There is controversy about the ideal timing of hormone replacement therapy (HRT) and duration of treatment. In this study we have examined intrapair differences in bone mineral density (BMD) in twins who were discordant for HRT use. Twin pairs in which only one co-twin had been exposed to HRT for more than 12 months continuously were selected from 365 postmenopausal monozygotic (MZ) and dizygotic (DZ) pairs recruited as part of the St Thomas' Adult UK Twin Registry of normal volunteers. BMD was measured by dual-energy X-ray absorptiometry at the lumbar spine and femoral neck. Intrapair differences in BMD between HRT users and non-users were compared. A total of 65 HRT-discordant pairs were identified, of which 36 were discordant for current HRT use (mean age: 55.3 years, median duration of HRT use: 36 months) and 29 were discordant for past HRT use (mean age: 60.4 years, median HRT duration: 30 months). Among current users BMD was consistently and significantly higher than in non-users at both sites (lumbar spine mean intrapair difference (IPD%): 12.3%, 95% confidence interval (CI): 7.1%, 17.5%; femoral neck IPD%: 8.6%, 95% CI: 3.4%, 13.7%). The intrapair differences were substantially smaller when past users and non-users were compared (lumbar spine IPD%: 2.4%, 95% CI: -3.7%, 8.6%; femoral neck IPD%: 0.4%, 95% CI: -5.3%, 6.0%). These differences remained little changed after adjusting for the potential confounding effects of the duration of HRT use, and intrapair differences in alcohol and tobacco consumption and physical exercise. The results confirm, in a closely matched design, the findings of other observational research that current use of HRT has a major effect on BMD at the lumbar spine and femoral neck. Past users of HRT do not, however, show the same benefits. The clinical implications of these findings are that HRT needs to be used continuously to influence BMD and that alternative treatments need to be considered in those who discontinue HRT.