The Association Between the Incidence Risk of Peripheral Neuropathy and PD-1/PD-L1 Inhibitors in the Treatment for Solid Tumor Patients: A Systematic Review and Meta-Analysis

Purpose: We conducted this study to determine the relationship between PD-1/PD-L1 inhibitors and the incidence risk of peripheral neuropathy in patients with solid tumors. Method: The process of the meta-analysis was performed by us according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines. Incidence of all-grade and grade 3–5 treatment-related peripheral neuropathy in patients with solid tumors were taken into account. Results: After screening and eligibility assessment, a total of 17 clinical trials involving 10,500 patients were selected for the final meta-analysis. The incidence risk of peripheral neuropathy for all grade was significantly lower in the PD-1/PD-L1 inhibitor group than that of the control group, either monotherapy (OR = 0.08, 95%CI:[0.03, 0.19]) or chemotherapy (OR = 0.05, 95%CI:[0.03, 0.11]). Similar incidence trend could also be seen for the incidence risk of grade 3–5 peripheral neuropathy. When PD-1/PD-L1 inhibitors were used in combination with chemotherapy, the incidence risk of peripheral neuropathy was higher than in the control chemotherapy group, whether it was all-grade (OR = 1.22, 95%CI:[1.00, 1.49]) or grade 3–5 degree (OR = 1.74, 95%CI:[1.03, 2.92]). Conclusion: Compared with chemotherapy, incidence risk of peripheral neuropathy related to PD-1/PD-L1 inhibitor was significantly lower than that of the chemotherapy group, while PD-1/PD-L1 inhibitor increased the incidence risk of peripheral neuropathy when it was combined with chemotherapy.

For drug-induced peripheral neuropathy, stopping the drug remained to be the primary treatment method (1)(2)(3)(4)(5)(6)(7)(8)(9)(10). However, for patients with malignant tumors, when severe drug side effects were encountered (12,13), careful consideration for stopping the drug should be taken into account. Because of the sudden stop of anti-tumor treatment, it was very likely to cause rapid progression of the tumor. When PD-1/PD-L1 inhibitors were used in combination with chemotherapy, it was particularly important to determine the cause of peripheral neuropathy and then decide which drug to be discontinued (15)(16)(17)(18).
To solve the above problems and clarify the association between incidence risk of peripheral neuropathy and PD-1/PD-L1 inhibitors, we designed this meta-analysis.

METHODS
The process of the meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Metaanalyses (PRISMA) guidelines (31).

Types of Enrolled Studies
According to the research design, the selected clinical studies must meet the following criteria: (1) Randomized controlled clinical trials would be prioritized, (2) PD-1/PD-L1 inhibitor was prescribed for at least one group of participants, (3)The control group was an anti-tumor drug or PD-1/PD-L1 in combination with an anti-tumor drug rather than a placebo, (4) Participants were diagnosed with solid malignant tumors rather than hematological malignancy, (5) Data on peripheral neuropathy were reported in the study, (6) the enrolled study was published in English.

Search Strategy
Original articles including PD1/PD-L1 inhibitor regimens for solid tumor patients were verified by a systematic search of PubMed. The reported date of the results was limited from Jan 22, 2013 to May 31, 2019. The following subject terms would be used in the literature search process: "cancer, " "tumor, " "PD1/PD-L1, " "nivolumab, " "Opdivo, " "pembrolizumab, " "Keytruda, " "Imfinzi, ","MK-3475, " "atezolizumab, " "Tecentriq, " "MPDL3280A, " "avelumab, " "Bavencio, " "durvalumab, " "BMS-963558." Studies limited in human beings, shown in full text, abstract, or poster form, were selected three investigators (Shuisheng Zhang, Yi Zhao, Qingshan Zhu) were appointed to check eligibility and duplicate independently by screening titles and abstracts of relevant studies. If data on peripheral neuropathy had not been reported, we would contact the corresponding author of the article to verify it again, or it would be precluded from the meta-analysis. The basic characteristics information included in the study would be summarized in Table 1.

Assessment of Study Quality and Publication Bias
Funnel plot, Egger's test and Newcastle-Ottawa scale, proposed by the Cochrane Collaboration, were taken to evaluate the bias (31)(32)(33)(34)(35). Three investigators (Shuisheng Zhang, Yi Zhao, Qingshan Zhu) were appointed to check the quality of all studies. The results, including random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective outcome reporting, would be summarized in a figure together.

Outcome and Exposure of Interest
The study name, year, phase, tumor type, PD-1 and PD-L1 inhibitor regimen, previous therapy regimen, number of evaluable cases, and number of peripheral neuropathy events were extracted from every enrolled study. Both all-grade and grade 3-5 peripheral neuropathy data were taken into account for the final comprehensive meta-analysis.

Assessment of Heterogeneity and Statistical Analysis
Cochrane's Q statistic and the I 2 statistic were taken into account for evaluating the heterogeneity among enrolled studies just as suggested by Moher et al. (31) and Higgins et al. (36). The grade of heterogeneity was calculated by the range of I 2 values. Heterogeneity was considered low, moderate or high according to I 2 values <25%, 25-50%, and >50%, respectively. Odds ratio (OR) value was reported to be a much more conservative evaluation parameter and might be more inclined to reveal a safety signal, as the method by which an OR is calculated provided a point estimate farther from unity than that provided by a HR. Odds ratio (OR), and 95% confidence interval (CI) would be calculated by random effect (RE) (37). Risk Ratio (RR) and Risk Difference (RD) were also calculated as secondary reference indicators for a more detailed interpretation of the results. P < 0.05 was considered to be of statistically significance. In order to clarify the correlation between peripheral neuropathy 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 drug. The software of Review Manager 5.3 was used for data consolidation and analysis. Statistical tests were all two-sided.

Literature Search Results
According to the searching principle set by our team, 505 related documents were retrieved on the PubMed website, and 58 related documents were found in other websites or published documents.

Risk of Bias
Study quality and risk of bias among enrolled studies were checked by Newcastle-Ottawa scale (35). Random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), and selective reporting (reporting bias) were assessed by three members of our team independently and summarized in Supplemental Figure 1. Publication bias, evaluated by Harbord , s test (31), was displayed by funnel plots (Supplemental Figures 2, 3, 5, 7, 9, 11).
We first performed a meta-analysis on the data of Group A, and the results of the analysis were summarized at the bottom of Figure 2A [OR = 0.08, 95%CI:[0.03, 0.19], I 2 = 69%, Z = 5.64 (P < 0.00001)] (19,21,23,(26)(27)(28)(29). Subgroup analysis was performed according to the different drug types in the control group and the experimental group, and the results were shown in Figures 2A1,A2, respectively. Moderate heterogeneity was found in Group A (I 2 = 69%). Subgroup analysis results suggested that the source of heterogeneity was the PD-L1 subgroup [ Figure 2A2; (21,23,26)]. The funnel plots of OR for Group A could be seen in Supplemental Figures 2A1,A2. Similar to the results of OR, RR and RD of Group A were displayed in Supplemental Figures 4A, 6A.The corresponding funnel plots were gathered in Supplemental Figures 5A, 7A.
When PD-1/PD-L1 drugs were compared with chemotherapy (Group B), the incidence of peripheral neuropathy was significantly lower than that of the control group, and the OR results are summarized in Figure 2B Figures 2C, 7C. The RR of Group C showed that the incidence risk of peripheral neuropathy in the PD-1/PD-L1 combined chemotherapy subgroup was significantly higher than that in the chemotherapy subgroup, and the P-value was statistically significant (Supplemental Figure 4C) [RR = 1.16, 95%CI:[1.01, 1.34], I 2 = 0%, Z = 2.13(P = 0.03)] (15)(16)(17)(18). The corresponding funnel plots of RR were gathered in Supplemental Figure 5C. No obvious heterogeneity was found among Group C (I 2 = 0%).
When PD-1/PD-L1 drugs were compared with chemotherapy (Group B), the incidence risk of peripheral neuropathy limited to grade 3-5 was significantly lower than that of the control group, and the OR results are summarized in Figure 3B [OR = 0.11, 95%CI:[0.03, 0.49], I 2 = 0%, Z = 2.92 (P = 0.004)] (20,22,30). The funnel plots of OR for Group B could be seen in Supplemental Figures 3B1,B2. Similar to the results of OR, RR, and RD of Group B were displayed in Supplemental Figures 8B, 10B.The corresponding funnel plots were gathered in Supplemental Figures 9B, 11B. No heterogeneity was found in Group B (I 2 = 0%) (20,22,30). The OR of Group C showed that the incidence risk of peripheral neuropathy in the PD-1/PD-L1 combined chemotherapy subgroup was significantly higher than that in the chemotherapy subgroup, and the P-value was statistically significant (Figure 3C) [OR = 1.74, 95%CI:[1.03, 2.92], I 2 = 0%, Z = 2.09 (P = 0.04)] (15)(16)(17)(18). The corresponding funnel plots of OR were gathered in Supplemental Figure 3C. No heterogeneity was found in Group C (I 2 = 0%). Similar analysis results could also be seen in Supplemental Figure 8C Figure 11C. Low heterogeneity related to RD was found in Group C (I 2 = 10%). Subgroup analysis revealed that the source of heterogeneity might be related to the inclusion of this clinical trial (15).

DISCUSSION
Peripheral neuropathy is a painful condition deriving from many and varied etiologies (38,39). Certain medications have been implicated in the iatrogenic development of drug induced peripheral neuropathy (DIPN) and include chemotherapeutic agents, antimicrobials, cardiovascular drugs, psychotropic, anticonvulsants, among others (39). Chemotherapy-induced peripheral neuropathy (CIPN), reported in several studies, especially for paclitaxel induced peripheral neuropathy, was common for cancer patients (40,41). CIPN was a dose limiting toxicity, negatively impacting both quality of life and disease outcomes (42). However, during the process of anti-tumor treatment, combinations of drugs that were unknown to cause CIPN were prescribed for cancer patients, and sequential treatment for recurrence with additional CIPN-inducing drugs would also be suggested (43). Therefore, it would be difficult for us to determine which specific drug was responsible for the occurrence of peripheral neuropathy, especially for some newly marketed targeted anti-tumor drugs without fully understanding of toxicities, such as PD-1/PD-L1 inhibitors and Brentuximab vedotin (3,(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30). To clarify the association between incidence risk of peripheral neuropathy and PD-1/PD-L1 inhibitors, we designed this meta-analysis.

CONCLUSIONS
Compared with chemotherapy, incidence risk of peripheral neuropathy related to PD-1/PD-L1 inhibitor was significantly lower than that of the chemotherapy group, while PD-1/PD-L1 inhibitor increased the incidence risk of peripheral neuropathy when it was combined with chemotherapy.

ETHICS STATEMENT
This study belongs to the type of data analysis and rearrangement, and does not involve human or animal related ethical issues.

AUTHOR CONTRIBUTIONS
YT had full access to all data in the study and all authors had final responsibility for the decision to submit for publication. ZS, SZ, XY, ND, and YT had the full data of the paper. MX, QZ, YL, LY, HS, JX, and YM were responsible for the collection of clinical data. ZS helped to gather online data and write the report.