Assessment of reporting biases

The review employed strategies to search and include relevant unpublished studies to reduce publication bias. The strategies included searching the gray literature, including conference proceedings (e.g., International AIDS Conference), and prospective trial registration database to over-come lime-lag bias. To investigate publication bias, if there were ≥ 10 trials included in an analysis, we planned to use a funnel plot to explore the possibility of small study effects-a tendency for smaller studies to report larger beneficial effects. However, this review did not use a funnel plot test because of fewer trials (< 10 trials) were included in an analysis, as the power of tests is too low to distinguish chance from publication bias [33].