Twenty articles reporting 27 validation (n >33,938) studies were included in the review. Nine articles were hospital-based, 12 were population-based and one was both.
All but one of the studies stated that the same reference standard was used regardless of the index test result. Seventeen studies described the reference standard in sufficient detail to permit its replication; three reported uninterpretable and intermediate test results.
The following diagnostic accuracy ranges were reported for the evaluated criteria:
Hanifin and Rajka (two studies): sensitivity 87.9 per cent to 96 per cent, specificity 77.6 per cent to 93.8 per cent based on clinical diagnosis as the reference standard.
UK diagnostic criteria (19 studies): sensitivity 10 per cent to 100 per cent, specificity 89.3 per cent to 99.1 per cent based on reference standards of clinical diagnosis, Hanifin and Rajka, UK, and Japanese Dermatology Association Criteria.
Schultz-Larsen (two studies): sensitivity 88 per cent to 94.4 per cent, specificity 77.6 per cent to 95.9 per cent based on clinical diagnosis and Hanifin and Rajka as the reference standard.
Diepgen (one study using three lists described by Diepgen): sensitivity 83 per cent to 87.7 per cent, specificity 83.9 per cent to 87 per cent based on clinical diagnosis as the reference standard
Kang and Tian (one study): sensitivity 95.5 per cent, specificity 100 per cent based on Hanifin and Rajka as the reference standard.
ISAAC (one study): positive predictive value 48.8 per cent, negative predictive value 91.1 per cent based on clinical diagnosis as the reference standard