In a 4 year clinical trial in 202 postmenopausal osteoporotic women receiving NaF at 75 mg/day or placebo (both groups received supplementary calcium at 1500 mg/day), we found (N Engl J Med 322:801, 1990) that NaF increased bone mineral density in the lumbar spine (LS-BMD) substantially but did not decrease vertebral fracture rate (VFR), and it increased the nonvertebral fracture rate. Additional analyses and extended observations are now available on 50 women from the NaF group followed for up to 6 years of treatment. In these women, LS-BMD increased linearly over the 6 years (median rate, 8.7%/year or 0.063 g/cm2/year). Because during the 4 year trial the NaF dosage was decreased (because of side effects) in 54 of the 101 women randomized to NaF, dose-response relationships could be evaluated. For the entire study population, serum F level correlated directly with increase in LS-BMD (r = 0.61, P < 0.001). When individual person-years of observation were grouped by deciles of LS-BMD, VFR (per 100 person-years) decreased to a nadir of 24 as mean LS-BMD for the group increased from 0.6 to 1.2 g/cm2 and then doubled to 52 in the group with mean LS-BMD of 1.6 g/cm2. Multivariate analyses and inspection of three-dimensional plots revealed a complex pattern in which VFR was influenced by interaction of several variables. When the effects of LS-BMD, changes in LS-BMD, and serum F were assessed simultaneously, VFR was seen to decrease with increasing LS-BMD except when the higher LS-BMD was associated with rapid rate of increase in LS-BMD or a large increase from baseline serum F. For some patients (noncompliers or nonresponders), serum F or LS-BMD failed to increase. Thus, it is possible that lower dosages of NaF produce moderate decreases in VFR.