KEY POINTS
- The study evaluated a prospectively maintained cohort of patients receiving postoperative whole-breast radiotherapy with 40 Gy in 15 fractions, without nodal irradiation or tumour-bed boost. After exclusions, 1,091 patients were available for modelling.
- Treatment used tangential 3D-CRT, and mean heart dose averaged 1.72 Gy. Coronary calcification was quantified automatically on planning CT using Agatston score, calcification volume, and maximum Hounsfield-unit density.
- During a median follow-up of 6.5 years, only 29 cardiac events occurred. Fifteen were classified as calcification-related—coronary disease, myocardial infarction, heart failure, or valvular disease—and 14 as less directly related to calcification, including arrhythmias, conduction disorders, pericarditis, and tachyarrhythmia.
- Calcification-related events were distributed similarly after right- and left-sided treatment (eight versus seven). Models combining age with a calcification metric achieved AUCs of approximately 0.79–0.80, while respiratory toxicity added no predictive value.
- Thirteen of the 14 less calcification-related events occurred after left-sided treatment. In the left-sided cohort, age plus fluorouracil–epirubicin–cyclophosphamide chemotherapy produced an AUC of 0.73; anthracycline treatment had an OR of 6.89 (95% CI 1.58–30.11).
- Adding grade ≥2 respiratory toxicity increased the AUC from 0.73 to 0.77. Respiratory events occurred in only 18 of 1,172 patients (1.5%), and the resulting estimates had wide uncertainty.
- Mean heart dose was not independently associated with either event category. This should not be interpreted as excluding a dose effect because modern heart doses were low, interpatient variation was limited, and cardiac-substructure doses were unavailable.
CLINICAL TAKEAWAY
Cardiac calcification visible on planning CT may identify patients whose risk is dominated by pre-existing atherosclerotic disease, while other cardiac events may reflect a more complex interaction with systemic therapy and thoracic irradiation. The low event count and post hoc classification make these findings hypothesis-generating rather than a validated risk model.