2. Materials and Methods
This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (17) and recommendations for optimizing literature searches for systematic surgical reviews (18). The protocol was prospectively registered on the Open Science Framework (https://osf.io/agzkf/).
Relevant clinical studies were retrieved from PubMed, SCOPUS, Embase, the Web of Science, and the Cochrane Central Register of Controlled Trials up to March 2022. The search terms were as follows: oral carcinoma, oral neoplasm, oral cavity neoplasms, neoplasm, oral cavity, oral cavity cancer, squamous cancer, SNB, END, prognosis, survival, hazard ratio, overall survival rate, and disease-free survival. Reference lists were searched to ensure that no relevant studies were missed. Two independent reviewers removed irrelevant studies (i.e., those that did not discuss prognostic factors and survival rates) by reviewing the title, abstract, and text.
The inclusion criteria were as follows: comparison between SNB and END in terms of the prognosis of early stage (T1–2) OSCC patients; survival data and prognostic predictions including HRs with 95% CIs and/or OS, disease-free survival (DFS), or disease-specific survival (DSS); human study published in English; and exclusion of advanced OSCC (cT3–4 or N1) patients and those on drugs that might affect OSCC development. The exclusion criteria were as follows: reviews; case reports; studies on other head and neck cancers such as nasopharyngeal, oropharyngeal, hypopharyngeal, or salivary cancer; and a lack of adequate prognostic data. The search strategy is shown in Figure 1.
All data were extracted by two independent reviewers, who also assessed study quality. Differences were resolved by panel discussion. We recorded the first author, year of publication, country, type of cancer, and number, age, sex, and T stage of the patients. For OS, DFS, and DSS, hazard ratios (HRs) with 95% confidence intervals (CIs) were either described (19-31) or calculated as described by Tierney et al.(32) and Parmar et al. (33). If both multivariate and univariate analyses were used to evaluate OS, the HRs and 95% CIs generated by multivariate analysis were extracted (34). The risk of bias in nonrandomized studies of interventions (ROBINS-I) and Cochrane risk of bias tool for randomized trials (RoB 2) were used to assess study quality in line with the GRADE guidelines (35, 36).
R statistical software (R Foundation for Statistical Computing, Vienna, Austria) was utilized for meta-analysis. Homogeneity was assessed using the Q statistic. The degree of heterogeneity was indicated by the I2 value (76–100% = high, 50–75% = moderate; 25–49% = low). Parameters with I2 values < 50% were analyzed using a fixed-effects model, while those with I2 values > 50% were analyzed with a random-effects model. Subgroup analyses were performed by follow-up period (3, 5, and 10 years). We used Begg’s funnel plots and the Egger linear regression test to evaluate publication bias. For sensitivity analysis, we removed each item individually to assess its contribution to the observed effect.