To evaluate the in vitro performance of the International Caries Detection and Assessment System (ICDAS) visual examination, bitewing radiography (BW), and the DIAGNOdent 2190, a pen-type laser fluorescence device (LFpen), in detecting caries around amalgam restorations on approximal surfaces. Approximal surfaces (N=136) of permanent posterior teeth (N=110) with Class II amalgam restorations were assessed twice by two experienced examiners using ICDAS, BW, and LFpen. The occurrence of proximal overhangs was also evaluated. The teeth were histologically prepared and assessed for caries extension. Different cutoff limits for the LFpen were used. Intraexaminer and interexaminer reproducibility showed moderate to good agreement for all the methods (weighted κ/intraclass correlation coefficient=0.40 to 0.87). The specificities at D1 (all visible lesions affecting enamel) and D3 (lesions extended into dentin) were, respectively, 0.41 and 0.82 for ICDAS, 0.70 and 0.82 for BW, and 0.77-0.89 and 0.88-0.94 for LFpen. The sensitivities were 0.80 and 0.52 for ICDAS, 0.56 and 0.51 for BW, and 0.04-0.23 and 0.01-0.02 for LFpen at D1 and D3, respectively. At the D1/D3 thresholds, the accuracy and the area under the receiver operating characteristic curve (Az) values were similar and statistically higher for ICDAS (0.65/0.68 and 0.633/0.688) and BW (0.64/0.68 and 0.655/0.719), respectively; whereas, LFpen presented lower accuracy (0.37-0.44/0.49-0.52) and Az (0.390-0.454/0.345-0.395) values. The occurrence of overhangs (26.8%) was shown to be irrelevant in determining the presence of secondary caries. The ICDAS and BW methods presented the best performance in detecting caries lesions affecting enamel and dentin on approximal surfaces of amalgam restorations.