ESTRO 2024 - Abstract Book

S1670

Clinical - Lung

ESTRO 2024

The analysis of historical data identified three distinct patient subgroups associated with risk of elevation of hs-cTnT from low to high: Group 1 (tumor location=right), Group 2 (tumor location=left/mediastinal and age<64), and Group 3 (tumor location=left/mediastinal and age≥64). We identified the most important treatment -related factor, mean heart dose (MHD), for designing the optimal radiation plan. We also determined several crucial risk factors to be incorporated into the personalized NTCP model, including the presence of pre-existing heart disease (PHD) and the baseline hs- cTnT (BTnT) status (≥10 ng/L decision threshold). We then built a hierarchical logistic regression model to quantify the personalized NTCP: Under the Bayesian framework, proper prior distributions were specified for the model parameters and the posterior samples of Pr(elevated hs-cTnT) were obtained using Markov Chain Monte Carlo methods to estimate the cardiac risk for each patient risk group. We have utilized the posterior distribution of Pr(elevated hs-cTnT) to determine whether the standard radiotherapy plan is excessively toxic or if a personalized MHD is needed to mitigate the cardiac risk. Additionally, we have designed a Bayesian optimal adaptive monitoring rule [7] using a Dirichlet-multinomial model to ensure that the PART plan we have developed can effectively reduce treatment-related cardiac toxicity. The Bayesian monitoring rule simultaneously monitors the above two co-primary endpoints with the go/no-go decision boundaries optimized under the null hypothesis of Pr(elevated hs- cTnT)>25% and Pr(grade≥2 cardiac AEs)>20% and the alternative hypothesis of Pr(elevated hs- cTnT)≤15% and Pr(grade≥2 cardiac AEs)≤15%. The optimal monitoring rule can achieve a 5% type I error rate of 5% and 95% study power. As of September 2023, 73 patients have been enrolled (50 and 23 in cohorts 1 and 2, respectively, Table 1). The first cohort of patients were treated using the standard radiation plan and the data collected in cohort 1 were used to optimize the parameters of the personalized NTCP model. The recommended maximum MHD based on the initial version of the PART strategy is given in Table 1. The PART plan was then implemented and will be further refined from cohort 2 underway currently. Preliminary data show that the incidence of hs-cTnT elevation during radiation was 34% (12 out of 35 evaluable patients) in cohort 1 and 11% (2 out of 18 evaluable patients) in cohort 2, with 91% of cohort 2 patients (21 out of 23) having the PART strategy triggered. The median treated MHD (interquartile range) in cohort 1 was 6.5 (2.2-12.3) and that in cohort 2 was 5.4 (2.0-8.4). Pr(elevated hs-cTnT)=logit -1 (a g +b g log(MHD)+c 1 PHD+c 2 BTnT)