The main characteristics of all included studies will first be narratively synthesized (i.e., year of publication, study period, study population, study design, aim of study, geographic location, duration of follow-up). A descriptive analysis of study characteristics will be undertaken to explore the heterogeneity of the studies. Summary statistics will then be used to describe study outcomes, including means or medians, and frequencies. Proportions with exact binomial 95% confidence intervals (95% CI) will be calculated for each outcome and presented in forest plots. We will calculate the between-study variance (tau-squared) and
p values from tests of between-study heterogeneity. We expect substantial between-study heterogeneity, and the focus of the subsequent analyses will therefore be on the identification and exploration of sources of heterogeneity. Finally, we will explore associations that may exist between proportions in countries, settings (e.g., urban, rural), and study outcomes (i.e., the rate of viral suppression and the level of drug resistance) using random intercept logistic meta-regression (binomial-normal) models. These models avoid the biases that arise when normal-normal models are applied to logit or arcsine-square root transformed proportions. Where appropriate, we will use the same models to calculate combined estimates of proportions. We will use the GRADE approach to rate the certainty of evidences as “high,” “moderate,” “low,” and “very low ”[
17]. Major findings on INSTI-resistance, drug susceptibility and linkage to DTG-therapy will be summarized in a table. This table will also present the quality of the evidence found, all sorted according to socio-demographic data, clinical, and laboratory parameters. Though the highest quality rating is for randomized trial evidences, they may be downgraded to moderate, low, or even very low-quality evidence. Rating will depend on limitations in study design and implementation, indirectness of evidences, unexplained heterogeneity, imprecision of result, and a high probability of publication bias [
17]. Evidences from sound observational studies (cohorts and case-control studies) will be graded as low quality [
17]. However, if such studies yield large effects and there is no obvious bias explaining those effects, we would rate the evidences as moderate or—if the effect is large enough—even high quality [
17]. Detailed interpretation of each evidence and its respective recommendation is provided in Additional file 3.