Methods: To perform a simulation of the diagnosis of crown margin caries, an in vitro model with extracted crowned teeth was established. A total of 32 approximal crown margin (1.5 mm subgingival) surfaces were evaluated by ten clinicians (n=320) using the conventional diagnostic methods - visual-tactile and radiographic examination (carious lesion present: yes/no). These evaluations were compared to the histological findings. The statistical validation included the sensitivity, specificity, positive (PPV) and negative predictive value (NPV), receiver operating characteristics (ROC), kappa score and logistic regression.
Results: Data analysis demonstrated deficiencies of the conventional methods for diagnosing caries at crowns margins. NPV data were more accurate when a carious lesion had lead to a cavitation (95-98%). Caries lesions located especially on distal tooth surfaces were poorly diagnosed (sensitivity 24-50%, ROC 0.47-0.55, logistic regression p <0.05). Logistic regression was used to describe factors influencing a false positive caries diagnosis by the examiner. Tooth surface location (mesial/distal) and the type of lesion significantly affected the odds ratio for false positive diagnosis (p <0.05). Whereas the diagnostic method, experience of the clinician (> or <3 years) and the order of the evaluation did not. An influence was found between the presence of a horizontal crown margin gap and the examiner's caries diagnostic (Specificity 90-100%, PPV 96-100%).
Conclusions: Visual-tactile and radiographic evaluations represented weak diagnostic accuracies for the detection of caries situated at 1.5 mm subgingivally placed approximal crown margins.