Modeling Normal Tissue Complication Probability of Radiation-Induced Alopecia Following Intensity-Modulated Radiation Therapy in Glioblastoma Patients

Abstract

Background: Glioblastoma multiforme (GBM) is an aggressive primary brain tumor requiring radiotherapy (RT), which often leads to radiation-induced alopecia (RIA), a distressing toxicity impacting quality of life. This study establishes the first normal tissue complication probability (NTCP) framework for RIA in GBM patients, comparing the Lyman-Kutcher-Burman (LKB) model and multivariate logistic regression. Materials and Methods: A prospective cohort of 41 GBM patients undergoing intensity-modulated RT (IMRT) was analyzed. Scalp contours were generated for dosimetric evaluation, and NTCP was modeled using the LKB framework (parameters: TD50 = 24 Gy, m = 0.47) and logistic regression. Dosimetric parameters and clinical variables were assessed. Model performance was evaluated via AUC-ROC, Brier score, and Hosmer-Lemeshow tests. Results: Grade 2 RIA incidence was 46.3 at 3 months and 31.7 at 6 months. Dosimetric parameters D-max and concurrent chemotherapy were significant predictors (P < 0.05). The logistic regression model outperformed the LKB model (AUC: 0.91 vs. 0.89), demonstrating superior discrimination and calibration. For NTCP < 50, the generalized equivalent uniform dose threshold was identified as 38 Gy. Conclusion: Multivariate logistic regression, integrating dosimetric and clinical factors, offers enhanced predictive accuracy for RIA compared to the LKB model. These findings emphasize the importance of optimizing IMRT plans to limit scalp dose and considering chemotherapy in risk stratification. Implementation of these models may improve patient-centered care by balancing tumor control and toxicity reduction.