We studied the relationship between change in bone turnover and vertebral fracture risk during raloxifene therapy using 3-year data from the MORE trial, where 2622 of the 7705 randomized women had measurement of bone markers at baseline and after 6 and 12 months participation. Change in bone turnover was significantly related to future risk of vertebral fracture, also after adjusting for baseline vertebral fracture status and BMD. Thus, for a decrease of 9.3 pg/l in serum osteocalcin after 1 year's raloxifene therapy, the odds ratio (OR) for a new vertebral fracture during 3 years was 0.69 (0.54-0.88), p = 0.003. Similarly, for a decrease of 5.91 microg/l in serum bone alkaline phosphatase, OR was 0.75 (0.62-0.92), p = 0.005. The change in BMD over 12 and 24 months was not related to fracture risk in any of the analyses. The strongest predictor for vertebral fracture was prevalent vertebral fracture--even during therapy. The predictive value of baseline BMD was in the same order of magnitude as bone turnover change during raloxifene treatment. In conclusion, the change in bone turnover is related to fracture risk during raloxifene therapy. In contrast the change in BMD is not related to fracture risk. The strongest predictor for vertebral fracture is prevalent vertebral fracture.