Track event model is superior to linear quadratic cell survival model for predicting TCP of SBRT treatments of NSCLC

PURPOSE: The linear-quadratic (LQ) model is widely used in clinical practice, particularly for estimating equivalent fractionation schemes that yield a given isoeffect. While its validity for low-dose per fraction treatments is well established for in-vivo applications, its accuracy in predicting outcomes for high-dose fractionated radiotherapy remains debated. In response, several mechanistic models have been proposed as alternatives to this empirical approach. Among them, the track-event model (TEM) uniquely incorporates a two-parameter structure that exhibits exponential behavior at high doses and a finite gradient at null dose -an effect supported by experimental evidence. In this study, we compared the predictive accuracy of the LQ and TEM models specifically for stereotactic radiotherapy regimens. MATERIALS AND METHODS: A population-based TCP model incorporating variability in radiosensitivity and tumor volume was used to fit clinical data from 15 NSCLC studies involving low-dose per fraction treatments, using a log-likelihood optimization approach. These fitted models were then used to predict responses at high-dose levels and evaluated against observed stereotactic treatment outcomes from 24 additional studies. Model performance was assessed using log-likelihood values, along with the Bayesian Information Criterion (BIC) and Akaike Information Criterion (AIC) for comparative analysis. RESULTS: When comparing observed treatment outcomes with model-based predictions, the LC-TEM formulation demonstrates significantly greater predictive accuracy than the LC-LQ model at high doses per fraction. This is reflected in the evaluation metrics: log-likelihood per point, AIC, and BIC values were -2532 vs. -4413, 5075 vs. 8837, and 5088 vs. 8850 for LC-TEM and LC-LQ, respectively. CONCLUSION: The TEM offers a more accurate alternative to the LQ cell survival model for translating between different dose fractionation schemes, particularly those involving high doses per fraction. As a mechanistic model, TEM retains the simplicity of its empirical counterpart by relying on only two parameters, while providing improved predictive performance.