More large trials needed to decide best duration of treatment with tamoxifen

EDITOR—I agree with Rea et al that we do not know the optimum duration of adjuvant tamoxifen treatment.1 The data that they presented, however, suggest that the trials they advocated—ATLAS (adjuvant tamoxifen longer against shorter) and aTTom (adjuvant tamoxifen treatment offers more?)—may still fail to resolve this issue. The most recent overview, which provides much indirect evidence, suggests that five years is better than two, but it also reports a non-significant trend for the incidence of endometrial cancer to rise with prolonged exposure to tamoxifen.2 The three randomised studies comparing five with 10 years suggest little added benefit for the extra years of tamoxifen; indeed one study reported a poorer disease free survival.3^–5

When the ATLAS and aTTom studies are analysed the patients recruited in the early years will inevitably contribute more patient years and thus more events. Rea et al's data suggest that a much higher proportion of these early patients received tamoxifen for less than three years, a trend reversed more recently. Hence if the recent overview data reflect the true situation the half of these early patients randomised to five extra years of tamoxifen will show a considerable benefit. In contrast, the published data suggest that the more recently recruited patients, most of whom received over five years' treatment, may not show a similar benefit. As later recruits, however, they will contribute fewer events and may therefore have little impact on the trial outcome. There is therefore the distinct possibility of the false positive result that an extra five years' tamoxifen improves outcome.

Given the increasing trend to give tamoxifen for five years (as evidenced by the data reported by Rea et al), patients might be given around 10 years' treatment because of misleading trial results. Any impact on the incidence of endometrial cancer of the extra five years will probably not have been seen, as the number of women in the ATLAS and aTTom studies is smaller than that reported on in the recent overview.1 Indeed, the extra five years could be hazardous if the trend towards increasing endometrial cancer with prolonged duration is in fact real but only fails to be significant because of small numbers.

Clearly, trials on the scale of ATLAS and aTTom are needed, but the issue of the heterogeneity of the duration of tamoxifen treatment before recruitment needs to be addressed in the study design.