Systematic Assessment of Risk of Fever in Solid Tumor Patients Treated With PD-1/PD-L1 Inhibitors: A Systematic Review and Meta-Analysis

Purpose: Drug-induced fever is frequently reported in cancer patients treated with anti-programmed cell death 1 (PD-1)/programmed cell death ligand 1 (PD-L1), and stoppage of the offending agent is the management of choice. However, given the complex management of cancer patients, this needs to be carefully studied. Therefore, we conducted a meta-analysis to estimate the risk of fever associated with anti-PD-1/PD-L1 in cancer patients. Methods: From May 2010 to 2020, an electronic search was conducted through PubMed for relevant studies. All clinical trials reporting fever in cancer patients treated with PD-1/PD-L1 inhibitors were included, while other designs were excluded. A manual search was also conducted to search for relevant articles. Outcomes included the risk of pyrexia and febrile neutropenia in the overall population and based on the grade of fever (all grades vs. grades 3–5). The Newcastle–Ottawa Scale was used to assess the quality of included studies. Results: Thirty-one articles, involving 27 clinical trials and 15,867 participants, were included. The increased risk of pyrexia for all grades is only found when PD-1/PD-L1 plus cytotoxic T lymphocyte-associated protein 4 (CTLA-4) was compared to CTLA-4 [odds ratio (OR) = 2.48, 95% CI: 1.17, 5.23]. The risk of febrile neutropenia for all-grade fever was significantly lower in the PD-1/PD-L1 group compared to that of chemotherapy alone (OR = 0.02, 95% CI: 0.01, 0.05). A similar trend in the risk of febrile neutropenia was also found for grades 3–5 (OR = 0.02, 95% CI: 0.01, 0.05). Conclusion: The increased risk of pyrexia for all grades could only be found when PD-1/PD-L1 plus CTLA-4 was compared with CTLA-4. Meanwhile, compared to chemotherapy, PD-1/PD-L1 inhibitors reduced the risk of febrile neutropenia.


INTRODUCTION
Fever, relating to drug, infections, or others, is frequently reported among cancer patients (1)(2)(3). In the course of antitumor therapy, drug-induced fever might be caused by chemotherapy (4), targeted therapy (5), or immunotherapy drugs (6). At present, in clinical practice, enough attention has been paid to drug-induced fever of cancer patients, and relevant guidelines have been developed in this regard (1)(2)(3). In order to control the fever of cancer patients in a timely manner, it is important to clarify the origin of the fever (1)(2)(3)(4)(5)(6). In terms of drug-induced fever, stopping the offending agent remains the first choice of treatment (1)(2)(3). However, due to the delicate medical care in cancer patients, the sudden stoppage of antitumor therapy is very likely to lead to the rapid progression of such tumors. Therefore, the decision to stop antitumor drug-induced fever needs to be studied thoroughly.
In many clinical trials, immunotherapeutic drugs, especially those directed at the programmed cell death 1 (PD-1) and programmed cell death ligand 1 (PD-L1) pathways , have achieved satisfactory clinical efficacy and safety profiles in cancer patients. That being said, various treatment-emergent adverse events (TEAEs) were gradually discovered and reported in individual studies, including fever . However, to date, there has been no published systematic literature review studying the incidence or risk of drug-induced fever in cancer patients treated with PD-1/PD-L1 inhibitors . Therefore, in order to provide clear evidence in this regard, we conducted the current systematic review and meta-analysis to report the overall risk of drug-induced fever in cancer patients treated with PD-1/PD-L1 inhibitors.

METHODS
This systematic review and meta-analysis was carried out in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (38).

Types of Enrolled Studies
Clinical trials involving hematological malignancies were excluded first. Phase III clinical trials reporting the incidence or the risk of fever in patients with PD-1/PD-L1-positive solid tumors were prioritized. Clinical trials of other phases would be placed in an alternative position. All clinical trials had to have a control group to be eligible for inclusion in our study. At least one piece of fever-related data had to be available for inclusion; otherwise, the paper would be excluded. Various definitions of fever, such as pyrexia and febrile illness, were used to search for eligible studies, and any study reporting any definition of Abbreviations: PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; PD-1, programmed cell death-1; PD-L1, programmed cell death ligand 1; CTLA-4, cytotoxic T-lymphocyte antigen 4; HR, hazard ratios; OR, odds ratio; RD, risk difference; CI, confidence interval; RE, random effect; NSCLC, nonsmall-cell lung cancer; SCLC, small cell lung cancer; OSCC, esophageal squamous cell carcinoma; HNSCC, head and neck squamous cell carcinoma; UC, urothelial cancer; BC, breast cancer; RCC, renal cell carcinoma. fever while being consistent with our eligibility criteria was included. We included articles that were originally published in the English language, while other trials published in other languages were excluded.

Evaluation of Study Quality and Publication Bias
Publication bias was assessed using Egger's regression test, while the quality of the included trials was assessed by the Newcastle-Ottawa Scale, which was proposed by the Cochrane Collaboration (38,(40)(41)(42)(43). The quality assessment of included clinical trials was also carried out by the above three reviewers (Dongmei Xu, Hongmei Liu, and Wentao Wang). The quality assessment included assessing the risk of bias in the following points: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective outcome reporting. All these points were evaluated together, and the evaluation results would be summarized in a single graph. Harbord's test was used to check publication bias for all enrolled clinical trials (44) A P < 0.05 was considered indicative of publication bias.

Exposure and Outcome of Interest
Basic information of enrolled studies, including the first author's name, year of publication, trial number, trial title, the specific name of the anti-PD-1/PD-L1 agent, status of previous therapies, treatment regimens, number of participants, and number of fever events (rate), was collected and summarized in Table 1. Various definitions and terms of fever indicators, such as pyrexia and febrile neutropenia, were collected and used for the final comprehensive analysis. Both all-grade and grade 3-5 fever data were used for the final meta-analysis.

Assessment of Heterogeneity and Statistical Analysis
Cochrane's Q statistic and the I 2 statistic were used to check the heterogeneity among analyzed studies, as proposed by Higgins et al. (38,39,43). The grade of heterogeneity was evaluated by the range of I 2 values (38,43). Heterogeneity was taken as low, moderate, or high according to I 2 values <25, 25-50, and >50%, respectively (39). Odds ratio (OR) and the corresponding 95% confidence interval (CI) would be calculated by random effect (RE) (45). A P < 0.05 was considered the cutoff value for statistical significance. In order to clarify the relationship between fever indicators (pyrexia and febrile neutropenia) and PD-1/PD-L1 inhibitors, we performed a large number of subgroup analyses based on the type of tumor, the treatment regimen, and the specific administered drug. The software (Review Manager 5.3) was used for data consolidation and analysis. Statistical tests were all two-sided (39).

Literature Search Results
According to our preliminary electronic search, a total of 651 articles discussing PD-1/PD-L1 inhibitors and cancer, including clinical trials, were identified on PubMed, and 65 related studies were added after conducting a manual search of articles. A total of 31 published articles (27 clinical trials), including 15,867 participants, were finally included in our review. The basic information of included studies is listed and summarized in Table 1 . The PRISMA flow diagram of our review is shown in Figure 1, while the risk of bias summary is presented in Figure 2  . Different stages of a clinical trial, named CheckMate 067, were reported by four articles (8,(34)(35)(36). After reviewing and evaluating the data of the four articles, only the most comprehensive data provided by one of four articles were used for the final comprehensive analysis (35). After assessment and screening for all included clinical trials, the data relating to fever were mainly displayed in two forms: pyrexia and febrile neutropenia. The two variables would be analyzed separately.

Risk of Bias
The overall risk of bias of all included studies is presented in a single graph (Figure 2, Supplementary Figure 2). Meanwhile, publication bias, evaluated by Harbord's test, shown in the form of funnel plots, was checked and displayed in Supplementary Figures 3-6 (38, 40-43).
The overall analysis results of group D revealed no statistically significant difference regarding pyrexia of all grades among the studied arms (OR = 1.05, 95% CI: 0.64, 1.71; Figure 3D) (8,9). The funnel plot is provided in Supplementary Figure 3D.
Inconsistent with the above findings, the overall risk of pyrexia for all-grade fever in patients in group E was significantly different in the studied arms (OR = 2.48, 95% CI: 1.17, 5.23; Figure 3E) (29,35). Moderate heterogeneity (I 2 = 48%) was found among the included clinical trials. The corresponding funnel plot is provided in Supplementary Figure 3E (29, 35).
In group C, only two clinical trials were enrolled in the final meta-analysis (29,35). The analysis results are summarized and displayed at the bottom of Figures 5C, 6C (29,35). No statistically significant differences were found between the studied arms. The corresponding plots are provided in Supplementary Figures 5C, 6C (29, 35).

DISCUSSION
With the development of antitumor immunotherapy, more and more clinical trials investigating immunotherapeutic drugs have been developed, and satisfactory clinical effects have been achieved . PD-1/PD-L1 inhibitors are currently the most widely used antitumor immunotherapeutic drugs in studied clinical trials, and they are also the most widely used antitumor immunotherapeutic drugs in clinical practice (7-37, 46, 47). In clinical trials related to PD-1/PD-L1 inhibitors, as well as in clinical work using PD-1/PD-L1 inhibitors for antitumor therapy, patients with drug-induced fever are frequently encountered and reported (7-37, 46, 47) In order to control the fever of cancer patients in a timely manner, it is important to clarify the origin of the fever (1-6). However, the specific role of PD-1/PD-L1 inhibitors in the etiology of fever is still unclear, especially when combined with chemotherapy for antitumor therapy. Therefore, when patients who are receiving combined antitumor therapy develop hyperthermia of grades 3-5, it is difficult for us to judge whether to stop the PD-1/PD-L1 inhibitor drugs to relieve the fever or not. Furthermore, due to the particularity of cancer patients, the sudden stoppage of antitumor therapy is very likely to lead to the rapid progression of the tumor. Therefore, the decision to stop or withdraw the offending drug (antitumor drug) needs to be carefully studied.
In an attempt to explore the relationship between PD-1/PD-L1 inhibitors and drug-induced fever, this meta-analysis was designed and performed.
According to the guidelines of PRISMA, after screening and eligibility assessment,27 clinical trials reporting on the use of PD-1/PD-L1 inhibitors, including 15,867 cancer patients, were enrolled for the final comprehensive analysis . The quality of all included studies was evaluated and summarized in a single graph (Figure 2, Supplementary Figure 2) and was considered to be better. Therefore, the analysis conclusions drawn from the data of these clinical trials are much more inclined to be true and reliable . The data relating to fever were mainly displayed in two forms: pyrexia and febrile neutropenia. When the extracted data were verified, it was found that the incidence of pyrexia reported in clinical trials was mostly grades 1-2, while the incidence of febrile neutropenia was almost all grades 3-5 . Therefore, febrile neutropenia is the type of fever that is most likely to lead to the stoppage of the offending antitumor agent.
Similar to the above results (Figures 3A,B), regardless of whether the control group was placebo or targeted therapy and regardless of whether PD-1/PD-L1 was used alone or in combination with targeted therapy (Figures 3C,D) (21,22,29,32,35), no statistically significant differences were found regarding the risk of pyrexia. In other words, PD-1/PD-L1 inhibitors did not result in a statistically significant increase in the risk of pyrexia. The results of the subgroup analysis suggested that the moderate heterogeneity of that analysis (I 2 = 49%; Figure 3C) might originate from the included PD-1 clinical trial on gastric or junction cancer ( Figure 3C) (22). Due to the small number of included clinical trials (reporting pyrexia in group D) to perform sufficient subgroup analysis, it was impossible for us to clarify the origin of low heterogeneity (I 2 = 29%; Figure 3D) (8,9). Different from the above groups, the risk of pyrexia for the experimental group (PD-1) was higher than that of the control group (PD-1 + CTLA-4) with a statistically significant difference (OR = 2.48, 95% CI: 1.17, 5.23; Figure 3E) (29,35). However, since only two clinical trials were enrolled in this group, the conclusion still needs to be verified by more clinical trials (29,35). Moderate heterogeneity (I 2 = 48%) was found among the included trials. The number of included trials in the above analysis was very small, and thus, it was impossible to conduct subgroup analysis to identify the origin of the resulting heterogeneity (29,35). The corresponding funnel plot was provided in Supplementary Figure 3E (29,35). No obvious publication bias was found.
The good safety and satisfactory clinical efficacy of PD-1/PD-L1 inhibitors have been reported by a large number of clinical trials . The safety of PD-1/PD-L1 inhibitors was further confirmed by the findings of our meta-analysis. The increased risk of pyrexia for all grades could only be found when PD-1/PD-L1 plus CTLA-4 was compared with CTLA-4 alone (29,35). Furthermore, no statistically significant differences in the risk of febrile neutropenia could be found in all studied groups, except for the PD-1/PD-L1 group, which was associated with a significantly lower risk of febrile neutropenia when compared with chemotherapy (7, 11, 13, 15, 16, 18, 19, 23-25, 28-31, 33, 35). It would be helpful for us to clarify the source of antitumor drug-induced fever and adopt the best treatment regimens for cancer patients.

CONCLUSIONS
The increased incidence risk of pyrexia for all grades could only be found when PD-1/PD-L1 plus CTLA-4 was compared with CTLA-4. Meanwhile, compared to chemotherapy, PD-1/PD-L1 inhibitors reduced the risk of febrile neutropenia.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
YT had the right to deal with all the data and was responsible for the decision to submit the manuscript for publication. HL, DX, WW, FS, and XY had the data of all included clinical trials. DX, HL, and WW were responsible for checking and evaluating the quality of the collected data. All authors contributed to the article and approved the submitted version.

FUNDING
This study was funded by the Medicine and Health Technology Development Plan of Shandong Province (2017WSB04081), which was led by HL.