Paired in vivo diagnostic accuracy and reliability of QrayCam Pro and QrayPen for occlusal and proximal caries in permanent posterior teeth

Abstract Background Quantitative light-induced fluorescence (QLF) provides fluorescence-loss and red-fluorescence signals that may behave differently across devices, surfaces, and lesion thresholds. This study compared QrayCam Pro and QrayPen for occlusal and proximal caries in permanent posterior teeth. Methods In this single-centre, paired in vivo diagnostic-accuracy and reliability study, 120 consecutively recruited adults underwent QrayPen and QrayCam Pro imaging. Multiple surfaces could be contributed by one participant. Surface-specific regions of interest in QA2 software generated maximum fluorescence loss (|ΔFmax|) and maximum red fluorescence (ΔRmax). Occlusal surfaces were classified using ICDAS and proximal surfaces using bitewing-based International Caries Classification and Management System (ICCMS)-oriented categories. Agreement was evaluated on 194 occlusal and 397 proximal paired measurements. Diagnostic analyses included 181 occlusal surfaces with complete ICDAS scores and all 397 proximal surfaces. AUC confidence intervals and paired AUC contrasts were estimated by 10,000 participant-cluster bootstrap resamples; Holm adjustment was applied across four paired contrasts within each surface-threshold family. Results At ICDAS ≥ 1, QrayCam |ΔFmax| and QrayPen |ΔFmax| yielded AUCs of 0.909 (95% CI 0.862–0.950) and 0.872 (0.821–0.919), respectively; |ΔFmax| exceeded ΔRmax within both devices (Holm-adjusted P < 0.001), whereas the QrayCam-QrayPen |ΔFmax| difference did not remain significant after multiplicity correction (adjusted P = 0.061). At proximal ICCMS ≥ 1, AUCs ranged from 0.646 to 0.697 and no paired contrast was significant after adjustment. QrayPen |ΔFmax| exceeded QrayPen ΔRmax at ICCMS ≥ 2 (adjusted P = 0.022) and ≥ 3 (adjusted P = 0.005); at ICCMS ≥ 3, QrayCam ΔRmax exceeded QrayPen ΔRmax (adjusted P = 0.011). Excluding the verified QrayPen ΔRmax value of 1962 changed individual proximal AUCs by no more than 0.012. Conclusions Metric selection influenced diagnostic performance at selected thresholds, but it was not uniformly more important than device selection. |ΔFmax| was more consistent for early occlusal detection and selected proximal thresholds, while device differences were generally small or threshold-specific. QLF should be used as an adjunct to structured visual examination and bitewing radiography rather than as a replacement. Trial registration Not applicable; this was an observational diagnostic-accuracy study and was not prospectively registered.