Risk of bias was assessed using the QUADAS-II checklist. Evidence quality was downgraded by 1 increment if the evidence was at high risk of bias and downgraded by 2 increments if the evidence was at very high risk of bias.

Inconsistency was assessed by inspection of the sensitivity and specificity forest plots, using the point estimates and confidence intervals. Particular attention was paid to values above or below 50% (diagnosis based on chance alone) and the 80% threshold set by the GC (the threshold above which would be acceptable to recommend a test). The evidence was downgraded by 1 increment if the individual studies varied across 2 areas (for example 50-80% and 80-100%) and by 2 increments if the individual studies varied across 3 areas.

The evidence was downgraded by one increment if the 95% confidence interval crossed one clinical decision threshold and by two increments if it crossed two clinical decision thresholds. The GC set the thresholds for sensitivity and specificity as 50% (no better than chance) and 80% (threshold to recommend a test). Imprecision was assessed on confidence intervals produced by WinBUGS;