliable and updated information (28). If the lesion is deep-seated, B-mode permits to plan the surgical corridor and if necessary to select the appropriate sulcus for a trans-sulcal approach. During surgical removal B-mode is repeated several times, to understand the dynamic surgical anatomy (e.g., inform on the distance to ventricles or vital structures) and to guide other ioUS modalities such as Doppler, CEUS, and elastography. In our experience repeated ioUS scan permits to be more confident in surgical resection and at the same time to be more efficient and safer.

At the end of surgical resection, a last scan is performed to identify potential residual tumor (10) (**Figure 3**). In case of doubt, more advanced imaging such as CEUS can be employed.

It has to be said that as surgical resection advances the ioUS image quality decrease (**Figures 1**, **3**). The sensitivity, specificity and derived values (positive predictive value, negative predictive value) are optimal before surgical resection and deteriorate as surgery proceeds (41). This is mainly due to surgical induced artifacts and edema and as a consequence several approaches have been proposed to overcome this limitation.

Navigating the US probe it is possible to compare the location of an hyperechoic area in ioUS with the tumor extension on pre-operative MRI; if the suspected tumor is outside the tumor area in the pre-operative imaging it is likely to be an artifact (25, 28, 37, 42–47) (**Figures 2**, **3**). Notably this approach can be only suggestive because even correcting the brain shift it is still not possible to correct brain deformation.

Selbekk et al. in 2013 proposed the use of a special coupling fluid to fill the surgical cavity. The hypothesis is that surgical artifacts are related to different acoustic coefficients of saline water and brain parenchyma thus inducing a bright artifact in surgical cavity wall. In order to overcome this limitation the Authors proposed a fluid with the same attenuation coefficient of human brain (48). Even if really interesting this approach is still experimental and is far from being routinely used in clinical practice.

Šteno et al. identified the cause of brightness artifacts in the column of water in the surgical cavity and as a consequence in the distance between probe and surgical bed (26, 36, 43). They proposed the application of miniature high frequency probes to scan the border of the surgical cavity from inside. The main limitation of this approach is the physical characteristics of these probes that are limited in field of view, lateral resolution, and US penetration.

In our experience the most reliable solution to overcome surgical induced artifacts and to discriminate between them, residual tumor and tumor induced edema is CEUS and we will analyze this application in the specific section below.

Tumor recognition in different phases of surgical resection is only one of the limitations of B-mode.

ioUS has to pay the steep learning curve and operator dependency mainly related to orientation and semiotics interpretation. Usually, neurosurgeons are accustomed to standard orthogonal planes (axial, coronal, and sagittal) while ioUS provides oblique planes dependent on probe location and orientation. In our experience fusion imaging with pre-operative MRI is extremely helpful to overcome this issue especially in case of an un-experienced operator (**Figures 2**, **3**). The continuous comparison with a familiar imaging allows to understand the orientation and the specific US semiotics, which is dynamic and influenced by surgical resection (10, 25, 28, 42–44, 46, 49). Another solution is 3D ioUS which allows to scan the surgical field obtaining a 3D volume in which is possible to navigate through a pointer in the standard planes (axial, coronal and sagittal) (10, 24, 25, 40–44). In our opinion this represent a really useful solution that can facilitates ioUS understanding and permits to navigate in an update 3D volume. At the same time, US scan, being a volumetric acquisition, bears less information than 2D US. Indeed 3D US does not allow to take advantage from the proprioceptive feedback and eye-hand coordination to reconstruct a real-time mental representation of the surgical field, as it is performed by sonologists in other corporeal regions to explore different relationships between structures (26).

#### Navigated ioUS

Cerbral US is not a familiar imaging for neurosurgeons. This is due to the impossibility to use US in pre- and post-surgical phases, whereas the opposite is true for diagnostic imaging such as MRI and CT, which own specific semiotics and orientation in three orthogonal planes. Furthermore, ioUS is peculiar for several reasons. Image orientation depends on the plane of insonation and consequently on probe orientation and position (**Figures 1**–**3**). Semiotics is specific and dynamic among the different phases of surgical resection. US does not permit to study intracranial space before bone removal and to plan craniotomy because of bone shielding (28) (**Figures 1**–**3**).

Fusion imaging permits to co-register ioUS and preoperative MRI for a continuous comparison of the two imaging modalities, enhancing US understanding and orientation. MRI provides known anatomical details and superimposing or visualizing side-by-side the modalities permits to interpret US orientation, semiotics and to understand its changes during time (**Figures 2**–**4**). Furthermore, different MRI datasets can be uploaded such as functional MRI, DTI, perfusion MRI, positron emission tomography in order to understand the location of vital structures, white matter tracts, or more aggressive areas in relation to tumor and real-time surgical situation (24, 25, 28, 41–44, 46). Brain shift can also be corrected relying on realtime US, updating neuronavigation for the most part of surgery while standard neuronavigation can be used to plan the surgical approach (28, 41, 43, 44, 46, 49, 50). Some groups have also demonstrated the possibility to correct brain shift and brain deformation in an automatic fashion through a software analysis taking into account landmarks position in ioUS and deforming MRI images accordingly (50–53).

Numerous Authors have demonstrated the clinical utility of navigated ioUS in glioma surgery allowing to maximize extent of resection and improving patients outcome (10, 19, 25, 40).

In our experience fusion imaging has demonstrated to be a reliable, accurate and useful technique especially for novice US users but also in complex cases and in experimental settings to compare or validate different US modalities in relation to MRI (**Figure 4**).

#### Doppler ioUS

Doppler US differs from standard B-mode not providing strictly anatomical but rather functional information. It relies on the Doppler effect. When a mechanical wave is reflected by a moving object this generates modifications of frequency and wavelength of echo-waves that can be studied allowing to infer information in regard of vessels blood flow (31). In routine practice, different sub-modalities of doppler are available, depending on which analysis is conducted it is possible to characterize blood flow in specific features (**Figure 5**). Color Doppler studies the presence of flow, its direction and velocity through the setting of a region of interest (ROI) (**Figure 5**). US scanners provides these information as color scale superimposed to standard B-mode (31, 54). Power Doppler is more sensitive to the amplitude of flow rather than direction and velocity (**Figure 5**). The image in encoded in a color scale depicting the total amount of Doppler signal, which in turn is dependent on the number of scattering molecules (in case of blood vessels mainly erythrocytes). This technique is extremely sensitive also to slow flow, typical of capillary district,

enhanced ultrasound in a case of left insular anaplastic astrocytoma. (C) Strain elastography in a case of left temporal glioblastoma; note the differences between the necrotic and the cystic areas and the interface with surrounding brain. Exploiting the continuous comparison between ioUS an pre-operative MRI is possible to understand US images and to infer about US and MRI correspondences. Legends are as follow: arrow heads: interface between tumor and brain; N: necrotic part of the tumor; C: cystic part of the tumor.

thus allowing, in some circumstances, to visualize also flow in sub-millimetric vessels (10, 31, 39, 54, 55) (**Figure 5**). Spectral Doppler requires identification of vessel of interest in B-mode through the setting of a proper ROI and indicating the vessel orientation. This analysis provides a detailed flow-velocity over time graph allowing to characterize vessels nature (e.g., artery, vein) and modification of flow during surgery (31, 54, 55) (**Figure 5**).

Each of these modalities has its specific indications and drawbacks. Color Doppler is informative in regard of vessels location and flow direction/velocity but at the same time it provides low spatial and temporal resolution and is severely limited by angle of insonation (**Figure 5**). Power Doppler is less dependent on angle of insonation and provides higher spatial resolution permitting to study also sub-millimetric vessels (**Figure 5**). On the other hand it is not informative on velocity and direction and suffers from low temporal resolution related to Doppler signal analysis. Lastly, spectral Doppler produces analyses with high temporal resolution allowing to characterize flow velocity pattern and changes in great detail while the drawbacks are the need to set a ROI, the absence of anatomical information and the important influence of angle of insonation (31, 54) (**Figure 5**).

In glioma surgery, probably, vessels study has less relevance if compared to extra-axial tumor such as skull base meningiomas. In any case, in specific condition this could be very helpful in orienting in the surgical field and in preventing post-operative deficits. Doppler imaging can aid in planning the surgical corridor and dura opening in case of medially located glioma (e.g., midline) according to bridging veins and sinus location (55). In case of deep seated gliomas surfacing on the basal cortex, Doppler allows to identify vessels position and to avoid undesirable damages (e.g., pericallosal arteries or middle cerebral artery branches (10, 49, 55). In specific conditions a vessel can represent a natural landmark for tumor margins location; if this situation is identified on pre-operative MRI, intra-operative Doppler could inform on vessel position thus aiding in achieving GTR avoiding unsafe excesses in eloquent areas (39). In this regard, Steno et al. reported on the feasibility to visualize lenticolo-striatal arteries in insular low grade glioma surgery. The Authors were able to visualize and preserve these small perforators thanks to last generation power Doppler (39).

In our experience Doppler imaging own several indications in glioma surgery, such as vessels flow characterization with spectral Doppler and repeated power Doppler scans in approaching vital structures (e.g., Sylvian vessels). At the same time, we are convinced that in most of the aforementioned applications contrast enhanced ultrasound is more informative and in our routine practice has replaced Doppler imaging.

#### Contrast Enhanced ioUS

For other imaging methods such as CT and MRI the use of contrast media is almost mandatory while it is less recognized for ioUS.

The use of CEUS during neuro-oncological procedures has been recently included in the guidelines from the European Federation of the Societies for Ultrasound in Medicine and Biology (EFSUMB), representing a paradigm shift for the use of US in neurosurgery (56).

Contrast enhanced ultrasound (CEUS) is an US modality which exploits a contrast agent (UCA) and a specific algorithm to study the cerebral vasculature down to the capillary bed (**Figures 4**, **6**–**8**). Nowadays, second-generation UCA are suspensions for venous administration of gas filled microbubbles stabilized by a phospholipid shell (MB) allowing for a dynamic and continuous imaging (10, 35, 55–59) (**Figure 7**). US scanner must be set to low-mechanical index acoustic power in order to induce MB oscillation (minimizing disruption) and consequently to produce a non-linear harmonic echo. Exploiting this feature, CEUS algorithm suppresses the linear US echo from tissue and display only the non-linear harmonic echo of MB thus producing a specific representation of MB distribution (10, 56, 59–62) (**Figures 4**, **6**–**8**).

Furthermore, MBs, being micron-sized, are not able to extravasate from vessels and behave as a purely intravascular contrast agent, allowing to study all districts of the vascular tree: arterial, venous, and capillary (10, 55–59) (**Figures 4**, **6**–**8**). CEUS is a dynamic modality which permits to visualize tumors by virtue of degree of vascularization, sharing features with other organs with a terminal circulation such as the kidney (**Figures 4**, **6**–**8**). It is possible to identify four phase of contrast enhancement (CE): arterial phase, peak of CE, parenchymal phase and venous phase (58, 59) (**Figure 7**). These phases are dependent on tumor vascularization and perfusion pattern and consequently are extremely informative on tumor biology. Our group has extensively studied CEUS application in neurosurgery with a special attention to glioma surgery (35, 37, 45, 55, 57, 58, 63–69). In our experience CEUS demonstrated to be able to (1) highlight tumors and their phases compared to brain parenchyma (35, 57, 69) (**Figures 4**, **6**–**8**), (2) characterize glioma grade (35) (**Figure 6**), (3) inform on vascularization and degree of perfusion (57, 69) (**Figures 4**, **6**–**8**), (4) show vascular rearrangement that take place with tumor removal (37, 55, 69) (**Figure 8**), (5) highlight residual tumor (37) (**Figure 8**), (6) aid surgical decision making through serial imaging assessment of surgical anatomy (10) (**Figures 4**, **6**–**8**).

GBM usually demonstrate a rapid contrast enhancement (CE) with an impetuous arterial phase (2–3 s), a prompt CE peak (3–5 s) followed by a short parenchymal phase and rapid venous phase (5–10 s). It is almost always possible to identify several feeders, which give a centripetal chaotic transit of MB. Venous phase highlights a diffuse drainage system with multiple medullary veins directed toward ventricles. After tumor removal, we observed with CEUS that medullary veins disappear and in some cases it is possible to see arterialized veins to change flow direction after resection (37, 57, 69). Characteristically it is possible to identify two CE patterns in GBM: (1) heterogeneous with nodular high CE spots interspersed by low-CE areas of necrosis and (2) peripheral rim CE surrounding a central core of necrosis without CE. In all cases GBM show a clearly demarcated border after UCA administration due to the different vascularization of tumor and healthy brain parenchyma (35, 57, 69) (**Figures 4**, **6**–**8**). We also demonstrated that CEUS is able to highlight the same tumor volume of pre-operative MRI with the same CE pattern thus permitting to visualize residual tumor among all surgical phases (37, 69) (**Figure 6**).

Anaplastic astrocytoma (AA) demonstrated a slower UCA dynamics with longer phases and CE duration. Even in AA, arterial feeders and venous drainage are visible in most of cases but in general less defined than in GBM. CE pattern is usually diffuse with in some cases few scattered areas of higher CE mixed with small hypoperfused areas. Tumor borders are visible but less sharply than in GBM (35, 57) (**Figure 6**).

LGG are characterized by two behaviors. Astrocytomas tend to resemble AA CE but with phases even slower. Arterial feeders are usually not identifiable, MB transit is organized and regular while venous drainage is diffuse through numerous capillaries and consequently not discernible. CE is diffuse with dotted appearance, only slightly higher than surrounding parenchyma and with blurred margins (35, 57) (**Figure 6**). On the other hand, oligodendroglioma CE has a tendency to be more rapid than in astrocytoma, owing faster arterial and venous phases which in any case are slower than in AA. CE pattern in homogeneous with sporadic intralesional cysts and calcification. Margins are better defined than in astrocytoma (35, 57).

Lastly, CEUS permits to identify neighbor vascular structures (both arterial and venous) allowing to follow their localization even at the end of the surgery when resection is on the margins and brain shift has made navigation inaccurate thus assuring a safer dissection (10, 55) (**Figures 4**, **6**, **7**).

#### Intra-Operative Elastosonography

Finger palpation has always been used in medicine. US elastography (ESG) represents the evolution of this approach. Applying a force to a tissue is possible to obtain a deformation that is related to its intrinsic mechanical characteristic, namely Young's E modulus (measure of the stiffness of a solid material) (70, 71). There are several techniques to measure and represent the elastic property of a tissue (70–72). Outside of the liver the most employed elastographic techniques are shear wave elastograhy (SWE) and strain elastography (SE). SWE belongs to

FIGURE 6 | CEUS representation of different glioma grades. (A,B) Right frontal low-grade glioma, (C,D) right frontal anaplastic astrocytoma, (E,F) left frontal glioblastoma. These images demonstrate the different degree and pattern of contrast enhancement among different glioma grades.

dynamic elastosonography and involves a focused US stimulus to induce a micrometric displacement to obtain share waves, which propagate orthogonally in the tissue (72). SWE provides both quantitative and qualitative information on tissue stiffness. SE is a quasi-static elastographic modality based on a mechanical stimulus to induce a tissue deformation which is measured by a high-frequency serial US acquisitions (70–72) (**Figures 4**, **9**). SE is more diffuse than SWE but again is capable only of qualitative measures.

Our group has focused the attention on SE demonstrating the feasibility in large-scale cohort of oncological-neurosurgery patients with the aims of lesions discrimination and

FIGURE 7 | Time-frame of contrast enhancement in a case of right temporal glioblastoma. GBM have a rapid arterial and venous phase. MBs transit is chaotic and the peak is extremely intense. The major arterial supplies and draining veins are clearly visible. Contrast enhancement pattern is irregular and heterogeneous CE with both nodular high-enhanced and hypoperfused areas. Tumor borders are better defined than in B-mode.

characterization. Our SE exam is usually conducted before dura opening, maintaining the probe stationary and exploiting brain pulsatility as described by other Authors (73). In 64 patients we did not observed damage or adverse effect and we were able to discriminate lesion volume. In glioma subgroup we observed that in most of cases SE provide a lesion representation superimposable to standard B-mode but with a sharper margins visualization (**Figures 4**, **9**). More importantly, SE demonstrated of being able to discriminate between LGG and HGG with a 85.7% of sensitivity and 94.7% of specificity (**Figure 9**). Indeed in most of cases LGG appear stiffer while HGG softer than surrounding brain parenchyma thus allowing to differentiate these tumor through an intra-operative US scan (**Figure 9**).

Our findings are aligned with the results obtained from other groups with SE and with SWE (10, 73–77). In any case, even if these results are really encouraging, elastosonography still must be considered an experimental technique.

### CONCLUSION

ioUS represents a pivotal adjunct to the existing surgical armamentarium, especially for a delicate application such as glioma surgery. ioUS is a polyvalent real-time imaging technique able to provide a great amount of information both anatomical and functional. Exploiting the advantages of each modality (B-mode, fusion imaging, Doppler-, CEUS, Elastography) it is possible to overcome several limitations of ioUS and to study glioma under various aspects. However, ioUS is an imaging technique that is rather demanding, requiring a specific training for each modality and in general for US semiotics, US physics and "knobology."

In our opinion ioUS should be part of a multimodal comprehensive approach for surgical guidance in glioma resection also encompassing other imaging and functional modalities in a synergistic and complementary fashion.

FIGURE 9 | Strain elastography (SE) scans in different glioma grades. (A,B) SE in a case of right parietal low grade glioma. (C,D) SE in a case of right fronto-parietal glioblastoma. SE is able to differentiate between LGG and HGG relying on their stiffness. LGG appear stiffer than brain whereas HGG softer. Furthermore, SE aid in identifying tumor borders and in distinguishing tumor and edema. (T: tumor, B: brain, arrow heads: interface between tumor and surrounding brain).

### AUTHOR CONTRIBUTIONS

MD, FP: study concept and design. FP, MD, and FD critical revision of the manuscript for intellectual content. FP, MD,

### REFERENCES


CC, MS, FL, AP, LM, AS, IV, and FD acquisition of data, data analysis, and interpretation. FP, MD, and FD: study supervision.


resection for glioma. Cochrane Database Syst Rev. (2018) 1:Cd012788. doi: 10.1002/14651858.CD012788.pub2


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Del Bene, Perin, Casali, Legnani, Saladino, Mattei, Vetrano, Saini, DiMeco and Prada. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Time-Tested Information System in Neurosurgical Oncology

Dima Suki\*, David M. Wildrick and Raymond Sawaya

Department of Neurosurgery, The University of Texas MD Anderson Cancer Center, Houston, TX, United States

The Brain and Spine Center at The University of Texas MD Anderson Cancer Center is a leading multidisciplinary referral center for patients with nervous system (NS) tumors. It has a wealth of clinical experience and an internationally recognized leadership role in the management of NS cancers. In that context, an informatics infrastructure that allows the archiving of both the prospective and retrospective characterization of patients, diseases, treatments, and outcomes is invaluable. We describe our experience with the Neurosurgical Oncology Database, a database that has provided valuable, extensive, and readily searchable data on multifaceted patient, tumor, and treatment characteristics for many years, successfully serving as an administrative and operational resource and as a resource for retrospective and prospective research endeavors.

Keywords: neurosurgical oncology, database, informatics, data management, information systems

#### Edited by:

Marcos Vinicius Calfat Maldaun, Hospital Sírio-Libanês, Brazil

#### Reviewed by:

Carsten Friedrich, University of Rostock, Germany Yoshua Esquenazi, University of Texas Health Science Center at Houston, United States

> \*Correspondence: Dima Suki dsuki@mdanderson.org

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology

Received: 27 August 2018 Accepted: 26 November 2018 Published: 17 December 2018

#### Citation:

Suki D, Wildrick DM and Sawaya R (2018) A Time-Tested Information System in Neurosurgical Oncology. Front. Oncol. 8:593. doi: 10.3389/fonc.2018.00593 INTRODUCTION

The Brain and Spine Center at The University of Texas MD Anderson Cancer Center (MD Anderson) is a major referral center for patients with nervous system (NS) tumors. It has a wealth of clinical experience and an internationally recognized leadership role in the management of NS cancers, both common and rare (1). As such, the department's leaders were keenly aware of their obligation to learn from their experience, and to apply this knowledge to improve disease classification, come up with novel therapy and management, and supply benchmark data that supports innovative experimental protocols (**Figure 1**). To meet this obligation, it was essential to have an informatics infrastructure that allowed us to archive and readily query both prospective and retrospective characterization of patients, diseases, treatments, and outcomes. The repository was intended to serve as an administrative and operational resource and as a resource for quality improvement, benchmarking, and educational purposes. It was also intended to serve as a valuable resource for research studies encompassing the epidemiology, natural history, characterization, and treatment of NS tumors and the multifaceted outcomes of patients with these tumors. The repository was endorsed as a top priority from day 1 and was afforded significant resources over time.

Establishing a database such as this one is a complex endeavor whose success and longevity necessitate a wide range of skills and resources on a long-term basis. This manuscript discusses the features of such a database and elements that are key to its success and longevity (**Table 1**).

## METHODS

The initial step in the development of the database was the formation of a multidisciplinary multilevel database task force. The task force was charged with identifying the uses of the database,

**58**

the various data items to be collected, the format and level of detail of these items, and resolving various multidisciplinary issues. The task force included:


A strong line of communication was secured among various team members. With the support of the task force, we applied a user-centered method to analyze tasks, workflow, and optimal interfaces for data entry, review, and mining. We then designed prototypes to map the results of user, task, and representation (interface) analyses and evaluated these prototypes. These steps were critical and fundamental to the final product.

TABLE 1 | Important considerations in planning, developing and maintaining a successful database.

	- a. Intra- or interdisciplinary
	- b. Commitment/involvement
	- a. Target patient population
	- b. Volume
	- c. Disease-, site- or treatment-specific
	- d. Cross-sectional vs. longitudinal
	- e. Database lifetime
	- f. Core dataset
	- a. Background/training
	- b. Exact role
	- c. Retention
	- d. Cost
	- a. The ALCOA data integrity test (Attributable, Legible, Contemporaneous, Original and Accurate)
	- b. Standardized data sources/definitions
	- c. Oversight plan
	- a. Institutional review/ethical board approval
	- b. Issues of consent/authorization vs. waiver
	- c. Applicable policies and regulations at all levels (e.g., Code of Federal Regulations; Good Clinical Practice; Health Insurance Portability and Accountability Act [HIPAA] and/or other applicable privacy laws; institutional information security policies; other policies and regulations)

### HARDWARE/SOFTWARE AND SYSTEM DESIGN (FIGURE 2)

The database is web based, which ensures easy access to users from various physical locations within the MD Anderson firewall. The software used has many of the features of other currently used data management programs, including (but not limited to) allowing complex data structures and complex logical checks. In addition, the software has superior security features and easy to navigate user and statistical software interfaces. It is backed by a premiere company and is periodically updated, thereby assuring the users state-of-the-art technology.

### DATABASE STRUCTURE (FIGURES 3–5)

The database is relational, as depicted in **Figure 3**, and is organized into screens of varying lengths and numbers of

fields. The main data entry screens are accessed through a tab at the top of the screen. Key tabs are the Demographics tab (with details on the patients' demographic information); Tumor History (with longitudinal details on the patients' tumor history, including radiographic and pathologic diagnoses, histology, grade, and metastases); Surgery (with specific subscreens depending on the type of case (brain, skull base, spine, peripheral, pain, other; metastasis vs. primary NS; and details on all procedures related to the nervous system tumor, hospital stay, symptomatology, complications); Chemotherapy; Radiation; and Imaging. **Figure 4** shows an example schema of brain procedurerelated tables, and **Figure 5**, an example of the brain procedure interface. The database allows for additions/modifications as needed via an approved process.

## PATIENT ELIGIBILITY

All patients undergoing a neurosurgery at MD Anderson, irrespective of diagnosis, have their data collected and stored in the database. The requirement for patient consent was waived by the institutional review board (IRB). The waiver relied on justifications that the research (in this case strictly data collection and storage) involves no more than minimal risk to the subjects, that the waiver would not adversely affect the rights and welfare of subjects, that the collection could not be practicably carried out without the waiver, and that whenever appropriate, the subjects will be provided with additional pertinent information.

## DATA COLLECTION

The patient is registered in the database at the time of their MD Anderson neurosurgery. Registration triggers the beginning of the review/data collection process for a given patient. Dictated notes for all patient encounters (clinic visits; surgery or other procedure; testing; or other reason), scanned documents, images, and other components of the patient's electronic medical record (EMR) are reviewed after registration and until postoperative day 30 or until events and treatments during the 30 days can be captured, whichever is later. Variables entered into the database include but are not limited to: patient demographic data; clinical parameters, such as Karnofsky Performance Scale score (and other functional measures), neurologic status, symptoms, diagnosis date, cancer site(s), tumor histology (and other pathology-specific characteristics), and imaging characteristics; treatments, such as surgery details (procedures, indications, intraoperative adjuncts used, blood loss and transfusions, and extent of resection, where applicable), radiotherapy details (type, date, dose, schedule), chemotherapy details (type, date, dose, schedule); and treatment outcomes including length of hospital and rehabilitation stay, complications/toxicities, and survival time. In addition to the imaging data obtained from the patients' medical records, a qualitative visual review of the image is performed to assess for entities such as presence or absence of tumor necrosis, contrast enhancement, cysts, hemorrhage, gliomatosis cerebri, and others, as well as confirmation of tumor location and tumor functional grade. A quantitative assessment of preoperative/postoperative necrosis, tumor, and cyst volumes and extent of resection is also performed using the Vitrea software.

All patients are uniquely identified in the database. An identifier is assigned, allowing information to be retrieved without any traceable link to the actual identity of the patient (except, of course, by the database administrator/users with permission). There are multiple levels of user privilege so that various classes of users only have access to information appropriate for their roles.

Every attempt is made to collect the data in a prospective fashion, e.g., operating room findings are documented by the neurosurgeons as soon as possible after the surgery. Also, whenever possible, data acquisition from other hospital sources (e.g., hospital registration database) is automated to decrease duplication of effort, reduce error level, and ensure a quick download. Other manual downloads include data from the MD Anderson Tumor Registry, Surgical Indexing, and others. Evaluation of data from additional sources for suitability for potential download is performed on a regular basis, though not always achievable.

## DATA ENTRY

The database contains a minimal number of text fields. Most fields are coded for consistency and ease of entry and to enhance search and retrieval capabilities. Data are entered through a point-and-click approach with drop-down menus. Entries are standardized by precise inclusion criteria and precise definitions noted in a data and database dictionary. The coding structures used have been designed for a maximum flexibility and precision of searches and data analysis. Where applicable, data on a given entity are recorded from an expandable hierarchical set of codes and linked to a number of relevant descriptors. For clarity and ease of data entry, the system was designed to show specific fields on a given screen only when applicable (e.g., fields related to the primary non-NS cancer history only appear for patients with a NS metastasis) or to automatically fill in fields in a hierarchy where appropriate. Standard diagnostic and procedural coding schemes (Systematized Nomenclature of Medicine [SNOMED] and Current Procedural Terminology [CPT] Systems) are also included as an additional coding methodology.

## DATA DICTIONARY

An extensive database dictionary has been developed to serve as a reference guide for database coordinators and other database users. The dictionary includes all details on all data collected, including a description of the fields, their source, definition, and allowable responses. As a simple example: "**Any Treatment field:** Pull Down Menu options are Yes; No; Unk. This field refers to any treatment to the primary cancer at any time up to the patient's first visit to the BTC neurosurgeon. Does not include treatment to the systemic metastases. Get information from history of present illness and past medical history in neurosurgeon's dictation or previous relevant patient dictations/scan documents/or medical chart." Additional notes or unique scenarios are highlighted as a guide for the staff. As an example: "**Note:** A family member with a primary outside the CNS that metastasized to the CNS should not be coded as one with a history of CNS cancer."

The dictionary ensures the consistency and validity of the data stored and of their interpretation. The complete version of the

dictionary is accessible within the database for easy access and is constantly updated to reflect changes.

### DATABASE GOVERNANCE

Strict written policies and procedures and standard forms are in place to govern every database-related aspect including database access and maintenance, data collection, entry, extracting, and quality assurance (**Table 2**). These are reviewed on a regular basis and modified as necessary. All requests for data are tracked and their status is documented.

### DATA QUERIES AND RETRIEVAL (FIGURE 6)

The database includes some built-in reports that are routinely generated from the data. An example of such a report is the number of surgical cases during a given time frame, stratified by surgical procedure, tumor site (brain, skull base, spine, and other), surgeon, or other. The database allows the direct e-mailing of such reports to designated individuals on specified dates. Non-routine or complicated queries are performed by the departmental database programmer on an as needed basis. Reports are formatted for viewing or analysis according to the needs of the user. The database has interfaces with commonly used statistical and data management software. This allows the quick export of pertinent patient data in a standard format. Approval by designated individuals is required for all data extractions. Data to be used for research purposes will only be retrieved and distributed according to a protocol approved by the IRB.

## DATA QUALITY

The most important and most difficult aspect of having a clinical database is ensuring the veracity of the data and its meeting of all applicable quality assurance (QA) standards. This issue is addressed at multiple levels:

1 Hiring of highly-qualified staff members for data extraction and coding. Given the complexity of medical data in general, and data related to the nervous system in particular, close attention was paid to the selection of the staff members, most of whom are non-practicing medical doctors, and to their training on aspects relevant to the database and data collection and coding. The database team includes 4.5 full time equivalents (FTEs) handling clinical data collection and

entry, imaging, data and database management, and general oversight.

	- a. A check of the daily database census against clinic and operating room schedules allows the capture of data on all eligible patients.
	- b. Complex built-in logic checks and validation rules limit a large number of data entry errors/missing data at the time of entry (**Figure 7**).
	- c. Multiple internal consistency checks of the data by designated quality assurance staff reveal missing and erroneous information not detected by the logic checks and validation rules.
	- d. A dedicated "to be resolved" sub-screen on all screens allows for flagging of questionable data and signaling the need for further review by appropriate staff. (**Figure 8**)

### SECURITY AND REGULATORY COMPLIANCE

The database is governed by an IRB-approved protocol. Issues of patient consent and authorization are addressed as appropriate. The database meets all applicable policies and regulations at various levels (Code of Federal Regulations; Good Clinical Practice; the Health Insurance Portability and Accountability Act [HIPAA, a United States law that sets privacy standards for the protection of patients' medical records and other health

#### TABLE 2 | Examples of database policies and procedures in place.

	- Account creation, deletion, modification policy
	- Account creation form
	- Account modification form
	- Account inactivation form
	- Change/correction process policy
	- Change/correction request form
	- Request for individual data retrieval–research purposes
	- Request for data counts–research purposes
	- Request for data counts/retrieval–non-research purposes
	- Request for individual data retrieval–non-research purposes
	- Request for data retrieval–preparation for research purposes

information]; information security policies; other institutional policies). Particularly with regard to HIPAA compliance, the following measures are taken:


#### RESULTS

At the time of preparation of this manuscript, the database had spanned a period of 25 years, with the latest structure being in place for 14 years. It currently houses historical and demographic data; data on disease, tumor, and patient characteristics; and perioperative and other treatment data on close to 27,000 patients [over 34,000 neurosurgical cases with a current annual accrual of around 1,800 cases. Since its inception, the database has been highly utilized within and outside the Department of Neurosurgery for various research, administrative, educational, and other purposes. It has been the source of data for numerous publications and presentations at scientific meetings. These publications and presentations encompass areas of epidemiology, tumor characterization, treatments, and treatment outcomes. As an example, the study by Lacroix et al. (2) evaluated the outcome of 416 patients undergoing varying degrees of glioblastoma resection and required extensive use of the Neurosurgery database. It confirmed the survival advantage of a near total resection (98% or more) compared with a lesser resection. This seminal study has resulted in renewed interest in advancing neurosurgical techniques for the treatment of malignant brain tumors. In 2014, Marko et al. (3) used data from this database to study 721 patients newly diagnosed with glioblastoma (from 1993 to 2010) to construct a mathematical model of factors affecting personalized survival. Their findings argued against a surgical management strategy based on rigid extent-of-resection thresholds and instead provided the first explicit evidence supporting a maximum safe resection approach to glioblastoma surgery. These findings were further bolstered by a study by Li et al. (4), who employed our database to study the influence of maximum safe glioblastoma resection in 1,229 patients. In what is probably the largest single-center series of glioblastoma patients with extensive tumor resections, their study supported the established association between extent of resection and survival and moreover, showed that going beyond a conventional 100% resection (of all contrast-enhancing tumor) by also removing a significant portion of the fluid-attenuated inversion recovery (FLAIR) abnormality region, when safely feasible, may prolong survival without significantly increasing overall or neurological postoperative morbidity. More recently, Al-Holou et al. (5) drew data from our database on 1204 patients with glioblastoma to show that relative to piecemeal resection of these tumors, circumferential perilesional resection is significantly and independently associated with improved outcomes. Clearly, studies such as these on a scale this large would have been impossible without a patient database as robust and extensive as ours, and one that ensured that data typically inconsistently recorded in the medical records, such as method of tumor removal, or data not typically available, such as volumetric perioperative analyses of all brain tumors is regularly and consistently documented. Other select publications are listed in **Table 3**].

Meetings with the data management team members are held regularly as well as on an ad-hoc basis. The meetings serve as a platform for dealing with problematic issues and identifying needs for modification. They aim at encouraging good data management practices and keeping communication open in a friendly unthreatening environment.

Initially, for the first few years, both an informed consent and a HIPAA authorization document needed to be signed by patients, but the IRB later approved waivers for these, given that waiver justifications for both consent and authorization were met.

Funding for the database over the years was provided by departmental funds, two MD Anderson Cancer Center institutional database grants, as well as by designated and undesignated philanthropic donor funds.


### DISCUSSION

Information management is the creation and application of processes directed toward the collection and review of data in a structured and effective manner (13). The ready accessibility of patient, disease, treatment, and outcome data from an informatics infrastructure repository is a major catalyst for the advancement of medical knowledge, as it helps in the rapid translation of clinical and laboratory discoveries into new and better treatments and therapies. As noted under the Methods section, the database is web based, which ensures easy access to users from various physical locations protected by the MD Anderson firewall. It is sufficiently flexible to support multiple tumor and treatment types, and multiple clinical case scenarios. It incorporates elements of a prospective data collection process as well as elements of a point-of-care data entry process (including ease of access and navigation). It provides the necessary and compatible backbone for downloading data from other MD Anderson Cancer Center institutional sources. This relational database allows for user-centered, screen-driven, structured, and efficient data entry, review, and mining, and a role-based security system. It ensures the highest data quality through a large number of external and internal consistency checks, complex built-in logic checks and validation rules, dedicated "to be resolved" fields on every screen, an audit trail, a detailed data dictionary, and extensive training of qualified data management staff, a crucial prerequisite to a high quality product (14–17). It is HIPAAcompliant.

The database task force was valuable to the success of the project. Task force members brought important clinical knowledge and expertise to the group. Traditionally, clinicians may have had little or no control over the development and implementation of patient information systems. Many databases are developed by programmers with a somewhat limited understanding of the clinical issues involved. A closer look at the traditional "Build it and they will come" approach reveals major limitations and an inherent risk of failure. A high level of input in the development process by intended users leads

preventing the input of erroneous data.

to a product that meets the users' needs, has a high level of relevance, acceptance, and utilization and thereby is successful. Participation by the faculty, then, is an element crucial to the longevity of this project.

A key difficulty was the development of a robust data dictionary that clarified the various field definitions, sources, or time frames. This key difficulty was also a key strength of the database, once developed. It continues to be a living breathing document, whereby various aspects are subject to review and adaptation.

Another key difficulty was sustaining the funding. The database is an expensive endeavor, and securing financial support of full-time staff members involved in the day-to-day database activities including data collection, mining, and quality assurance is no small feat. A key strength was the hiring of highly qualified staff members and their retention. Many of the staff members have had a longevity of more than a decade with the department. During times of hardship, and these times are inevitable with a database that spans a long period of time, focus on the core data set and the "bread and butter" was crucial.

In addition to the well-known challenges of developing clinical databases in general (securing the acceptability of end-users, allowing for simple data entry and retrieval methods, and possessing large storage capacity and adequate security safeguards) (18–23), this database posed some unique challenges stemming from its unique aims. These challenges included: (1) Its wide scope (research and administrative purposes, among others); (2) the variety of tumors involved (primary vs. metastatic; brain, spine, skull base, peripheral, and each its own separate entity); and (3) the complexity and multidisciplinary nature of the management approaches, entailing differences in definitions and documentation processes and methodology.

Nevertheless, our group encompassed a wide range of expertise and valuable accumulated experiences. Both of these strengths were essential in overcoming the challenges and securing a successful endeavor. Furthermore, the perceived usefulness and relevance of the database rank high among intended users from various specialties and levels, thereby affording us strong and crucial acceptance and support necessary to overcome future challenges. Although the database represents an efficient and effective approach to handling the data management needs of a sizable multispecialty treatment center, its design and programming methodology can readily be adapted to other healthcare settings.

Single center databases such as the one described in this manuscript are invaluable in terms of the breadth and depth TABLE 3 | Select case series publications from the neurosurgery oncology database, underscoring the opportunity to investigate the impact of novel surgical techniques and novel variables (e.g., objective measure of extent of resection, perilesional resection) on varied neurosurgical outcomes in large cohorts of patients with nervous system tumors.


of information they provide and the research opportunities they present. They are crucial in providing well-annotated and consistently defined and coded data. They have a critical role to play in advancing knowledge on the topics of nervous system tumors and neurosurgical oncology and should be promoted where feasible and sustainable. That said, these databases can be hard and costly to develop and maintain (an element often underestimated during the planning phases), though the cost varies widely depending on factors such as the general infrastructure in place at the healthcare facility, the medical record system in use, the patient load, the database scope, and the caliber of the data management staff involved, to mention a few. These databases may also be limited by potential referral biases and generalization issues. National databases and registries are another valuable source of information and have allowed for valuable scientific studies. In the US, the Central Brain Tumor Registry of the United States (CBTRUS); the Surveillance, Epidemiology, and End Results (SEER); and the National Cancer Data Base (NCDB) are among the major centralized databases for brain tumor information. Each has its unique attributes, as well as its limitations: The CBTRUS includes primary tumor incidence but lacks patient follow-up data. The SEER database has incidence and follow-up data, but only on malignant brain tumors. The NCDB is the largest non-population-based US database, which identifies all newly diagnosed primary brain tumors, both benign and malignant, and has extensive patient treatment and outcome data, but because its data primarily come from hospitals accredited by the American College of Surgeons (and few unaccredited hospitals), this does not allow for tumor incidence rate estimation. None of these three public databases contain reliable or complete information on infrequently coded surgical details such as piecemeal vs. en bloc resection, dural entry and other similar variables, or outcomes such as objective postoperative tumor volume and extent of resection. In this era of big data and with the technologic advances at hand, combining data from multiple sources, single center databases, as well as national registries and databases, administrative datasets, and others, allows for a more powerful and wideencompassing analysis of a wide array of data at every level. But an extensive discussion of this is outside the scope of this manuscript.

### FUTURE DIRECTIONS

With the advent and widespread use of the EMR, the growing wealth of electronic data encompassing all aspects of humanity, and staggering advances in technology and artificial intelligence making their way at an unprecedented and previously unimaginable pace and scope, we are pressed to pay a close look at where we are and how we should proceed in this new era. For example: (1) Wider utilization of data from various sources, such as patient-reported outcomes, the sequencing and "omics" data which have revolutionized the understanding of nervous system tumors and become essential in any research endeavor, and other data not currently in the database, and (2) linking the database to the institutional biobank will be essential to a better understanding of all relevant aspects at play. (3) Wider adaptation of available technology in an appropriate fashion to replace manual labor and connect various sources of data will decrease cost and increase efficiency. Steps that will add those features to the current ones are currently underway.

## AUTHOR CONTRIBUTIONS

The concept of the Neurosurgical Oncology Database was originated by RS and DS. The original manuscript was written by DS, with additions to the manuscript and edits provided by both RS and DW.

## FUNDING

This work was supported by Department of Neurosurgery funds, by two database grants from The University of Texas MD Anderson Cancer Center, and by gifts from philanthropic donors, especially the Anne C. Brooks and Anthony D. Bullock III Distinguished Chair in Neurosurgery endowment fund awarded to RS.

## REFERENCES


of aggressive multimodal management. J Neurosurg. (2015) 123:781–8. doi: 10.3171/2014.10.JNS141037


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Suki, Wildrick and Sawaya. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Quality of Life in Brain Cancer: Clinical Validation of the Mexican-Spanish Version of the EORTC QLQ-BN20 Questionnaire

Bernardo Cacho-Díaz <sup>1</sup> , Nydia A. Lorenzana-Mendoza<sup>1</sup> and Luis F. Oñate-Ocaña<sup>2</sup> \*

<sup>1</sup> Unidad de Neurociencias, Instituto Nacional de Cancerología, Mexico City, Mexico, <sup>2</sup> Subdirección de Investigación Clínica, Instituto Nacional de Cancerología, Mexico City, Mexico

Background: Overall survival (OS) of patients with Brain Cancer (BC) is slowly increasing. The disease itself and its treatments deeply impact patient Health-related quality of life (HRQL). Therefore, valid and reliable instruments are needed. In this study, the Mexican-Spanish version of the QLQ-BN20 instrument is psychometrically and clinically validated.

#### Edited by:

Sujit Prabhu, University of Texas MD Anderson Cancer Center, United States

#### Reviewed by:

Carsten Friedrich, University of Rostock, Germany Yoshua Esquenazi, University of Texas Health Science Center at Houston, United States

> \*Correspondence: Luis F. Oñate-Ocaña lfonate@gmail.com

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Neurology

> Received: 28 July 2018 Accepted: 11 January 2019 Published: 30 January 2019

#### Citation:

Cacho-Díaz B, Lorenzana-Mendoza NA and Oñate-Ocaña LF (2019) Quality of Life in Brain Cancer: Clinical Validation of the Mexican-Spanish Version of the EORTC QLQ-BN20 Questionnaire. Front. Neurol. 10:40. doi: 10.3389/fneur.2019.00040 Methods: Patients with brain cancer (BC) (primary or metastatic) evaluated at a tertiary cancer center, were invited to respond to the questionnaire, as well as the core-module QLQ-C30. Tests to demonstrate the instrument's internal consistency, the association of HRQL scales with clinical variables and OS were investigated.

Results: One hundred and nineteen patients were included in this cohort: 77 women and 42 men (mean age, 46.2 years). Patients answered both instruments in < 30 min. Good convergent [all correlation coefficients (CC) > 0.37] and discriminant validity was observed and was associated with significant overlap (CC 0.007–0.68). All four multi-item scales of QLQ-BN20 also demonstrated good reliability (Cronbach α > 0.7). Several scales of the QLQ-BN20 were significantly associated with performance status and a modified Recursive Partition Analysis. Of the possible scale correlations, 40 of 161 (24.8%) scales in both instruments, were significantly (directly or inversely) correlated. Visual disorders, Motor dysfunction, Seizures and Weakness of the legs presented association with OS (p < 0.05).

Conclusion: The Mexican-Spanish version of the BN20 instrument is valid and reliable and can be used in clinical trials in patients with BC. Some HRQL scales were associated with OS and could therefore be incorporated in future studies of prognostic models.

Keywords: health-related quality of life, brain neoplasms, surveys and questionnaires, validation study, prognostic factors

## INTRODUCTION

Brain cancer (BC) constitutes a heterogeneous group of diseases, accounting for 1–2% of all primary cancers in adults (1). These neoplasms (primary or secondary) are characterized by severe and complex symptoms, usually associated with a poor prognosis (1, 2). There is no definite cure for most patients. Therefore, a reasonable primary aim of treatment is to extend survival with effective symptom relief (2, 3).

**70**

Efforts with emerging new therapeutic strategies are mainly focused on prolonging survival (4). However, BC is associated to symptoms and complications that negatively impact patients' Health-related quality of life (HRQL). BC may directly provoke disabling symptoms including headache, sensory-motor dysfunction, seizures, mood disorders, personality changes, and cognitive dysfunction (5). Consequently, the clinical benefits of treatment should be evaluated not only according to the classical outcome measures (objective response or survival) but also by ensuring HRQL improvement, and must be weighed against treatment side-effects (5–7).

The clinical consequences of disease can be identified by physical examination and evaluated with neurological and neuropsychological tests. Patients' opinions on their own HRQL differ substantially from the opinions of proxies or healthcare personnel. Hence, two instruments have been developed to measure HRQL in the specific case of BC: the FACT-Br, Peds-FACT-BrS, and the European Organization for Research and Treatment of Cancer (EORTC) quality of life questionnaire (QLQ) BN20 (8–11). The MDASI-BT is a symptom inventory designed to measure symptoms and not for HRQL assessment (12).

The QLQ-BN20 measures focus more specifically on function and symptoms, while the FACT-Br assessments cover more psychosocial aspects of the disease. Therefore, the QLQ-BN20 is superior when assessing the treatment outcome and may provide more information in trials that focus on functional endpoints, whereas FACT-Br could be more useful in patients with positive functional capacity but psychosocial concerns, although no instrument is superior to the other (13). Both must be considered as complementary.

The QLQ-BN20 instrument comprises 11 symptom scales that cover the more common complains in patients with BC and it was designed to be used with the core questionnaire QLQ-C30. This core instrument comprises five functional scales, and nine symptom scales plus a global HRQL scale. Both have been translated and validated into several languages and have been extensively used in the medical literature. However, available information on the subject published in Latin-American countries is scarce. The aim of this study was to validate the Mexican-Spanish version of the QLQ-C30 and QLQ-BN20 questionnaires in patients with BC.

### MATERIALS AND METHODS

#### Patients

Patients treated at the Neuro-oncology Unit of the Instituto Nacional de Cancerología (INCan) in Mexico City from February 2005 to October 2014, were invited to participate in this study and respond the questionnaires. Inclusion criteria were: Literate individuals of any gender or age, who had a clinical diagnosis of primary or secondary BC. The diagnosis of BC was established by computed tomography scan and/or magnetic resonance imaging. The clinical history was obtained, as well as blood cytology and chemistry, tumor markers, and chest-X ray. The Karnofsky and ECOG status performance scales were assessed. The Institutional Review Board and Ethics committees approved the study protocol (registration codes 014/007/CCI and CEI 865/14). Patients signed the informed consent form in which the purpose of the study and a safety protection policy were detailed and accompanied the questionnaires.

### Instruments

The EORTC QLQ-C30 consists of 30 items, which are organized into five functional scales (physical, role, emotional, cognitive, and social functioning); three symptom scales (fatigue, nausea and vomiting, and pain); one global health status scale; and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties). The QLQ-BN20 includes 20 items, which are organized into four multi-item scales (future uncertainty, visual disorders, motor dysfunction, and communication deficit) and seven single items (headache, seizures, drowsiness, hair loss, itchy skin, weakness of the legs, and bladder control) (11). The validated Mexican-Spanish version of the QLQ-C30 was used (14). The adapted Mexican-Spanish version of the QLQ-BN20 was pilot-tested in 10 patients with BC to identify the adequacy of the translation. All patients responded the questionnaire without assistance, stated that the questions were clear and easy to understand and complete. Both questionnaires were used with permission of the EORTC Quality of Life Group, and they supervised the entire process.

#### Statistical Analysis

Scale scores were calculated by linear transformation of raw scores into a 0–100 score, with 100 representing best global health, best functional status, or worst symptoms, as described by EORTC (15), and the summary score was also calculated (16). Correlation assessment was obtained with the Spearman Correlation coefficient (CC). Convergent validity was determined calculating the CC between each item and items belonging to their own scale, and the success criteria was CC >0.3. Divergent validity was evaluated calculating the CC between each item and items belonging to other scales, and the success criteria was CC < 0.3. Cronbach's α was used to measure multi-item correlation, and the success criteria was ≥0.7. In addition, clinical validity was evaluated by the extent to which scores were able to discriminate among groups of patients who differed in terms of their clinical status. Patients were classified according to treatment intent, the number of metastases (single, multiple or carcinomatosis) and according to Recursive Partitioning Analysis (RPA) (17). Due to the heterogeneity of primary BC and the limited availability of specific prognostic scores for each of the different types of BC, for the purpose of the study, patients with primary BC were classified using the same RPA approach, as if it were metastatic, considering the absence of neoplasm in sites other than the CNS, age (>65 years) and a Karnofsky score >70 (18, 19).

The Kruskal–Wallis test was used to analyze differences between groups. Scale scores were categorized by terciles. The correlation between the different QLQ-C30 and QLQ-BR20 scales was explored to identify differences and clinical overlapping. Overall survival (OS) was considered as the period of time from diagnosis of BC metastases or primary BC to death. The association of HRQL and OS was evaluated using the Kaplan-Meier method, and differences were tested with the Log rank method. Hazard ratios (HR) were calculated using the Cox model. Sample-size calculation was based on the proposal by Tabachnik and Fidell (20); a minimal ratio of 5 patients per item was required (20 × 5), i.e., a sample size of 100 patients. Any probability value of 0.05 or less was considered significant, and two-tailed statistics were applied in all cases. The SPSS for Mac version 23 software was used for computations (IBM, Inc., Armonk, NY, USA).

### RESULTS

### Patients

One hundred twenty-seven (127) patients were invited to participate, but eight did not consent. Therefore, 119 patients with BC were included in the study. There were 77 women (64.7%) and 42 men (35.3%), with a mean age of 46.18 years (SD, 15.8; range 17–80). Brain metastases were found in 76 patients (63.9%); among these, the most common primary cancer sites were: 25 breast (32.9%), 18 lung (23.7%), four ovary (5.3%), four non-Hodgkin lymphoma, three Hodgkin's lymphoma (3.9%), three cervix-uteri, three renal, three testicle, two melanoma (2.6%), two endometrial, one acute myeloid leukemia (1.3%), one gastric, one prostate, one metastases to spinal cord from a treated medulloblastoma, one meningeal metastases from a treated medulloblastoma, one rectum, one nasopharynx, one adenoid cystic and one basocellular skin cancer. Primary BC was found in 43 patients (36.1%) and the most frequent diagnoses were: 12 meningioma (27.9%), six astrocytoma (13.9%), five CNS primary germinal neoplasms, four primary CNS lymphoma, four medulloblastoma (9.3%), four high-grade glioma, three pituitary macroadenoma, two oligoastrocytoma, one craniopharyngioma, one ependymoma, one gliosarcoma, one hemangiopericytoma, one oligodendroglioma, one meningeal sarcoma and one


n, number of patients; SD, standard deviation; %, is the percentage of patients with floor or ceiling value.

TABLE 2 | Convergent and discriminant validity of scales of the EORTC QLQ-C30 and QLQ-BN20 instruments in patients with (primary or secondary) brain cancer (n = 119).


n, number of patients; <sup>a</sup>Spearman correlation coefficients; <sup>b</sup> success criteria for item own-scale correlations (>0.3); <sup>c</sup> success criteria for item other-scale correlations (<0.3); <sup>d</sup>Cronbach α values. All correlation coefficient values are absolute values.

pinealoblastoma. Among patients with primary BC, one developed breast cancer as a second primary, with brain metastases and meningioma; one patient had a retroauricular mucoepidermoid carcinoma with a metachronous astrocytoma and meningioma; in two patients with primary medulloblastoma, one had meningeal metastases and the other developed spinal cord involvement.

Six patients had a 100% KPS, 39 90%, 41 80%, nine 70% and 27 60% or below KPS. All patients answered both questionnaires in <30 min, and there were 6 missing values (0.1%) in the 50 items of both instruments, including 5,950 possible responses.

### Reliability and Internal Validity

Descriptive statistics of the HRQL data are presented in **Table 1**. Most scales of both instruments have a zero floor and a 100 ceiling values, whereby mean scores mainly represent highfunctional and low-symptom values. The summary of multi-trait scaling analyses is depicted in **Table 2**; good convergent and discriminant validity is observed for most scales. All multi-item scales presented good own-scale correlations. Divergent validity revealed low correlation for other-scale items but also frequent overlapping. The Cognitive and Pain scales of the QLQ-C30 did not show a Cronbach's α coefficient >0.70, but all four multi-item scales of the QLQ-BN20 did.

### Clinical Validity

Many scales of both instruments were significantly associated with clinically relevant factors. **Table 3** describes the association of three categories of ECOG with scale scores of both instruments; nine of Sixteen (including summary score), and four of Eleven scales of the QLQ-C30 and QLQ-BN20, respectively, TABLE 3 | Mean scale scores of QLQ-C30 and QLQ-BN20 depending on the Eastern Cooperative Oncology Group (ECOG) performance status grading in patients with (primary or secondary) brain cancer (n = 119).


n, number of patients; numbers represent means (in parentheses are standard deviation values); p, probability values obtained by Kruskal-Wallis test. Bold values correspond to statistical significant values.

showed significant associations. **Table 4** shows the association of RPA categories (for metastatic and primary BC) with the mean scale scores of both instruments; three of 16 (including summary score), and three of 11 scales of the QLQ-C30 and QLQ-BN20, respectively, did not yield significant associations.

#### Correlations Between Instruments

Forty-two of 176 (23.8%) possible (16 × 11) bivariate correlations between QLQ-C30 and QLQ-BN20 scales were significant (p < 0.05). As expected, correlation between functional and symptom scales were usually negative. The correlation matrix is shown in the **Table 5**.

#### Survival

Median follow-up of the cohort was 4.49 years (SD 3.38) (range 0.21–13.8). During this period, 79 patients (66.4%) died from progressive or recurrent disease. Median OS was 3.98 years (95% CI 2.99–4.97). The bivariate association of HRQL and OS were explored; the physical, role, social, fatigue, nausea/vomiting, pain, dyspnea, appetite loss, financial difficulties scales, and the summary score were associated with OS. Of the QLQ-BN20 scales, Visual disorders (HR 1.01 [95%CI 1.002–1.018]), Motor dysfunction (HR 1.011 [95% CI 1.003–1.019]), Seizures (HR 1.013 [95% CI 1.004–1.023]), and Weakness of the legs (HR 1.013 [95% CI 1.007–1.02]) were associated with OS. The Kaplan-Meier OS curves depending on Visual disorders, Motor dysfunction, Seizures and Weakness of the legs scales are depicted in **Figure 1**.

#### DISCUSSION

In this study, the Mexican-Spanish version of the QLQ-BN20 instrument along with its core instrument QLQ-C30 has been TABLE 4 | Mean scale scores of QLQ-C30 and QLQ-BN20 depending on the RPA class in patients with (primary or secondary) brain cancer (n = 119).


n, number of patients; numbers represent means (in parentheses are standard deviation values); p, probability values obtained by Kruskal-Wallis test. Bold values correspond to statistical significant values.

psychometrically and clinically validated as we found them reliable and valid.

The main traditional outcome measures in oncology research are the frequency of objective responses after therapy, progression-free survival, or OS. However, many brain neoplasms are incurable, and maintenance or improvement of patients' HRQL are, at least, as important as increases in the progression-free survival or OS. On the other hand, a patient-centered approach complementing the decision-making process in Neuro-Oncology is feasible and desirable. Most patients with BC can participate actively in the decisions on their management options if relevant information is presented in a clear and reasonable manner. When informed, most patients are able to identify concepts of HRQL, the capability to maintain functional independence and the influence of treatment on survival as the most relevant factors in determining their decision (21). As physicians, we must be prepared to facilitate this process.

In a 20-year period, only five Randomized clinical trials (RCT) included HRQL evaluations as primary or secondary outcome measurements. However, the quality of reporting HRQL data has not considerably improved (22). In these contexts, the availability of valid instruments to accurately measure HRQL is mandatory. In general terms, the psychometric characteristics of our study were similar to the original report and other validation reports (23, 24).

The original QLQ-BN20 instrument was developed in multilingual and multicultural settings in Europe (including European-Spanish) and has proven to be valid and reliable (11, 25). Our study is the first validation protocol of the instrument in the Mexican-Spanish language, performed in a Cancer Center in Latin America. Similar psychometric findings


n, number of patients; numbers represent means (in parentheses are standard deviation values); FU, future uncertainty; VD, visual disorders; MD, motor dysfunction; CD, communication deficit; HA, headache; SZ, seizures; DW, drowsiness; IS, itchy skin; HL, hair loss; WL, weakness of legs; BC, bladder control.

Non-significant correlations are shown in bold numbers (p > 0.05).

FIGURE 1 | Kaplan-Meier survival curves of the association of health-related quality of life and overall survival constructed depending on (A) Visual disorders, (B) Motor dysfunction, (C) Seizures, and (D) Weakness of legs scales of the QLQ-BN20 instrument (n = 119).

**76**

are described in other validation studies but the association of several HRQL scales and relevant clinical variables has not been previously reported (11, 23–26). In known-group comparisons, the association of the QLQ-BN20 instrument with the ECOG performance status scale, RPA and OS are described.

The main pitfalls of our study are that responsiveness to cancer treatment was not investigated as long as we have performed one HRQL evaluation for each patient, and our relatively small sample size (n = 119), when compared with the original report from the EORTC (n = 891) (11).

Floor and ceiling values are 0–100 in all scales except in the Cognitive scale of the QLQ-C30 (**Table 1**). In general terms, most patients report high functional and low symptom scales (**Table 1**), reflecting a population with a recent diagnosis and low disease burden. Convergent and divergent validity is adequate for both instruments, as shown in **Table 2**, and as similarly reported in the other four validation studies (11, 23–26). The reliability of the Cognitive scale of the QLQ-C30 is below 0.7 and this finding is similar to the Korean validation study (24), while the other three validation studies did not mention the values of the QLQ-C30 scales (11, 23, 25, 26). Low Cronbach α values in terms of Cognitive scale's reliability are frequent in the literature in patients with diverse types of cancer. Examples of this finding include the original EORTC report of the QLQ-C30 instrument (27), and the original Mexican-Spanish validation study of the QLQ-C30 instrument (14). All multi-item scales of the QLQ-BN20 presented fair Cronbach α values as in the other three validation reports (11, 23–26). Certain QLQ-BN20 mean scale scores revealed important associations with ECOG, RPA and OS (**Tables 3**, **4** and **Figure 1**). No comparison is possible because to our knowledge, these findings have not been previously reported.

Although, there is currently no available cure for advanced BC, survival rates have been increasing over the last few years, so this tool is useful in assessing the development of an effective treatment that improves HRQL (5). The QLQ-BN20 instrument has been cited in 42 publications in PubMed since 2009 and has been used in clinical trials to measure HRQL in BC patients undergoing chemotherapy (Ch) and/or radiotherapy (RT). A recent study compared two treatment outcomes in glioblastoma patients: one received RT alone, while the other group of patients was treated with RT plus adjuvant temozolomide; QLQ-C30 and QLQ-BN20 were used to assess the patients at follow-up and results showed similar HRQL in the two groups with minimal differences in the nausea/vomiting and constipation scales, which were worse in the Ch / RT group than in the RT only group. Nevertheless, the use of adjuvant temozolomide therapy further prolonged patient survival compared to the RT only group (28).

In clinical trials, statistically significant changes in HRQL can be observed by increasing the sample size or in the scenario of multiple comparisons (such as comparison of multiple HRQL scale scores). However, these changes may not be clinically relevant. The meaning of the minimal clinical important difference is pertinent in the design of clinical trials, when proposing an adequate sample size and in the correct interpretation of results. The minimal clinical important difference can be defined as the smallest difference in the mean score which is clinically important (as in between groups or paired comparison designs). In a recent study, a decrease of 6.1 units or 13.8 units was required to represent clinically relevant deterioration of the Seizures or Weakness of legs variables, respectively (29).

In another study of BC patients, 5.2 units change represented the minimal clinically important deterioration in the motor dysfunction scale. Similarly, 9.1 units change represented clinically important improvement in the communication deficit (30).

In general terms, the authors consider that any 10-unit change or difference in the mean score represents a clinically important difference.

Most patients do not report problems with the cognitive functioning scale. This problem may result from their sociocultural background. At the INCan hospital, we mainly treat patients with a low income, illiteracy, a low education level and poor working possibilities.

Distinguishing patients with glioma from those with meningioma was not tested in this study because of the great variability of histopathology diagnoses in the cohort. This is a validation study, so we did not test the impact of different treatments on HRQL. This question and others could be investigated in future research studies, including the usefulness of these instruments in revealing subtle differences associated to novel treatments in randomized clinical trials.

#### CONCLUSION

In conclusion, the Mexican-Spanish version of the QLQ-BN20 instrument is a valid and reliable test that can be used in clinical studies that include patients with primary or metastatic BC. Some HRQL items were associated with the OS and could be used as prognostic factors or might contribute to assemble prognostic models as aids in treatment trade-offs.

### AUTHOR CONTRIBUTIONS

BC-D and LO-O conception and design of the study and analysis and interpretation of data and draft and revise the article. NL-M recruiting patients, acquisition of data, cleaning of the database. All authors have approved the final version to be submitted.

### FUNDING

The INCan financially supported this study in its entirety.

### ACKNOWLEDGMENTS

Many thanks to Professor Deborah Dion Aleman-Hoey for her kind help and English language editorial review.

### REFERENCES


metastases: an analysis of 1,960 patients in the RTOG database. Int J Radiat Oncol Biol Phys. (2008) 70:510–4. doi: 10.1016/j.ijrobp.2007.06.074


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Cacho-Díaz, Lorenzana-Mendoza and Oñate-Ocaña. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Temozolomide Induced Hypermutation in Glioma: Evolutionary Mechanisms and Therapeutic Opportunities

Paul Daniel <sup>1</sup> , Siham Sabri 2,3, Ahmad Chaddad1,4, Brian Meehan<sup>5</sup> , Bertrand Jean-Claude3,6, Janusz Rak 3,5† and Bassam S. Abdulkarim1,3 \* †

<sup>1</sup> Division of Radiation Oncology, Department of Oncology, McGill University, Montréal, QC, Canada, <sup>2</sup> Department of Pathology, McGill University, Montréal, QC, Canada, <sup>3</sup> Research Institute of the McGill University Health Centre, Montréal, QC, Canada, <sup>4</sup> The Laboratory for Imagery, Vision and Artificial Intelligence, École de Technologie Supérieure, Montréal, QC, Canada, <sup>5</sup> Department of Pediatrics, McGill University, Montréal, QC, Canada, <sup>6</sup> Department of Medicine, McGill University, Montréal, QC, Canada

#### Edited by:

Marcos Vinicius Calfat Maldaun, Hospital Sírio-Libanês, Brazil

#### Reviewed by:

Kamalakannan Palanichamy, The Ohio State University, United States Sunit Das, St. Michael's Hospital, Canada

\*Correspondence: Bassam S. Abdulkarim bassam.abdulkarim@mcgill.ca

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology

> Received: 26 November 2018 Accepted: 16 January 2019 Published: 04 February 2019

#### Citation:

Daniel P, Sabri S, Chaddad A, Meehan B, Jean-Claude B, Rak J and Abdulkarim BS (2019) Temozolomide Induced Hypermutation in Glioma: Evolutionary Mechanisms and Therapeutic Opportunities. Front. Oncol. 9:41. doi: 10.3389/fonc.2019.00041 Glioma are the most common type of malignant brain tumor, with glioblastoma (GBM) representing the most common and most lethal type of glioma. Surgical resection followed by radiotherapy and chemotherapy using the alkylating agent Temozolomide (TMZ) remain the mainstay of treatment for glioma. While this multimodal regimen is sufficient to temporarily eliminate the bulk of the tumor mass, recurrence is inevitable and often poses major challenges for clinical management due to treatment resistance and failure to respond to targeted therapies. Improved tumor profiling capacity has enabled characterization of the genomic landscape of gliomas with the overarching goal to identify clinically relevant subtypes and inform treatment decisions. Increased tumor mutational load has been shown to correlate with higher levels of neoantigens and is indicative of the potential to induce a durable response to immunotherapy. Following treatment with TMZ, a subset of glioma has been identified to recur with increased tumor mutational load. These hypermutant recurrent glioma represent a subtype of recurrence with unique molecular vulnerabilities. In this review, we will elaborate on the current knowledge regarding the evolution of hypermutation in gliomas and the potential therapeutic opportunities that arise with TMZ-induced hypermutation in gliomas.

Keywords: glioma, recurrence, MGMT, hypermutation, temozolomide

## INTRODUCTION

Glioma refers to a group of malignant brain tumors comprised of oligodendroglioma, anaplastic astrocytoma, and glioblastoma (GBM) (1). Amongst gliomas, GBM is the most commonly diagnosed malignant primary brain tumor making up 54% of all gliomas and 16% of all primary brain tumors (2). GBM is also the most lethal brain tumor with a median survival of only 15 months (2–4). GBM can be further stratified into IDH1 wild-type (90%) and IDH1 mutant (10%). Patients with IDH1 mutations are thought to comprise secondary GBM as they are enriched for ATRX mutations similar to that found in lower-grade glioma and patients survive for a median survival of 31 months which is consistent with lower-grade glioma (5).

**79**

The emergence of next-generation sequencing and characterization of the genome, epigenome, and transcriptome of GBMs revealed an increasing level of disease complexity. For example, it is now known that GBM is not a homogenous disease but comprises 3 distinct subtypes; Proneural (PN), Mesenchymal (MES), and Classical (CL) which are able to classify both primary and recurrent disease (6–9). Comparison of these molecular subtypes shows that each subtype is enriched for unique molecular alterations. PN is enriched for aberrations in the gene expression of platelet-derived growth factor receptor alpha (PDGFRA) and TP53 mutations, whilst MES and CL are enriched for neurofibromatosis type I (NF1), and epidermal growth factor receptor (EGFR) alterations, respectively (8, 9). Similarly, each subtype has also been found to be associated with a specific local immune tumor microenvironment. By comparing immune signatures between each subtype, MES tumors were found to be enriched for macrophages and neutrophil signatures (6). In contrast, PN tumors exhibit suppression of CD4<sup>+</sup> T-cell signature while CL tumors were enriched for dendritic cell signatures (6). Studies showing that an individual tumor is comprised of cells from multiple subtypes (i.e., intra-tumoral heterogeneity) have added additional complexity to this picture of heterogeneity (10, 11).

The current treatment regimen for primary GBM implemented since 2005 involves surgical resection followed by concurrent chemoradiation (12). This aggressive upfront trimodal regimen improved 2-years survival from 10% for treatment with radiotherapy (RT) alone to 27% compared to treatment with RT and Temozolomide (TMZ) (12). Despite this improvement, recurrence following treatment remains inevitable, typically occurring within months following completion of treatment at first diagnosis and is ultimately lethal as recurrent disease shows limited response to further chemoradiation. To date, there are currently no known therapies, which provide substantial survival benefit to GBM patients at recurrence, urging investigation into alternative treatment options. Understanding the mechanisms underlying response and emergent resistance to chemotherapy is therefore of utmost importance to inform decisions about the next generation of therapies.

### THE CYTOTOXIC EFFECT OF TEMOZOLOMIDE

TMZ, the main chemotherapy utilized for glioma, is an alkylating pro-drug which methylates DNA at the O<sup>6</sup> position of guanine (13). During DNA replication, the maintenance of this methyl-adduct causes a mismatch pairing of guanine with thymine rather than cytosine leading to genomic instability and eventually cell death (13, 14). Two major mechanisms oppose the cytotoxic action of TMZ (**Figure 1A**). O<sup>6</sup> methylguanine-DNA methyltransferase (MGMT), a suicide enzyme able to sequester the methyl-adduct from O<sup>6</sup> guanine through covalent transfer, effectively repairs the alteration prior to replication (15). Consistent with this role for MGMT in driving resistance, MGMT promoter methylation, which is an indirect measure of the ability for cells to express the MGMT protein, is one of the strongest predictors of response to TMZ. Comparison of patient cohorts treated with chemoradiation revealed that those with MGMT methylation survive a median on 21.7 months compared to a median survival of just 15.3 months for patients without MGMT methylation (16).

In the absence of MGMT expression, resultant base mismatches invoke the mismatch repair (MMR) pathway. MMR proteins including MSH2, MSH6, MLH1, and PMS2 recognize and bind to the mismatched guanine and cause cells to enter a cycle of DNA repair (13, 17). Mismatches in newly synthesized daughter DNA strands are repaired whilst methyl adducts persist on parental DNA in the absence of MGMT. This leads to a cycle of futile repair followed by mismatching which eventually induces DNA double strand break formation, cell arrest and death (13). MMR capacity is therefore essential to repairing TMZ-induced toxicity. Consistent with this, comparison of MMR protein expression in 80 matched primary and recurrent GBM specimens treated with chemoradiation revealed consistent downregulation of MMR repair genes in recurrent GBM, highlighting the importance of MMR in dictating response to TMZ (18).

### RECURRENT GLIOMA AND EMERGENCE OF HYPERMUTATION

The failure of chemoradiation, culminating with the inevitability of recurrence and acquisition of a chemo-resistant phenotype have spurred investigations into novel approaches toward treating recurrent disease. An emerging paradigm for finding effective treatments for recurrent GBM is targeted therapy, where treatment is specifically directed against driver alterations necessary for maintenance of malignant phenotypes. Increased availability and reduced cost of sequencing allowed interrogation of the molecular landscape of disease and identification of clinically relevant "subtypes" spurring further interest in targeted therapies. Notably, subtyping of disease states has been a major advancement in simplifying inter-tumoral heterogeneity whilst facilitating the identification of a targetable subset of patients sharing common molecular features. A recent remarkable finding from several longitudinal observational studies comparing pre- and post-treatment glioma has now established that at least two distinct genomic outcomes exist at recurrence; hypermutant and non-hypermutant recurrence (7, 19–23). For hypermutant recurrent tumors, the hallmark identifiers include (i) fold increases in subclonal mutations across the whole genome, (ii) enrichment of the C:G>T:A mutational signature indicative of TMZ mutagenesis, and (iii) gain of inactivating mutations in MMR pathway components (7). In comparison, non-hypermutant recurrent tumors do not exhibit any of these features but instead maintain a similar level of tumor mutational burden (TMB) compared to the primary tumor. Greater understanding of the processes which dictate emergence of these subtypes and the underlying molecular mechanisms responsible for maintenance of malignant features will likely be essential for the identification of targeted therapies against each subtype.

Attempts to understand the processes responsible for emergence of a hypermutant state have thus far been limited to observational studies of glioma, from primary to recurrent states. Although initially observed to be associated with malignant transformation of low-grade to high-grade glioma, TMZ-induced hypermutation has now been observed to occur in grade IV GBM, albeit occurring at a lower frequency compared to low grade (19, 21, 22). Specifically, recent reports from the largest observation study to date found that whilst only ∼10% of GBM display hypermutation at recurrence, a much higher proportion of low grade gliomas emerge as hypermutant following treatment with TMZ (7, 22). This differential capacity for low and high-grade glioma to evolve toward hypermutation raises several questions related to the exact mechanisms which dictate predisposition to undergo mutagenesis.

### O-6-METHYLGUANINE-DNA METHYLTRANSFERASE (MGMT)

To date, MGMT promoter methylation is the strongest correlative feature which predicts GBM hypermutation at recurrence (7). Unsurprisingly, the importance of MGMT in preventing mismatch pairing during DNA replication due to removal of methyl adducts is likely responsible for this relationship to hypermutation. It is interesting to note however that reports vary regarding the correlation between MGMT protein expression and MGMT promoter methylation, making it plausible that methylation status of MGMT may play a surrogate role as a biomarker of a state predisposed toward hypermutation (24, 25). One potential alternative is the observation that MGMT promoter methylation may be indicative of a global hypermethylated phenotype. Indeed, epigenetic features have now been found to impact the ability for DNA repair to take place, as evidenced by the differential rates of mutagenesis predicated by chromatin accessibility (26). Supporting the impact that epigenetic features can have upon hypermutation, low grade gliomas exhibit a global higher hypermethylated state compared to GBM whilst undergoing hypermutation at a much higher frequency despite exhibiting similar proportions of MGMT promoter methylation at first diagnosis (27). Whether hypermethylated subtypes of GBM, defined by IDH1 mutation and a global CpG island methylation phenotype G-CIMP, undergo hypermutation at a higher rate than IDH1 wild-type non-G-CIMP GBM, which are comparatively hypomethylated is not yet known. This would provide critical information regarding the role that epigenetic status plays in dictating the emergence of hypermutation (28).

### MISMATCH REPAIR PROTEINS

Along with MGMT methylation, acquired mutations in MMR genes have been strongly correlated with hypermutation (7, 19, 21). As described earlier, MMR represents the main mechanism by which mismatched bases are repaired and downregulation of these genes is a common mechanism by which gliomas acquire resistance to TMZ. Indeed, minimal loss in MMR genes MSH2 and MSH6 have been shown to be sufficient to provide substantial survival benefit to cells (29). It is important to note however that whilst MMR gene downregulation is common to recurrent glioma and may represent a convergent mechanism of acquired chemoresistance, MMR pathway mutation seems specifically enriched in hypermutant recurrence(7).

One explanation for why this may occur is the observation that MMR proteins are involved in alternate repair pathways which may contribute to resistance to mutagenesis. For example, non-homologous end joining (NHEJ) and homologous recombination (HR) are alternative DNA repair pathways which may drive repair of double strand breaks (DSB) following TMZ treatment. The MMR protein MSH6 directly binds to KU70, a regulatory subunit of the DNA-dependant protein kinase involved in NHEJ and HR repair (30). Similarly, MSH2 is a critical component of the BRCA1 associated genome surveillance complex that recognizes and initiates response to abnormal DNA structures (31). Consistent with the role of MMR is alternate DNA repair pathways, MSH6 knockout cells display impaired NHEJ typified by accumulation of persistent DSBs (30). Together these findings suggest that the complete loss of function of MMR proteins may be necessary for reduced capacity to repair DNA defects, allowing emergence of hypermutation.

#### ADDITIONAL INFLUENCES UPON THE EMERGENCE OF HYPERMUTATION

The majority of GBM tumors will recur as non-hypermutant displaying no overt signs of widespread mutagenesis following TMZ treatment (7). This occurs despite MGMT expression being reportedly stable between primary and recurrent disease (32), suggesting additional features are responsible for protecting from the mutagenesis imposed by TMZ. TMZ cytotoxic effects are dependent on the active proliferation of cells and rapid DNA replication to induce mismatches recognized by MMR machinery, which ultimately leads to either repair, death, or hypermutation (33). This implies that slower proliferation or dormancy would be protective against the acquisition of genomic hypermutation from TMZ treatment (**Figure 1B**).

There is accumulating evidence suggesting that solid tumors are driven by a minor population of stem cells (34). These multipotent stem cells give rise to rapidly proliferating progenitor cells, which are the major drivers of immediate tumor growth. Importantly, stem cells are phenotypically distinct from more differentiated progeny, characterized by their multipotent differentiation capacity, slower proliferation and greater drug efflux activity (34, 35). Together these likely contribute to the reportedly enhanced ability for stem cells to survive chemotherapy compared to their more differentiated progeny. Consistent with this, recent evidence has demonstrated that brain tumor stem cells (BTSCs) are responsible for initiating recurrence (36). Following completion of therapy, surviving stem populations exit dormancy and drives repopulation of the tumor mass. Notably, in this hierarchical model of tumor growth, only mutagenesis acquired by these stem cell populations will be represented in the recurrent tumor. Stem cells which emerge from dormancy following treatment result in recurrence with no observable hypermutation, an outcome independent from MGMT expression. It should also be noted that chemotherapy has been proposed to promote acquisition of stem-like features by differentiated cells (37, 38). As such, the traditional hierarchical model of differentiation and growth is likely insufficient to fully describe the disease state. Further investigation of this process of dedifferentiation is needed to understand its impact on the acquisition of hypermutation state at recurrence.

An additional feature which may impact the emergence of hypermutation is the concurrent use of radiation alongside TMZ for glioma. Indeed, radiation has been shown to be able to induce the expression of MGMT (39). Similarly, radiation has the capacity to drive transient growth arrest which as outlined above, may provide temporary resistance to TMZ induced mutagenesis (40). Whilst sufficiently powered datasets are not yet available for recurrent GBM to make definite conclusions regarding this, what is clear is that hypermutant tumors are able to emerge in patients which receive concurrent radiation and TMZ. We predict that generation of animal models exploring the modalities in clonal xenografts and assessment of incidence of hypermutation from using individual and combined treatment modalities will be able to elucidate the exact role of radiation in hypermutation and facilitate additional exploration into alternate means to drive evolution toward specific outcomes.

### THERAPEUTIC OPPORTUNITIES IN HYPERMUTANT RECURRENT GBM

Both the diversity of genomic alterations as well as the underlying mechanisms which facilitate acquisition of hypermutation make it likely that the approach toward treating hypermutant tumors may be completely different from that of non-hypermutant recurrence. For example, the mutation of MMR genes is specifically observed in hypermutant but not non-hypermutant recurrent tumours (7, 22). Similarly, an increased neo-antigen load linked to the higher global tumor mutational burden is observed exclusively in hypermutant tumors. Exploitation of these unique features of recurrent GBM may provide the means to personalize patient treatment.

### CHECKPOINT INHIBITORS

Immune checkpoint inhibitors (ICI) have shown great promise in the treatment of many diseases. High mutational burden has been identified as the best predictor of response to this treatment option, regardless of disease (41). This has culminated in the recent approval of the Programmed Death Protein 1 (PD-1) inhibitor pembrolizumab for use in all MMR-deficient or microsatellite instability (MSI)-high tumors. Of note, whilst the link between MMR-deficiency and hypermutation has been observed in several longitudinal studies (7, 20, 22), somewhat counter intuitively MSI is not associated with TMZ-mediated hypermutation in GBM (22). Regardless, case reports have suggested the capacity for ICI to be used for GBM in cases with hypermutation. For example, the treatment of GBM patients with germline POLE mutations driving a biallelic mismatch repair deficiency (bMMRd) phenotype characterized by hypermutation of the genome with pembrolizumab was reported to drive radiologically measured tumor regression (42, 43). However, the poor outcomes from the most recent Checkmate 143 trial (NCT02017717) which tested nivolumab and ipilimumab has moderated expectations with no improvement of survival in patients diagnosed with recurrent GBM. Importantly, this trial did not integrate any biomarkers of immunotherapy response such as mutational burden or T-cell inflamed gene expression profiling and as such, retrospective analyses following completion of this trial will likely be able to identify a subset of patients with the potential to respond to immunotherapy.

One feature of glioma which is thought to limit the impact of immunotherapy is the "cold" tumor immune environment of this disease. The blood brain barrier (BBB) comprises a system of pericytes, endothelial cells, astrocytic processes, and basement membrane, which has long been thought to prevent the movement of immune cells to the brain and contribute to the low immunogenicity of glioma. However, leukocyte trafficking across the BBB has been known to play an essential role in the control of several neurodegenerative and infectious diseases (44). Similarly, the paradigm of an intact BBB enforcing an immune privileged environment in GBM is increasingly being challenged as more evidence accumulates demonstrating that the BBB can be severely disrupted during disease in addition to observation of prevalent immune infiltration into the tumour (6, 45). As such, it is likely that the failure of immunotherapy is not limited by the BBB but instead due to intrinsic features of the local immune microenvironment such as immunesuppressive M2 macrophages or T-cell exhaustion which prevent robust immune surveillance and reactivity (46). Targeting these immunomodulatory cells is currently underway and may pave the way toward increasing the efficacy of immunotherapy in GBM.

### COMBINING IMMUNOTHERAPY WITH RADIOTHERAPY

A large number of studies have investigated the complex interaction between radiation and the immune system, which has led to the emergence of the radio-immunobiology (47). The most prominent example of this interaction is observation of the "abscopal effect" where irradiation against a primary tumor results in the regression of metastatic clones in disparate areas of the body. This is now understood to be mediated by the immune system: as tumor neo-antigens are released by dying cells, they are taken up by dendritic cells followed by systemic activation of T-cell responses which continue in a feedforward fashion leading to tumor control (48). It should be noted that induced immune response is greatly influenced by dose given per fraction, where increasing dose generates decreasing immune responses. For example, RT has been found to cause upregulation of immunosuppressive factors such as PD-L1 (49). Similarly, immunosuppressive M2 macrophages are more radioresistant than immune-promotive M1 macrophages and targeted radiation has been observed to cause a shift toward a M2 microenvironment in GBM (50). The complex role of radiation has now led to design of companion radio-immunotherapeutics (RIT) which acts to restrict the immunosuppressive effects of radiation. Furthermore, additional parameters such as the total dose of radiation, dose per fraction, and chronological sequencing of radiation and ICI warrant further investigation as to how they can be utilized to sustain immune cell infiltration into the tumor.

### RADIATION AND HYPERMUTATION AS A SYNTHETIC LETHAL COMBINATION

In addition to the immunological relevance of hypermutation, the co-occurrence of MMR gene mutation in TMZ-driven hypermutant tumors may offer additional opportunities for exploitation (7, 20, 22). For example, PMS2 knockout cells demonstrate a 4-fold increase in mutations following treatment with radiation compared to their wild-type counterparts (51). Models of population fitness suggest that an increased mutation rate can be beneficial up to a point, beyond which further mutagenesis becomes detrimental due to accumulation of deleterious alterations. Accordingly, mouse models of genomic instability demonstrate a decreased tumor growth upon elevation of mutational burden (52). In the context of hypermutation, increasing the number of mutations in hypermutant cells following RT treatment may lead to reduced fitness, making them less aggressive and more amenable toward additional treatments.

### CONCLUSION AND FUTURE PERSPECTIVES ON HYPERMUTATION

It is now accepted that hypermutation represents a distinct subtype of recurrent glioma. However, several questions remain unresolved before we can start to understand the impact of hypermutation at recurrence. What is the mechanism behind emergence of hypermutation and is this process targetable? Is hypermutation in GBM associated with better or worse outcome for patients? Does immunotherapy represent a valid therapeutic approach for hypermutant tumors and can this be combined with radiotherapy? We predict that studies seeking to identify the underlying molecular features of hypermutant and nonhypermutant recurrent subtypes will likely pave the way for novel treatment approaches for recurrent GBM.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### REFERENCES


## ACKNOWLEDGMENTS

This work was performed with support from Varian Medical Systems. Thank you to Henri R. Shibata and the Cedars Cancer Foundation for fellowship support of PD.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Daniel, Sabri, Chaddad, Meehan, Jean-Claude, Rak and Abdulkarim. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Health Related Quality of Life in Adult Low and High-Grade Glioma Patients Using the National Institutes of Health Patient Reported Outcomes Measurement Information System (PROMIS) and Neuro-QOL Assessments

#### Edited by:

Luis Souhami, McGill University, Canada

#### Reviewed by:

Mariko D. DeWire, Cincinnati Children's Hospital Medical Center, United States David D. Eisenstat, University of Alberta, Canada

#### \*Correspondence:

Shawn L. Hervey-Jumper Shawn.Hervey-Jumper@ucsf.edu

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Neurology

> Received: 30 September 2018 Accepted: 19 February 2019 Published: 15 March 2019

#### Citation:

Gabel N, Altshuler DB, Brezzell A, Briceño EM, Boileau NR, Miklja Z, Kluin K, Ferguson T, McMurray K, Wang L, Smith SR, Carlozzi NE and Hervey-Jumper SL (2019) Health Related Quality of Life in Adult Low and High-Grade Glioma Patients Using the National Institutes of Health Patient Reported Outcomes Measurement Information System (PROMIS) and Neuro-QOL Assessments. Front. Neurol. 10:212. doi: 10.3389/fneur.2019.00212 Nicolette Gabel <sup>1</sup> , David B. Altshuler <sup>2</sup> , Amanda Brezzell <sup>2</sup> , Emily M. Briceño<sup>1</sup> , Nicholas R. Boileau<sup>1</sup> , Zachary Miklja<sup>2</sup> , Karen Kluin3,4, Thomas Ferguson<sup>2</sup> , Kaitlin McMurray <sup>2</sup> , Lin Wang<sup>2</sup> , Sean R. Smith<sup>1</sup> , Noelle E. Carlozzi <sup>1</sup> and Shawn L. Hervey-Jumper <sup>2</sup> \*

<sup>1</sup> Department of Physical Medicine and Rehabilitation, University of Michigan, Ann Arbor, MI, United States, <sup>2</sup> Department of Neurosurgery, University of Michigan, Ann Arbor, MI, United States, <sup>3</sup> Department of Neurology, University of Michigan, Ann Arbor, MI, United States, <sup>4</sup> Department of Speech-Language Pathology, University of Michigan, Ann Arbor, MI, United States

Health related quality of life (HRQOL) measures have become increasingly important in the management of glioma patients in both research and clinical practice settings. Functional impairment is common in low-grade and high-grade glioma patients as the disease has both oncological and neurological manifestations. Natural disease history as well as medical or surgical treatment can negatively influence HRQOL. There are no universal standards for HRQOL assessment in glioma patients. In this study, we examine patient perspectives on functional outcome domains and report the prevalence of impairments rates using the National Institutes of Health (NIH) Patient Reported Outcomes Measurement Information System (PROMIS) and Neuro-QOL item banks as measures of HRQOL. Retrospective analysis of a prospectively collected dataset involving 79 glioma patients reveals that quality of life concerns are the most important consideration behind making decisions about treatment in 80.7% of patients. The prevalence of functional impairment by PROMIS and NEURO-QOL assessment is high, ranging from 28.6% in the physical function domain to 43.9% in the cognitive function domain. Pain and anxiety related to physical decline is higher in LGG patients compared to HGG patients. Aphasia severity also impacts HRQOL. The results of this study suggest that the PROMIS and NEURO-QOL assessments may be important HRQOL metrics for future use in larger clinical research and clinical trial settings.

Keywords: glioma, health-related quality of life, language, neuro-rehabilitation, PROMIS, Neuro-QOL, astrocytoma, glioblastoma

### INTRODUCTION

Gliomas are the most frequent primary brain tumor in adults (1). There are currently more than 700,000 people living with a primary central nervous system tumor in the United States. Despite relatively low incidence, gliomas result in a disproportionate share of cancer morbidity and mortality. Brain tumors account for the highest number of years of life lost when compared to non-CNS cancers (2). Despite treatment with maximal safe surgical resection with or without adjuvant chemoradiation, overall survival has remained largely unchanged. Survival is approximately 14 months for glioblastoma and 6 to 15 years for those with WHO II and III glioma (LGG) depending on the genetic profile of the tumor.

Health-related quality of life (HRQOL) metrics have become increasingly important in brain tumor research alongside standard patient outcome measures such as progression-free and overall survival. There are several validated HRQOL assessments used in clinical practice and clinical trials research. Continued efforts to develop and implement HRQOL measurements are needed as research study and clinical endpoints. The relationship between HRQOL and survival in adult glioma is poorly understood. The World Health Organizations' (WHO) International Classification of Functioning Disability and Health (WHO 2010) defines HRQOL based on the following functional domains: physical, social, emotional well-being, and relational (3–5). In the glioma patient population, both disease progression and treatment related effects have been shown to negatively impact HRQOL (6–9).

While HRQOL metrics continue to become incorporated in clinical practice and clinical trials research, there is no consensus regarding assessment measures. The objective of this study was to evaluate patient perspectives on functional domain affecting health related quality of life. We then applied the National Institutes of Health (NIH) Patient Reported Outcomes Measurement Information System (PROMIS) and Quality of Life in Neurological Disorders (Neuro-QOL) instruments as subjective HRQOL patient-reported outcome (PRO) measures in an adult low- and high-grade glioma patient population. Although there are some overlapping domains, PROMIS was developed for use across the general population with multiple chronic health conditions, whereas Neuro-QoL was focused on developing measures that represent HRQOL domains that are specific to neurological disorders (specifically stroke, Parkinson's disease, multiple sclerosis, child and adult epilepsy, amyotrophic lateral sclerosis, and muscular dystrophy). Therefore, the calibration samples and content, while often overlapping, are different for each tool. These tools, PROMIS and Neuro-QoL, may be useful in assessing patient HRQOL and may be an important component of a multidisciplinary treatment approach for glioma patients. Many HRQOL factors may be common to patients with any cancer diagnosis (e.g., pain, emotional distress, sleep disturbance, etc.) and adequately assessed with PROMIS measures; yet, some domains of HRQOL are likely to be uniquely impacted by neurological changes associated with glioma, including cognitive and behavioral functioning in daily life, and may therefore be better captured by Neuro-QOL. To the authors' knowledge this is the first study employing the use of PROMIS and Neuro-QOL prospectively in a cohort of adult low and high-grade gliomas.

### METHODS

The study design involved retrospective analysis of a prospectively collected HRQOL single institution data registry. Participants were recruited at the time of an initial clinic visit following the diagnosis of a presumed glioma. Patients remained enrolled in the study after histopathologic confirmation of a new WHO grade I–IV glioma. Exclusion criteria included age <18 and language and/or neurocognitive dysfunction limiting patient ability to complete PROMIS and Neuro-QOL questionnaires. Aphasia was assessed by the Boston Diagnostic Aphasia Examination (BDAE).

All patients were administered the Montreal Cognitive Assessment (MoCA), a screening instrument developed to estimate global cognitive ability in the service of detecting mild cognitive impairment and dementia (10). Several studies have demonstrated its utility in brain tumor populations, wherein it has been shown to have superior sensitivity compared to other screening instruments (11), is correlated with quality of life measures (12), and predicts median overall survival (13).

To examine patient preferences on functional domains and HRQOL, structured interviews were conducted focusing on how patients frame functional and cognitive domains with their disease experiences based on methodology established by Mortensen and Jakobsen (14, 15). Analysis of these semistructured interviews was used to identify those functional domains considered important to individual glioma patients, which were then developed into a study questionnaire using a Likert scale to identify each domain as extremely important, important, neutral, somewhat important, or not at all important.

Study participants completed PROMIS version 1.0 and Neuro-QOL version 1.1 as HRQOL measures. Examined HRQOL functional domains included Neuro-QOL cognition, PROMIS physical functioning, and PROMIS ability to participate in social roles and activities; impairment domains include PROMIS pain, sleep, fatigue, depression, anxiety, and Neuro-QOL emotional/behavioral dyscontrol. PROMIS Physical Function assesses self-reported (not actual) ability to perform with one's lower extremities (e.g., walking), upper extremities (e.g., dexterity), back, and neck, and to engage in instrumental activities of daily living (16). PROMIS Anxiety assesses anxiety symptoms, including hyperarousal and fear (17, 18). PROMIS Depression measures feelings of worthlessness and sadness among other symptoms of depression. PROMIS Fatigue measures the intensity and impact of fatigue on quality of life (17, 18). PROMIS Sleep disturbances measures perceived quality, adequacy, and satisfaction with sleep as well as difficulties falling asleep and staying asleep (19, 20). PROMIS Ability to participate in social roles and activities measures one's reported ability to participate and be involved in social roles and activities (17, 18). PROMIS Pain interferences assesses the impact of pain on physical, emotional, and recreational activities (17, 18). PROMIS Pain intensity instrument assesses how much a patient hurts. The pain intensity short form is global (i.e. not site specific) and universal rather than disease specific (16). Neuro-QoL Cognitive function measures perceived executive functioning and memory difficulties (20, 21). Neuro-QOL Emotional and behavioral dyscontrol assesses emotionality and impulsivity (20, 21) Normalized mean t-scores for each domain were standardized to 50. For functional domains, higher scores indicate less distress (score >50 more desirable); for impairment domains, higher scores indicate more distress, higher scores indicate more distress (score <50 more desirable). Analyses were conducted using SAS 9.4 statistical software. Independent t-tests were conducted to assess differences between LGG (WHO grades I-II) and HGG (WHO grades III-IV) groups. Prevalence of impairment was assessed in the study population where patients who scored >1 standard deviation beyond the normative mean was considered impaired.

Language assessments were performed by a certified Speech pathologist using the Boston Diagnostic Aphasia Examination Severity rating (BDAE) (22, 23). All assessments were performed in a noise controlled clinical examination room according to standard protocol. BDAE aphasia severity scores reflect the ability to communicate wants, needs, ideas with or without help from listener. BDAE severity scores ranging from 1 to 2 were considered severe aphasia (1 = severe, 2 = moderatelysevere). Scores 3–5 were categorized as mild-moderate aphasia (3 = moderate, 4 = mild, 5 = trace) (22). Study inclusion required BDAE aphasia severity score ≥1 (22, 23).

Univariate analyses were conducted to assess differences between HGG and LGG for each of the PROMIS and Neuro-QOL measures. Partial eta-squared (η 2 ) effect sizes were examined to determine the proportion of variance in HRQOL that was accounted for by tumor-grade (small = 0.01, moderate = 0.09, and large = 0.25) (24). An independent t-test was performed to compare PROMIS and Neuro-QOL scores among patients according to language dysfunction categorized as mild aphasia (BDAE 3-5) vs. severe aphasia (BDAE 1–2).

### RESULTS

Seventy-nine patients were eligible for inclusion. Of the 79 patients, 58 had HGG and 21 had LGG. Average patient age was 52 years (SD = 15.6). Global cognitive status was not different between LGG and HGG patient cohorts by the Montreal Cognitive Assessment (MOCA) (mean HGG = 21.8, LGG = 24.8; P = 0.114). Additional population characteristics are found in **Table 1**.

### Patient Perspectives on Functional Domains of Importance

Study subjects were surveyed about the importance of quality of life concerns vs. survival on medical decision making at the time of diagnosis. Among LGG patients, 81.3% indicated treatment strategies based on quality of life concerns, while 78.6% of HGG patients were concerned primarily with quality of life (P = 0.69) (**Table 2**). The functional domains of greatest concern and importance were language, motor, and memory. There were no differences between LGG and HGG patients in these domains; however, HGG patients placed higher importance on creativity/problem-solving and art domains compared to LGG patients (P = 0.009).

### HRQOL Functional and Impairment Domains Using PROMIS and Neuro-QOL

Prevalence of impairment for the HGG cohort was elevated for PROMIS physical functioning (46.6%), NEURO-QOL cognitive dysfunction (43.9%), PROMIS ability to participate in social roles and activities (28.6%), and PROMIS anxiety (27.6%) (**Figure 1**). Clinical impairment rates for the LGG cohort were elevated for PROMIS pain interferences (38.1%), PROMIS physical functioning (28.6%), PROMIS sleep disturbance (28.6%), and NEURO-QOL cognition (23.8%). There were no significant differences between PROMIS and Neuro-QOL PRO scores between HGG and LGG groups, with the following exceptions: PROMIS pain intensity, in which patients with LGG experienced greater pain-related intensity relative to patients with HGG (t-score: HGG 1.76 ± 2, LGG 3.29 ± 3; P = 0.01) and greater distress from declining physical function among patients with HGG (t-score: HGG 41.83 ± 12.59, LGG 47.74 ± 12.16; P = 0.05) (**Table 3**).

### Aphasia Severity Impacts HRQOL Functional Domain

In a subgroup of 26 patients with dominant hemisphere gliomas within the perisylvian frontal, parietal, and temporal lobes, 7 had a BDAE aphasia severity score of 1–2 (severe) and 19 had a BDAE score of 3–5 (mild). The mean BDAE severity scores for both HGG and LGG cohorts were 3 ± 1 and 5 ± 0.5, respectively (P = 0.004). Aphasia severity had a moderate association with greater distress on PROMIS measures of anxiety (r = −0.51; P = 0.0074) and NEURO-QOL cognition (r = 0.55; P = 0.0033) (**Table 4**).

## DISCUSSION

HRQOL measurements have become an increasingly important measure in the care of glioma patients. There is a need for reliable patient quality of life assessment measures which are easy to use and clinically relevant for both patients and clinicians. Assessing HRQOL PRO measures in glioma patients can be a challenge because of self-reporting difficulties in this population due to functional and cognitive impairments (25). The PROMIS survey was developed to measure PRO measures for patients with a variety of chronic diseases. PROMIS as a subjective assessment tool for glioma patients has been validated and compared to the more commonly used European Organization for Research and Treatment of Cancer (EORTC-30) and Caregiver Quality of Life Cancer (CQOLC) scales (26). To our knowledge, only one pilot study in a small cohort of 10 patients has described the use of PROMIS as a HRQOL assessment tool in adult highgrade glioma patients (26). Here we are the first to compare

#### TABLE 1 | Patient demographics and clinical characteristics.


Bold values mean significant p value (<0.05).

HRQOL using PROMIS and Neuro-QOL between adult LGG and HGG patients.

Perhaps unsurprisingly, 79.3% of all patients reported that they value quality of life over survival at the point of diagnosis. After cross sectional analysis of PROMIS and NEURO-QOL data, we found that LGG patients experienced more pain intensity and greater distress from declining physical function when compared with HGG patients. The biologic and psychological correlates to explain these differences are unclear; however, this information carries significance when caring for patients and determining clinical trial efficacy. Cognitive dysfunction is more commonly found in HGG patients; therefore distress measures in PRO domain such as pain intensity might be reported at different rates in LGG and HGG patients. We also found a high rate of impairment in the PROMIS functional domains assessed in our study population, again indicating that patient functional wellness should be carefully considered in an individualized treatment approach. Future studies may compare these prevalences to other cancer patient populations. Our results also demonstrate that aphasia severity is associated with increased anxiety and cognitive distress. We find a higher prevalence of severe aphasia in HGG patients relative to LGG, which may be due to selection bias due to small sample size or differences in intrinsic tumor biology.


Glioma patients suffer from a wide range of possible neurological and functional limitations which influences quality of life and survival. Aphasia and cognitive disorders are more prevalent in patients with WHO III and IV tumors. Cognitive dysfunction, as determined by global cognitive task performance, occurs in 35.9% of HGG patients and 23.7% of patients experience aphasia throughout their disease trajectory (27). It is therefore of little surprise that our Neuro-QoL analysis determined that 43.9% of HGG patients experience distress from impairment of cognitive function (**Figure 1**). Similar results are seen for distress from physical function in HGG patients. Despite the absence of identifiable oncological differences between our LGG and HGG cohorts, pain intensity scores were higher in LGG patients (**Figure 1**). This could be caused at least in part by the increased rate of cognitive dysfunction resulting in under reporting of pain in HGG patients. These differences bring to light important considerations when interpreting PRO in the adult glioma population. Looking beyond survival, when designing clinical trials, is critical given the extensive burden of symptoms experienced by glioma patients. Furthermore, it cannot be assumed that LGG and HGG patients experience the same symptoms and distress profile.

Patient reported outcome (PRO) measures are used in clinical practice as a mechanism to understand the natural history of disease or as a health measure of clinical change. There are few publications focused on thresholds constituting meaningful clinical change. Clinical judgment must be applied for the interpretation of clinically meaningful PRO. Defining the magnitude of change that is clinically important is necessary and there's a growing body of evidence for this important area of study. There are several terms for clinically relevant HRQOL change, including, minimally important difference (MID). "True" differences do not exist in HRQOL assessments and the magnitude of a score is an estimate which must be interpreted with clinical judgment (28). There is no empirical literature on which to base MID estimate; therefore, many use a half standard deviation (5 points on a T score metric). However clinical significance has been illustrated at a lower threshold (28, 29). MID for the adult glioma population are currently unavailable and a topic of future study. For example, patients with advanced stage cancer illustrate fatigue PROMIS MID of 3.0– 5.0, pain interference MID of 4.0–6.0, and physical function MID of 4.0–6.0 (29). It is important to note that MID estimates vary based on cross sectional and longitudinal analysis. Furthermore, these assessments of clinical significance are averages across subjects; therefore, individual patients may require more or less to be clinically meaningful. The objective of this study was to evaluate adult glioma patient perspectives on functional domain affecting health related quality of life and apply cross sectional analysis of the PROMIS and Neuro-QOL instruments as subjective PRO measures in an adult low and high-grade glioma patient population. MID estimates were beyond the scope of this initial study which was focused on characterization of disease. Moving forward we hope to define MID and clinical relevance in the adult glioma population.

Other study limitations include the single institution small sample size which prohibited stratification of patients by

TABLE

2


Functional

domain

of

importance

and

quality

of

life

concerns

influencing

medical

decision-making

in

adult

patients

with

low-

or

high-grade

glioma.

**90**

who scored >1 standard deviation beyond the normative mean.

TABLE 3 | Comparison of HRQOL scores for low- and high-grade glioma groups.


SD, standard deviation. Bold values mean significant p value (<0.05).

additional potential confounders including tumor location, volume, or burden of disease at the time of assessment. Given our small sample size, within the LGG cohort we do not see the expected distribution across male and female patients. This difference does not reach statistical difference; however, it's not in line with expected results for the general population (30). Gender differences may contribute to variations in health outcomes. These differences in PRO have been reported primarily with pain and pain related disorders; however, it is certainly possible that gender differences impact this dataset focused on adult glioma patients (31). Furthermore, pain intensity interpretation is limited given that PROMIS is specifically focused on global pain making the distinction between headaches and neuropathy impossible. It is well known that both patient perspectives and HRQOL PRO measures vary with time (32). For this reason, our current analysis focused solely on HRQOL at the time of initial



Bold values mean significant p value (<0.05).

diagnosis with the goal of longitudinal analysis throughout the course of disease to better understand how responses change with time. This and other limitations will be mitigated by increasing the sample size in future studies. Additionally, while subjective patient PRO measures are valuable, it should be noted that they are excellent HRQOL measures of distress but not dysfunction (25). Objective measures of function should also be incorporated into patient assessment.

#### CONCLUSIONS

HRQOL measurements have become increasingly important in glioma research and clinical practice. There has been limited and slow progress in developing effective treatments for glioma

### REFERENCES


patients. The natural history of the disease in addition to treatment related side-effects can also negatively impact patient function and HRQOL. Treatment of glioma patients should focus on both prolonging life in addition to maintaining quality of life. The PROMIS and NEURO-QOL are two measures, which are valuable for quantifying patient reported HRQOL. The current study will hopefully lead to the use of these tools in more robust clinical research and practice settings.

#### ETHICS STATEMENT

This study was performed in accordance with the University of Michigan institutional ethics committee (IRB- HUM00092238). The protocol was approved by the UM-IRB and all subjects gave written informed consent in accordance with the Declaration of Helsinki.

#### AUTHOR CONTRIBUTIONS

NG: manuscript preparation, data analysis, data interpretation; DA, NB, EB, LW: manuscript preparation, data analysis; AB, ZM: data acquisition, data analysis; KK: data acquisition, manuscript preparation; TF, KM: data acquisition; SS, NC: concept, study design, manuscript preparation; SH-J: concept, study design, data acquisition, data analysis, manuscript preparation, manuscript editing.

### FUNDING

Senior author (SH-J)—Alfred A. Taubman Foundation and Mark Trauner Brain Research Fund.


a range of chronic conditions. J Clin Epidemiol. (2016) 73:89–102. doi: 10.1016/j.jclinepi.2015.08.038


information system survey as a quality-of-life instrument for patients with malignant brain tumors and their caregivers. Nurs Clin North Am. (2015) 50:679–90. doi: 10.1016/j.cnur.2015.07.009


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Gabel, Altshuler, Brezzell, Briceño, Boileau, Miklja, Kluin, Ferguson, McMurray, Wang, Smith, Carlozzi and Hervey-Jumper. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Impacts of Surgery on Symptom Burden and Quality of Life in Pituitary Tumor Patients in the Subacute Post-operative Period

Mark R. Waddle<sup>1</sup> , Mollie D. Oudenhoven<sup>2</sup> , Casey V. Farin<sup>3</sup> , Allison M. Deal <sup>4</sup> , Riane Hoffman<sup>5</sup> , Hojin Yang<sup>4</sup> , Jennifer Peterson<sup>1</sup> , Terri S. Armstrong<sup>6</sup> , Matthew G. Ewend<sup>5</sup> and Jing Wu<sup>6</sup> \*

*<sup>1</sup> Department of Radiation Oncology Mayo Clinic, Jacksonville, FL, United States, <sup>2</sup> Department of Pediatrics, University of Colorado, Denver, CO, United States, <sup>3</sup> Department of Neurology, Duke University, Durham, NC, United States, <sup>4</sup> Lineberger Comprehensive Cancer Center Biostatistics Core Facility, UNC, Chapel Hill, NC, United States, <sup>5</sup> Department of Neurosurgery, University of North Carolina, Chapel Hill, NC, United States, <sup>6</sup> Neuro-Oncology Branch, National Cancer Institute/National Institutes of Health, Bethesda, MD, United States*

#### Edited by:

*Sujit Prabhu, University of Texas MD Anderson Cancer Center, United States*

#### Reviewed by:

*David D. Eisenstat, University of Alberta, Canada David M. Peereboom, Cleveland Clinic Lerner College of Medicine, United States*

> \*Correspondence: *Jing Wu jing.wu3@nih.gov*

#### Specialty section:

*This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology*

> Received: *30 August 2018* Accepted: *01 April 2019* Published: *23 April 2019*

#### Citation:

*Waddle MR, Oudenhoven MD, Farin CV, Deal AM, Hoffman R, Yang H, Peterson J, Armstrong TS, Ewend MG and Wu J (2019) Impacts of Surgery on Symptom Burden and Quality of Life in Pituitary Tumor Patients in the Subacute Post-operative Period. Front. Oncol. 9:299. doi: 10.3389/fonc.2019.00299* Background: Pituitary tumors are rare but are associated with significant symptoms that impact patients' quality of life (QOL). Surgery remains one of the most effective treatment options for long term disease control and symptom benefit, but symptom, and quality of life recovery in the subacute period has not been previously reported. This study aimed to better understand the impact of surgery on patients' symptom burden and QOL in the subacute post-surgical period.

Methods: Twenty-three adult patients with pituitary tumors undergoing surgical resection at University of North Carolina Cancer Hospital were enrolled in this study. M.D. Anderson Symptom Inventory Brain Tumor Module, European Organization for Research and Treatment of Cancer QLQ-C30 and QLQ-BN20 questionnaires were collected pre- and 1-month post- surgical resection and differences were analyzed for individual and groups of symptoms and QOL using Wilcoxon signed-rank tests.

Results: Twenty adult patients had both pre-operation and post-operation follow-up visits; 60% had functional pituitary adenomas. Seven symptoms including fatigue, memory, vision, numbness, speaking, appearance, and weakness were significantly improved at the 1-month post-operation visit while one symptom, sleep, worsened. Global Health Status/QOL measurements was improved minimally from 63 (SD 25) at pre-operation to 67 (SD 22) at 1-month post-operation without statistical significance.

Conclusions: This study demonstrated a rapid improvement of many symptoms in the subacute post-operative period in pituitary tumor patients. Disturbed sleep was identified as the only symptom to worsen post-operatively, encouraging potential prospective interventions to improve sleep, and subsequently improve the QOL in pituitary tumor patients following surgical intervention.

Keywords: pituitary, adenoma, surgery, QOL, subacute, post-operative

### INTRODUCTION

Pituitary tumors are relatively rare primary central nervous system (CNS) tumors in adults (1) but make up 10% of all neurosurgical interventions in the United States (2). They are often an incidental finding in autopsies and brain imaging studies (3). Although these tumors are usually benign, pituitary tumors have a significant impact on a patient's health and have been shown to cause cognitive dysfunction (4, 5), visual deficits (6–8), headaches (7), and an overall decline in quality of life (QOL) at the time of diagnosis (3, 8, 9). Multiple studies have demonstrated the adverse clinical effects of these tumors at presentation, yet there are little data to describe how these symptoms, and thus QOL, are impacted with treatment. Surgery, the primary definitive treatment for these tumors, carries risk of damage to surrounding structures such as the internal carotid artery and optic nerve, CSF leak, and/or hormone abnormalities (10). Van der Klaauw's group showed that patients with all subtypes of pituitary tumors experienced decreased QOL as far as 10– 15 years after treatment when compared with healthy controls, but the added impact of surgery on these outcomes was not described (11, 12). Other studies have shown improved sinonasal functioning in the long term, but have mixed results in the subacute period (13). Further, there is currently limited evidence for the impact of surgery on many other important quality of life metrics which may impact a patient's ability to function (13–15).

The burden of patient symptoms and QOL is an essential consideration and has been increasingly recognized in literature as a primary end point for both benign and malignant tumors (16–18). The World Health Organization defines QOL as "an individual's perception of their position in life in the context of culture and value systems in which they live and in relation to their goals, expectations, standards, and concerns" (7). The importance of the symptom and QOL endpoint in oncology care has resulted in a multitude of validated scales and surveys which are routinely used in research and in practice. The goal of this study is to characterize the common symptoms, side effects, and overall QOL of patients with pituitary tumors before and after surgical intervention in the subacute period with the use of validated QOL instruments. Additionally, we hope this study will help identify potential biologic underpinnings and guide early intervention, symptom screening following intervention, and to aid in general clinical management to further improve QOL in patients with pituitary tumors who undergo surgical resection.

#### MATERIALS AND METHODS

#### Patient Population

This is a prospective study using validated questionnaires of symptom burden and QOL. The study was undertaken in accordance of Good Clinical Practice guidelines and the Declaration of Helsinki. All eligible patients were provided with IRB-approved consent forms and all participating patients provided written informed consent. Approvals for the study protocol and consent forms were obtained from the Institutional Review Board (IRB) at the University of North Carolina at Chapel Hill (UNC).

Eligible patients were 18 years old or above, English speaking, with adequate mental capacity to fill out questionnaire, and no other malignancy that required active anti-neoplastic treatment in the past 3 years. All patients had histological diagnosis of pituitary tumor at UNC from June 2011 to March 2014.

#### Study Design

With UNC IRB approval, all eligible patients were provided with informed consent forms. Patients were given the questionnaires including M.D. Anderson Symptom Inventory Brain Tumor Module (MDASI-BT), European Organization for Research and Treatment of Cancer (EORTC) QLQ-C30 and QLQ-BN20, at their pre-operative clinic visit within 7 days prior to surgery. Longitudinal follow up was completed with the same preoperative surveys given post-operatively at 1 month after surgery, which was defined as the "Sub-acute post-surgical period." Patients without histological diagnosis of pituitary tumors were considered non-evaluable after the pathology reports were reviewed.

### Outcome Assessment

Three extensively validated symptom inventory questionnaires were selected to assess patient's symptom burden and QOL: the MDASI-BT, EORTC QLQ-C30, and EORTC QLQ-BN20. Baseline and one-month post-operative symptom burden and QOL scores were recorded for each questionnaire. Changes relative to baseline were calculated for each patient and those values were averaged and reported for symptom scores and QOL.

### MDASI-BT

MDASI-BT consists of 28 questions which can be completed in an average time of 5–10 min. It is composed of questions rated on an 11-point scale (0–10) to indicate the presence and severity of each symptom in the last 24 h, with 0 being "not present" and 10 being "as bad as you can imagine." A total of 28 questions include 22 symptom questions (13 core symptoms, 9 braintumor-specific symptoms) (19, 20) and six interference with life questions, which are further divided into activity and moodrelated items. The 22 symptom specific questions on MDASI-BT also measure six underlying constructs: (1) an affective factor comprised of distress, fatigue, sleep, sadness, and irritability, (2) a cognitive factor comprised of difficulty understanding, remembering, speaking and concentrating, (3) focal neurological deficits factor, including seizure, numbness, pain, and weakness, (4) treatment-related symptoms such as dry mouth, drowsiness, and appetite, (5) generalized disease status symptoms, including change in vision, change in appearance, change in bowel patterns, and shortness of breath, and (6) GI related factors, including nausea and vomiting (20). Symptoms on the MDASI-BT are those common in the brain tumor population as well as those associated with cancer therapies. The MDASI-BT has evidence of content and construct validity, discriminant validity by performance status and disease progression, and internal consistency (20).


*KPS, Karnofsky Performance Status; GH, Growth Hormone.*

#### EORTC QLQ-C30 and QLQ-BN20

EORTC QLQ-C30 (Version 3.0) is composed of 30 questions organized into a global health status/QOL scale that include 5 functional scales (physical, role, emotional, cognitive, and social), 3 symptom scales (fatigue, nausea and vomiting, and pain), and a number of single items assessing additional difficulties (dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties).

EORTC QLQ-Brain Cancer Module (EORTC QLQ-BN20) is a supplement to the EORTC QLQ-C30 specifically designed for brain tumor patients and consists of 20 questions scored as four multi-item scales (future uncertainty, visual disorder, motor dysfunction, and communication deficit) and seven single-item symptom scales (headaches, seizures, drowsiness, hair loss, itchy skin, weakness of legs, and bladder control).

The responses to both questionnaires were scored as outlined in the QLQ-C30 scoring manual to a score from 0–100 where a higher score represents a high/healthy level of functioning and a high quality of life, but a high level of symptomatology/problems, depending on the question (21, 22).

#### Statistical Analysis

Descriptive statistics are provided for patient characteristics and questionnaire results. Wilcoxon signed-rank tests are used to evaluate if changes in symptom burden and QOL were significantly different compared to baseline for both the symptom burden and QOL, and unadjusted p-values are reported. All analysis was done using SAS software v9.3 (Cary, NC).

#### RESULTS

#### Patient Characteristics

A total of 23 patients diagnosed with a pituitary tumor at UNC hospital from June 2011 to March 2014 completed pre-operative TABLE 2A | MDASI symptom severity score and standard deviation for each symptom at pre-operation and 1 month post-operation.


*Symptoms with a statistically significant change (p* < *0.05) relative to pre-operative values are marked with an arrow, indicating the direction of change. ( indicates less symptoms, indicates more symptoms). The most severe symptoms for each time point are marked with bold text.*

questionnaires. Of these, one subject withdrew, one did not have surgery, and one subject was lost to follow up. The remaining 20 subjects underwent first time surgery for a pituitary tumor and completed the 1 month follow up questionnaire in the subacute surgical period.

Age of all evaluable patients ranged from 24 to 77 years with a median age of 49 years and 50% were male. The baseline Karnofsky Performance Score (KPS) ranged from 80 to 90. Functional adenomas were most common representing 60% of all studied patients including 30% prolactinomas and 20% GH secreting adenomas (**Table 1**).

#### Symptom Burden Questionnaire Results

MDASI-BT symptom severity scores are shown in **Table 2A**. Pre-operative MDASI-BT questionnaire results showed the most severe symptom was fatigue, with mean TABLE 2B | MDASI symptom severity main category and subcategory scores, and standard deviation for each symptom group at pre-operation and 1 month post-operation.


*Symptoms with a statistically significant change (p* < *0.05) relative to pre-operative values are marked with an arrow indicating the direction of change. ( indicates less symptoms, indicates more symptoms). The most severe symptoms for each time point are marked with bold text.*

(standard deviation) severity scores of 5.0 (SD 2.7). Seizure was the least severe symptom with mean scores of 0.2 (SD 0.5).

At 1-month post-operation, only one symptom, disturbed sleep, significantly worsened with a mean symptom severity score of 4.8 (SD 3.2, p = 0.03), a 55% increase from the baseline value of 3.5. Seizures also worsened by 166% from 0.15 (SD 0.5) to 0.4 (SD 1.2), and dry mouth slightly worsened from 1.9 (SD 3.0) to 2.3 (SD 2.7), both of which were small absolute changes and statistically non-significant. For the same time point, significant improvements were found in fatigue, impaired memory, vision, numbness/tingling, difficulty speaking, appearance, and weakness. The symptom that improved the most was vision, with a symptom severity score of 1.0 (SD 1.0), an improvement of 71% from the baseline. The percentage change of symptom severity score from baseline for all measured items can be seen in **Figure 1**.

The mean symptom severity scores for all symptoms preoperation and 1-month post-operation were 2.3 (SD 1.2) and 1.7 (SD 1.2), respectively. In the interference with life category, three areas significantly improved from pre-operation to 1 month post-operation: general activity, enjoyment of life, and walking (p < 0.05).

Aggregate scores and their changes were calculated for the two MDASI-BT main categories and six MDASI-BT sub-categories, shown in **Table 2B**. Interference with life was significantly improved at 1-month post-operation, with a 43% improvement from the baseline. When the 22 symptom specific questions were analyzed in subgroups measuring six underlying constructs, four subgroups, including cognitive, neurologic, generalized disease, and gastrointestinal related factors were significantly improved by 41, 32, 55, and 79% from baseline, respectively, at 1-month post-operation follow-up (p < 0.05). Affective related factors did not show significant improvement at 1-month follow-up. The

percent changes from baseline for the six subgroups are shown in **Figure 2**.

#### Quality of Life Results

EORTC QLQ-C30 and QLQ-BN20 questionnaire results are shown in **Table 3**. Pre-operatively, the most severe symptoms were headaches, drowsiness, fatigue, pain, and insomnia. Headaches were most severe with a symptom severity score of 53 (SD 35). The least severe was seizures with a symptom score of 0.

By 1-month post-operation, the symptom that improved the most was visual disorders, which improved by 67%, while insomnia worsened the most, by 33%. Headaches remained the most severe symptom at 47 (SD 40). Statistically significant improvements were shown in cognitive functioning (p < 0.01) and visual disorders (p < 0.01) from pre-operation to 1-month post- operation.

Global status of health/QOL was stable from pre-operative baseline to 1-month post-operation with scores of 63 (SD 25) and 67 (SD 22), respectively, an improvement of 5%.

#### DISCUSSION

This study is consistent with prior studies and demonstrates that the majority of patients are symptomatic prior to intervention (11, 12). However, we successfully showed a novel finding that

in this group of otherwise healthy individuals, many symptoms improve rapidly with surgical intervention, as outlined in the result section above.

The initial improvement in vision was expected. Vision loss is a well-known symptom of pituitary tumors due to compression of the overlying optic chiasm and nerves which improves rapidly following surgical decompression and removal of the tumor (6). However, many other symptoms including fatigue, memory, appearance, difficulty speaking, numbness, and appetite showed improvement in the subacute post-operative period in this study, which is a finding with little to no prior data for comparison in this patient population.

Two interference items, walking and enjoyment of life also improved. In fact, the interference with life category and four of six subcategories for symptom burden in MDASI-BT (cognitive, neurologic, generalized disease, and gastrointestinal) showed statistically significant improvement at 1-month postoperation, as shown in **Table 2B**, demonstrating a multitude of improvements in symptom burden and quality of life. These findings may be useful in counseling patients' suffering from these symptoms or clinicians looking to weigh the risk/benefit ratio of surgery for a given patient.

Patients' functioning improved as well. Self-reported measures of cognitive functioning improved significantly from pre-operation to 1-month post-operation, while role functioning, emotional functioning, physical functioning and social functioning remained the same.

Our findings are unique. A prior study by Glicksman et al. showed that at 3 months post-operatively there was a significant improvement of sino-nasal symptoms per the Sinonasal Outcome Test (SNOT-22), which demonstrated improvements in Rhinologic, Extranasal, Ear/facial, Psychological, and Sleep domains, that continued to improve over a 2 year period (23). However, this study failed to report on either acute or subacute changes in QOL. Another study by McCoul et al. (13) reported generalized symptom burden at 3 weeks and showed transient overall worsening of symptoms at this time point, driven largely by site specific factors such as intranasal edema, crusting, and hyposmia, but reported minimal to no improvement in other quality of life endpoints until 6–12 weeks post-operatively (13). However, these studies were both limited by the quality of life metrics they reported on and did not test for symptoms which showed dramatic improvements in our study, such as memory, appearance, difficulty speaking, numbness, appetite, cognitive function, walking, and enjoyment of life. Therefore, our findings are encouraging that, despite a possible worsening of surgery related symptoms in the subacute surgical period after sino-nasal resection shown in prior studies, many symptoms that can have a dramatic impact on a patient's QOL did improve rapidly. Additionally, we found that the MDASI-BT in this patient population was an excellent test to accurately define quality of life changes which are pertinent and important. The MDASI-BT has evidence of content and construct validity, discriminant validity by performance status and disease progression, and internal consistency. For this reason, we encourage its use for patients undergoing treatment of pituitary adenomas. However, more extensive research and experiences of evaluating pituitary patients with current QOL questionnaires may lead to a pituitary specific QOL instrument, which could better serve this patient population.

One symptom, disturbed sleep, did show worsening in the subacute period 1-month post-operation and was the most severe symptom following surgery in our studied patients. Disturbed sleep increased from a low range severity before surgery to nearly high range severity after surgery, an increase of 55%. Similarly, insomnia increased in severity by 33% in the EORTC-C30 questionnaire testing, but without statistical significance. Impairments in sleep have been previously described in patients with cancer, and may correlate with decreased total and free



*Symptoms with a statistically significant change (p* < *0.05) relative to pre-operative values are marked with an arrow indicating the direction of change. ( indicates less symptoms,*

*indicates improved functioning) The most severe symptoms and the lowest level of functioning for each time point are marked with bold text.*

cortisol levels, due to a disrupted hypothalamus-pituitaryadrenal (HPA) axis (24). Our patients had similar sleep impairments, and our findings suggest that pituitary surgery for pituitary tumors may temporarily further disrupt the HPA axis, possibly worsening an already impaired sleep-wake cycle. Alternatively, the patient's daily routine may be temporarily altered leading to poor sleep hygiene, and the patients may be on treatment such as steroids which are known to have side effects such as disturbed sleep. Regardless of the underlying cause, disturbed sleep is likely a symptom that could benefit from routine screening possibly with sleep studies, avoidance of steroids if possible, and/or prophylactic treatment postoperatively, and prior studies are encouraging in that this symptom significantly improves by 6 months to 12 months post-operatively (23).

Another troubling symptom on the secondary questionnaire testing was headaches, which was the most severe symptom reported in both pre-operative and post-operative surveys. This symptom marginally improved from baseline with an improvement of 11% at 1-month post-operation, although not statistically significant. While this is a symptom well known to clinicians treating pituitary tumors, the severity and lack of change with treatment indicate that, on average, patients are experiencing significant distress from this symptom. Continued attention should be given to symptomatic relief of headaches even after resection, as surgical treatment is not expected to dramatically improve this problem, at least in this studied population.

This study has several limitations. The primary limitations, and ones common to studies of pituitary tumors, was that the sample size obtained was not large and thus only powered to detect large differences, and that limited follow-up data is available to assess for further longitudinal effects. Although the study took place at a large academic hospital with a robust pituitary program, the number of patients undergoing first time pituitary surgery was not as large as predicted, and there were several gaps in patient accrual. Additionally, several patients were excluded for not speaking English, as described in the inclusion criteria. As a result, there were 20 patients followed to 1-month post-operation. While a size of 20 subjects was adequate to find statistical significance in very large changes, it is likely that many more subtle changes were not identified. As an example, sleep disturbance was found to have a statistically significant worsening after surgery on the MDASI-BT, whereas insomnia was worse, but not statistically significant on EORTC QLQ-C30. This is likely a matter of sample size resulting in two validated questionnaires yielding similar results, yet only one reaching statistical significance. Future studies could be improved by collaboration with multiple other institutions to increase accrual and by having questionnaires and consent forms available in multiple languages.

Finally, we were unable to conduct a sub-group analysis between pituitary tumor types, hormone status, or patient demographics. Studies have shown symptoms vary between pituitary tumor groups and hormone status. For example, appearance score has been found to be the worst in patients with acromegaly, and patients with Cushing disease secondary to pituitary adenomas have the most impaired QOL (7, 12). Subgroup analysis would allow physicians to identify certain patient groups with more severe symptoms to better target symptom control or prevention.

Despite the above limitations, we believe that this study has important clinical implications. This study successfully prospectively characterizes specific symptoms which have previously not been investigated in pituitary adenomas. A strength of this study is that, to our knowledge, it is the first prospective study examining symptom burden in a number of measures following pituitary tumor surgery. Additionally, the use of comprehensive and very well validated questionnaires, MDASI-BT, EORTC-C30, and EORTC-BN20, allowed a complete assessment of symptoms in this disease group. We successfully identified sleep disturbance and headaches as symptoms that can be targeted clinically either with improved monitoring, pre-emptive treatment, and/or prospective future study. The study results also identified several other symptoms which significantly improved after the surgery. Future

interventional studies may include multi-centered collaboration with well-established research infrastructure for the investigation of symptom burden and health related quality of life. We hope these data will aid clinicians both in pre-operative and postoperative patient counseling, provide clinicians with improved awareness of troublesome symptoms associated with surgery, and continue the progress being made in the quality of life of brain tumor patients.

#### ETHICS STATEMENT

The study was undertaken in accordance of Good Clinical Practice guidelines and the Declaration of Helsinki. All patients

#### REFERENCES


provided their written informed consent. Approvals for the study protocol and consent forms were obtained from the Institutional Review Board (IRB) at the University of North Carolina at Chapel Hill (UNC). Eligible patients were of or above 18 years old.

### AUTHOR CONTRIBUTIONS

JW, TA, ME, CF, MO, JP, and MW contributed to the design and implementation of the research, to the analysis of the results and to the writing of the manuscript. AD and HY contributed to the statistical analysis. RH contributed to design and implementation of the research, to the analysis of the results and to the writing of the manuscript.

prospective study. J Neurosurg. (2012) 117:498–506. doi: 10.3171/2012. 6.JNS111066


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Waddle, Oudenhoven, Farin, Deal, Hoffman, Yang, Peterson, Armstrong, Ewend and Wu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Radiomics in Glioblastoma: Current Status and Challenges Facing Clinical Implementation

Ahmad Chaddad<sup>1</sup> , Michael Jonathan Kucharczyk <sup>1</sup> , Paul Daniel <sup>1</sup> , Siham Sabri 2,3 , Bertrand J. Jean-Claude3,4, Tamim Niazi <sup>1</sup> and Bassam Abdulkarim1,3 \*

<sup>1</sup> Division of Radiation Oncology, Department of Oncology, McGill University, Montreal, QC, Canada, <sup>2</sup> Department of Pathology, McGill University, Montreal, QC, Canada, <sup>3</sup> Research Institute of the McGill University Health Centre, Glen Site, Montreal, QC, Canada, <sup>4</sup> Department of Medicine, McGill University, Montreal, QC, Canada

Radiomics analysis has had remarkable progress along with advances in medical imaging, most notability in central nervous system malignancies. Radiomics refers to the extraction of a large number of quantitative features that describe the intensity, texture and geometrical characteristics attributed to the tumor radiographic data. These features have been used to build predictive models for diagnosis, prognosis, and therapeutic response. Such models are being combined with clinical, biological, genetics and proteomic features to enhance reproducibility. Broadly, the four steps necessary for radiomic analysis are: (1) image acquisition, (2) segmentation or labeling, (3) feature extraction, and (4) statistical analysis. Major methodological challenges remain prior to clinical implementation. Essential steps include: adoption of an optimized standard imaging process, establishing a common criterion for performing segmentation, fully automated extraction of radiomic features without redundancy, and robust statistical modeling validated in the prospective setting. This review walks through these steps in detail, as it pertains to high grade gliomas. The impact on precision medicine will be discussed, as well as the challenges facing clinical implementation of radiomic in the current management of glioblastoma.

#### Keywords: brain, cancer, diagnosis, glioma, MRI, radiomics, glioblastoma

### INTRODUCTION

Glioblastoma (GBM) is the most common astrocytic primary brain malignancy, with an annual incidence of 2–3 cases per 100,000 adults in North America and Europe (1, 2). The standard of care for newly diagnosed GBM combines maximum safe resection followed by chemo-radiation and adjuvant courses of temozolomide (TMZ) (3). The median overall survival is poor at 14.6 months and 5-year survival rates are under 10% following standard of care treatment. If patients tolerate the chemoradiotherapy without progression, they may be considered for tumor-treatment fields. Even in this setting, the survival is still limited at a median of 20.9 months (4). Given these poor outcomes, there is hope that up-and-coming therapies will show benefit in the randomized setting (5, 6). It will be essential to ascertain which patients can benefit from these therapies, highlighting the need for efficacious tools to offer personalized medicine.

#### Edited by:

Marcos Vinicius Calfat Maldaun, Hospital Sírio-Libanês, Brazil

#### Reviewed by:

Sunit Das, St. Michael's Hospital, Canada Riccardo Soffietti, University of Turin, Italy

\*Correspondence: Bassam Abdulkarim bassam.abdulkarim@mcgill.ca

#### Specialty section:

This article was submitted to Neuro-Oncology and Neurosurgical Oncology, a section of the journal Frontiers in Oncology

> Received: 26 November 2018 Accepted: 23 April 2019 Published: 21 May 2019

#### Citation:

Chaddad A, Kucharczyk MJ, Daniel P, Sabri S, Jean-Claude BJ, Niazi T and Abdulkarim B (2019) Radiomics in Glioblastoma: Current Status and Challenges Facing Clinical Implementation. Front. Oncol. 9:374. doi: 10.3389/fonc.2019.00374

**101**

Magnetic resonance imaging (MRI) is the preferred imaging modality for both the diagnosis and monitoring of central nervous system (CNS) malignancies (7). It provides a massive amount of information to clinicians. Unfortunately, clinicians are typically restricted to qualitative descriptors or subjective quantitative assessments to articulate changes in imaging. The resulting clinical evaluations have a significant potential for bias.

Clinicians immensely value non-invasive approaches that can direct patients to the correct therapeutic approach in an objective fashion. This begins at diagnosis, where various molecular factors differentiate the diagnosis between lowgrade glioma, high-grade glioma, or GBM (8). Such factors may also be predict the efficacy of a systemic agent (9, 10). This information requires tissue, introducing patient morbidity, an additional procedure, and a variety of expensive molecular assessments.

Radiomics has demonstrated remarkable progress in demonstrating that it may be a tool that can derive this information. Radiomics is a field of biomedical imaging using advanced non-invasive assessments of complex imaging characteristics within the MRI images that are too complex for a human to appreciate (11–14). These characteristics are known as features. Imaging features have been associated with a CNS tumor's histological features (14), progression (15) grade (16), or even overall survival (17–21). Radiomics analysis thus hosts a major role in producing novel non-invasive biomarkers acquired from a test—MRI—that is already routinely acquired from patients as part of the standard of care.

### RADIOMICS METHODOLOGY

A standard pipeline of radiomic analysis has been described by several studies in the past (**Figure 1**) as mentioned previously by several studies (12, 13, 19, 21–23). This review discusses recent studies in the development of MRI-based radiomics analysis in relationship to this pipeline. For CNS malignancies, the literature discusses the most significant cause of diagnostic and management dilemmas—low and high-grade glioma. To facilitate an understanding of the process, there are sections on the: (1) preprocessing and image acquisition for developing a radiomic model; (2) segmentation/labeling of the cancer; (3) identification of relevant features types that may relate to the molecular properties of the tumor (14, 24) and (4) statistical modeling to describe a radiomic profile's relationship with a clinical outcomes. Given the number of variables at each step, collaboration is essential. Radiologists and oncologists must ensure that the appropriate regions are being assessed and the right questions are being asked. Molecular scientists must communicate the relevant genetic and proteomic characteristics that will influence a patient's clinical course. Engineering teams must determine what information can be reliably extracted from the images and then adapt the machine learning to fashion a reliable model. Consultation with statisticians will allow for a methodological approach allows for a potentially statistically significant solution.

#### Image Acquisition

MRI radiomics has repeatedly shown the ability to differentiate low and high-grade glioma, which have different management strategies (https://www.nccn.org/) and a remarkably different prognosis (25–28). One reason this data can be more rapidly generated is that there is a wealth of clinical information available—glioma patients have regular MRIs throughout their lives. However, reproducibility is a significant issue at different stages of the radiomics pipeline. The issues begin at image acquisition. Different academic groups acquire their MRI images to different settings at the first step of the pipeline. This is one reason that radiomic analysis collaboration has been limited between research groups. Standardization offers a rational solution to overcome this barrier.

### Standardization

Potential variations in images are often secondary to the MRI scanner model, including image resolution (i.e., pixel size and slice spacing), image contrast, slice thickness, patient position, and further variations introduced by different reconstruction algorithms. When generating or applying a radiomics model, standardization must occur so the data can be assumed that it was extracted from similar settings. To accomplish this, volume datasets are usually re-sampled to a common voxel resolution of 1 mm<sup>3</sup> and an image size of 256<sup>3</sup> (or 512<sup>3</sup> ) voxels.

A common further step is normalizing the intensities within each volume image to the [0,1] or [0,255] range. Less commonly adopted normalization approaches have included gaussian and Z-score normalization. For example in Ellingson et al. (29), Gaussian normalization was the best normalization technique for image intensity correction. The need for standardization would be reduced if radiomic analysis could be performed with data acquired at a single geographical site. However, a single site would only provide a limited dataset. Thus, several studies have augmented their datasets through the use of multiple sites and an imputation technique to facilitate standardization (30).

The lack of standardization is a recognized problem. The Quantitative Imaging Biomarker Alliance offers an expert consensus after reviewing the available data. This group offers insightful guidelines for standardization that should be heavily considered in present and future studies. Such guidelines will be dynamic. Radiomic features may change from site to site or have new ways to be extracted or MRI image acquisition may change. Standardization in either of these contexts will a challenge in the future. Ongoing communication between institutions and robust reporting of new methodological approaches will be essential to groups studying radiomics.

### Segmentation of Brain Tumors

Accurate labeling of brain tumors in the images is required for radiomic analysis. It first involves defining the tumor volume, known as the region of interest (ROI), so it can then have its radiomic features extracted. The act of employing clinical, pathological and imaging features to mark out the ROI on the two-dimensional MRI images is called either the segmentation or labeling process. Segmentation is performed by clinicians—typically a radiologist or oncologist. The process

is subject to inter-rater variability, as the ROI definition will inevitably differ between clinicians. An approach to overcome this variation is different clinicians each generating their own ROI. The geographical regions common between the different ROIs is considered the true tumor mask. This tumor mask is then matched with the corresponding brain images to then extract the imaging features (i.e., radiomic features).

Since MRI generates several image sequences, the registration step involves matching the mask to the relevant MR series (ex. T1 weighted, T1-post contrast, T2 weighted, FLAIR), a welldescribed process (20, 31). Many tools used to delineate the ROI, such as the publicly accessible 3D Slicer (32), require slice by slice labeling on each series to ensure accuracy and precision (21). For efficiency and to minimize both inter- and intra-user variability, several studies have explored segmentation to all relevant MRI sequences without registration across the sequences (33). Registration distortions between MRI series may limit this approach (34). Distortion could cause incorrect localization of the ROI, directing the radiomic analyses to the incorrect MRI-defined anatomy. More investigation is required to allow for the trans-sequence application of user-generated segmentation data.

To overcome user variability in registration, (semi-)automated segmentation has been explored in various studies (35–38). Strong signals for a successful model, a promising Dice Similarity Coefficient (DSC) of 80%, have been reported with fully automated segmentation based on an adaptive algorithm with multi-level of thresholding (38). When deep learning radiomics (DLR) was applied to multiple tumor regions, the ability to label the tumor subregions achieved a DSC of 90% (35). DLR has become a success story for machine learning integral to limiting user variability. The use of DLR's convolutional neural networks (CNNs) to the various steps of the radiomics pipeline is elegantly described elsewhere (39). As to fully automated segmentation, further validation is required. Success here could enable the rapid integration of radiomics into personalized medicine.

### Radiomic Features Extraction

Extracting radiomic features is the first step in analyzing the segmented image. The features themselves are measures of the heterogeneity within the ROI (40). The degree to which these different features are present is a radiomics feature cluster, perhaps better conceptualized as an ROI's radiomic signature. There are different types of features, the most common and presently relevant are outlined in the **Table 1**.

#### Feature-Analyses

Once the features have been extracted, statistical modeling can highlight relationships between the extent a given feature is present and a clinical characteristic. There are various methodologies to analyze this, including minimizing the number of features likely to contribute to the statistical analysis. Feature selection methods (60) or reducing dimensionality in another fashion can accomplish this minimization. This has included sorting features by their minimum redundancy maximum relevance, mutual information, principal component analysis feature rank or the importance of features in other classifier models (31, 61–65). Once the features that are potentially relevant for analysis are determined, they are typically subject to assessments of their significance (e.g., Wilcoxon test, Kruskal-Wallis, log-rank, etc.) and correlation (e.g., Spearman rank, Pearson). These forms of univariate analysis determine if a feature is a significant predictor for the selected clinical outcome, with significance typically being defined as either a p < 0.01 or 0.05. The p-values should be corrected by the Bonferroni or Holm-Bonferroni procedure to limit the influence of random chance, including the false discovery rate (21, 60, 66).

### Multivariate Analysis and Model Building

Multivariate analysis fills an essential role in separating seemingly relevant features on univariate analysis from those that are likely independent predictors for the clinical outcome being assessed for. This is a critical step, limiting non-contributory features from influencing our eventual final statistical model (67). Once these features are selected from multi-variate analysis, the radiomics team must determine how many of their finite number of clinical

#### TABLE 1 | Features extraction techniques used in radiomic analysis.

Histogram features: These are first-order statistics computed from image's histogram of voxel/pixel intensities. Histogram features (e.g., average, standard deviation, skewness, kurtosis, energy and entropy) encode the voxel intensities and the shape of the data's distribution (41, 42). In non-CNS malignancies, these features have been associated with histological features, subtype and grade (43, 44).

Texture features: Texture features use second order statistics to characterize the spatial relationship between voxel intensities, describing the local spatial arrangement of intensities in the image. The features encode several matrices that represent the special intensity distribution in several ways. Not included in the list below are also texture features based on several conventional techniques that have been predictive of clinical outcomes, such as: as scale-invariant feature transform (SIFT), histogram of oriented gradients (HOG), fractal texture analysis (FTA) and local binary patterns (LBP) (45–47). Elsewise, the most common texture features are:

Gray-level co-occurrence matrix (GLCM)—the most commonly used texture feature. Considering only voxels within a specific range of gray values, it produces a matrix of the spatial relationships of pairs of voxels (48).

Joint intensity matrix (JIM)—evaluates the spatial relationships of pairs of voxels within given intensity ranges across different MRI different sequences. This is in contrast to GLCM, which is restricted to a single MRI sequence (21).

Neighborhood gray-tone difference matrix (NGTDM)—a description of the differences in signal intensity, or gray-tone, between each voxel and its neighboring voxels (49). It has been used in several topics of images analysis and classifications (45).

Neighboring gray-level dependence matrix (NGLDM)—Similar to NGTDM, is computed from the gray tone relationship between every element in the image and all of its neighbors at a certain distance (50, 51).

Gray-level run length matrix (GLRLM)—A matrix of all the voxels within the same gray level value (52).

Multiscale texture features: These features have been derived from filters, such as the Laplacian or Gaussian filter (53), that serve as a generic differential operator. Multiscale texture features provide an excellent description of local image variations, such as edges or blobs. The ROI's image is filtered in a multiscale way—from fine to coarse texture—that can be quantified by parameters like entropy (31, 54). The wavelet decomposition of an image generates multiscale texture images based on multiband frequencies, a radiographic characteristic called a detail. Each of these bands has a scale of the texture inside the image. A quantifier function then evaluates the texture of the images, using the resultant value as an input for a classifier model (42, 55).

Deep features: These features are derived from deep neural networks, the process of which is well-described in a recent review (56). To accomplish this, a pre-trained network must be established prior to texture extraction. As a case study from the literature, (1) ImageNET was pretrained to identify textures, (2) the CNN analyzed a fully connected layer of ImageNet, deriving 4,096 texture features, then (3) these features were used an input for a classifier model, which could also incorporate a CNN (as described in this review's Radiomics Analysis step) (39). However, CNNs require numerous examples to develop a reliable model. In general, studies implementing CNNs require more patients than the number of features being analyzed. Achieving this sample size can be a challenge, so alternative methods of model generation are needed for many studies. One such example reported the conditional entropy from a texture of the CNN's feature map. This was a reliable alternative when implemented into a random forest classifier, instead of another different standard CNN model (57).

Shape features: Shape features describe the 3D (or 2D) geometrical composition of the ROI considered the size (e.g., volume), form (e.g., sphericity, solidity, major length axis) and tumor location. As with traditional radiological assessment, shape is a characteristic that does relate to tumor characteristics with radiomics as well (19, 58, 59).

cases will be used to produce/train their model and how many need to be reserved to validate the model.

Increasing the size of the training cohort will increase the model's accuracy. Thus, typically 70–80% of the dataset is used for the training stage. Alternatively, if an external dataset is available, then all the datasets can train the model. This is the preferred scenario, allowing for a demonstration of external validity. If the datasets are limited in size, k-folds cross-validation can mitigate some of the statistical concerns (31, 68, 69).

Machine learning changes the available options. If unsupervised, the program can utilize different methods (e.g., k-means, nearest neighbors) to partition the features into different groups, then compare the relationships of the different features within their group—not the clinical data. After this is completed, the ability of the different groups to predict the clinical outcome is assessed, even though the clinical data did not contribution to the model's development (70). In comparison, supervised machine learning techniques (e.g., support vector machine, Bayes model, neural network nearest neighbors, random forests) will place varying numbers of the pre-determined relevant features into groups. Then their relative contribution to the model's ability to predict for the clinical outcome is altered until the most reliable combination of weightings is determined. Random forest classifier is a simple model that automatically selects the relevant features. Furthermore, random forest has shown the great ability to predict for survival (71) and endure an imputation technique to account for censored patients (31). Alternatively, the least absolute shrinkage and selection operator (LASSO) Cox regression model has also been reported reliably predict for survival in glioma (72, 73).

A third option is semi-supervised machine learning, wherein some complete clinical data is provided to the program generating the model, but other data is complete. For example, the program would have a range of radiomic features that it knows correspond to high grade glioma and a range of radiomic features that belong to an unknown clinical entity. Thus, all the dataset is used for a training step. The validation step is then a question if the program can correctly identify the unlabeled data. This process has been used to suggest brain tumor histology and prognosis (74).

#### PROGRESS OF RADIOMICS IN GBM

Radiomics has provided key insight into critical features of GBM, as advanced radiomic analysis seek to establish reliable associations between key clinical features and those features derived from images. For example, radiomics has been used to predict for clinical, proteomic (e.g., Ki-67 expression), genomic (e.g., IDH1 status) and transcriptomic characteristics (75–77). This evolution of the radiomics field has been titled multi-omics or radiogenomics, dependent on the source (21, 78–80). This will be part of the future of radiomics, as these details are pertinent to physicians due to their influence on treatment and prognosis (8). In addition, recent advancements have been made in defining radiomic subtypes. By utilizing T1 and FLAIR sequencing, researchers were able to define three distinct imaging subtypes—rim enhancing, irregular and solid. Each subtype represents a distinct phenotype enriched in unique molecular alterations such as MGMT methylation and EGFRvIII mutations (37). Continued advancements in defining tumor heterogeneity using imaging features may offer a complimentary means with which to characterize GBM and provide personalized treatments for patients.

Radiomics analysis has the capacity to answer critical questions facing clinicians such as the discrimination between pseudoprogression and progressive disease in GBM patients. For example, combining the diffusion tensor imaging and dynamic susceptibility contrast MRI features can improve accuracy treatment response and may aid in individualized treatment of patients with GBM (81). Recently, a deep radiomics model used the MR images with clinical features demonstrate the capacity to predict the PsP from progression for patients with GBM (82). While, another study showed that the radiomics analysis is not able to distinguish between true-progression and PsP (83). However, many of these steps exist in an early developmental stage. Combining all such information into an artificial intelligence model would be a promising direction to advance personalized medicine.

#### INTRATUMOURAL HETEROGENEITY AND RADIOGENOMICS

Perhaps the greatest utility of radiomics in the management of gliomas lies in the application of radiogenomics. Radiogenomics implements radiomics analysis to predict specific genetic characteristics. Classically, gliomas have been managed based on their grade—a histopathological characterization made by specialized physicians (neuropathologists) to articulate the likely behavior of the malignancy. Over the past two decades, molecular assessment of the tumor's genome, protein expression, and epigenetic state have become more common as the relevance of these features to outcome and/or therapeutic response is being increasingly understood (84). Given the relative abundance of high quality MRI data which accumulates over time during standard of care for glioma patients (85– 87) radiomics offers a potentially efficient and non-invasive method of tumoral evaluation (37, 88, 89). Indeed, recent efforts have generated radiomic signatures to predict the majority of information sought by classical histopathological and modern molecular assessments including: isocitrate dehydrogenase mutations (79, 90–92), 1p/19q codeletion loss of heterozygosity (24, 92, 93), O<sup>6</sup> -methylguanine-DNA methyltransferase promoter methylation (45, 94) and ATRX mutations (95). This has culminated in recent findings demonstrating a conserved radiomic signature can predict CD8+ T-cell infiltration and response to immunotherapy (96).

However, intratumoral heterogeneity significantly confounds both molecular and histopathological assessments as the entirety of a tumor cannot be assessed by neuropathologists. Disparate clonal populations may be minimally represented in histopathological sampling introducing sampling errors and limiting relevance for informing treatments (97–101). Radiomics offers an opportunity to overcome this limitation as analysis is performed upon the complete tumor enabling spatial mapping of distinct genetic features. In addition, radiomics offers the means to provide quantitative values (e.g., % of tumor mutated) rather than binary designations (e.g., mutant or not) to describe molecular features which may have important implications for predicting response to therapies. Utilizing co-clinical models, researchers are starting to establish radiomic signatures which are closely associated with specific molecular features in an attempt to describe intratumoral heterogeneity (102). Further development of pre-clinical models and correlation with clinical datasets will be essential to drive this field forward toward improving the utility of radiomics for diagnosis in GBM.

### FUTURE RADIOMICS

Radiomics needs massive amounts of biomedical data, socalled "Big data (103)," to validate it's deep-learning approaches and expanding applications. The development of strong public datasets has empowered these approaches, with such initiatives including The Cancer Genome Atlas (TCGA) (85), The Cancer Imaging Archive (86), and The Quantitative Imaging Network (87). However, there is still the barrier of segmentation such as acquiring clinician input to identify the relevant ROIs. While the clinician will still be sought as the gold standard, deep-learning strategies have the potential to define ROIs without the bias of human segmentation (104). To accomplish this, even larger datasets will be required—further emphasizing the need for reliable Big Data. These strategies have begun in part, but developing validated models to all the clinically relevant questions will simply require more data (105, 106).

The potential applications for radiomics is expanding, with logistical and technical challenges needing to be overcome prior to true clinical deployment. We view these as: (1) expanding what is included in and the access to Big Data, (2) establishing common criteria from image acquisition to feature definitions, (3) agreement on the clinical questions that radiomics must address, and (4) developing a clinically implementable and prospectively validated statistical model to answer those questions.

### CONCLUSIONS

This review explained how the vast amount of radiological data not used by the clinicians managing CNS malignancies can be used to generate radiological signatures that can predict the characteristics of these brain tumours. In a step-by-step process we outlined how this data can be used to predict for numerous pertinent biological outcomes. With constant progress in deeplearning processes and expanding public access to Big Data, radiomics has the potential to non-invasively address numerous clinical questions or support clinical decision making. There are numerous future directions for radiomics, but a continued focus on ensuring there is public access to large databases of clinical and radiological correlated data will be instrumental to seeing those directions leading to a desirable destination.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

AC conception and design. AC, PD, and MK drafting the manuscript and review of the literature. MK, PD, BJ-C, TN, SS, and BA critical revision of the manuscript.

### ACKNOWLEDGMENTS

Research supported by grant from Varian Medical System.

transformation of multispectral images. Front Oncol. (2018) 8:96. doi: 10.3389/fonc.2018.00096


high-grade gliomas from MR imaging. Clin Cancer Res. (2018) 24:1073–81. doi: 10.1158/1078-0432.CCR-17-2236


report and review of the literature. J Neurooncol. (2001) 55:167–71. doi: 10.1023/A:1013873431159


106. Hosny A, Parmar C, Quackenbush J, Schwartz LH, Aerts HJWL. Artificial intelligence in radiology. Nat Rev Cancer. (2018) 18:500–10. doi: 10.1038/s41568-018-0016-5

**Conflict of Interest Statement:** Author PD's fellowship was partly funded by a grant from Varian Medical System. The funder had no role in study design, data collection and analysis, or preparation of the manuscript.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Chaddad, Kucharczyk, Daniel, Sabri, Jean-Claude, Niazi and Abdulkarim. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.