Systematic review of prognostic models for cardiomyopathy and heart failure applicable to survivors of adolescent and young adult cancer.

Abstract

12037 Background: Survivors of adolescent and young adult (AYA) cancer who receive cardiotoxic chemotherapy and/or radiation are at risk of developing cancer therapy-related cardiac dysfunction (CTRCD), including asymptomatic reduction in left ventricular ejection fraction and symptomatic heart failure. Early detection and intervention using risk-adapted surveillance strategies can reduce morbidity and mortality. While numerous risk prediction models (RPM) have been developed and/or validated for CTRCD in pediatric or older adult cancer populations, it is unclear whether they can be applied to survivors of AYA cancer. Methods: We undertook a systematic review of cardiovascular RPM (including CTRCD) development/ validation studies in survivors of cancer diagnosed at any age. We searched MEDLINE, EMBASE, and Web of Science with additional hand searching until November 2024. Two reviewers screened abstracts and full texts; we included studies that used real-life patient data to predict asymptomatic or symptomatic CTRCD (per European Society of Cardiology guidelines) ≥1 year from diagnosis and after completing therapy. We excluded abstracts, non-English studies, and RPM which used data not routinely accessible in outpatient clinics. We extracted study data and applied the Prediction model Risk of Bias ASsessment Tool for risk of bias (RoB) and applicability to AYA cancer survivors. When not reported, we estimated AYA (age 15-39) proportions using cancer incidence patterns. We used descriptive statistics to evaluate studies, models, included risk factors, and participants overall and by proportion of AYA. Meta-analysis was not possible given limited overlap in the identified models. Results: We screened 12740 abstracts and 249 full text articles; of these, 100 studies underwent a second full text screen to identify CTRCD models. We identified 22 studies (7 enrolled > 20% AYA) which developed and/or validated 64 models (32.8% machine learning) to predict CTRCD (54.7% symptomatic) in 129077 cancer survivors (10.7% AYA) using clinically available data. Nine (14.1%) models validated existing RPM (e.g. those developed in non-cancer populations); the remainder were newly developed models, with age at diagnosis, diabetes, hypertension, and anthracycline dose the most common predictors ( > 50% of models). Most models (n = 54) were at high RoB, primarily due to concerns with analytical reporting. Only 2 studies/8 models were rated as both low RoB and high applicability to AYA survivors. Conclusions: Several RPM for subclinical and overt CHF are available for pediatric and adult cancer survivors, with varying applicability to AYA cancer survivors. Adherence to recommended statistical reporting methods is poor and RoB high, further limiting utility. Additional development and validation of high-quality RPM for heart failure in AYA cancer survivors is warranted.