A better understanding underlying radiation (RT) response after breast-conserving surgery (BCS) is needed to mitigate over-treatment of DCIS. The hazard ratio (HR) measures the effect of RT but assumes the effect is constant over time. We examined the hazard function adjusted for adherence to surveillance mammography to examine variations in LR risk and the effect of RT over time.
Crude hazard estimates for the development of LR in a population cohort of DCIS treated by BCS ± RT were computed. Multivariable extended Cox models and hazard plots were used to examine the association between receipt of RT and risk of each outcome adjusted for baseline covariates and adherence to mammography.
Population cohort includes 3262 women treated by BCS; 1635 received RT. Median follow-up was 13 years. LR developed in 364 women treated by BCS alone and 274 treated with RT. LR risk peaked at 2 years, declined until year 7, and then remained steady. The peak hazard of LR was associated with adverse features of DCIS. Early LR risk was attenuated in patients treated with RT but late annual risks of LR and invasive LR were similar among the two treatment groups. On multivariate analysis, RT was associated with a reduction in early LR risk (HR = 0.52, 95% CI 0.43-0.63, p < 0.0001) but did not reduce the risk of late LR (HR = 0.89, 95% CI: 0.67, 1.19, p = 0.44) (interaction, p = 0.002).
The effect of RT is not uniform over time and greatest in the first 7 years after BCS for DCIS, which can guide future research to understand mechanisms underlying RT response and optimize future management of DCIS.