In a population-based cohort of 4694 women with invasive breast cancer, operated upon with breast conserving surgery (BCS) in 1981--1990 and followed through to 1997, we studied how this technique had been adopted into clinical practice, especially with reference to the use of radiotherapy (RT). Our main aim was to see whether there was a drift in the risk of local recurrence and breast cancer death over time. During the 30,151 person-years of observation in the cohort, there were 582 local recurrences, 456 breast cancer deaths and 438 deaths due to other causes. Postoperative RT was given to 70.2%, but usage increased over the period. The women not receiving RT were mostly elderly, but also in women <70 years, 20.4% did not receive RT. The risk for local recurrence after RT were 7.6 and 17.8% at 5 and 10 years, respectively. Without RT, more than 30% had a local recurrence at 10 years. Thus, the choice not to irradiate failed to target women at a low risk. In a multivariate Cox analysis taking tumour size, nodal status, age at operation and RT into account, there was a trend for a higher risk of local recurrence in the later time period, relative hazard 1.5 (95% confidence interval (CI) 1.0--2.1). Corrected survival was 93.3 and 85.2% at 5 and 10 years, respectively.