Logistic regression equation to screen for vertical root fractures using periapical radiographs

Abstract

To develop a predictive equation to screen for vertical root fractures (VRFs) by numerically evaluating the shapes of radiolucent areas on the periapical radiographs of endodontically treated maxillary incisors and premolars. 41 pre-operative periapical radiographs of maxillary incisors and premolars with radiolucent areas at root apices were used. Out of 41 teeth, 18 had a fractured root (VRF group) and 23 had a non-fractured root (non-VRF group). The periapical radiolucent area of each tooth was traced out by six examiners on a personal computer and two indices, "Complexity" and "Radial SD", were measured. For each index, the difference between the VRF and non-VRF groups and the interexaminer differences were analysed with two-way ANOVA at 5% significance level. Multiple logistic regression analysis was used to develop a predictive equation and the probability of VRF in all samples was calculated. A receiver operating characteristic (ROC) curve was constructed to select the optimal cut-point. Each sample was predicted as "VRF" or "non-VRF" with this cut-point. For both "Complexity" and "Radial SD", the VRF group showed significantly greater values than the non-VRF group (P<0.05). With a cut-point derived from the ROC curve, sensitivity, specificity and efficiency of VRF were 0.68, 0.80 and 0.75, respectively. VRF teeth have more complicated radiolucent areas compared with non-VRF teeth. By evaluating the shapes of radiolucent areas, a logistic regression equation to screen for VRF was calculated and this equation could contribute to the diagnosis of VRF.