Development of multivariate NTCP models for radiation-induced hypothyroidism: a comparative analysis

Area of Science:

• Radiation Oncology
• Medical Physics
• Endocrinology

Background:

• Hypothyroidism is a common late side effect of cervical radiation therapy.
• Accurate prediction of radiation-induced hypothyroidism (RHT) is crucial for patient management.

Purpose of the Study:

• To develop and validate multivariate normal tissue complication probability (NTCP) models for RHT.
• To compare the predictive performance of developed models with existing RHT NTCP models.

Main Methods:

• Retrospective review of 53 patients treated with chemo-radiotherapy for Hodgkin's lymphoma.
• Analysis of thyroid gland dose distribution and clinical factors using Spearman's rank correlation and multivariate logistic regression with bootstrapping.
• Model performance evaluation using Area Under the Curve (AUC) and external validation.

Main Results:

• A two-variable NTCP model (V30(%) and gender) achieved an AUC of 0.87.
• A three-variable NTCP model (V30(cc), thyroid gland volume, and gender) showed improved prediction (AUC = 0.85).
• The three-variable model demonstrated superior performance on an external cohort (AUC = 0.914).

Conclusions:

• The developed three-variable NTCP model offers enhanced prediction capability for RHT.
• This model improves RHT risk assessment in both the initial patient cohort and an external validation group.
• The findings support the use of absolute thyroid volume, overall thyroid volume, and gender for predicting RHT.