Assessment of the Clinical Trials Safety Profile of PD-1/PD-L1 Inhibitors Among Patients With Cancer: An Updated Systematic Review and Meta-Analysis

Background Understanding the safety and adverse event profiles of PD-1/PD-L1 inhibitors is important in guiding cancer immunotherapy. Consequently, we designed this meta-analysis to evaluate the safety of PD-1/PD-L1 inhibitors in clinical trials involving cancer patients. Methods Four safety indicators comprising treatment-related adverse events, death, discontinuation of therapy and grades 3–5 adverse events were evaluated using the random effect model. The quality of enrolled trials was assessed using the Newcastle Ottawa Scale (NOS). Results Forty-four clinical trials were included in the final meta-analysis. Compared with chemotherapy, the risk of death due to the use of PD-1/PD-L1 inhibitors was much lower than that experienced in the control group (OR = 0.65, 95%CI: [0.47, 0.91], I2 = 0%, Z = 2.52 (P = 0.01)). Similar observations were apparent regarding the other three indicators of safety and also when the use of PD-1/PD-L1 inhibitors alone is compared with the combined use of PD-1/PD-L1 and CTLA-4. When used together with chemotherapy, PD-1/PD-L1 inhibitors increased the incidence of the adverse events as compared to the use of chemotherapy alone. Increased risks for adverse events were also noticed with the use of PD-1/PD-L1 inhibitors over the use of a placebo. Conclusion The use of PD-1/PD-L1 inhibitors alone is associated with a better safety profile compared to either the use of chemotherapy or the use of PD-1/PD-L1 inhibitors with other anticancer regimens.


INTRODUCTION
Cancer immunotherapies, including immune checkpoint inhibitors (ICIs) and adoptive cell therapy, have revolutionized the treatment landscape and improved the survival prognosis for most cancer patients (1). Among these, PD-1/PD-L1 inhibitors are the most common type of immunosuppressants used in the treatment of solid tumors (1)(2)(3)(4). PD-1/PD-L1 inhibitors can block the interaction between tumor cells and T cells, restore the immune recognition function of T cells, and then eliminate tumor cells (1)(2)(3)(4). The unique anti-tumor mechanism of PD-1/ PD-L1 inhibitors means that the toxicities caused by these agents are also different from other traditional anti-tumor drugs (1).
Although PD-1/PD-L1 inhibitors have shown remarkable clinical benefits in the treatment of tumors, the spectrum of immune-related adverse events (irAEs) that affect body organs are a major concern with the use of these agents (5,6). Serious adverse events are a frequent limitation in the use of PD-1/PD-L1 inhibitors among cancer patients (5)(6)(7)(8)(9). It, therefore, behooves clinicians to conduct adequate and elaborate systematic assessment of potential recipients of these therapies, to ensure that the benefits outweigh the potential risks in the use of PD-1/PD-L1 inhibitors. In view of the limitations of previous meta-analyses regarding the safety and toxicity of PD-1/PD-L1 inhibitors (10)(11)(12), and the availability of recent information from results of clinical trials, we designed this study to reassess the safety of PD-1/PD-L1 inhibitors in cancer chemotherapy.

METHOD
This study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) (13).

Selection Criteria for Clinical Trials
All randomized, open-label, controlled clinical trials with efficacy and safety data of PD-1/PD-L1 inhibitors were explored. Although Phase III clinical trials were given priority, the phase of clinical trials was not the primary inclusion criterion. Malignancies were limited to solid tumors and, as such, hematological tumors were excluded from the study. The four safety indicators evaluated in the meta-analysis were: a) treatment-related death, b) treatment-related adverse events leading to discontinuation of therapy, c) treatment-related grades 3-5 adverse events and d) any treatment-related adverse events.
Articles were only included if they were published in English between 09 July 2013 and 19 Sep 2020. Three researchers were designated to independently scrutinize all the data and where there was duplication of clinical trials in selected articles, only one was used for the final analysis.

Assessment of Study Quality and Publication Bias
The Cochrane Collaboration tool was used to assess risk of bias in randomized trials (14), while the Funnel plot and Egger's test were applied to evaluate publication bias (15). Three researchers independently checked the quality of all the included clinical trials. The quality assessment comprised evaluating: a) Selection bias (random sequence generation and allocation concealment), b) Performance bias (blinding of participants and personnel), c) Detection bias (blinding of outcome assessment), d) Attrition bias (incomplete outcome data) and e) Reporting bias (selective outcome reporting).

Outcome and Exposure of Interest
Our primary assessment indicators were the incidence of PD-1/ PD-L1 inhibitors-induced "treatment-related death" and "treatment-related adverse events leading to discontinuation". "Treatment-related grades 3-5 adverse events" and "any treatment-related adverse events" were also recorded. The basic characteristics and information on all the enrolled clinical trials were collected, including the first author's name, year of publication, trial number, trial title, trial phase, the specific name of the anti-PD-1/PD-L1 agent, treatment regimens, whether treatment was first-line or not, tumor types and the number of participants, treatment-related death and treatmentrelated discontinuation.

Assessment of Heterogeneity and Statistical Analysis
Heterogeneity of all the eligible trials was evaluated using Cochrane's Q statistic and the I 2 statistic as reported by Higgins and colleagues (13,16). Publication bias was checked using the Harbord test (16). Using the I 2 value, heterogeneity was regarded as low (<25%), moderate (25-50%) or high (>50%). Odds ratio (OR) and the corresponding 95% confidence interval (CI) were calculated using the random effect (RE) (17). Data analysis was conducted using Review Manager 5.3 and all statistical tests were two-sided with a value of P <0.05 considered statistically significant. Subgroup analysis was performed according to the tumor type, treatment regimen and PD-1/PD-L1 inhibitor used.

Incidence of Treatment-Related Adverse Events Leading to Discontinuation
The risk of treatment-related adverse events leading to discontinuation of therapy in the use of PD-1/PD-L1 was significantly lower than witnessed with the group that received chemotherapy [OR = 0.

DISCUSSION
PD-1/PD-L1 inhibitors have been playing an increasingly important role in anti-tumor therapy (1,5,6,8). While these agents have been reported to achieve gratifying clinical anti- tumor efficacy, they are beset by a growing list of diverse treatment-related side effects . As more clinical trials have been completed in recent years, it is critical that information about the safety and efficacy of PD-1/PD-L1 inhibitors are updated to provide the latest guidance in the administration and use of these therapeutic agents (1,5,6,8,. The need to provide the most recent information on the safety and adverse effect profiles of PD-1/PD-L1 inhibitors motivated the current meta-analysis. Following the selection criteria, 44 clinical trials reported by 53 articles were included in the meta-analysis . High risk of attrition bias was noticeable due to articles with incomplete data (Figure 2) (22,31,43,51,54,55,(58)(59)(60).
Our meta-analysis found that PD-1/PD-L1 inhibitors were generally distinguished in having a more favorable safety profile as compared to chemotherapy, across the four safety indicators applied to the analysis. Similarly, stratified investigation also revealed that between them, PD-L1 inhibitors were associated with fewer cases of adverse events as compare to PD-1 inhibitors, especially when considering the incidences of treatment-related adverse events leading to discontinuation of therapy or death. This observation is contrary to the conclusion reached in the mirror principle based meta-analysis (71). As there lacked randomized controlled trials between PD-1 and PD-L1 , the differences in the adverse event profiles between these two groups of agents were controversial as well as inconclusive (71). High heterogeneity was found across three evaluation indicators ( Figures 4A; 5A and 6A) and the subgroup analyses suggested the role of the tumor types and the inherent quality of the data in this observation (18-21, 27, 33, 61). Notably, however, there was no obvious publication bias in the articles (Supplementary Figures S3A; S4A; S5A and S6A). In addition, the trend in adverse events was repeated when PD-1/PD-L1 inhibitors were compared with combinational use with CTLA-4 ( Figures 3D; 4D; 5D and 6D) (21,22,50,(57)(58)(59)(60)(61). The combined results from the above analyses led us to the conclusion that PD-1/PD-L1 inhibitors display better safety characteristics than chemotherapy or the combined use of PD-1/PD-L1 with CTLA-4. Although PD-1/PD-L1 inhibitors, when prescribed in combination with chemotherapy, increased the occurrence of the four classes of adverse events (Figures 3B; 4B; 5B and Figure  6B) (19,(41)(42)(43)(44)(45)(46)(47)(48)(49), the increase was only statistically significant regarding grades 3-5 treatment-related adverse events [OR = 1.28, 95%CI:(1.05, 1.57), I 2 = 63%, Z = 2.43 (P = 0.01); 6B] (19,41,(43)(44)(45)(47)(48)(49). The high heterogeneity (I 2 = 63%) was tied to the NSCLC group (I 2 = 47%; Figure 6B) (43,44). The failure to note any meaningful differences with the other groups ( Figures  3B; 4B and 5B) might be due to the limitation of data. In order to draw more conclusive statistically significant analysis, more clinical trial results need to be considered.
We experienced similar challenges and limitations in the attempt to carry out subgroup analysis based on the treatment regimen and safety indicators, due to insufficient volumes of data. The observed trends and potential differences within the various subgroups need to be verified by using more clinical trials data.
In summary, our meta-analysis indicates that there is a better safety profile in the use of PD-1/PD-L1 inhibitors as compared to either chemotherapy or the use of combined regimens incorporating PD-1/PD-L1 inhibitors. The PD-1/PD-L1 inhibitors, however, had a worse adverse event profile over placebo. The present study, therefore, suggests caution and awareness of the occurrence of treatment-related adverse events when PD-1/PD-L1 inhibitors are used solely or in combination with other interventions. Clinicians should be aware that should adverse events occur in combinational treatment, withdrawing PD-1/PD-L1 inhibitor may not be the first approach to alleviate severe drug-related toxicities. This meta-analysis provides insights into important considerations to bear in mind when using PD-1/PD-L1 inhibitors and what adverse events to anticipate.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

AUTHOR CONTRIBUTIONS
YT, AH, YY, QD, and QW collected the data. YT, AH, YY, and QD performed data cleaning and analysis. YT drafted the manuscript. YS and LW reviewed the manuscript for scientific soundness. All authors contributed to the article and approved the submitted version.