To the Editor.
— Steinberg et al1 estimate that there is a 30% increase in the risk of breast cancer after 15 years of estrogen use. This conclusion is critically dependent on the model selected for the relationship between risk and duration of use. Their assumed model is as follows: risk α k1duration (or log [risk] = duration × log [k1]). An equally plausible model (shown to fit the data on smoking and lung cancer) is the following: risk α durationk (or log [risk] = k2 × log [duration]). In both equations k1 and k2 are constants.From Table 2 in the paper by Steinberg et al, the value of k1 is 1.018 (adding 1.0 to the proportional increase in risk of 0.018). This predicts an RR after 5 years' use of 1.09(1.0185).The second model predicts lower estimates for the excess