The purpose of this study was to evaluate whether premenopausal women's voluntary unsupervised aerobic and step training could maintain the skeletal benefits obtained by an 18-month supervised high-impact training, and if so, to what extent. Thirty women of the original 39 study subjects (i.e., persons who completed the preceding 18-month randomized training intervention and who volunteered to continue the training on their own for a further 8 months) and 19 women of the 45 original control subjects (i.e., persons who volunteered to continue as controls) were included. The study group trained an average of twice per week and the training consisted of regular aerobic and step classes provided by local fitness centers. Areal bone mineral density (BMD, g/cm 2 ) was measured from the lumbar spine, femoral neck, trochanter area of the femur, distal femur, patella, proximal tibia, calcaneus, and dominant distal radius at baseline and after 18 and 26 months. During the extended 8-month follow-up, the BMD of the study group increased more at the femoral neck (the intergroup change was +0.9% at 18 months and +2.8% at 26 months, p = 0.004 for the change between 18 and 26 months) and remained at the 18-month level at the distal femur, patella, proximal tibia, and calcaneus. In these sites, the statistically significant changes during the entire 26 months of training were 1.7-4.0% in the training group as compared with the changes of −0.9-1.5% in the control group. In the lumbar spine, BMD decreased from the 18-month level in both groups. In conclusion, the significant BMD increases that were obtained by supervised 18-month high-impact training were effectively maintained with subsequent unsupervised regular aerobic and step classes (twice per week). The finding emphasizes the effectiveness and feasibility of self-controlled aerobic and step exercises in the primary prevention of osteoporosis among healthy premenopausal women. (J Bone Miner Res 1999;14:125-128)