A Clinical Prediction Model of Asymptomatic Left Ventricular Dysfunction in Survivors of Childhood Cancer in Ontario, Canada
Author(s)
Roaa Shoukry, BSc1, Valerie Liu, BSc2, Danielle Weidman, MD, FRCPC, MHSc2, Sameera Ahmed, MSc2, Paaladinesh Thavendiranathan, MD, MSc, FRCPC2, Aasthaa Bansal, MS, PhD3, David C. Hodgson, MPH, FRCPC, MD2, Petros Pechlivanoglou, MSc, PhD1;
1The Hospital for Sick Children, Toronto, ON, Canada, 2University Health Network, Toronto, ON, Canada, 3University of Washington, Seattle, WA, USA
1The Hospital for Sick Children, Toronto, ON, Canada, 2University Health Network, Toronto, ON, Canada, 3University of Washington, Seattle, WA, USA
Presentation Documents
OBJECTIVES: Childhood cancer survivors (CCS) treated with anthracycline chemotherapy or radiation are at increased risk of developing cardiomyopathy and heart failure. A clinical prediction model can identify survivors at high risk for asymptomatic left ventricular dysfunction (ALVD), which precedes heart failure, supporting medical decision-making and improving CCS health outcomes. A clinical prediction model for ALVD risk in CCS was developed and evaluated using patient data from Ontario, Canada. Its results will inform a cost-utility analysis over a lifetime horizon to estimate the incremental costs and health outcomes of different cardiac screening schedules for CCS.
METHODS: Longitudinal data from Princess Margaret Cancer Centre in Ontario, Canada were used in a landmark analysis. The cohort comprised CCS who were treated with anthracyclines or chest radiation between 0-20 years and received at least one post-treatment echocardiogram between 1980/01/01 to 2022/05/31. ALVD was defined by ejection fraction (EF) percent < 51% in males and < 53% in females. Landmark times were set every two years over 22 years to train the model, with two-year predictions of ALVD risk. A Cox proportional hazards model was then fitted to interval-censored data. Calibration and time-dependent dynamic AUCt scores were used to assess predictive performance at each landmark time.
RESULTS: In this cohort, 94 survivors (n = 901) experienced ALVD. Fixed predictors included age at cancer diagnosis, time between diagnosis and first screening, and cardiomyopathy risk from treatment. Time-varying predictors included age, number of past screenings, and EF at screening, with interactions with landmark times (linear and quadratic). The model produced AUCt scores of 77.3% (95% CI 72.6, 81.9).
CONCLUSIONS: This prediction model demonstrated high predictive performance of ALVD in CCS using standard clinical parameters. The time-varying hazard ratios generated will be used in the microsimulation-based cost-utility analysis for routine echocardiographic screening in cancer survivors in Canada.
METHODS: Longitudinal data from Princess Margaret Cancer Centre in Ontario, Canada were used in a landmark analysis. The cohort comprised CCS who were treated with anthracyclines or chest radiation between 0-20 years and received at least one post-treatment echocardiogram between 1980/01/01 to 2022/05/31. ALVD was defined by ejection fraction (EF) percent < 51% in males and < 53% in females. Landmark times were set every two years over 22 years to train the model, with two-year predictions of ALVD risk. A Cox proportional hazards model was then fitted to interval-censored data. Calibration and time-dependent dynamic AUCt scores were used to assess predictive performance at each landmark time.
RESULTS: In this cohort, 94 survivors (n = 901) experienced ALVD. Fixed predictors included age at cancer diagnosis, time between diagnosis and first screening, and cardiomyopathy risk from treatment. Time-varying predictors included age, number of past screenings, and EF at screening, with interactions with landmark times (linear and quadratic). The model produced AUCt scores of 77.3% (95% CI 72.6, 81.9).
CONCLUSIONS: This prediction model demonstrated high predictive performance of ALVD in CCS using standard clinical parameters. The time-varying hazard ratios generated will be used in the microsimulation-based cost-utility analysis for routine echocardiographic screening in cancer survivors in Canada.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
CO37
Topic
Clinical Outcomes
Topic Subcategory
Relating Intermediate to Long-term Outcomes
Disease
SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory), SDC: Oncology, SDC: Pediatrics