Patients admitted from May 2013 to November 2018 harboring a radiological diagnosis of giant insular gliomas (i.e., a mass lesion involving all four insular zones) and candidate for resection were included. All patients gave written informed consent to the surgical procedure, covered by Ethical-Committee IRB-1299 Humanitas Research Hospital.

MR images were performed on a Philips Intera 3.0T pre-operatively, within 48 h of and 1 month after surgery. The protocol for lesion morphological and dimensional assessment and EOR estimation included: a) axial three-dimensional-FLAIR; b) post-gadolinium three-dimensional-T1-weighted fast-field-echo (MPRAGE); c) DWI and ADC imaging.

To calculate tumor volume (in cm³) and EOR, volumetric scan images were analyzed with semi-automatic segmentation using iPlanCranial software (BrainLab, Germany) by four blinded investigators (MRo, LG, MCN, TS); FLAIR hyperintense or T1-weighted gadolinium-enhanced signal abnormalities were included in the lesion load for non-contrast enhancing lesions or high-grade gliomas, respectively. The EOR corresponded to the percentage of preoperative tumor volume resected: (preoperative volume − postoperative volume)/preoperative volume (15). It was classified as: Gross Total Resection (GTR, EOR = 100%); Subtotal Resection (EOR <100–90%); Partial Resection (EOR <90%). In the analysis, Subtotal and Partial Resections were considered together.

Individual vascular anatomy and vessels relation to tumor mass were evaluated on preoperative post-gadolinium three-dimensional-T1-weighted images, ischemia on immediate postoperative DWI; the amount of DWI signal was considered significant when more than a small punctum or small rim, i.e., focal areas >2 ml in volume around the resection cavity or in other subcortical areas were detected. The location of the DWI abnormalities and eloquent subcortical sites involvement were evaluated; a subcortical site was considered eloquent when located within the course of tracts mediating language, cognitive, motor, praxis, or visual functions.

All surgical resections were pursued up to functional boundaries (motor, language, haptic, visual, and cognitive), aiming to achieve a Gross Total Resection (GTR) whenever possible, without any patient or tumor a priori selection. Surgery was performed under either asleep-awake-asleep anesthesia or general anesthesia, with the aid of Mapping and Monitoring technique (16). Mapping includes Low- and High-Frequency (LF, HF) direct electrical stimulation (DES) (17). Monitoring includes EEG, electrocorticography (ECOG), motor evoked potentials (MEPs), somatosensory evoked potentials (SEPs) recordings. Patients were positioned supine, with the shoulder elevated of 30° and the head tilted of 30° toward the tumor opposite side. A large fronto-temporal flap was designed. The craniotomy, performed under general anesthesia, exposed the tumor area and an amount of surrounding tissue, including functional cortical landmarks, i.e., the ventral pre-motor area (vPM) and the motor cortex (M1) ( Figure 1 ). Cortical mapping was initially performed with HF stimulation (Train of Five, ISI 3) delivered by a monopolar probe to identify M1, specifically the hand knob, looking for the cortical motor threshold (cMT, i.e. the lowest current intensity which produced the lowest MEP response) for three hand muscles (Abductor Pollicis Brevis, Extensor Digitorum Comunis, Abductor Digiti Minimi) (18). The cortical strip was then placed over this area for continuous ECoG and MEP recordings. In asleep–awake–asleep procedures, the working current interfering with counting was then established over vPM. The awake phase consisted of both a cortical and a subcortical step. In the cortical step, LF stimulation delivered with a bipolar probe mapped functional cortical sites; negative cortical sites were identified to design the safe cortical entry zone for corticectomy, over the frontal and temporal lobes. In the subcortical step, with the aid of subcortical mapping by LF stimulation, functional limits were sought at the periphery of the opercular windows outlined, starting from the corticectomies. Positive sites (discrete regions where DES produced transient functional disturbance) where marked and secured with patties, representing the surgical resection limits. In this way, a functional disconnection of tumor frontal and temporal portion was progressively obtained. The neuro-psychological tasks performed were tailored to the region of stimulation. Language mapping was performed with naming and semantic association tasks. Praxis mapping using the Hand Manipulation Task (hMT) (19). Cognitive mapping (attentive and executive functions and memory) was eventually deployed (20). To preserve visual function during temporal lobe resection the Visual field Task was utilized (21). The last phase, proper tumor removal, was performed under general anesthesia also in asleep–awake–asleep procedures. Frontal and temporal portions of the tumor were resected according to the functional boundaries previously defined. By this maneuver, exposure of the insular lobe was achieved either through the frontal or the temporal window. Afterward, the insular portion of the tumor was removed with the aid of HF motor mapping by monopolar probe, up to a 2–3 mA sMT was eventually reached, and continuous MEP, SSEP, and EEG monitoring. Dissection was discontinued when MEP recordings showed either fluctuation of at least 2 out of the 3 recorded hand muscles or 50% amplitude reduction. Irrigation with saline and increase in arterial pressure in these circumstances were applied to let MEP recover.

According to the different surgical strategies adopted, the series was categorized in two groups: 1) 28 patients treated before 2015, among which less than 60% of the surgeries (all dominant hemisphere lesions) were performed under awake settings, mostly with mapping limited to motor function and language; 2) 67 patients treated after 2015, among which all patients, if clinically feasible and irrespective of hemisphere involved, were operated under asleep-awake-asleep regimen (>75%) with an extensive mapping involving also cognitive, haptic, and visual tasks.

Examinations were performed before, on the 5th day, 1–3 months, and 1-year after surgery. Morbidity was assessed by evaluating postoperative new-onset neurological deficits and the occurrence of intraoperative complications (e.g., intra-cavitary bleeding or ischemia). Intraoperative mortality was also reported.

The neuropsychological profile was evaluated with the Milano-Bicocca-battery (22), assessing five neurocognitive domains: language, memory, visuospatial abilities and praxis, attentive and executive functions, and fluid-intelligence. Each test score was corrected for patient age and educational level to obtain a normalized value and converted to an equivalent score on a scale of five values from 0 to 4 (0 = pathological, 1–4 = different degrees of normal performance) (23). The division of the corrected scores into five different regions, each corresponding to an equivalent score, was based on the distribution of the scores observed in a control population; the first region (equivalent score = 0) corresponds to the worst 5th percentile results observed in the control population, while the region on the opposite end (equivalent score = 4) corresponds to the corrected scores in the control population between the median and the maximum score. The other three regions (1, 2, and 3) are defined between these two extreme regions so that the ranges of the corrected scores observed are evenly spaced between the normality reference threshold and median. This conversion in equivalent scores allowed the comparison of test scores among different subjects. For the analysis, pathological versus normal test scores were considered.

Quality of Life was assessed as global performance status, ability and time to return to work, and by the Short Form (36) Health Survey questionnaire (SF-36) and Hospital Anxiety and Depression Scale (HADS) (24). To provide a single score in the SF-36, we assumed the “General Health scale” (GH scale) score as an indicator of the QoL. The GH scale scores of patients under analysis were compared to the average score recorded in the Italian general control population; patients with a score one standard deviation below the average score of the Italian control group were classified as patients with a low perception of their QoL (25). The HADS Italian version cut-off point of 10 was adopted to identify patients with emotional disorders (26).

Factors possibly associated with EOR and functional outcome included in the analysis are illustrated in Table 1 .

Peri-insular vascular structures, assessed on preoperative 3.0T volumetric T1-weighted post-Gadolinium scans, included: 1) lenticulostriate arteries (LSA complex) that branch off the M1 segment (less frequently M2) of the MCA; the relation of LSA to tumor mass was categorized as “LSA mesial to the tumor” when tumor mass got in proximity to or medially displayed the LSA and their origin without invading these vessels, and as “LSA within the tumor” when tumor mass encased the LSA. 2) Other deep perforators, branches that stem at the origin of the AChA from the internal carotid artery (ICA) and supply lateral part of the geniculate body, posterior two-thirds of the posterior limb of the internal capsule, most of the globus pallidus, the origin of the optic radiations, and the middle third of the cerebral peduncle; the relation of deep perforators to tumor mass was categorized as “tumor invasion” when these branches were encased by tumor volume, or as “no relation” when tumor mass medially displaced the deep perforators without engulfing them. 3) Opercular branches of the MCA (M2–M3 segments), specifically from the superior trunk, and running over the insular cortex: orbitofrontal, prefrontal, and precentral arteries; the relation of these structures was categorized as “all involved” by tumor when all the opercular branches were within the tumor mass, versus “not all involved” when at least one of the opercular arteries (anterior or precentral) was not involved by the tumor.

Chi-Square and Fisher’s exact test evaluated the association among factors. Continuous variables were categorized according to cutoffs reported in  Table 1 . The predictive value of the significant variables resulting from the univariate analysis was further assessed by multivariate regression analysis. Exact McNemar’s test was used for matched comparisons. Tests were performed with SPSS version 24.

During the investigated period, out of 343 patients with insular gliomas consecutively admitted to our unit, 95 (27.7%) had a giant insular glioma ( Table 1 ). Mean tumor volume was 76.74 ml. Most patients presented with seizures (73.7%); 43 out of 70 patients had poor seizure control, despite AEDs. Mild language and/or motor deficits were registered in 28 patients (29.5%).

A transcortical approach was always adopted. When awake mapping was performed (70 cases, 73.6%) the procedure was well tolerated (reported patient fatigue intraoperatively in 2.1%, absence of mapping abortion). Severe hand muscles MEP changes were observed in four cases; two of these recovered after surgical manipulation discontinuation, arterial pressure increase, and saline irrigation. The other two patients presented clinically relevant (contralateral hemiparesis) ischemic insult documented also on post-op DWI. All patients, at the 1-year evaluation, recovered to normal baseline neurologic examination.

Mean EOR was 92.3%, median 99%; a GTR was obtained in 70 patients (73.7%).

The achievement of GTR was not associated with any pre-operative clinical factor, except for seizure control in the pre-op that was significantly associated with a higher chance of GTR ( Table 1 ). LSA relation to tumor mass, differently from current literature (27), was irrelevant. Instead, tumor involvement of the other deep perforators, opposed to medial displacement of these vessels by tumor mass, was associated with a minor chance of achieving a GTR (p = .012). EOR was associated with the type of brain mapping (p = .010): asleep–awake–asleep procedures with extended brain mapping with cognitive, haptic, and visual other than merely language testings, displayed a higher proportion of GTR.

Globally, the procedure was safe, and no intraoperative complications were recorded. No peri-operative mortality was documented.

In the pre-op, 28 patients presented with neurological disturbances ( Tables 1 , 2 ). Immediately after surgery, half of patients developed new-onset postoperative deficits. The function most affected was language (32.6% of total, 75% of patients with dominant lesions). At one month, more than half of these patients recovered. At one year, language deficits persisted in three patients only (3.2% of the total, 6.3% of patients with dominant lesions) and affected speech-production in two and speech-production and comprehension in one patient. No motor deficits were reported. Two patients (2.1%) presented persistent superior quadrantanopia.

Forty-three patients had no or poor seizure control before surgery, despite AEDs. After surgery, poor seizure control was observed in seven patients only (Engel Outcome Scale class IVA and IVB) (Exact McNemar’s test, pre-op versus post-op, p = .000). Thirty-eight of 43 (88.4%) patients with no preoperative seizure control were seizure-free 1-year after surgery (Engel Outcome Scale class I); no increase in AEDs dose-number was reported. EOR was associated with seizure control (p = .001); six patients (24%) with subtotal/partial resection experienced poor seizure control compared to only one patient (1.5%) of those with GTR ( Table 1 ).

Pre-operative neuropsychological scores were more affected than neurological performance ( Table 2 ). Forty-eight patients (50.5%) presented pathological scores in at least one of the domains analyzed, most in memory (48.4%) and language (33.3%). Five days after surgery, most patients declined in at least one domain. Substantial recovery resulted at 3-month evaluation in all domains, but language and attentive and executive functions; language presented the slowest recovery rate. At 1-year evaluation, however, only six patients (6.3%) presented new-onset deficits in language, all in production (naming and/or verbal-fluency), three also in comprehension.

QoL was evaluated considering the global performance status and ability and time to return to work. Seventy-four patients returned to work at 3 months, 92 at 6 months. Only 72 patients maintained the same working level. QoL evaluation (SF-36 and HADS scores) exhibited the same neuropsychological profile trend ( Table 3 ).

Neurological and neuropsychological outcomes and EOR were not associated with most of the clinical or radiological factors under analysis ( Table 1 ). Pre-op seizure control, brain mapping strategy, and tumor relation to deep perforators were significantly associated to the EOR. To assess the predictive value of these variables on EOR, a backward stepwise binomial logistic regression was performed. The logistic regression model was statistically significant, c(3) = 20.236, p <.0001. The model explained 28% (Nagelkerke R) of the variance in EOR and correctly classified 78% of the cases. Specifically, all the three variables were statistically significant (pre-op seizure control, p = .006, Exp(B) = .216; relation to other deep perforators, p = .021, Exp(B) = 3.370; brain mapping, p = .012, Exp(B) = .249).