1007 - Radiation Oncology Workflow-Integrated Coronary Artery Disease Risk Calculator for Pediatric and Adolescent Cancer Patients

03:10pm - 03:15pm ET

Room 256

Presenter(s)

Rebecca Howell, PhD - MD Anderson Cancer Center, Houston, TX

R. M. Howell¹, Q. Liu², T. G. Meyers¹, A. C. Paulino³, C. C. Pinnix³, S. Smith⁴, C. Owens⁵, N. Esiashvili⁶, M. Roth⁷, D. Noyd⁸, E. J. Chow⁹, G. T. Armstrong¹⁰, J. E. Bates¹¹, D. A. Mulrooney¹², and Y. Yasui¹²; ¹Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, TX, ²Department of Public Health Sciences, University of Alberta, Edmonton, AB, Canada, ³Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, ⁴The University of Texas MD Anderson Center, Houston, TX, ⁵The University of Texas MD Anderson Cancer Center, Houston, TX, ⁶Department of Radiation Oncology, Winship Cancer Institute of Emory University, Atlanta, GA, ⁷Department of Pediatrics, The University of Texas MD Anderson Cancer Center, Houston, TX, ⁸University of Washington/Seattle Children's Hospital, Seattle, WA, ⁹University of Washington, Seattle, WA, ¹⁰St. Jude Children's Hospital, Memphis, TN, ¹¹Department of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, GA, ¹²Department of Epidemiology and Cancer Control, St. Jude Children's Research Hospital, Memphis, TN

Purpose/Objective(s):

To develop and validate radiation oncology (RO) workflow-integrated personalized pediatric/adolescent cardiac substructure models to predict future risk of coronary artery disease (CAD).

Materials/Methods:

We developed piecewise exponential models in the Childhood Cancer Survivor Study (CCSS) and externally validated in the St. Jude Lifetime (SJLIFE) to predict severe/life-threatening CAD (events/participants: 549/19,962 CCSS; 129/4,389 SJLIFE). Mean cardiac substructure dose (MCSD) models used lasso regression with 10-fold cross-validation (100 repeats) to select dosimetric predictors from mean doses to coronary arteries, atria, valves, ventricles, and heart V5/V20; we included those retained in =95% iterations. Comparison models used mean heart dose (MHD). Attained age, sex, race/ethnicity, diagnosis age, and anthracycline dose were forced variables. To assess age/lifestyle-acquired cardiovascular risk factors (CVRF: obesity, smoking, hypertension, dyslipidemia, and diabetes), models were refit among those that survived =20 yrs post-diagnosis. An in-silico study estimated CAD risk for 22 children/adolescents treated with mediastinal IMRT and matched simulated IMPT plans. Risk was also simulated for worst-case scenarios where survivors develop all five CVRF in the decades after treatment. A Python-based calculator was implemented in the RayStation treatment planning system.

Results:

MHD models achieved internal/external AUC of 0.736/0.686; CAD rate doubled per 10-Gy MHD (rate ratio (RR): 2.0, 95% CI: 1.87-2.14; p<0.001). MCSD models with left ventricle and left main coronary artery doses statistically significantly improved prediction (internal/external AUC: 0.744/0.694; p=0.016 vs MHD), with RR per 10-Gy of 1.35 (95% CI:1.11-1.62) and 1.49 (95% CI:1.32-1.67), respectively (p<0.002). The =20-year post-diagnosis MHD models achieved AUC 0.757/0.712; Adding MCSDs did not improve fit. Obesity, smoking, hypertension, dyslipidemia, and diabetes increased CAD rates by factors of 1.36, 1.32, 2.80, 1.05, and 2.07, respectively.

In-silico, MCSD models predicted median absolute CAD rates of 5.22 at age 50 per 1,000 person-yrs (IQR:4.43-7.62) for IMRT, reduced by 35% with IMPT (3.89; IQR:2.87-4.55). MHD models predicted lower rates (IMRT: 3.64; IQR:3.02-6.31) with ~15% reduction using IMPT (3.15; IQR:2.75-4.57). The calculator reports the higher of MHD- or MCSD-risk to support conservative decision making. For reference, the predicted CAD rate is 2.49 (IQR:1.69-3.28) among a same sex matched control group (CCSS siblings without cancer). Among =20-year survivors, if five CVRF were to develop verses none, predicted CAD rate would increase ~5-fold (IMRT RR:4.66, IQR:2.89-6.25; IMPT RR: 4.67, IQR:2.90-6.26).

Conclusion:

We developed/validated RO workflow-integrated CAD risk calculators enabling patient-specific, real-time estimates of late CAD risk based on MCSD/MSD, anthracycline dose, demographics, and age/lifestyle-acquired CVRF.