Diagnosis and Management of Hematological Adverse Events Induced by Immune Checkpoint Inhibitors: A Systematic Review

There has been less volume of literature focusing on the Immune-related Hematological Adverse Drug Events (Hem-irAEs) of Immune Checkpoint Inhibitors (ICPis) in cancer patients. Furthermore, there has been no consensus about the management of hematological toxicity from immunotherapy in the recently published practice guidelines by the European Society for Medical Oncology (ESMO). We conducted a systematic review of case reports/series to describe the diagnosis and management of potentially rare and unrecognized Hem-irAEs. We searched Medline, OVID, Web of Science for eligible articles. Data were extracted on patient characteristics, Hem-irAEs, and management strategies. We performed quality assessment using the Pierson-5 evaluation scheme and causality assessment using the Naranjo scale. Our search retrieved 49 articles that described 118 cases. The majority of patients had melanoma (57.6%) and lung cancer (26.3%). The most common Hem-irAEs reported with ICPis (such as nivolumab, ipilimumab, and pembrolizumab) were thrombocytopenia, hemolytic and aplastic anemias. Less reported adverse events included agranulocytosis and neutropenia. Steroids were commonly used to treat these adverse events with frequent success. Other used strategies included intravenous immunoglobulins (IVIG), rituximab, and transfusion of blood components. The findings of this review provide more insights into the diagnosis and management of the rarely reported Hem-irAEs of ICPis.


INTRODUCTION
In the past decade, the enthusiasm for connecting the immune system and malignancy has expanded. Exploiting the host's immune system to treat cancers depends on immune surveillance: the ability of the immune system to identify foreign neo-antigens and target them for elimination (1). Immune checkpoint receptors, i.e., cytotoxic T-lymphocyte-associated protein 4 CTLA4 antibody ipilimumab, and programmed cell death protein-1 (PD-1) are critical for the physiological responses of the immune system. Checkpoint signaling triggers immune tolerance of T-cell activation to avoid autoimmunity and the adverse effects of excessive inflammatory responses. Tumor cells utilize these mechanisms to avoid destruction by the immune system (2).
In August, 18, 2010, the FDA approved the CTLA-4 ipilimumab antibody as the first ICPi for the treatment of metastatic melanoma (3). The filing was based on results from the primary analysis of the pivotal MDX010-020 trial, which were published online in the New England Journal of Medicine and presented in June 2010 during a plenary session at the 46th Annual Meeting of the American Society of Clinical Oncology (3). Despite its approval, ICPis have not been widely used except in the last 2 years. Recently, PD-1 inhibitors were approved for the treatment of non-small cell lung cancer (NSCLC) (4). Following their approval, these immunotherapeutics became integral parts of the treatment protocols against melanoma and NSCLC. Furthermore, they have shown promising responses [objective response rates (ORRs)] against different cancers, including mismatch repair deficient colorectal cancer (60%) and Hodgkin's disease (65-85%) (5).
Although the side effects of immunotherapy are less than chemotherapeutic agents (4), immunotherapy still may cause dermatological (reticular, maculopapular erythematous rash, and mucositis), gastrointestinal (diarrhea and colitis), hepatic (elevation of liver enzymes in serum), and endocrine adverse effects (involving pituitary, adrenal, or thyroid glands). This is because the immune response triggered by these drugs is not completely tumor-specific (6). The management of their adverse events usually includes various forms and regimens of corticosteroids (7).
With the expanding use of ICPis in clinical practice, more rare side effects are being discovered. Some Hem-irAEs were described, including immune thrombocytopenia, autoimmune hemolytic anemia, agranulocytosis, or pure red cell aplasia (8). The evidence focusing on the Hem-irAEs of ICPis is scarce. Moreover, there is no consensus on the management of hematologic toxicity from immunotherapy in the recently published practice guideline by ESMO (9). We aimed to evaluate the published literature on this topic and summarize the successful management approaches of the rare side effects.

Data Sources and Searches
We commenced this study in May 2018 and included all available updates published since 2008 till the present time.
We conducted literature search using different databases: Medline, OVID, and Web of Science. Furthermore, we searched the gray literature; conference proceedings; using Web of Conferences, Open Grey up to January 2019. We searched the bibliographies of relevant studies for any eligible case reports/series up to January 2019. The flow of the article selection process is presented in the graphical abstract as Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) figure. We used no time limit to date.
Initial screening of the eligible articles was done independently by two authors NO and NE. The articles were screened first based on their titles and abstracts, and then the full text was reviewed to decide the eligibility. Any conflict was solved by a third author KE. Only full-text articles published in peer-reviewed journals were retrieved for review according to the following criteria. AA, MY, AH, SE contributed to data analysis. 10. Use of non-FDA approved ICPis up to the date of data extraction.

Data Extraction and Quality Assessment
Data was extracted by NO and NE, then was revised by KE. The extracted data included type of cancer, ICPis, number of cases, Hem-irAEs, onset of the adverse events, management of Hem-irAEs, and management outcomes. We used the Pierson-5 evaluation scheme (57) to assess quality of case reports based on 5 domains: documentation, uniqueness, educational value, objectivity, and interpretation as shown in Table 2. Each domain is scored, for example (0, 1, or 2 points, the upper score is 10 points). When a case report scores 9-10 points, the report contributes to the literature; a 6-8 points indicates validity and clinical value of the report are doubtful; a 5 points or less indicates insufficient quality. The assessment was carried out by KE and a random sample was crosschecked by NO and NE.

Causality Assessment
Each case report was assessed according to Naranjo scale (58) for causality as shown in Supplementary Table 1. Naranjo scale was used for causality assessment of the case reports, that allows categorical classification of adverse events as "definite, " "probable, " "possible, " or "doubtful" based on the answers to 10 questions. One investigator KE carried out the assessment and NO and NE randomly re-checked it.

Patient Characteristics
Patient characteristics are described in Table 3

Nivolumab
Seventeen case studies (out of 49) reported Hem-irAEs with nivolumab in 20 cases (13 lung cancer, 5 melanoma, 1 cutaneous squamous cell carcinoma, and 1 glioblastoma). Anemia was reported in 7 cases; two had aplastic anemia and five had   hemolytic anemia. Thrombocytopenia was reported in five cases. Bone marrow failure was reported in three cases, pancytopenia in one case, neutropenia in one case, red cell aplasia in one case, hemophagocytic syndrome in one case, agranulocytosis in one case and acquired hemophilia A in one case. Treatment was reported for all patients. Resolution of the adverse events was reported in 11 cases (55%) and treatment was ineffective in 8 cases (40%). One case showed partial and transient response to treatment. In the 11 cases that showed response, the most common treatment for Hem-irAEs was IV corticosteroids, however, IV romiplostim, platelets transfusion, IVIG, and oral steroids were used. Many patients had to discontinue nivolumab with the treatment used.
Another two-case series reported Hem-irAEs with nivolumab in 27 cases. An increase in the absolute eosinophil count was reported by Bernard-Tessier et al. (54). No treatment was mentioned in this report. Delanoy et al. (56) reported neutropenia, anemia, thrombocytopenia, pancytopenia, bicytopenia, pure red cell aplasia with nivolumab, pembrolizumab, and atezolizumab. Twenty one patients had resolved symptoms with oral steroids, IV steroids, IVIG, and rituximab.

Pembrolizumab
Twelve reports described Hem-irAEs with pembrolizumab in 13 cases (7 melanomas, 4 lung cancer, and 1 bladder cancer). In these cases, hemolytic anemia was reported in five cases and thrombocytopenia in two cases. Neutropenia, pancytopenia, red cell aplasia, hemophagocytic lymphohistiocytosis, and Evan's syndrome were reported in one case each. Adverse events were resolved in 11 cases. Steroids (whether IV or oral) were used in all the managed cases, and IVIG was used in five cases.

Combination of Ipilimumab-Nivolumab
This combination of ipilimumab and nivolumab, used to treat metastatic melanoma, was associated with Hem-irAEs in 6 cases (5 reports). Thrombocytopenia, aplastic anemia, and hemolytic anemia were reported in two cases each. The adverse events were resolved in 5 cases. One case died with refractory aplastic anemia. Rituximab was a commonly used treatment; one patient with thrombocytopenia recovered after 4 doses of rituximab following IVIG failure. The second occurrence of hemolytic anemia in one of the cases resolved with rituximab use.

Durvalumab
A fatal allo-and immune-mediated thrombocytopenia was reported with durvalumab use in one NSCLC case. Platelet transfusion, polyvalent immunoglobulins and steroid treatments were used to treat the patient without improvement.

Avelumab
One patient with metastatic Merkel cell carcinoma developed lethal immune thrombocytopenia (ITP) after avelumab administration. Treatments with steroids and IVIG were ineffective and the patient died after 1 month from initial diagnosis.
Concerning the treatment of Hem-irAEs reported, steroids were the most commonly used (80/118, 67.7%), with a failure rate of (16/80 = 20%) out of 118 cases. Other treatment options included IVIG, rituximab, and combination of the three options at varying doses. Table 2 shows quality assessment of the extracted citations using Pierson-5. The number of case reports is based on five domains: uniqueness, documentation, objectivity, interpretation, and educational value. Every domain is scored with 2 points, the upper score is 10 points. Naranjo scale was used for causality assessment of the case reports, that allows categorical classification of adverse events as "definite, " "probable, " "possible, " or "doubtful" based on the answers to 10 questions.

Quality Assessment
Fifty-four case reports were retrieved from the literature and assessed. Out of the 54 reports, 5 (9.2%) could not be assessed, since the data presented were insufficient for assessment for 4 of them, while 1 study was an observational study. Seven cases (12.9%) were rated as "of insufficient quality for publication" because they scored 5 or less. The second case reported in Shiuan et al. (50) got zero score in the five domains. Twenty-six studies (48.1%) were assessed as "reader should be cautious about validity and clinical value of the report" because they scored 7-8. Twenty-one cases (38.8%) were rated as "likely to be a worthwhile contribution to the literature" as they scored 9-10.

Causality Assessment
Eight studies were ranked as "possible" adverse drug reaction, scoring 3 (one study) and 4 (7 studies). Two studies were not assessed because their data were insufficient. Sixteen studies were ranked as "probable" adverse drug reaction as they scored between 5 and 8. No cases were ranked as "definite" or "doubtful" adverse drug reaction. For pembrolizumb case reports (13 reports), 8 of them (61.5%) were assessed as probable Hem-irAEs. Next to pembrolizumab, nivolumab (20 reports), 12 of which (60%) were assessed as probable, then comes ipilimumab (14 reports), 8 of which (57%) were assessed as probable. For the combination of ipilimumab and nivolumab (6 reports), 3 of them (50%) were assessed as probable. Finally, only one case report was assessed for durvalamab where the causality assessment yielded as a possible Hem-irAEs.

DISCUSSION
Immunotherapeutics are increasingly used in cancer patients. However, adverse events can limit their use and may result in serious adverse outcomes, including death. While some adverse events have been well-described in clinical trials (e.g., dermatitis and colitis), other inflammatory and autoimmune manifestations are reported. Case reports can provide vital clues and signals to identify rare but serious events and can generate hypotheses that can direct ongoing scientific research. We conducted a systematic review of case reports/series of patients treated with checkpoint blockade to identify the scope of rare Hem-irAEs that may occur with these therapies. We included publications that had adequate description of the clinical manifestations of the patients reported.
This systematic review showed thrombocytopenia, hemolytic and aplastic anemias as the most commonly associated with ICPis use, i.e., nivolumab, ipilimumab, and pembrolizumab. Less reported adverse events included agranulocytosis and neutropenia. Steroids (either intravenous or oral) were commonly used to treat these adverse events with frequent success. Other strategies used IVIG, rituximab and transfusion of blood components.
The mechanisms of the recorded adverse events in the included articles remain elusive. The most plausible theory is activation of T-cells, leading to the secretion of different cytokines from T-helper cells and consequent tissue infiltration with cluster of differentiation 8 (CD8) T-cytotoxic cells (59). Another suggested mechanism was immune-mediated dysfunction in hematopoietic cell maturation and proliferation, yet, the exact intermediate mechanism is unknown (20). The response to steroids in the majority of these cases potentiates the theory of immune-mediated mechanisms that occur centrally (in the bone marrow) or peripherally (in the circulation).
We used the Naranjo scale to infer causality of the reported adverse event to the used ICPi drug. Although data were not available for some reports, we showed possible or probable causality in several included reports. In some of these reports, the ICPi was the only new treatment introduced and the events diminished after the drug withdrawal. Further, the temporal relationship between ICPis administration and the occurrence of the adverse effect implicates these drugs. Hem-irAEs are known to occur within 12 to 16 weeks of treatment initiation (60).
As reflected from the causality assessment results, the majority of cases reported were "probable"; being at the near top of the causality continuum of the Naranjo scale (just before definite). Consequently, the association between ICPis and Hem-irAEs cannot be ignored.
This review provides insights into the proper management strategies for Hem-irAEs. Previously, it was thought that cancer patients receiving immunotherapy should not receive immunosuppressive drugs. This view has significantly changed over the past few years and the use of immunosuppressive agents has been proven not to impair the efficacy of ICPis (61). Corticosteroids should be the first resource and some reports highlighted the benefit of high dose steroids therapy. In grade 3/4 adverse events, the ICPis should be discontinued and steroids can later be tapered off in 4 to 6 weeks with close monitoring of blood counts (7). Other immunosuppressive drugs as IVIG, rituximab or tumor necrosis factor antagonists may also be effective. In case the immunosuppressive therapy is prolonged, immunization against pneumocystis is recommended (4).
Definitions of the side effects in the registries of rare events are poor. Therefore, we focused on the qualitative features such as demographic characteristics of patients, diagnosis and management. We did not perform quantitative analysis of these case reports because risk analysis was not possible. Randomized clinical trials were not related to our objective and were excluded in this systematic review. Limiting the inclusion criteria to studies published in English was challenging. However, a former analysis showed that this language limitation does not usually alter the study results (62).
Future case reports/series should follow a standardized approach in reporting their patients characteristics and findings. Further attention should be given to Hem-irAEs in ICPis randomized controlled trials to provide higher quality data in this regard. Moreover, the mechanisms of these adverse events should be investigated on the molecular and cellular levels to specify more effective pharmacological interventions. The management of Hem-irAEs in patients receiving ICPis needs evidence-based guidelines to inform future practice and research in this area.
Concerning the factors that may have predisposed patients to the adverse effects, there was no clear pattern for age. Patients characteristics were heterogenous for age with high interpatient variability with median age of 54 years and wide range 32-85 years. For gender, most patients were males (n = 73, 61.8%); although the percentage is not conclusive, it warrants further investigations and more research.
There was no predictor for the response to treatment. However, steroids were the most commonly used option. This can be explained secondary to its relative availability, low cost, and physicians' experience compared to other options. Furthermore, steroid was not always successful (20% failure rate) which implies seeking other treatment options and keeping patients on steroids for Hem-irAEs closely monitored.

CONCLUSION
Although rare, Hem-irAEs are serious adverse events that may be associated with checkpoint blockade therapy. Depending on the grade of the adverse event, the ICPi therapy may be discontinued and steroid therapy should be initiated. Steroids were the most commonly management strategy with considerable failure rate. There were no detected underlying factors predicting the outcome to steroid therapy. Other promising management strategies for some events include IVIG, rituximab, and transfusion of blood components.

FUTURE RESEARCH RECOMMENDATION
Further research should focus on the plausible mechanisms contributing to these adverse events, to develop more specific management strategies.

DATA AVAILABILITY STATEMENT
Datasets are available on request from the authors.

AUTHOR CONTRIBUTIONS
NO and NE extracted eligible articles. KE-F conducted initial screening of the eligible articles. Any conflict was solved by KE-F. The assessment was carried out by KE-F. A random sample was cross checked by NO and NE. AA, MY, AH, and SE contributed to the analysis. DJ, AA, AB, and AN contributed to writing of the manuscript and discussion. SD contributed to the discussion and reviewing the scientific background. All authors approved the article for submission.