A meta-analysis was conducted to reach a pooled estimate of the diagnostic accuracy of the SCOFF. The 15 selected studies represented a total of 882 cases and 4350 controls. The main criterion for inclusion was that the primary study had provided diagnostic classification with both a diagnostic reference and with the SCOFF (with five items and a cut-off point of two). The pooled estimates were .80 (sensitivity) and .93 (specificity). The moderator variables gender and type of measure for the diagnostic reference (interview versus psychometric tests) account for part of the observed variability. For diagnostic references based on interviews the estimate of the efficacy improves significantly. For the studies that match this criterion the sensitivity is .882 and the specificity .925 (diagnostic odds ratio, 92.19). The main conclusion was that the five questions of the SCOFF constitute a very useful screening tool, in several languages; it is highly recommended for screening purposes.
