Dr Barton is an employee of Procter & Gamble and has received stock and corporate appointments from the company. Dr Grauer is a full-time employee of Procter & Gamble. All other authors have no conflict of interest.
Relationship Between Pretreatment Bone Resorption and Vertebral Fracture Incidence in Postmenopausal Osteoporotic Women Treated With Risedronate
Article first published online: 16 DEC 2003
Copyright © 2004 ASBMR
Journal of Bone and Mineral Research
Volume 19, Issue 2, pages 323–329, February 2004
How to Cite
Seibel, M. J., Naganathan, V., Barton, I. and Grauer, A. (2004), Relationship Between Pretreatment Bone Resorption and Vertebral Fracture Incidence in Postmenopausal Osteoporotic Women Treated With Risedronate. J Bone Miner Res, 19: 323–329. doi: 10.1359/JBMR.0301231
- Issue published online: 2 DEC 2009
- Article first published online: 16 DEC 2003
- Manuscript Accepted: 10 SEP 2003
- Manuscript Revised: 5 JUL 2003
- Manuscript Received: 6 JUN 2003
- bone resorption;
- fracture risk
It is unclear whether the antifracture efficacy of bisphosphonates depends on pretreatment bone turnover. We analyzed the risedronate phase III clinical programs using the urinary excretion of deoxypyridinoline (uDPD) as an index of pretreatment bone resorption rates. Risedronate reduced incident vertebral fractures in women with postmenopausal osteoporosis independent from pretreatment bone resorption.
Introduction: Earlier studies on postmenopausal osteoporosis have suggested that the therapeutic efficacy of antiresorptive therapies might be influenced by pretreatment bone turnover. Because all of these studies have used bone mineral density (BMD) as the primary endpoint, it remains unclear whether this association holds true for incident fractures.
Materials and Methods: This study aims to answer this question in a post hoc analysis of a subset of the risedronate phase III clinical programs, using the urinary excretion of deoxypyridinoline (uDPD) as an index of pretreatment bone resorption (PBR). A total of 1593 women with postmenopausal osteoporosis that had baseline uDPD values and paired spinal radiographs available were pooled, in similar proportions, from the risedronate multinational and North American VERT, and from the risedronate HIP trials. Patients from treatment and placebo groups were stratified by the uDPD premenopausal normative median. The four resulting groups were balanced for age, years since menopause, body mass index, baseline femoral neck BMD, and number of prevalent fractures, but baseline lumbar spine BMD was significantly higher in patients with low PBR rates.
Results: In all groups, the proportion of patients with new vertebral fractures was higher in patients with baseline uDPD levels above the normative median. The incidence of vertebral fracture was significantly lower in groups assigned to risedronate compared with placebo. This effect was independent of PBR: in patients with high PBR, the relative risk (RR) of vertebral fracture after 1 year of risedronate was 0.28 (p = 0.03 compared with controls, absolute risk reduction 7.1%). In patients with low PBR, the RR of fracture after 1 year was 0.33 (p < 0.001, absolute risk reduction 4%). After 3 years, the RR of fracture was 0.52 (p = 0.042, absolute risk reduction 8.3%) in patients with high PBR, and 0.54 (p = 0.002, absolute risk reduction 7.1%) in patients with low PBR. Results were similar after adjusting for age, baseline lumbar spine BMD, and prevalent fractures. The number needed to treat to avoid one vertebral fracture at 12 months was 15 in the group of patients with high PBR and 25 in patients with low PBR. Risedronate significantly increased lumbar spine BMD. During the first year of treatment, women with high PBR gained lumbar spine BMD at a faster rate than patients with low PBR. Treatment-by-PBR status interactions were not significantly different over time.
Conclusion: The efficacy of risedronate to reduce incident vertebral fractures in women with postmenopausal osteoporosis is largely independent of pretreatment bone resorption rates.
VERTEBRAL FRACTURES ARE considered to be the most frequent manifestation of postmenopausal osteoporosis. During the past years, a number of independent predictors of osteoporotic fracture risk have been identified, among them low bone mass(1) and prevalent fracture status.(2, 3) Several large prospective studies have demonstrated that increased rates of bone remodeling are also associated with a higher risk of nonvertebral and vertebral fractures, independent of bone mineral density (BMD), age, and disability.(4, 5) In these studies, patients with low BMD and/or high bone turnover were shown to be at the highest risk for osteoporotic fractures.
These observations may also provide a rationale for the use of bone markers in the selection of therapy. Ideally, therapeutic interventions should be tailored to the individual patient's profile. A range of factors could potentially influence treatment outcomes, such as age and life expectancy, disability, comorbidity, prevalent fracture status, BMD, calcium intake, and possibly, pretreatment bone turnover. Several studies point to a relationship between the rates of baseline bone turnover and the subsequent change in BMD during antiresorptive treatment. Civitelli et al.(6) demonstrated that, after 12 months of treatment with subcutaneous calcitonin, osteoporotic patients with high bone turnover had significantly greater changes in lumbar spine (but not in hip) BMD than similar patients with low bone turnover. Later, similar associations were found in healthy and osteoporotic subjects treated with hormone replacement therapy (HRT)(7–9) and alendronate.(10) Thus, while there is some evidence suggesting an association between pretreatment bone turnover and therapy-induced changes in BMD, it remains unclear whether this relationship also translates into a change of fracture incidence during antiresorptive treatment.
This study therefore investigates the relationship between pretreatment bone resorption (PBR) and fracture incidence in a post hoc analysis of a subset of patients from the three largest trials of the risedronate phase III clinical programs, using the urinary excretion of deoxypyridinoline (uDPD) as a marker of PBR rates.
MATERIALS AND METHODS
A subgroup analysis was conducted on pooled data from the three largest clinical trials conducted to determine risedronate's efficacy to prevent fractures—the multinational and the North American VERT (Vertebral Efficacy with Risedronate Therapy) and the HIP (Hip Intervention Program) studies.(11–13) In the VERT trials, 2457 postmenopausal women with two vertebral fractures or one vertebral fracture and a low lumbar spine BMD (T-score < −2 SD) were randomized to receive risedronate 5 mg daily or placebo. In the HIP trial, 6346 postmenopausal women 70-79 years of age with low femoral neck BMD (T-score < −3 SD) or at least 80 years of age with at least one nonskeletal risk factor for hip fracture were randomized to receive placebo or risedronate 5 mg daily. Subjects from the trials received 1000 mg/day elemental calcium and up to 500 IU/day vitamin D if baseline levels were low. Detailed descriptions of the trial populations and results can be found elsewhere.(11-13)
The primary population of interest in this study was the subset of 1593 patients who had baseline measurements of uDPD and took at least one dose of study medication (i.e., placebo or risedronate 5 mg daily) and had baseline and at least one post-baseline spinal radiograph. Patients from the VERT trial and patients with a baseline femoral neck T-score ≤ −2.5 SD from the HIP trial were included because BMD measurements were not required for patients in the HIP study group ≥80 years of age. In the group with low PBR rates, 320 patients (56%) were from the VERT trial and 249 patients (44%) were from the HIP trial. For the high PBR group, these numbers were 610 (60%) and 414 (40%), respectively.
Urine samples for the determination of collagen cross-links were collected from a subset of patients recruited by those centers participating in the bone turnover marker assessments. Urine was collected as a second morning void in the fasting state, aliquoted, and kept at −70°C until analysis.
For each trial, baseline uDPD levels were measured in a single batch by high-pressure liquid chromatography at the same central laboratory (Quest Diagnostics Inc., San Juan Capistrano, CA, USA) as described previously.(14) Results were corrected for urinary creatinine and expressed as nanomoles deoxypyridinoline per millimole creatinine.
The premenopausal normative data used in this study were based on the normal range developed by the laboratory in which the analyses had been performed (Quest Diagnostics Inc.). This reference range was obtained by investigating a population of 49 randomly selected North American, white, healthy females, 21–40 years of age. The interassay variation of the assay was determined to be 8.8% for subjects with uDPD levels within the normal range, and 9.4% for subjects with the elevated uDPD levels. The intra-assay variation was 6.6% in normal controls, and 8.0% in individuals with elevated uDPD levels.
BMD at the lumbar spine (L1-L4), femoral neck, and femoral trochanter was measured by DXA using Hologic and Lunar densitometers (Hologic, Bedford, MA, USA; Lunar Inc., Madison, WI, USA). Baseline lumbar spine T-scores were calculated using unstandardized BMD measurements if all four vertebrae were deemed intact, and standardized BMD values were generated using published equations to adjust for machine type. Baseline femoral neck T-scores were based on data from the NHANES III survey.
Identification of vertebral fractures
Lateral thoracolumbar (T4-L4) radiographs were obtained at baseline and annually during the studies. Prevalent and incident new vertebral fractures were diagnosed quantitatively and semiquantitatively, as described elsewhere.(11)
A patient's PBR status was defined as being “low” or “high” by comparing their baseline (i.e., pretreatment) uDPD value with the median normative value of 15.4 nmol deoxypyridinoline/mmol creatinine derived from the premenopausal normative data set. The study data (age, years since menopause, baseline body mass index, and baseline BMDs) were normally distributed. Demographic and baseline characteristics were summarized across treatment groups and PBR status using descriptive statistics.
The incidence of new vertebral fractures was estimated and compared between treatment groups and PBR status using time-to-first fracture methodology, adjusting for trial, consistent with previous analyses(11–13) (Kaplan-Meier estimate, Cox-regression, stratified log-rank test). Percent change in BMD from baseline at common time-points (i.e., 6, 12, 18, 24, and 36 months) across treatment groups and PBR status were summarized using descriptive statistics. ANOVA, adjusting for trial, was performed to obtain estimates of the treatment effect (mean, SE, and 95% CI). The main effects (i.e., difference between treatment groups and between PBR status) were investigated at the 5% level of significance.
The treatment-by-trial interaction was investigated at the 10% level of significance for both incident new vertebral fracture and BMD percent change from baseline. As all tests were not statistically significant, implying that the treatment effect was consistent between trials, all analyses were performed on the pooled data.
The treatment-by-PBR status interaction was investigated at the 10% level of significance to evaluate whether the treatment effect was consistent in patients with low and high bone turnover.
Sensitivity analyses were performed for both incident new vertebral fracture and changes in BMD to provide confidence in the original analyses. These analyses comprised adjustment for clinically important baseline data, such as age, BMD, and prevalent vertebral fractures.
Characteristics of the study population
The demographic characteristics of the study population at baseline are summarized in Table 1. Sixty-five percent of all subjects had a uDPD excretion above the normative median of 15.4 nmol/mmol creatinine. When stratified both by treatment modality (placebo versus risedronate) and PBR status (above versus below the normative median), the groups were well balanced. There was no statistical difference between any of the groups in regard to age, years since menopause, body mass index, femoral neck BMD, and number of prevalent fractures. Lumbar spine BMD was similar between treatment groups but was significantly higher in patients with uDPD values below than above the normative median (T-score: −2.4 ± 1.3 SD versus −2.8 ± 1.3 SD, p < 0.001).
Relationship between PBR and vertebral fractures
The total number of patients who had uDPD measurements at baseline, took at least one dose of study medication, and had evaluable paired spinal radiographs was 1196 at 1 year and 1223 at 3 years.
In the placebo group, the cumulative proportion of patients with new vertebral fractures at 1 and 3 years was higher in patients with baseline uDPD levels above the normative mean than in those with uDPD values below the normative median (Table 2). Compared with the low PBR group, the relative risk (RR) of sustaining a new vertebral fracture was approximately twice as high during the first year for patients with high PBR (RR = 2.1, 95% CI = 1.0–4.1, p = 0.039). There was no difference in risk over the entire 3-year period (RR = 1.4, 95% CI = 0.9-2.2, p = 0.118). In the risedronate-treated patients, the effect of baseline PBR status on new vertebral fractures did not reach statistical significance at any time-point (year 1: RR = 2.2, 95% CI = 0.7-6.9, p = 0.159; 0-3 years: RR = 1.3, 95% CI = 0.8-2.3, p = 0.287; Fig. 1).
The 1- and 3-year incidence of vertebral fracture was significantly lower in the groups assigned to risedronate compared with placebo (Table 2), and this effect was essentially independent of PBR (Fig. 2). After 1 year of treatment with risedronate, the RR of a new vertebral fracture was 0.28 (95% CI = 0.09–0.89; p = 0.03) in the low PBR and 0.33 (95% CI = 0.18-0.62; p < 0.001) in the high PBR group. After 3 years of treatment, the respective RRs were 0.52 (95% CI = 0.30-0.92, p = 0.042) and 0.54 (95% CI = 0.36-0.80, p = 0.002). Both the treatment-by-PBR status interaction (0-1 years: p = 0.808; 0-3 years: p = 0.980) and the trial-by-treatment interaction (0-1 years: p = 0.432; 0-3 years: p = 0.656) were not statistically significant.
To account for any differences at baseline between patients with high and low PBR rates, the primary analyses were repeated after adjusting for baseline lumbar spine BMD, age, and prevalent fractures. Data were also re-analyzed after omitting 118 patients from the VERT trials who, despite the inclusion/exclusion criteria, were found, after adjudication, to have had no prevalent vertebral fractures and a baseline lumbar spine BMD T-score above −2.5 SD. Results from both these re-analyses were in all aspects similar to results described above (data not shown).
The influence of underlying fracture risk on the number of patients needed to treat (NNT) to avoid one vertebral fracture is shown in Table 2. At 1 year, the NNT was substantially lower in the group of patients with uDPD values above the normative median than for their counterparts with low PBR rates. This result is mostly because of the higher fracture incidence and therefore greater underlying risk in the placebo group with high PBR rates, but was less pronounced when analyzing the entire 3-year study period.
Relationship between PBR and BMD
Irrespective of PBR status, risedronate significantly increased lumbar spine BMD from baseline compared with placebo (Fig. 3). However, during the first year of risedronate treatment, the magnitude of change in lumbar spine BMD was significantly greater in patients with baseline uDPD levels above the normative median (p = 0.045 at 6 months; p = 0.008 at 12 months). In contrast, during the first year of placebo treatment, the magnitude of change in lumbar spine BMD was not statistically significant between the two PBR groups (p = 0.595 at 6 months; p = 0.734 at 12 months). Nevertheless, when comparing risedronate-treated patients with placebo-treated controls, the treatment-by-PBR status interactions were not significantly different over time. When the analyses were repeated, adjusting for baseline lumbar spine standardized BMD, these inferences did not change.
At the hip, risedronate statistically significantly increased BMD from baseline compared with placebo, independent of PBR status. Again, the treatment-by-PBR status interactions were not statistically significant (data not shown). When adjusting for baseline lumbar spine standardized BMD, the inferences did not change.
To our knowledge, this is the first study investigating the influence of PBR rates on the incidence of vertebral fracture during antiresorptive treatment. Several important observations were made. First, risedronate significantly reduced the risk of new vertebral fractures independent of baseline bone resorption rates. Second, in both the placebo and the risedronate-treated groups, the proportion of patients with new vertebral fractures was higher in women with PBR rates above than those below the normative median. Third, during the first 12 months of treatment with risedronate, the number of patients that need to be treated to avoid one vertebral fracture was lower in the group of patients with uDPD values above the cut-off used in this analysis. Finally, risedronate significantly increased lumbar spine BMD compared with placebo, and this effect seemed primarily independent of baseline bone resorption.
Individualized treatment is an important aim of any patient-centered medical approach, and this principle should be applied to the patient with postmenopausal osteoporosis. In fact, the availability not only of different antiresorptive compounds, but also of distinct therapeutic principles such as antiresorptive and anabolic treatments, calls for a differentiated approach in each patient. Theoretical considerations have included pretreatment bone turnover as a potential determinant in such individualized strategies, hypothesizing that patients with accelerated bone turnover will derive greater benefits from antiresorptive treatment (which lowers bone turnover), whereas patients with low bone turnover are more likely to gain through anabolic treatments. This study indicates that, in women with established osteoporosis, antiresorptive treatment with risedronate reduces the risk of new vertebral fractures independent of pretreatment uDPD levels.
This study differs in many aspects from the previously published evidence.(6–10) First, until this point, the only studies to test the influence of pretreatment bone turnover on therapeutic outcomes have had BMD as an endpoint. It has been recognized, however, that changes in BMD during antiresorptive treatment only partly explain the effects of drugs such as risedronate,(15) alendronate,(16) or raloxifene(17) on fracture incidence. While there is evidence that baseline bone turnover affects the accrual of BMD during antiresorptive treatment regimens (see also below), our study indicates that, at least for risedronate, this association does not translate into an increased risk for fracture.
Second, none of the previous studies have been conducted for more than 2 years. This study, including analyses of 1- and 3-year data, suggests that, while there seems to be slight differences in BMD accrual over time (see below), the effect of PBR rates on subsequent fracture rates remains unchanged over 12 and 36 months.
Third, most of the previous studies were conducted in early postmenopausal women,(6–8) whereas the population included in this study had a mean age of 72 years. Notably, changes in lumbar spine BMD seem to be more pronounced in some of these earlier studies.(6) Thus, our results do not exclude that PBR rates may affect fracture outcome during antiresorptive treatment in younger individuals.
Fourth, measurements of bone turnover and chosen cut-offs differ between the various studies. While some investigators have used studies of whole body retention kinetics,(6, 9) other investigators and the present analysis used measurements of biochemical markers of bone resorption such as the aminoterminal telopeptide of type I collagen (NTx)(7, 8) or deoxypyridinoline. However, one would not expect large differences in results based on different methodologies because good correlations between whole body retention kinetics and markers of bone turnover has been described.(18) A number of studies have shown that patients with high baseline bone turnover had greater changes in lumbar spine BMD than patients with lower baseline bone turnover, independent of the methodology used to quantify or classify bone turnover rates.(6–10)
This study uses the premenopausal normative median as the cut-off to stratify subjects into those with “high” and “low” bone resorption rates. This obviously is an arbitrary decision and the case for other cut-offs or means of stratification can be made. In the present analysis, application of the normative median divides the study population so that approximately one-third is below and two-thirds are above that cut-off. Gonelli et al.(9, 10) have used the 95th percentile of99TC-methylene-diphosphonate whole body retention as the cut-off, which resulted in relative proportions of “high” and “low” turnover patients similar to the one reported here. Other groups have looked at a population split defined by quartiles of NTx excretion comparing the highest and the lowest quartiles in the given study population.(7, 8) While all of these approaches have their specific advantages and disadvantages, they are unlikely to fundamentally change the outcome of the analyses itself. Our choice of the normative median was driven by considerations of clinical relevance and the potential to reapply the cut-off to patients of similar age and disease stage.
The finding that the risk for osteoporotic fractures during the first 12 months increases with increasing bone resorption rates is consistent with previous observations.(4, 5) Within the placebo group, the difference in fracture risk between “low” and “high” PBR groups was significant, to the extent that patients with a baseline uDPD excretion above the cut-off had a 2-fold higher risk of sustaining a new vertebral fracture during the first year of the study than patients with a value below this limit. Although a similar trend was seen in the risedronate-treated patient group, the number of vertebral fractures in this group was much lower and the difference between the risedronate-treated subgroups was not statistically significant. Together with the evidence discussed above, this observation supports our assumption that the population examined in this study, and the type of statistical analysis, are comparable with the populations and results of other, independent studies.
The number of subjects who have to be treated to prevent one vertebral fracture during the first 12 months was less in the “higher” risk group (i.e., the group with high bone resorption at baseline). It is important to note that, while results in regard to NNT have implications for cost-effectiveness, the findings of this study do not support treating patients solely based on bone turnover as treatment efficacy was not significantly influenced by baseline bone resorption status.
The treatment effect of risedronate on BMD (i.e., risedronate over placebo) did not differ between the high and low PBR groups. However, it was interesting to note that within the risedronate-treated groups, patients with higher PBR rates gained significantly more lumbar spine BMD in the first 12 months than those with lower bone resorption rates. This influence of baseline bone turnover on initial (i.e., 1 year) lumbar spine BMD response to treatment is consistent with previous studies. Civitelli et al.(6) demonstrated that 1 year of treatment with subcutaneous calcitonin resulted in substantially greater gains in lumbar spine, but not in hip BMD, when treating patients with high bone turnover as opposed to subjects with low bone turnover (as determined by whole body99TC-methylene-diphosphonate retention). Interestingly, while in our study BMD accrual differed between PRB groups at the level of the lumbar spine, no such effects were seen at the hip. Similar observations at 12 months have been reported for HRT(7–9) and alendronate.(10) Two of these studies reported significant differences in 2-year lumbar spine BMD between high and low bone turnover groups on HRT(9) and alendronate(10); this difference seems to be because of the initial 12 month differences in BMD gain rather than continued differences after 12 months. Our 3-year study provides further evidence for an early but not continued influence of baseline bone turnover status on BMD response to treatment.
A limitation of this study is the fact that it is a post hoc analysis of pooled data and that data were pooled from three distinct clinical trials. Thus, the balance between treatment and control groups of known and unknown determinants of fracture risk and BMD achieved by the randomization of the initial trials may have been compromised in the pooled analysis. However, this fact is unlikely to affect the results of the study, as there were no significant differences in demographic characteristics and prevalent fractures at baseline between groups based on treatment and bone resorption status. There were, however, differences in baseline lumbar spine (not hip) bone density between the high and low PBR groups. Again, these differences are unlikely to have biased the results because they would not influence changes in BMD and there were no significant differences in baseline BMD between treatment and placebo groups. Also, additional sensitivity analyses suggested that the difference in baseline lumbar spine BMD did not affect the essence of the uncorrected results. Another limitation of the study is the use of only one marker of bone resorption when there are many others available. Thus, although uDPD is considered a sensitive and reliable marker of bone resorption, one should be careful in generalizing our results to other markers of bone turnover. Also, normative data for uDPD were derived from a North American white population only, when the patient population was recruited from North America, Europe, and Australia. Finally, we used the normative median to stratify our population in a dichotomous fashion when other methods of stratification were possible (e.g., by tertiles or quartiles). However, there are no agreed criteria by which to define “low” and “high” bone turnover groups and a dichotomous approach seems clinically most relevant.
In summary, the present analysis provides evidence that treatment with risedronate reduces the risk of vertebral fractures in women with postmenopausal osteoporosis regardless of PBR rates assessed in comparison with the premenopausal normative median.
We thank Prof Ego Seeman for helpful comments on this manuscript.
- 12001 Identification and fracture outcomes of undiagnosed low bone mineral density in postmenopausal women: Results from the National Osteoporosis Risk Assessment. JAMA 286:2815–2822., , , , , , , ,
- 21991 Pre-existing fractures and bone mass predict vertebral fracture incidence in women. Ann Intern Med 114:919–923., , ,
- 32001 Risk of new vertebral fracture in the year following a fracture. JAMA 285:320–323., , , , , , , , , , ,
- 41996 Case-control analysis of bone resorption markers, disability, and hip fracture risk: The Rotterdam study. BMJ 312:482–483., , , , , , ,
- 51996 Increased bone turnover in late postmenopausal women is a major determinant of osteoporosis. J Bone Miner Res 11:337–349., , ,
- 61988 Bone turnover in postmenopausal osteoporosis. Effect of calcitonin treatment. J Clin Invest 82:1268–1274., , , , , ,
- 71997 The predictive value of biochemical markers of bone turnover for bone mineral density in early postmenopausal women treated with hormone replacement or calcium supplementation. J Clin Endocrinol Metab 82:1904–1910., ,
- 81997 Hormone replacement therapy in postmenopausal women: Urinary N-telopeptide of type I collagen monitors therapeutic effect and predicts response of bone mineral density. Am J Med 102:29–37., , , , , , , , , ,
- 91997 The usefulness of bone turnover in predicting the response to transdermal estrogen therapy in postmenopausal osteoporosis. J Bone Miner Res 12:624–631., , , , ,
- 101999 Bone turnover and the response to alendronate treatment in postmenopausal osteoporosis. Calcif Tissue Int 65:359–364., , , , , ,
- 111999 Effects of risedronate treatment on vertebral and nonvertebral fractures in women with postmenopausal osteoporosis: A randomized controlled trial. Vertebral Efficacy With Risedronate Therapy (VERT) Study Group. JAMA 282:1344–1352., , , , , , , , , , ,
- 122000 Randomized trial of the effects of risedronate on vertebral fractures in women with established postmenopausal osteoporosis. Vertebral Efficacy with Risedronate Therapy (VERT) Study Group. Osteoporos Int 11:83–91., , , , , , , , , ,
- 132001 Effect of risedronate on the risk of hip fracture in elderly women. Hip Intervention Program Study Group. N Engl J Med 344:333–340., , , , , , , , , , ,
- 141989 Excretion of pyridinium cross-links of collagen in ovariectomized rats as urinary marker for increased bone resorption. Calcif Tissue Int 44:343–347., ,
- 152001 A method to assess the proportion of treatment effect explained by a surrogate endpoint. Stat Med 20:3175–3188., ,
- 162002 Improvement in spine bone density and reduction in risk of vertebral fractures during treatment with antiresorptive drugs. Am J Med 112:281–289., , , , , ,
- 172002 Relationships between bone mineral density and incident vertebral fracture risk with raloxifene therapy. J Bone Miner Res 17:1–10., , , , ,
- 181999 Measurement of biochemical markers of bone resorption. In: SeibelMJ, RobinsSP, BilezikianJP (eds.) Dynamics of Bone and Cartilage Metabolism. Academic Press, San Diego, CA, USA, pp. 411–426.,