Treatment of Postmenopausal Osteoporosis with Slow-Release Sodium Fluoride: Final Report of a Randomized Controlled Trial

It seems logical to use fluoride in osteoporosis, because fluoride can stimulate osteoblastic proliferation and new bone formation [1, 2]. However, clinical trials with fluoride have yielded mixed results because excessive exposure to fluoride may cause abnormal bone formation, microfractures, and gastric bleeding [3, 4]. Thus, treatment with a high dosage of plain sodium fluoride did not decrease the spinal fracture rate despite markedly increasing vertebral bone density, and it increased the rate of appendicular fractures and microfractures [4]. To overcome the complications associated with sodium fluoride, we have advocated the cyclical, intermittent use of a lower dose of less bioavailable, slow-release sodium fluoride and continuous supplementation with calcium citrate [5, 6]. This treatment has been shown to maintain serum fluoride concentrations within the narrow therapeutic window [7, 8], thus avoiding toxic peaks in serum [9], and to stimulate the formation of normally mineralized bone [5, 10] with an improved intrinsic quality of cancellous bone [11-13]. We previously reported the results of an interim analysis [6] of a placebo-controlled randomized trial (median duration of treatment for fracture analysis, 2 years). Here, we present the final report of that trial (median duration of treatment for fracture analysis, 3 years). Methods Clinical Data Demographic and baseline presentations were described in the interim report [6]. We recruited 110 women with postmenopausal osteoporosis into the trial. All had radiologic evidence of osteopenia and osteoporosis; one or more vertebral fractures believed to be nontraumatic; and no secondary cause of bone loss. They were randomly assigned to one of two groups and stratified according to estrogen treatment. All study personnel were unaware of group assignment while data were being gathered. Ninety-nine patients completed at least 1 study cycle (1 year of actual treatment). The demographic or baseline presentations of these 99 patients did not differ according to treatment group [6] (Table 1). The two groups were similar in age, time since menopause, dietary calcium intake, height, weight, and number of spinal fractures. Both groups had moderate to severe osteoporosis: The average L2-L4 bone density was approximately 30% less than of a normal 30-year-old woman, and each group had a median of two spinal fractures at baseline. Table 1. Baseline Characteristics* Treatment Patients in the fluoride group received slow-release sodium fluoride (Slow Fluoride, Mission Pharmacal Co., San Antonio, Texas), 25 mg twice daily, orally before breakfast and at bedtime in repeated 14-month cycles (12 months receiving treatment followed by 2 months not receiving treatment). They also received calcium citrate (Citracal, Mission Pharmacal), 400 mg calcium twice daily, before breakfast and at bedtime continuously throughout the study. Those in the placebo group received placebo (identical in appearance to Slow Fluoride but containing excipient only [provided by Mission Pharmacal]) on the same time schedule. The Mission Pharmacal Company had no role in the design of the study or in data retrieval, analysis, or interpretation. Thirteen of 48 patients in the fluoride group and 16 of 51 patients in the placebo group received concurrent treatment with estrogen. Nine of the 29 patients treated with estrogen were recruited at the primary site at Dallas; the other 20 were enrolled and evaluated at the Scott and White Clinic, Temple, Texas. Fracture Quantitation Before treatment and at 12 months of each cycle, a lateral spine roentgenogram was obtained for the assessment of spinal fractures. In the interim analysis [6], prevalent fractures (fractures present at baseline) were identified with the aid of radiology reports. For this final report, prevalent fractures were also analyzed using a computer program that calculated the vertebral dimensions of clearly unaffected vertebrae from landmarks (anterior and posterior corners and midpoints). By comparing these dimensions with published normal values [14], we obtained a correction factor. Using this correction factor, we estimated idealized vertebral dimensions before a fracture had occurred for the remaining vertebrae in the given baseline radiograph. A reduction in any height of more than 20% (from idealized to actual) accompanied by a decrease of at least 10% in vertebral area represented a prevalent fracture. Incident spinal fractures (fractures occurring during the trial) were identified as described previously [6], using a computer-derived method. A reduction in any vertebral height of more than 20% accompanied by a decrease in vertebral area of more than 10% from one year to the next constituted a fracture [15]. A new incident fracture was a fracture that occurred during treatment in a previously unaffected vertebrae. A recurrent fracture was one that developed on a previously fractured vertebra. Bone Mass Measurements The use of different densitometers prompted us to calculate and use percentage changes per year rather than absolute values. The method for calculating changes in L2-L4 bone mineral content and bone density of the femoral neck and the radial shaft was described previously [6]. Safety Variables Serum fluoride concentrations were measured before the morning dose of the test drug at 0, 3, 6, 9, and 12 months of each cycle, and they were analyzed using an ion-specific electrode. At the same visits, a history was taken for gastrointestinal and musculoskeletal side effects. A microfracture was defined clinically as moderate to severe lower-extremity pain that persisted for more than 6 weeks despite a reduction in treatment dose and objectively as changes on bone scan or radiograph. The relation of each side effect to treatment was assessed. A symptom was considered to be related to treatment if it was moderate to severe in intensity, had no other cause, had newly appeared and persisted during the treatment phase, or had disappeared during the withdrawal period or with dose reduction. It was considered to be unrelated if it was present at baseline or during the late withdrawal phase, or if it had newly appeared but was not persistent. The severity and frequency of side effects were also quantitated as adverse symptom scores. We identified 10 gastrointestinal items (symptoms such as nausea, vomiting, and diarrhea), 4 rheumatic items (pain in the foot, knee, hip, and other joints), and 3 skeletal items (pain in the lower, mid-, or upper back). Each item was given a numerical value of 1 to 3 for frequency (infrequent, frequent, or very frequent) and a numerical value of 1 to 3 for severity (mild, moderate, or severe). Side-effect score was the product of the value for frequency and the value for severity for each item. Thus, a constant, severe back pain yielded a score of 9 (3 3). A gastrointestinal score was derived for each patient by adding the scores of the 10 gastrointestinal items for all relevant visits and dividing the sum by the number of visits. A similar computation was done to derive rheumatic and skeletal scores for each patient. Statistical Analysis The data for incident spinal fractures were compared between the two groups-using three methods. Individual Vertebral Fracture Rate For each patient, the individual vertebral fracture rate was obtained by dividing the total number of new fractures by the duration of treatment. Because the data were skewed, this rate was compared between the two groups using the Wilcoxon rank-sum test. Fracture-Free Rate This rate was the percentage of patients without new fractures, unadjusted for covariates. The two groups were compared using the log-rank test to account for differential follow-up. Survival The Cox proportional-hazards regression model [16] was constructed to estimate the relative risk for a new spinal fracture while adjusting for covariates (treatment group, age, prevalent spinal fractures, years since menopause, height, weight, estrogen treatment, and stratum of baseline L2-L4 bone density). Time (in years) to the first fracture was considered to be the survival time. Analyses of fracture rates and logistic regression were also done [6]; the data are not presented because findings were similar to those obtained using the above methods. The arithmetic difference in height from baseline to the end of treatment for each patient was compared between groups using a two-sample t-test and a two-way analysis of variance with the following factors: 1) treatment [fluoride vs placebo] and 2) fracture status (fracture-free vs one or more new or recurrent fractures). For each patient, we calculated the percentage change per year for L2-L4 bone mineral content and bone density of femoral neck and radial shaft. The individual mean change for each patient was calculated as the average of yearly changes. The group mean was obtained by averaging the individual means. One-sample t-tests were then used to compare the percentage change to zero for each year or for the mean. Comparisons between groups were made using two-sample t-tests. Missing data precluded implementing a repeated-measures analysis of variance. For related adverse events, the frequency of each event was compared between the two groups by using the Fisher exact test. Adverse symptom scores were compared between the groups by using the Wilcoxon rank-sum test and within the groups by using the Wilcoxon signed-rank test. For nonvertebral fractures, the exact tests based on the binomial distribution using person-year data were used to compare the two groups. Most analyses were done using BMDP Statistical Software (BMDP, Los Angeles, California). Programs for analyzing person-time data were developed by the authors. Data are presented as mean SD unless otherwise indicated. All reported P values are two-sided. Results Duration of Treatment The total duration of follow-up, including withdrawal periods, was 193 patient-years in th