We previously reported [5, 6] a nonrandomized trial in which intermittent slow-release sodium fluoride with continuous calcium citrate supplementation stimulated the formation of normally mineralized bone and decreased the spinal fracture rate without serious complications. In 1986, we initiated a randomized trial using slow-release sodium fluoride plus calcium citrate compared with placebo plus calcium citrate in 99 patients with postmenopausal osteoporosis. The trial is ongoing, with an average duration of treatment in the two study groups of 2.44 and 2.14 cycles (14 mo/cycle) and with 15 patients completing the intended 4 cycles of treatment. Although the study is not expected to be completed until August 1996, this interim analysis was done in response to a request for an update at the Fourth International Symposium on Osteoporosis [7]. The trial will be completed in order to determine if the treatment effect is sustained. We believe that the interim analysis will not affect the conduct of the remainder of the trial.

Methods

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Clinical Data

Recruited into the trial were 110 fully ambulatory white women with postmenopausal osteoporosis, all of whom were referred for symptomatic osteoporosis by practicing physicians because of an inadequate response to conventional therapy or the unwillingness of physicians to care for them. No other ethnic groups were enrolled, probably because of the rarity of postmenopausal osteoporosis and the nature of the referred patients in the study areas. The entry criteria were as follows: postmenopausal state, radiographic evidence of osteoporosis, and one or more vertebral fractures believed to be nontraumatic. Exclusion criteria were as follows: the presence of conditions causing bone loss such as hyperparathyroidism, adrenocorticosteroid excess, thyrotoxicosis, chronic diarrheal state or malabsorption, renal tubular acidosis, renal impairment (endogenous creatinine clearance less than 0.7 mL/min per kg); previous treatment with diphosphonate, calcitonin, or fluoride; active peptic ulcer disease; and skeletal fractures that could not be quantified for anatomic or technical reasons. Those taking pharmacologic doses of vitamin D preparations were accepted if they had discontinued the drug for at least 6 months. At recruitment, 31 patients were taking estrogen (treatment initiated after osteoporosis was diagnosed). Thirteen patients had documented recurrent spinal fractures while receiving estrogen. The total duration of estrogen therapy represented about a third of their postmenopausal state (mean, 8 years). This treatment, usually consisting of conjugated estrogen, 0.625 mg/d given continuously or intermittently (25 d/mo), and intermittent progesterone (5 mg/d for 10 d/mo), was continued during the trial. Patients receiving estrogen had similar baseline demographic characteristics as patients not receiving estrogen and were considered at increased risk for further fractures.

Randomization and Treatment Scheme

Participants were randomly assigned to the two treatment groups, stratified according to estrogen treatment (untreated or estrogen-treated). Patients in the treatment group received slow-release sodium fluoride (Slow Fluoride; Mission Pharmacal Company, San Antonio, Texas), 25 mg twice daily given orally before breakfast and at bedtime intermittently in repeated cycles of 14 months (12 months receiving treatment, followed by 2 months off treatment) and calcium citrate (Citracal, Mission Pharmacal Company) as 400 mg of calcium twice daily before breakfast and at bedtime. In the placebo group, medication identical in appearance to Slow Fluoride that was devoid of sodium fluoride was given on the same time schedule along with calcium citrate at the same dose and schedule.

Study Protocol

Patients were evaluated in an outpatient setting before treatment and at 3, 6, 9, 12, and 14 months of each cycle. At each visit, a careful history was taken for gastrointestinal and musculoskeletal side effects defined as symptoms that newly appeared or increased from baseline without apparent cause and persisted more than a month during treatment or disappeared during withdrawal. Where severe lower-extremity pain lasted more than 2 weeks, the protocol required a bone scan followed by radiographs for detection of microfracture; however, none was required. In addition, systematic multichannel analysis of venous blood, complete peripheral blood count, and 24-hour urinary calcium were determined at each visit; serum parathyroid hormone, reticulocyte count, and 24-hour urinary hydroxyproline were measured before and at 6 and 12 months of each cycle. Serum fluoride was measured before the morning dose of slow-release sodium fluoride at 0, 6, and 12 months of each cycle. Before treatment and at 12 months of each cycle, a lateral spinal radiograph was obtained to detect spinal fractures; moreover, bone mineral content of the L2 to L4 vertebrae, bone density of the distal third of the radius of the nondominant forearm, and bone density of the femoral neck were measured.

Quantitation of Spinal Fractures

Lateral spinal films taken before treatment and at 12 months of the first cycle were compared in order to determine "new" and "recurrent" fractures occurring during the first cycle of treatment. For each vertebra from T3 to L5, landmarks (anterior and posterior corners and midpoints) were recorded using an electrostatic digitizing board (Scriptel Corporation, Columbus, Ohio) with a coefficient of variation of 1.5%. A computer software program developed by one of the authors was used to compute changes in vertebral heights and area, and to calculate the magnification error between the two sets of radiographs. After correction for the magnification error, if any, a decrease in height of more than 20% of anterior, middle, or posterior height, accompanied by a decrease in area of more than 10% in a previously unaffected vertebra, was considered a "new" fracture [8], or if the decrease was in a previously fractured vertebra, it was considered to be a "recurrent" fracture. Identical criteria were used to identify new and recurrent fractures occurring during the second cycle (by comparing 26 months with 12 months), during the third cycle (by comparing 40 months with 26 months), and during the fourth cycle (by comparing 54 months with 40 months).

Moreover, spinal films taken after the last cycle of follow-up were compared not only with the immediately preceding films but also with earlier radiographs. The same procedure was followed for the identification of new and recurrent fractures. Thus, it was possible to detect fractures occurring during two or more cycles ("cumulative" fractures) that escaped disclosure by previous cycle-to-cycle analysis.

Bone Mass Measurement

During the course of this study, two instruments were used to measure the L2 to L4 bone mineral content and the femoral neck bone density. A dual-photon x-ray absorptiometer (1.5 version, Lunar Radiation, Madison, Wisconsin) was used initially. It was replaced by quantitative digital radiography (Hologic, Waltham, Massachusetts) that yielded different absolute values for bone mass. The following procedures were adopted to accommodate problems imposed by the measurement of bone mass by two different densitometers. First, in estimating the extent of spinal bone loss at baseline, a given patient's L2 to L4 bone density obtained by either method was compared with the mean value for a normal 30-year-old woman established for the corresponding instrument. Second, in quantifying changes in the L2 to L4 bone mineral content and the femoral neck bone density produced by treatment, results were expressed as a percentage change for each cycle rather than as absolute values. When the same densitometer was used at the beginning and the end of a given cycle, the experimentally derived values were used to calculate the percentage change in bone mineral content or bone density. For a given cycle with the initial bone mass obtained by the Lunar method and the final bone mass measured by the Hologic technique, a correction factor was applied to convert the Lunar-derived value to the latter value. In patients who initially had bone mass determined by the Lunar method, a concurrent analysis by the Hologic instrument was done before converting to the latter technique for subsequent follow-up measurements. Thus, a correction factor could be calculated.

The radial shaft bone density was obtained throughout the study by a single-photon absorptiometer (Norland, Ft. Atkinson, Wisconsin) [9]. For follow-up measurements, actual experimentally derived bone densities were used to calculate the percentage change in bone density for each cycle. The coefficient of variation for the L2 to L4 bone mineral content using the Lunar or Hologic method was 1%, whereas that for the femoral neck bone density and the radial shaft bone density was 1% to 2%.

Biochemical Analysis

The blood screen was done as SMA-20 (Smith-Kline Laboratory, Dallas, Texas). The serum parathyroid hormone level was analyzed by the whole-molecule, immunoradiometric assay (using a kit from Nichols Institute, San Juan Capistrano, California). The serum fluoride level was measured using an ion-specific electrode. The urinary calcium level was determined by atomic absorption spectrophotometry, and the hydroxyproline level was analyzed using the Hypronosticon kit (Organon Teknika Corporation, Durham, North Carolina).

Statistical Analysis

Top Methods Statistical Analysis Results Discussion Author & Article Info References

The interim analysis was done because of the previously mentioned request for an update, the attainment of the initially estimated patient sample size, and the recommendation by the review body of the research grant to evaluate the power and determine the need for an extension if necessary. Data retrieval and analysis were done by independent examiners without identification of patients by name or knowledge of treatment allocation [10]. Even after completion of analysis, the exact randomization code was not disclosed to patients, to physicians, or to paramedical personnel involved in their care because most patients have not completed the trial. No further interim analysis is planned while the trial is continued.

Demographic variables were compared between the placebo and slow-release sodium fluoride groups using the Wilcoxon rank-sum test. Serum fluoride levels and other biochemical variables were compared with pretreatment levels using the paired t-test and were compared between groups with two-sample t-tests. The fractional change in the bone mineral content or the bone density for each group was compared with zero using one-sample t-tests and was compared between the two groups with two-sample t-tests. The Fisher exact test was used to compare side-effect frequencies [11].

The group vertebral fracture rate for new, recurrent, or all fractures was computed as the ratio of the total number of fractures to the total number of cycles for each group; differences between groups were assessed using exact binomial probabilities [12]. This analysis assumed each vertebra to be independent regardless of patient identity. Vertebral fractures were also quantitated by two other methods. The individual new vertebral fracture rate, defined as the number of new fractures per cycle of study, was calculated for each patient; median rates between the two groups were compared with the Wilcoxon rank-sum test [11]. The fracture-free rate, defined as the percentage of patients remaining free of new fractures, was compared between the two groups with the Fisher exact test [11].

A logistic regression model was constructed with occurrence of at least one new fracture during treatment dichotomized for use as the dependent variable [13]. This analysis was done to assess whether treatment group (as an independent variable) was related to the occurrence of new fractures after adjusting for other prognostic variables. The following covariates were evaluated in the logistic regression: treatment group, age, weight, height, body mass index, duration after menopause, estrogen treatment, number of initial fractures, the L2 to L4 bone density, and duration of treatment. Adjusted odds ratios were derived from the logistic regression coefficients.

After considering various approaches [10], the 0.01 level of significance was selected to adjust for multiplicity of testing. Exact P values are reported where possible. Before implementing parametric tests, assumptions of normality and homogeneity of variance were tested with the Wilk-Shapiro test and the Levene test, respectively. Results are expressed as mean ±SD and median (10th to 90th percentiles) unless otherwise indicated. Statistical analysis was done using BMDP, Epistat, and Prophet statistical software.

Results

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Baseline Presentation

Among 110 recruited patients, 11 patients (5 from the placebo group and 6 from the slow-release sodium fluoride group) withdrew from the study before completing 1 year of treatment because of lack of interest or transportation problems in 10 and stroke in 1 (from the placebo group). The remaining 99 patients completed at least 1 year of treatment, allowing measurement of critical data (for example, bone density and spinal radiographs). There were 51 patients in the placebo group and 48 in the slow-release sodium fluoride group (Table 1). Sixteen patients in the placebo group and 13 in the slow-release sodium fluoride group were taking estrogen. The remaining 70 patients (35 in each group) did not take estrogen. The median duration of treatment was 2 cycles/patient (mean, 2.14 ±0.98 cycles/patient) in the placebo group and 2 cycles/patient (mean, 2.44 ±0.99 cycles/patient) in the slow-release sodium fluoride group (P = 0.13).

Among 99 patients who completed at least 1 year of study, the two groups were similar in their demographic characteristics and baseline biochemical presentation (Table 1). Patients in both groups had moderate-to-severe spinal osteoporosis because the mean L2 to L4 bone density was approximately 30% below that of a normal 30-year-old woman and the median value for spinal fractures at baseline was 2 (with the mean value nearly 3). The absolute values for radial shaft bone density were similar in the two groups and were approximately 70% of that of a normal 30-year-old woman [9]. In results presented below, data from 99 patients completing at least 1 cycle of treatment were used.

Changes in Serum Fluoride Levels

In the placebo group, the mean fasting serum fluoride concentration was normal at 2.11 ±0.39 µmolars/L (< 5 µmolars/L, which is probably the therapeutic threshold [14]) at the start of the study (Figure 1). It remained below that level throughout treatment. In the slow-release sodium fluoride group, the mean serum fluoride concentration at 6 and 12 months of each cycle (when slow-release sodium fluoride was given) was substantially greater than the corresponding value of the placebo group (P < 0.001), ranging 5 to 10 µmolars/L, and was higher than the value at baseline (P 0.006). At 6 and 12 months of all cycles, 58% of individual fluoride values were within 5 to 10 µmolars/L in the slow-release sodium fluoride group compared with 1% in the placebo group. However, before treatment (0 month) and at 14 months of each cycle (after 2 months of withdrawal from slow-release sodium fluoride), the mean serum fluoride concentration in the slow-release sodium fluoride group was below 5 µmolars/L and was not statistically different from that of the placebo group.

View larger version (23K): [in this window] [in a new window]   Figure 1. Effect of treatment on the mean fasting serum fluoride level. Dashed horizontal lines indicate the "therapeutic window" [14]. Vertical dashed lines indicate the period of fluoride withdrawal. In each cycle (of 14 months), slow-release sodium fluoride or placebo was given during the first 12 months. *P = 0.006; P < 0.001. Values are presented as mean ±SE.

Change in Bone Mass

In the placebo group, the L2 to L4 bone mineral content did not change statistically during the first cycle, second cycle, or the combined third and fourth cycles (Figure 2). Data for cycles 3 and 4 were combined because observations were limited for adequate statistical analyses. However, it increased substantially by 4% to 6% per cycle in the slow-release sodium fluoride group (P < 0.001). The changes in the slow-release sodium fluoride group were substantially greater than in the placebo group at each of the corresponding time periods (P < 0.001).

View larger version (25K): [in this window] [in a new window]   Figure 2. Effect of treatment on the L2 to L4 bone mineral content and on the bone density of the femoral neck and the radial shaft. For each cycle, the percentage change of the value from the immediately preceding cycle or baseline was calculated. Thus, a cycle-to-cycle change, rather than a cumulative change, is depicted. The symbol above the bars indicates a significant change from zero, whereas the symbol above the bracket shows a significant change between the placebo and slow-release sodium fluoride (SR-NaF) groups. * P = 0.019; P < 0.001. BD = bone density; BMC = bone mineral content.

In the placebo group, the mean values for femoral neck bone density increased during all cycles; however, changes compared with baseline were not statistically different (P = 0.17 to 0.40). In the slow-release sodium fluoride group, the femoral neck bone density increased by 4.1% and 2.1% during the first and second cycles, respectively (P = 0.001 and 0.02) (Figure 2). At each cycle, no statistical difference was noted in femoral neck bone density between the two groups. The radial shaft bone density did not change statistically in any cycle in either group, nor was it statistically different between the two groups at corresponding points in the cycles.

Vertebral Fractures

In the placebo group, 26 new spinal fractures occurred during an average of 2.14 cycles/patient, whereas only 10 new fractures were found during 2.44 cycles/patient in the slow-release sodium fluoride group (Figure 3). For all cycles combined, the group vertebral fracture rate for new fractures in the slow-release sodium fluoride group was substantially less than that of the placebo group (0.085/patient cycle compared with 0.239/patient cycle, P = 0.006) (Table 2). For all vertebral fractures (new and recurrent), the new spinal fracture rate of the slow-release sodium fluoride group was also substantially less than that of the placebo group (0.137/patient cycle compared with 0.284/patient cycle, P = 0.02). However, the group vertebral fracture rates for recurrent fractures were not statistically different between the two groups (0.051/patient cycle compared with 0.046/patient cycle, P > 0.2).

View larger version (41K): [in this window] [in a new window]   Figure 3. Appearance of new vertebral fractures during placebo administration or slow-release sodium fluoride treatment. Each line represents a separate patient. The exact location of the circle on the line does not necessarily indicate the actual time of fracture occurrence; rather, it reflects the time of skeletal radiologic examination. \#9679; =a separate new vertebral fracture; = a "cumulative" fracture occurring during two or more cycles, escaping detection on a cycle-to-cycle analysis; numbers designated by E before the line indicate estrogen-treated patients.

The individual vertebral fracture rate was substantially greater in the placebo group than in the slow-release sodium fluoride group (0.204/patient cycle compared with 0.057/patient cycle, P = 0.02) (Table 2). The fracture-free rate was higher in the slow-release sodium fluoride group than in the placebo group (83.3% compared with 64.7%, P = 0.04).

For the logistic regression model, the patients were divided into two groups on the basis of occurrence of a new fracture. After adjusting for possible confounding variables (for example, number of initial fractures, duration of treatment, weight, and height) using logistic regression, the positive influence of treatment on the occurrence of new fractures persisted. The adjusted odds ratio for the group (slow-release sodium fluoride compared with placebo for presence or absence of new vertebral fractures) was 0.220 (95% CI, 0.058 to 0.827) after adjustment for all covariates examined compared with 0.221 (0.072 to 0.676) after adjustment for group and duration alone.

Estrogen treatment was not an independent predictor of new vertebral fractures (P > 0.2 from multivariate logistic regression analysis). In the slow-release sodium fluoride group, 35 non-estrogen-treated patients had a comparable baseline fracture number (L2 to L4 bone density and vertebral fractures) as the 13 estrogen-treated patients. Moreover, non-estrogen-treated patients did not differ statistically from estrogen-treated patients with respect to serum fluoride levels (at 6 and 12 months of treatment cycles, average of individual means of 5.60 µmolars/L in the nonestrogen group compared with 5.21 µmolars/L in the estrogen group, P > 0.2), increment in the L2 to L4 bone mineral content (average of individual means of 4.8% per cycle in the nonestrogen group compared with 6.3% per cycle in the estrogen group, P > 0.2), or individual vertebral fracture rate during treatment (0.052/patient cycle in the nonestrogen group compared with 0.071/patient cycle in the estrogen group, P > 0.2) (Table 2).

Table 2 lists fracture data for the whole group of 99 patients, for the 70 non-estrogen-treated patients, and for the 29 estrogen-treated patients. In the non-estrogen-treated patients and in the whole group, the group receiving slow-release sodium fluoride showed a lower individual vertebral fracture rate and group vertebral fracture rate (for new spinal fractures) and a higher fracture-free rate than the placebo group. No statistical differences related to estrogen treatment were noted between the slow-release sodium fluoride and placebo groups in estrogen-treated patients, possibly because of the small sample size.

Withdrawal and Safety

Among 99 patients completing at least 1 cycle, 8 patients in the placebo group and 7 in the slow-release sodium fluoride group withdrew from the study before completing 4 cycles of the trial. Reasons for withdrawal were transportation problems in 5 (2 patients receiving placebo and 3 receiving slow-release sodium fluoride), lack of interest in 5 (3 receiving placebo and 2 receiving slow-release sodium fluoride), death because of unrelated causes in 3 (2 receiving placebo and 1 receiving slow-release sodium fluoride), and side effects in 2 (1 from each group).

Side effects were assessed in all 110 recruited patients. Even though the duration of follow-up was shorter in the 11 patients who withdrew from the study before completing 1 cycle than in the remaining 99 patients, the mean treatment duration among early withdrawals was 6 mo/patient for both the slow-release sodium fluoride and placebo groups. Minor gastrointestinal side effects (dyspepsia, anorexia, nausea, or vomiting) were encountered in 2 patients (3.6%) in the placebo group and 3 patients (5.6%) in the slow-release sodium fluoride group. No one in either group had gastrointestinal bleeding, microfractures, or hip fractures. One patient (1.9%) taking slow-release sodium fluoride had joint pain and swelling of undetermined cause. Two patients (3.6%) in the placebo group had nonaxial fractures (proximal humerus and metatarsal bone) and 2 (3.7%) taking slow-release sodium fluoride developed fractures of the fibula, the metatarsal bone, and the pelvis, all from falling down. One patient (1.8%) receiving placebo withdrew from the study because of side effects (nausea and vomiting) and 1 (1.9%) taking slow-release sodium fluoride withdrew because of pelvic fracture. The overall side-effect frequencies between the two groups were not statistically different (P > 0.2).

Other Tests

Levels of serum calcium, phosphorus, alkaline phosphatase, parathyroid hormone, creatinine clearance, peripheral blood hemoglobin, as well as the hematocrit and reticulocyte count remained within normal limits during treatment and did not differ between the two groups. Stool culture results for occult blood were positive in six patients (three placebo and three slow-release sodium fluoride), four of whom had a history of hemorrhoids. The mean values for urinary calcium were slightly higher (by 0.5 to 1.5 mmol/d) and those for urinary hydroxyproline were lower (by 30 to 60 µmol/d) during treatment compared with baseline in both groups. However, the changes were significant only during some periods, and values were not generally statistically different between the two groups.

Discussion

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Interim results from this randomized trial provide evidence that intermittent treatment with slow-release sodium fluoride with continuous calcium citrate supplementation compared with calcium citrate alone is effective in preventing spinal fractures. The inhibition of new spinal fractures by slow-release sodium fluoride was shown by a decreased individual spinal fracture rate (mean of individual rates) and group spinal fracture rate (total number of fractures per group) as well as by a higher fracture-free rate. Although slow-release sodium fluoride decreased the new spinal fracture rate, it did not affect the recurrent spinal fracture rate. The results suggest that once trabecular continuity is disrupted by fractures, slow-release sodium fluoride treatment does not prevent further fractures.

As in nonrandomized trials [5, 6], the slow-release sodium fluoride therapy was safe, and the side-effect profile was comparable with placebo. No one developed gastrointestinal bleeding, microfractures, or hematologic abnormalities. This may have been because serum fluoride levels were kept well below 10 µmolars/L, the proposed toxic threshold [5, 14].

Moreover, no one taking slow-release sodium fluoride developed hip fractures, and the appendicular fracture rate of the slow-release sodium fluoride group was similar to that of the placebo group. The density of the radial shaft, composed largely of cortical bone, remained stable. Thus, treatment with slow-release sodium fluoride plus calcium citrate may not be deleterious to cortical bone as suggested by other clinical trials [3]. This conclusion is supported by a previous histomorphometric and ultrasound analysis of biopsied cortical bone [15, 16].

Our results with slow-release sodium fluoride clearly differ from those seen in a recent study of plain sodium fluoride [3]. In the latter study, skeletal fluoride uptake was probably considerably higher because of continuous administration of a more bioavailable fluoride preparation at a higher dosage [17]. Thus, the increase in lumbar bone mass was much more marked (8.5% per year rather than 4% to 6% per year found in this study) [3]. However, the radial shaft bone density decreased, suggesting that the rapid growth of trabecular bone was achieved at the expense of cortical bone, unlike in this study. Moreover, the amount of calcium absorbed from the intestinal tract may have been insufficient to adequately mineralize newly formed bone matrix, possibly because of the rapidity of matrix synthesis and provision of calcium as less bioavailable calcium carbonate [18, 19]. Thus, newly formed bone during treatment with plain sodium fluoride may have been poorly mineralized [20] and mechanically defective [3]. This could account for the reported ineffectiveness of sodium fluoride on the spinal fracture rate and for the increased appendicular fracture rate and the frequent complications seen with plain sodium fluoride [3]. Thus, the clinical response to fluoride seems to depend on the formulation, the dose, and the mode of delivery of fluoride [7]. It may also be influenced by the type of calcium supplementation.

Our study had some limitations. First, not all patients completed the intended 4 cycles of the trial. However, the relative risk for spinal fractures remained low during all 4 cycles in the slow-release sodium fluoride group compared with the placebo group. Moreover, a continued increase was noted in the vertebral bone mass in later cycles of slow-release sodium fluoride treatment. If the finding of Gallagher [21], in a nonrandomized trial, that a decrease in the spinal fracture rate is particularly marked during the third and fourth year of treatment is confirmed, we should see an even greater favorable clinical response as more patients complete 4 cycles.

Second, we found no modification in the response to slow-release sodium fluoride and calcium citrate by those patients taking estrogen. However, many patients were referred to us after developing fractures while they were taking estrogen or had only taken it short-term; they may not be representative of estrogen-treated patients overall. Moreover, the efficient bioavailability of calcium citrate administered along with slow-release sodium fluoride could have diminished the antiresorptive action of estrogen.

Interim results from this randomized trial validate findings from a nonrandomized trial [6] and from the laboratory [6, 15, 16], suggesting that slow-release sodium fluoride is clinically useful in the management of established postmenopausal osteoporosis.

Slow Fluoride is an investigational drug studied under the Food and Drug Administration's investigational new drug application number 20 612. The University of Texas Southwestern Medical Center, the sponsor of the investigational new drug application, holds all rights to this drug. The Mission Pharmacal Company, the manufacturer of Slow Fluoride, has provided the drug free of charge for this trial but has not offered other research support. Dr. Pak is the principal investigator of the investigational new drug application. None of the investigators has equity in the Mission Pharmacal Company, receives direct compensation, or serves on the Board of the company.

Presented in part at the Fourth International Symposium on Osteoporosis in Hong Kong on 29 March 1993.