Development and validation of a normal tissue complication probability model for acquired nasal cavity stenosis and atresia after radical radiotherapy for nasopharyngeal carcinoma.

PURPOSE: Curative radiotherapy for nasopharyngeal carcinoma (NPC) can lead to acquired nasal cavity stenosis and atresia (ANCSA). As the first study to investigate risk factors of ANCSA in a large cohort of NPC patients, this article aims to develop and validate a multivariate normal tissue complication probability (NTCP) model to predict the development of ANCSA and to establish a nomogram for clinical use. METHODS AND MATERIALS: The retrospective cohort was comprised of 548 NPC patients treated with radical radiotherapy. The cohort was randomly divided into training and validation groups. Least absolute shrinkage and selection operator regression was performed for variable selection from the clinical and dosimetric characteristics in the training group. A multivariate NTCP model and a nomogram were established for the prediction of ANCSA development. Discrimination and calibration were tested using receiver operating characteristic (ROC) curves and calibration tests, respectively, for both groups. RESULTS: ANCSA was observed in 132 (24.1%) of 548 patients with NPC who underwent radical radiotherapy. The median time to ANCSA detection after treatment was 2.8 months (range, 0.0-57.7 months). Five potential predictors, including choanal invasion, low white blood cell count, high C-reactive protein level, high serum amyloid A level, and high V70Gy of the nasal cavity, were selected to develop the NTCP model based on 365 patients in the training group. The model had a fairly good discriminative power according to the ROC analysis in both the training (area under ROC curve = 0.79, 95%CI: 0.73-0.84) and validation (0.73, 0.64-0.82) groups. The calibration power was tested using the calibration test in the training (E-max = 0.069, E-avg = 0.015, p = 0.977) and validation (E-max = 0.057, E-avg = 0.032, p = 0.747) groups. CONCLUSIONS: We developed and successfully validated an NTCP model for early prediction of ANCSA in patients with NPC after radical radiotherapy. This could help clinicians assess the risk of ANCSA before the initiation of follow-ups and ensure appropriate and timely management of this complication.