Validation of 3D intraoral scanner use for detecting caries lesions around dental restorations: A diagnostic accuracy study

Abstract Objectives This study aimed to validate the use of a 3D intraoral scanner (IOS) for detecting secondary caries lesions around restorations using image-based CARS-derived severity categories, compared with visual-tactile clinical examination. Materials and methods A diagnostic validation study was conducted at the university clinical facilities, from July 2023 to September 2024, including adult patients with amalgam or composite resin restorations on anterior and posterior teeth. The restorations were assessed using visual-tactile clinical examination, adopted as the clinical reference standard, and IOS image-based assessment. Sensitivity, specificity, predictive values, binary-classification AUC, and clinical–IOS agreement were estimated. Confidence intervals for diagnostic accuracy measures were obtained using participant-level cluster bootstrap resampling. Results A total of 299 restorations in 41 patients were assessed. IOS sensitivity varied, with 75% (95% CI: 66.2–82.0) at the initial lesion threshold, 40% (95% CI: 20.8–57.7) at the moderate lesion threshold, and 57.1% (95% CI: 33.3–100.0) at the severe lesion threshold. Specificity was consistently high, with 84.1% (95% CI: 78.7–89.4), 97.4% (95% CI: 95.6–99.2), and 99.7% (95% CI: 98.7–99.7) at the initial, moderate, and severe lesion thresholds, respectively. Binary-classification AUC values were 0.79 (95% CI: 0.75–0.84), 0.69 (95% CI: 0.58–0.78), and 0.78 (95% CI: 0.66–1.00) at the initial, moderate, and severe lesion thresholds, respectively. Positive predictive value was highest at the severe lesion threshold (80%), while the negative predictive value was especially high at this threshold (99%). Exact clinical-IOS agreement was 77.9%, whereas linearly weighted observed agreement was 91.1%, with a linearly weighted kappa coefficient of 0.53. Conclusions The 3D intraoral scanner is a complementary tool for assessing secondary caries lesions, showing high specificity across lesion thresholds. Sensitivity varied according to lesion severity and was less precise at the severe threshold due to the small number of severe lesions. Future research should increase sample sizes and explore combining IOS with other diagnostic methods to improve detection, especially for moderate lesions and higher caries prevalence. Clinical Relevance Conventional IOS images may assist in documenting and assessing visible changes around restorations. Clinical interpretation should consider the observed sensitivity limitations and incorporate visual-tactile examination and radiographic assessment when indicated.