Search Strategies, Study Selection, and Search Results

To assess the risk of bias of studies, we used the Risk Of Bias In Non-randomised Studies of Interventions (ROBINS-I) (Sterne et al. 2016) for nonrandomized controlled studies, and for randomized controlled trials (RCTs), we used the Cochrane Risk of Bias 2 tool. Two independent reviewers assessed the risk of bias at the study level and also considered rating bias at an outcome level if methodological limitations might affect different outcomes in a different way (e.g., lack of blinding might increase the risk of bias for quality of life but not for overall mortality). We assigned a “high risk of bias” rating to studies that had very serious limitations in design or conduct that might invalidate findings regarding all or individual outcomes. Disagreements between the two reviewers were resolved by discussion and consensus or by consulting a third member of the team. Risk of bias diagrams were generated using the Risk-Of-Bias VISualization (robvis) tool (McGuinness and Higgins 2021; see Appendix E).

We summarized all included studies in narrative form and in summary tables that tabulate the important features of the study populations, design, intervention, outcomes, setting, country, and results. If we found three or more similar studies addressing an outcome of interest, we considered quantitative analysis (i.e., meta-analysis) if studies were similar (in population, interventions, comparators, and outcomes). For all analyses, we used random-effects models (restricted maximum likelihood random effects) to estimate pooled effects. To determine whether quantitative analyses were appropriate, we assessed the clinical and methodological heterogeneity of the studies under consideration following established guidance (Gartlehner et al. 2012). If we conducted meta-analyses, we assessed statistical heterogeneity in effects between studies by calculating the chi-squared statistic and the I² statistic (the proportion of variation in study estimates attributable to heterogeneity). We examined potential sources of heterogeneity using sensitivity analyses. When quantitative analyses were not appropriate (e.g., due to heterogeneity, insufficient numbers of similar studies, or insufficiency or variation in outcome reporting), we synthesized the data qualitatively.

We graded the certainty of evidence of relevant outcomes based on current Grading of Recommendations Assessment, Development and Evaluation (GRADE) guidance (Balshem et al. 2011). Developed to grade the overall certainty of a body of evidence, this approach incorporates five key domains: 1) risk of bias, 2) inconsistency, 3) indirectness, 4) imprecision of the evidence, and 5) reporting bias. It also considers other optional domains that may be relevant for some scenarios. These included plausible confounding that would decrease the observed effect and strength of association (i.e., magnitude of effect) or factors that would increase the strength of association (i.e., dose-response effect). Two reviewers assessed each domain for each selected outcome and resolved differences by consensus discussion. We documented all decisions regarding up- or down-grading the certainty of evidence to ensure transparency. We used GradePro to develop summary of findings tables for the guideline panel.

A2–14 describes the grades of certainty of evidence, which reflect the certainty of the body of evidence regarding a specific outcome.