CT imaging-based nomogram for predicting early-stage glottic cancer recurrence following transoral laser microsurgery

Abstract

ABSTRACT Objective: To explore the differences between clinical features and computed tomography (CT) findings of early-stage glottic cancer (EGC) with or without recurrence after transoral laser microsurgery (TLM) and to establish a preoperative nomogram to predict postoperative recurrence. Methods: The clinical and CT features of 168 consecutive patients with EGC with or without recurrence were analyzed retrospectively. Multivariate logistic regression analysis was used to determine the independent predictors of recurrence. A nomogram was constructed to preoperatively predict recurrence. To assess the nomogram’s performance, the C-index and calibration plot were used. Results: EGCs with and without recurrence differed significantly in T-stage, depth, and normalized CT values in the arterial phase (NCTAP) and venous phase (NCTVP) (all P < 0.05). T-stage, depth, and NCTVP were independent predictors of recurrence in EGCs (all P < 0.05). The C-index (0.765, 95% confidence interval: 0.703–0.827) and calibration plot showed that the nomogram has good prediction accuracy. Nomograms based on T-stage and CT variables provided numerically predicted recurrence rates and were better than those based on only T-stage (C-index of 0.765 vs. 0.608). Conclusions: Using clinical and CT variables, we developed a novel nomogram to predict the recurrence of EGC before TLM, which may be a potential noninvasive tool for guiding personalized treatment.