Methods
Protocol and registration
The systematic review and meta-analysis was conducted under the principles of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2009 checklist (15) and registered at Prospective Register of Systematic Reviews (PROSPERO) (CRD42020223177). Tianjin First Central Hospital Research Ethics Committee had confirmed that no ethical approval was required.
Eligibility criteria
We screened literature following PICOS questions. The population included patients with HNC; the intervention included treatment either PD-1/PD-L1 immunotherapy alone or combining conventional therapies (e.g., chemotherapy, radiotherapy, and cetuximab); the comparison included standard‐of‐care therapy; the outcome included OS, PFS, AE, ORR, and DCR; the study design included RCTs and prospective single-arm clinical trials. The language type was English, and the types of articles excluded were conference abstracts, reviews, letters, personal opinions, book chapters, trial protocols, systematic reviews and meta-analyses because of the lack of complete and reliable data. Publications that reported duplicated data from the same clinical trials were also excluded, and the most complete of them were included.
Information sources, search, and study selection
We searched in the electronic databases Cochrane Library, Embase, PubMed, and Web of Science until November 20, 2020. The search strategies were uploaded in Supplementary File 1 . Duplicate references were removed with reference manager software (EndNote®), and the rest of the references were personally screened. We performed the first screen based on titles and abstracts and excluded the references that do not meet the eligibility criteria after getting access to full-text articles on the second screen.
Data extraction
Data from full‐text articles were extracted by us independently. Study characteristics (the trial name, publication year, phase of studies), baseline characteristics of patient populations (amount of participates, amount and proportion of males, age, HPV, smoke status, Eastern Cooperative Oncology Group (ECOG) performance status score), study design (type and dosage of PD‐1/PD‐L1 immune inhibitors), outcomes (OS, PFS, ORR, DCR, the median time to response, the median duration of follow-up and response,) and amount of any‐grade and grade ≥3 AEs in the articles were collected from individual studies.
Data synthesis and statistical analysis
OS and PFS were the primary outcomes, while the secondary outcomes included ORR, DCR, any grade, and grade≥3 AE. Survival data from the figures in the references were extracted by Engauge Digitizer 4.1 software(16), and ln[Hazard Ratio](ln[HR]) and Seln[HR] of survival curves were calculated by the Microsoft Excel program file(17). Cochrane Collaboration’s Review Manager®5.4 software (RevMan5.4) was used to conducted meta-analyses. Inconsistency indices (I²) were used to test the studies’ substantial amount of heterogeneity; fixed-effect models were performed when I²≤ 50 %. Otherwise, heterogeneity was considered among the studies, either through sensitivity analysis or subgroup analysis to explore heterogeneous sources, and random effect models were performed after eliminating evident clinical and methodological heterogeneity. Summary measure was the odds ratio (OR). P < 0.05 indicated a statistically significant difference.
Risk of bias
The risk of bias of individual studies was assessed by the Cochrane collaboration’s Risk of Bias Tool based on the following seven domains: selection bias, selection bias, performance bias, detection bias, attrition bias, reporting bias, and other biases. Each risk of bias was defined as low, unclear, or high by Xin Huang and Jie Zhang independently.
The quality of evidence and the strengths of recommendations were evaluated following the Grading of Recommendation, Assessment, Development, and Evaluation (GRADE) guidelines (18, 19) through informatics tools from GRADEpro (20) based on study design, risk of bias, inconsistency, indirectness, imprecision, and other considerations. Egger’s test evaluated latent publication bias in Stata 16 software.