Treatment-Related Adverse Events of Combination Immune Checkpoint Inhibitors: Systematic Review and Meta-Analysis

Background: Although clinical practice guidelines for the management of Immune Checkpoint Inhibitor (ICI)-related adverse events have recently been published, precise and nuanced toxicity data for combination ICI therapy are lacking. Therefore, herein we have conducted a systematic review and meta-analysis of published clinical trials on combination ICI to synthesize the treatment-related adverse event (TRAE) profile of combination ICI therapy. Methods: PUBMED, EMBASE, and the Cochrane Database/EBM were searched for eligible studies. Clinical trials evaluating combination immune checkpoint inhibitor therapy in advanced unresectable cancer were included in the analysis based on prespecified criteria. Risk of bias across studies was evaluated using Begg's funnel plot and Egger's regression test. The summary outcomes were pooled risk ratios (RR) and the logit-transformed proportion for incidence data. Results: A total of 18 studies comprising 2,767 patients across 10 cancer types were included in the final analysis. Combination ICI was associated with a slightly higher risk of all-grade adverse events (RR 1.07 [95% CI 1.03–1.11]) and markedly greater risk of grade 3 or higher adverse events (RR 2.21 [95% CI 1.57–3.10]) compared to monotherapy ICI. Subgroup analyses showed significant differences in risk of grade 3 or higher adverse events between treatment types (PD-1 + CTLA-4 and PD-L1 + CTLA-4), among cancer types, and among dosing regimens (N1I3, N3I1, and D20T1). The incidence of all-grade adverse events was 0.905 [95% CI 0.842–0.945], and the ratio of grade 3 or higher events to all-grade adverse events was 0.396 [95% CI 0.315–0.483]. The most common all-grade TRAEs were diarrhea/colitis, fatigue/asthenia, nausea/vomiting, rash, and pruritis. Conclusion: Combination ICI therapy has a significantly different treatment-related adverse event profile compared to monotherapy.


INTRODUCTION
Within a short span of time, with increasingly frequent use of immune checkpoint inhibitors across different types of cancer, knowledge and experience with immune checkpoint inhibitor (ICI)-related adverse events has also been increasingly accumulating (1)(2)(3)(4)(5)(6). Based on these accumulated data, clinical practice guidelines have been published to improve the management of these adverse events (7). Several questions remain unanswered, however, on the treatment-related adverse events (TRAEs) of ICI, especially in the setting of combination therapy. For instance, although there is a general consensus that combination ICI therapy results in higher risk of TRAEs compared to ICI monotherapy, data are unclear on whether this risk differs with different ICI combinations or across different cancer types and whether there are notable associations with certain ICI combinations and specific TRAEs (7,8). Furthermore, it is unclear whether the severity and frequency of adverse events are synergistic or just additive. Therefore, a systematic review of such adverse event data is necessary to guide informed decisions, both in clinical trials and in the clinic, for both clinicians and patients. Herein we conduct a systematic review and meta-analysis of the incidence of all-grade, grade 3 or higher, and individual TRAEs in combination ICI therapy vs. ICI monotherapy with the goal of synthesizing an accurate, precise, and comprehensive toxicity profile.

Study Selection
Eligibility assessment was performed independently in a nonblinded standardized manner by two reviewers. Disagreement between the two reviewers was resolved by discussion and consensus. Study inclusion criteria comprised the following: (1) randomized clinical trials; (2) studies in humans; (3) contains adverse event data. Exclusion criteria comprised the following: (1) review articles, meta-analyses, case reports, and case series; (2) conference abstracts and presentations. A data extraction form was developed a priori, and two reviewers in tandem conducted data extraction, with the final results reviewed by a third reviewer. If overlapping data were identified, the most recent or comprehensive study was included. Disagreement was resolved by discussion among the three reviewers. The  (9) country where the study was held; (10) adverse event data including total patients, number of total and severe adverse events, and total and severe adverse events for six selected specific TRAEs.

Statistical Analysis
Pooled risk ratios (RR) and 95% confidence intervals (CI) were calculated using a meta-analytical method that weighed the logarithm of the RR by the function of its variance for each study. Pooled incidence was calculated using the logit-transformed proportion equal to the log of xi/(ni-xi). A random-effects model or a fixed-effects model was chosen based on whether heterogeneity was significant. The Hartung-Knapp adjustment was used to fit the random-effects model, a continuity correction of 0.5 was used in studies with zero cell frequencies, and the Sidik-Jonkman estimator was used to derive tau 2 . Among the selected studies, only those studies containing both combination ICI and monotherapy ICI arms were included for the calculation of the pooled RR, whereas all studies were included in the calculation of the pooled incidence of selected TRAEs. If a study  Frontiers in Oncology | www.frontiersin.org contained more than one monotherapy arm, the combination arm was compared twice separately with each monotherapy arm. Subgroup analysis was conducted based on treatment combination, cancer type, control arm drug, and drug regimen. Publication bias was assessed using Begg's funnel plot and Egger's test, where a p-value <0.05 was considered statistically significant. Meta-analysis was performed using the package "metafor" of the R-project.

Study Selection
The initial database search yielded 3,813 studies. After selection of studies based on title and abstract review, 19 studies remained for full review, to which two studies were added on the basis of reference search and three studies were excluded based on our pre-specified criteria. The rationale for the addition and exclusion of each study has been summarized (Figure 1).

Subgroup Analyses
Subgroup analyses were conducted based on cancer type, treatment combination, treatment regimen, and monotherapy arm for grade 3 or 4 adverse events (Figures 4-7). Significant differences were found for all subgroup analyses except analysis by monotherapy arm for grade 3 or 4 adverse events; on the other hand, no significant differences were seen for all-grade adverse events ( Table 2).   (Tables 3, 4).  (Figure S1).

DISCUSSION
To the best of our knowledge, our meta-analysis is the largest, most comprehensive study on the TRAEs of combination ICI therapy to date. Previous meta-analyses have analyzed the adverse events of combination therapy as a minor subsection of the analysis, for a single tumor type, or for a specific adverse event site (3,5,6). Herein, we demonstrate that combination ICI therapy is associated with markedly greater risk of grade 3 or higher adverse events and marginally greater risk of allgrade adverse events compared to monotherapy. Furthermore, our study shows there are significant differences in risk of grade 3 or higher TRAEs when analyzed by type of cancer, treatment combination, and treatment dosing regimen. For grade 3 or 4 adverse events, subgroup analysis by type of cancer was limited by the small number of studies per type of cancer. Also, the studies included for each type of cancer differed in treatment combination. The risk of grade 3 or higher TRAEs tended to be higher in the melanoma subgroup compared to other cancer type subgroups despite that this subgroup comprised more studies with the N1I3 rather than the N3I1 regimen. This finding is at odds with the notion that anti-CTLA-4 agents in general are known to be more toxic than anti-PD-1 agents (9). This finding may be explained by a feature intrinsic to the tumor type or by between-study differences. Further accumulation of nivolumab + ipilimumab toxicity data in NSCLC is needed to clarify the reason for this difference.
Recent studies have suggested that the clinical response and toxicity of immunotherapy may be altered depending on patient-specific characteristics such as sex and ethnicity (10,29). Although our analysis did not conduct a subgroup analysis based on patient characteristics due to the limited number of studies, future studies are warranted on the analysis of differences in toxicity in combination ICI therapy based on patient characteristics.
Further subgroup analysis by treatment combination showed a significant difference between the risk of grade 3 or higher TRAE of anti-PD-1 + anti-CTLA-4 vs. anti-PD-L1 + anti-CTLA-4, with higher toxicity in the anti-PD-1-containing group, which is consistent with prior meta-analysis studies of checkpoint inhibitor toxicity. This finding is explained mechanistically by PD-L2 sparing in the context of PD-L1 inhibition. Whether this mechanism translates into lower anti-tumor activity in head-tohead comparisons of anti-PD-1 vs. anti-PD-L1 remains to be answered, and future studies, including clinical trials or network meta-analyses, exploring this question are warranted.
Our pooled incidence analysis suggests that all-grade TRAEs occur in about nine out of 10 patients treated with combination ICI therapy and that four out of ten patients experience grade 3 or higher events. Previous studies suggested that the risk of grade 3 or higher immune-related adverse events ranges from 15 to 42% in anti-CTLA-4 monotherapy, 5 to 10% in anti-PD-1 therapy, and 1 to 7% in anti-PD-L1 therapy (7). The findings of our study are consistent with these previous data, as the incidence of toxicity seems to be driven by the dose of anti-CTLA-4 in the combination therapy regimen. Future studies are necessary to determine whether the toxicity in combination therapy is synergistic or additive.
The results of this study show that an increased risk of adverse events, especially of those that are grade 3 or 4, should be taken into consideration when assessing the benefits of combination therapy. Immune therapy, even as monotherapy, has shown remarkably durable responses in multiple types of cancers, including melanoma, non-small cell lung cancer (NSCLC), renal cell carcinoma, solid-tumors positive for mismatch-repairdeficiency or high in microsatellite-instability (dMMR/MSI-H), and gastric or esophageal cancer (30). However, subsets of patients within these types of cancer and other types of cancer such as mismatch-repair-proficient or microsatelliteinstability-low (pMMR/MSI-L) colorectal cancer have primary   (12,17,26). Nevertheless, as longer-term OS and progression-free survival (PFS) data become available for combination ICI trials, a more formal risk-benefit analysis should be undertaken. Our study has several notable strengths. First, this is the first study to our knowledge to have compared different ICI combination regimens for toxicity with sufficient power and potential generalizability across cancer types. Second, all-grade, high-grade, and individual adverse events were all analyzed in this study for in-depth analysis of combination ICI therapy toxicity. Third, this meta-analysis included only randomized clinical trials of combination therapy. Because all randomized trials of combination ICI therapy were completed relatively recently and obviously came after monotherapy trials, reporting of adverse events had evolved compared to the earlier trials of monotherapy and are thus expected to be more accurate and standardized, meaning as a whole that this meta-analysis of only combination therapy trials is comprised of higher quality data compared to previous meta-analyses containing monotherapy trials.
A number of limitations should be noted in this study. Significant heterogeneity is seen across included studies, as with any other meta-analyses. The sources of heterogeneity include but are not limited to cancer types, trial phases, number of previous treatments, criteria or threshold for reporting adverse events, and therapeutic dosages. Some degree of heterogeneity was tolerated for the sake of inclusivity in this study. Furthermore, extensive subgroup analyses were conducted to enhance the sensitivity of this analysis.
In conclusion, combination ICI is associated with significantly higher incidence of severe adverse events compared to ICI monotherapy, and the risks of certain adverse events are markedly increased compared to others. With the publication of ongoing combination ICI therapy trials and new toxicity data, future studies should control for treatment dosages to confirm cancer-type-specific differences in toxicity and clarify PD-L1-containing regimen toxicity. Furthermore, analysis of therapeutic combinations not studied herein should be carried out, especially combinations containing pembrolizumab, given its growing number of indications in different cancer types.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
RP contributed substantially to the conception, design of study, acquisition of data, quality control, data analysis and interpretation, statistical analysis, manuscript preparation, editing, and review. LL contributed substantially to the data acquisition, quality control of data and algorithms. CC and IR contributed substantially to the data acquisition and quality control. AS contributed substantially to the conception and design of study, manuscript preparation, editing, and review.