PURPOSE Radiation-associated cardiac disease is a major cause of morbidity/mortality among childhood cancer survivors. Radiation dose-response relationships for cardiac substructures and cardiac diseases remain unestablished. METHODS Using the 25,481 5-year survivors of childhood cancer treated from 1970 to 1999 in the Childhood Cancer Survivor Study, we evaluated coronary artery disease (CAD), heart failure (HF), valvular disease (VD), and arrhythmia. We reconstructed radiation doses for each survivor to the coronary arteries, chambers, valves, and whole heart. Excess relative rate (ERR) models and piecewise exponential models evaluated dose-response relationships. RESULTS The cumulative incidence 35 years from diagnosis was 3.9% (95% CI, 3.4 to 4.3) for CAD, 3.8% (95% CI, 3.4 to 4.2) for HF, 1.2% (95% CI, 1.0 to 1.5) for VD, and 1.4% (95% CI, 1.1 to 1.6) for arrhythmia. A total of 12,288 survivors (48.2%) were exposed to radiotherapy. Quadratic ERR models improved fit compared with linear ERR models for the dose-response relationship between mean whole heart and CAD, HF, and arrhythmia, suggesting a potential threshold dose; however, such departure from linearity was not observed for most cardiac substructure end point dose-response relationships. Mean doses of 5-9.9 Gy to the whole heart did not increase the risk of any cardiac diseases. Mean doses of 5-9.9 Gy to the right coronary artery (rate ratio [RR], 2.6; 95% CI, 1.6 to 4.1) and left ventricle (RR, 2.2; 95% CI, 1.3 to 3.7) increased risk of CAD, and to the tricuspid valve (RR, 5.5; 95% CI, 2.0 to 15.1) and right ventricle (RR, 8.4; 95% CI, 3.7 to 19.0) increased risk of VD. CONCLUSION Among children with cancer, there may be no threshold dose below which radiation to the cardiac substructures does not increase the risk of cardiac diseases. This emphasizes their importance in modern treatment planning.