There is considerable concern about the adverse effects on the skeleton of loss of menstrual function as a result of athletic activity, as well as uncertainty as to how it should be managed clinically. In a pilot intervention study 34 elite middle and long-distance runners, aged 18-35 years, with menstrual irregularity due to their athletic activity were randomized to three groups: (A) to receive hormone replacement therapy (HRT) and 1000 mg calcium per day (n = 10), (B) to receive 1000 mg calcium per day (n = 14), (C) a control group who received no treatment (n = 10). Bone mineral density (BMD) was measured in the left hip and lumbar spine (L2-4) using dual-energy X-ray absorptiometry. Results were first analyzed according to whether menstruation returned, either naturally or secondary to HRT (EU), and compared with those from subjects who remained amenorrheic (AM). During the first year BMD increased in the EU group in Ward's triangle (3.8%) and the lumbar spine (4.1%; both P < 0.05). BMD fell in the AM group in all regions and the between-group differences were 5.6% (p < 0.02) in Ward's triangle, 5.8% (p < 0.02) in L2-4 and 3.9% in the trochanter (p < 0.05). An 'intention to treat' analysis was then performed. It was found that the mean relative improvement at 1 year in spinal BMD was only 1.5%, due to return of menses in some of the controls and withdrawals from treatment in the treatment group. In consequence, a trial designed to show, with 80% power and 5% significance, a measurable benefit in lumbar spine BMD resulting from allocation to HRT treatment would require about 1150 athletes with amenorrhea or oligomenorrhea. These numbers could be reduced substantially to 380 subjects by confining the trial to completely amenorrheic athletes, who in this study were less likely to regain menses. For these and other logistical reasons, an HRT trial in amenorrheic athletes could only be successfully organized through international collaboration. This study illustrates the major effects of treatment withdrawals and instability of menstrual status on the design of longitudinal studies on the bony effects of menstrual dysfunction prior to menopause.