Outcomes and analysis
Data from included studies were analysed for the CB DOT and the clinic DOT intervention. If studies additionally included patients undergoing other types of DOT supervision (e.g. DOT provided by private physicians) or SAT, we only included the patients receiving CB DOT or clinic DOT.
The primary outcome was the proportion of patients who were successfully treated; the secondary outcome was the proportion of patients who were lost to follow-up. Meta-analyses were performed for both outcomes.
Review Manager Version 5.3 (Nordic Cochrane Centre, Copenhagen, Denmark) was used to perform the meta-analyses. As the studies were from different countries and in different permutations of CB DOT and clinic DOT, we expected heterogeneity beyond that explained by random differences between patient groups in the individual studies. Thus, a random effects meta-analysis for the data was deemed more appropriate than assuming fixed variation. Confidence intervals (CIs) were set at 95% and data were presented using odds ratios (ORs) calculated via the Mantel-Haenszel method. Heterogeneity was quantified using the I2value described by Higgins and colleagues [23]. An I2 of 0% indicates no study heterogeneity, whilst progressively higher values represent greater inter-study heterogeneity.
Assessment of study quality and risk of bias
Studies were assessed for quality using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) criteria [24]. For RCTs, risk of bias was assessed using guidelines provided by the Cochrane Collaboration and those parts of the guidelines that were relevant were also used to conduct a risk of bias assessment for the included non-randomised studies [25].