Introduction 1 5 6 7 8 10 11 12 9 10 12 13 11 14 15 16 17 18 Methods Study design and patients 15 18 19 20 2 13 Patients treated with estrogen and estrogen-related drugs within 3 months of study entry or for more than 1 month within 6 months of study entry were excluded from the original studies. All patients received 1,000 mg calcium and, if needed, 500 IU vitamin D daily. Adverse event information was collected at the study visits made every 3 months during the trials. The investigator recorded adverse events reported by the patients, as well as adverse events observed on examination of the patient. Identification of fragility fractures Fragility fractures were defined as a composite of a patient’s incident morphometric vertebral and osteoporosis-related nonvertebral fractures (i.e., six fracture types including clavicle, humerus, wrist, pelvis, hip or leg fractures). This endpoint was chosen to include all radiographically confirmed fractures. With respect to the nonvertebral fractures, the selection of the nonvertebral fracture sites (clavicle, humerus, wrist, pelvis, hip or leg) followed the pre-specified definition of osteoporosis-related nonvertebral fractures of the risedronate trial program, which has been reported in all regulatory and previous scientific reports from this program. All osteoporosis related nonvertebral fractures required radiographic confirmation. The term “leg” fracture as used in the risedronate trials is a summary term, including the treatment emergent adverse event codes for fibula, tibia, subtrochanteric fracture, and leg, but not those for ankle or distal tibia fracture, the terms by which ankle fractures are usually reported. Statistical analyses The primary analysis focused on investigating the effect of risedronate to reduce fragility fractures (both vertebral and nonvertebral) in postmenopausal women who had no radiographic vertebral fractures at baseline and were considered osteopenic based on a femoral neck T-score ranging between −1 SD and −2.5 SD. For consistency with previous risedronate studies, the statistical analysis used time-to-first-event methodology. Cumulative fracture incidence was estimated using the Kaplan–Meier method, and a Cox proportional hazards regression model was used to estimate the relative fracture risk reduction, via the hazard ratio (HR), between the risedronate 5 mg and placebo groups. The regression model was stratified for each trial to allow for separate underlying hazard functions. In order to determine the acceptability of combining data from all trials, potential heterogeneity assessments were performed using the Cox regression model to assess the trial-by-treatment, age-by-treatment and baseline femoral neck BMD-by-treatment interactions. Fracture efficacy for nonvertebral osteoporosis-related fractures and new vertebral fractures were also investigated. A sensitivity analysis was conducted to investigate whether the exclusion of patients with lumbar spine T-score <−2.5 SD would influence the outcome of the primary analyses. As the inclusion criteria for the primary analysis were low bone mass at the femoral neck (BMD T-score between −1 and −2.5 SD), we felt it was important to test whether the exclusion of patients with an osteoporotic BMD at another site would significantly affect the results. Adverse events, serious adverse events, withdrawals, deaths, and upper gastrointestinal adverse events were summarized by treatment group. Differences in the proportions of patients between the treatment groups were tested using the Fisher’s exact test. Results Patients 1 Table 1 Baseline characteristics of patients with a femoral neck T-Score ≥ −2.5 SD and ≤ −1 SD without prevalent vertebral fractures   All patients Excluding lumbar spine T-score <−2.5 SD Placebon n = 309 Risedronate 5 mg n = 311 Placebon n = 146 Risedronate 5 mg n = 147 a 64 (7.5) 64 (7.5) 63 (8.3) 65 (7.0) a, b −1.84 (0.396) −1.85 (0.406) −1.77 (0.418) −1.76 (0.398) a 159.8 (6.92) 159.8 (6.05) 160.6 (6.14) 160.4 (6.32) a 65.9 (11.92) 66.8 (11.19) 69.8 (12.91) 69.6 (12.27) 2 a 25.9 (4.54) 26.2 (4.36) 27.2 (5.1) 27.1 (4.78) Patients with prevalent OP-related nonvertebral fractures 56 (18%) 59 (19%) 34 (23%) 31 (21%) a b Potential heterogeneity assessments revealed that there were no significant interactions for trial-by-treatment ( = 0.495), age-by-treatment ( = 0.981) or baseline femoral neck BMD-by-treatment ( = 0.188) Effect on fractures 1 Fig. 1 Reduction of fragility fracture risk in patients with femoral neck T-score between −1 and −2.5 SD and no prevalent vertebral fractures  When separated according to nonvertebral and vertebral fractures, the cumulative nonvertebral fracture incidence was 5.4% and 0.4%, respectively, for placebo and risedronate (HR = 0.09, 95%CI = 0.01–0.71 ( = 0.022), and the cumulative vertebral fracture incidence was 4.2% and 1.8%, respectively, for placebo and risedronate (HR = 0.44, 95%CI = 0.11–1.78 ( = 0.249). Sensitivity analyses When patients with baseline LS spine T-scores below −2.5 SD were excluded, results consistent with those from the primary analysis population were observed, although the reduced sample size widened the confidence intervals and led to a loss of statistical significance. For the 293 patients who had both femoral neck and LS spine T-scores between −2.5 and −1 SD, the hazard ratio for fragility fractures was 0.22 (95%CI = 0.03–2.02,  = 0.182), a magnitude of effect similar to that in the primary analysis population. Safety 2 15 18 Table 2 Summary of adverse events (AEs) for patients with a femoral neck T-score ≥ −2.5 SD and ≤ −1 SD and noprevalent vertebral fractures (BMD and VERT Trials) Category Number of patients (%) -value Placebo ( = 309) Risedronate 5 mg ( = 311) AEs 290 (93.9%) 293 (94.2%) 0.87 Serious AEs 40 (12.9%) 56 (18.0%) 0.10 AEs associated with death  2 (0.6%) 0 (0.0%) 0.25 UGI AEs  77 (24.9%) 81 (26.0%) 0.78 Withdrawn due to AEs  38 (12.3%) 26 (8.4%) 0.11 -values were from Fisher’s exact test. UGI: upper gastrointestinal Discussion 21 17 18 22 23 24 20 25 19 20 8 13 19 26 27 28 29 30 31 31 Limitations of the current study include the selection of initial populations in the VERT and BMD trials based on the presence of risk factors for osteoporosis (including low BMD and prevalent vertebral fractures). Because the original trials were designed to examine the efficacy of osteoporosis treatments in an osteoporotic population, rather than to determine representative incidence of fracture in a general osteopenic population, the data provided in the present analysis may reflect an overestimation of the fracture risk in the general postmenopausal population. 32