Objectives: It is not generally possible to assess diagnostic accuracy in d
ental surveys as no histological 'gold standards' are available, therefore
examiner agreement tends to be used as a proxy for accuracy. The aim of thi
s study was to investigate, using extracted teeth in arch models, the in vi
tro validity of a diagnostic system to assess caries at the D-1 (enamel and
dentine) and D-3 (dentine) diagnostic thresholds, for epidemiological purp
oses. Method: Two groups of 10 dental examiners trained in the use of the D
undee Selectable Threshold Method for caries diagnosis (DSTM) each examined
(on two occasions) 160 extracted permanent molar and premolar teeth set in
arch models in phantom heads according to the codes and criteria of the DS
TM. The teeth were subsequently radiographed and sectioned to provide valid
ation of the diagnoses. Results: Intra-examiner agreement according to the
kappa statistic was substantial. In general terms the results of the in vit
ro validation exercise demonstrated significantly higher sensitivity values
at the D-1 diagnostic threshold than were found at the D-3 diagnostic thre
shold with a consequent toss of specificity. Conclusions: The results of th
is in vitro validation exercise demonstrate that at the D-1 diagnostic thre
shold the sensitivity of the DSTM was greater than at the D-3 threshold ind
icating no loss of diagnostic accuracy at the D-1 threshold.