Data extraction and statistical analysis
An electronic data-collection form was used to extract the following data from each of the included studies: author, year of publication, study design, country and period of conduction, number of patients, demographic characteristics, staging, grading of the tumours, type of imaging used to stage the neck, criteria used for elective neck dissection and dissected levels if specified, number of cN0 patients who underwent elective neck dissection and cases of occult metastases identified (pN+/cN0). The authors of the selected studies were contacted in order to gather missing information about individual patient data and attempt to perform subgroup meta-analysis. Two authors (DB, AV) independently assessed the quality of the included studies with the Newcastle-Ottawa Scale(9).
The pooled proportion of occult metastases and corresponding 95% confidence interval (CI) were calculated according to random-effects models of DerSimonian and Laird(10), using the logit transformation and weighting through the inverse variance method. Statistical heterogeneity among studies was evaluated using the I² and t2statistics. Influence analysis was performed when pooled proportions were estimated from five or more studies: pooled proportion was calculated by omitting one study at a time. Publication bias was assessed through a funnel plot(11).
The results of the meta-analysis were presented graphically using forest plots, plotting the individual paper, pooled proportions and corresponding 95% CI. Analyses were conducted using R 3.6, and statistical significance was claimed for p < 0.05 (two sided).