3.4.4. Data synthesis and meta-analysis

A standardized overall approach was specified a priori for each systematic review, to account for significant heterogeneity in results expected between studies. First, data were synthesized separately for each commercial IGRA and by the World Bank country income classification (LMIC versus high-income countries) as a surrogate for TB incidence. Second, heterogeneity was visually assessed using forest plots, and the variation in study results attributable to heterogeneity was characterized (I-squared statistic) and statistically tested (chi-squared test). Third, pooled estimates were calculated using random-effects modelling, which provides more conservative estimates than fixed-effects modelling when heterogeneity is present. For each individual study, all outcomes for which data were available were assessed. First, forest plots were generated to display the individual study estimates and their 95% CIs. Pooled estimates were calculated when at least three studies were available in any subgroup, and individual study results were summarized when fewer than four studies were available. Standard statistical packages were used for analyses.

Book navigation