The aim of this review was to evaluate whether CBCT is reliable for the detection of root fractures in teeth without root fillings, and whether the voxel size has an impact on diagnostic accuracy. The studies published in PubMed, Web of Science, ScienceDirect, Cochrane Library, Embase, Scopus, CNKI and Wanfang up to May 2014 were the data source. Studies on nonroot filled teeth with the i-CAT (n=8) and 3D Accuitomo CBCT (n=5) units were eventually selected. In the studies on i-CAT, the pooled sensitivity was 0.83 and the pooled specificity was 0.91; in the 3D Accuitomo studies, the pooled sensitivity was 0.95 and pooled specificity was 0.96. The i-CAT group comprised 5 voxel size subgroups and the 3D Accuitomo group contained 2 subgroups. For the i-CAT group, there was a significant difference amongst the five subgroups (0.125, 0.2, 0.25, 0.3 and 0.4mm; P=0.000). Pairwise comparison revealed that 0.125mm voxel subgroup was significantly different from those of 0.2, 0.25 and 0.3mm voxel subgroups, but not from the 0.4mm voxel subgroup. There were no significant differences amongst any other two subgroups (by α'=0.005). No significant difference was found between 0.08mm and 0.125mm voxel subgroups (P=0.320) for the 3D Accuitomo group. The present review confirms the detection accuracy of root fractures in CBCT images, but does not support the concept that voxel size may play a role in improving the detection accuracy of root fractures in nonroot filled teeth.