Background And Purpose: In patients with locally advanced non-small cell lung cancer (LA-NSCLC) post-radiotherapy, mean heart dose (MHD) and the percent of left anterior descending (LAD) coronary artery receiving ≥15 Gy (LADV15) are associated with major adverse cardiac events (MACE). We developed a MACE prediction model in this population.
Materials And Methods: Total 701 patients with LA-NSCLC treated with curative-intent radiotherapy reviewed, split by diagnosis date into "development" (n = 500) and later (n = 201) "test" cohorts. Development patients were analyzed using a multivariable Cox-proportional hazard model with backward elimination scheme (Bonferroni-adjusted α = 0.025). Potential predictors were selected a priori: age, coronary heart disease (CHD), Framingham Risk, hypertension, MHD, LADV15, intensity modulated radiotherapy use, and CHD and LADV15 interaction (CHD:LADV15). Cardiac doses as quadratic, square root, and logarithmic (ln[X + 1]) forms were explored. Models were internally validated with bootstrapping.
Results: Final model incorporated CHD, Hypertension, Logarithmic LADV15, and CHD*ln[LADV15 + 1] (CHyLL; β coefficients: 5.51, 1.28, 1.48, -1.36; all p < 0.025; bootstrapping c-index: 0.80; test cohort c-index: 0.76). Possible risk score range: 0-8.11. MACE incidence was 6.8% and 23.6% at 48 months (p = 0.041), and survival rates were 51.6% and 35.0% (p = 0.099), in the low-risk (score <5.00) and high-risk (score ≥5) test groups, respectively. Using the model, calculated LADV15 constraints for patients without CHD were 11.3% and 28.3% for those with and without hypertension, respectively, to remain low-risk.
Conclusions: Pre-existing CHD, hypertension, and LADV15 were important factors in predicting MACE after radiotherapy. CHyLL has the potential to estimate personalized LADV15 constraints based on cardiac risk factors and acceptable MACE thresholds.
Download full-text PDF |
Source |
---|---|
http://dx.doi.org/10.1016/j.radonc.2022.02.010 | DOI Listing |
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!