Seventeen studies met the inclusion criteria (5,356 participants, range 17 to 1,831): 14 prospective cohort studies and three retrospective.
From the frequentist meta-analysis, overall sensitivity of the tests was 51% (95% CI 41% to 60%) with a corresponding specificity of 99% (95% CI 94% to 99%). From the bivariate model, the interaction between sensitivity and specificity was 0.32 (95% CI -0.64 to 0.89); LR+ was 34.5 (95% CI 12.7 to 93.6), LR- was 0.48 (95% CI 0.39 to 0.60), PPV was 94% (95% CI 85% to 98%), NPV was 82% (95% CI 79% to 85%) and DOR was 71.6 (95% CI 26.3 to 194.6).
Although summary estimates of sensitivity and specificity from the bivariate model were not provided in the text, the point estimate on the SROC curve presented showed similar results to those obtained from the frequentist meta-analysis, with a particularly wide 95% prediction region for sensitivity.
Results of sensitivity analyses were reported. The authors stated that there was some evidence of publication bias.