BMD Is Reduced in HIV-Infected Men Irrespective of Treatment
"....We show in this study that there was a high prevalence of osteoporosis in the HIV+ patients. Low BMD has been reported in HIV+ patients by several groups, but our data showing a high prevalence of osteoporosis (16%) and osteopenia (66%) in all HIV+ male patients are among the highest reported thus far.....
.....Our finding show that low bone density is present in HIV-infected males before treatment and that it is neither exacerbated nor cured by the treatment.....Multivariate analysis showed that the Z-score of HIV-infected patients was significantly correlated to BMI....and to the lowest BMI recorded since the onset of the disease....but not to how long they had been HIV+, the treatment they received, duration of the treatment, CD4 and viral load, whether they smoked, or amount of calcium in their diet. Weight accounted for 21% of the bone density.....Compared with control, bone alkaline phosphatase, a marker of bone formation, was significantly lower in the untreated patients......Because adipocytes and osteoblasts differentiate from a common precursor, it has been suggested that there could be some link between reduced bone density and abnormal fat repartition.....Beside low BMI, other osteoporosis risk factors such as smoking, low physical activity, low calcium intake, and periods of immobilization could account for decreased bone density...
....In conclusion, our data show that low bone density presents early in HIV+ men and is associated with both high cytokine levels before treatment and a risk factor for common osteoporosis (low weight). However, these factors do not explain the high prevalence of osteoporosis in HIV+ men. Further studies are needed to explain why this decrease in bone density does not improve over time with treatment. Larger populations of untreated HIV+ patients followed longitudinally are required to discover whether the length of time before beginning treatment is a risk factor for osteoporosis."
Journal of Bone and Mineral Research, March 2004:19:402-409 (doi: 10.1359/JBMR.0301246)
C Amiel, 1 A Ostertag, 2 L Slama, 1 C Baudoin, 2 T N'Guyen, 1 E Lajeunie, 2 L Neit-Ngeilh, 1 W Rozenbaum, 1 MC De Vernejoul2
1Department of Infectious Disease, Hôpital Tenon, Paris, France;
2INSERM U349, Department of Rheumatology and Biochemistry, Hôpital Lariboisière, Paris, France.
ABSTRACT
Osteoporosis has be reported to be a complication of active antiretroviral therapy of HIV infection. We studied 148 HIV-infected men stratified according to their treatment. Our data show that these patients have an average 9% decreased BMD, irrespective of their treatment. Low body mass index and high resorption markers were associated with low bone density.
Introduction: Osteoporosis has been reported in HIV-infected (HIV+) patients, and it has been suggested that it may be linked to protease-inhibitor treatments (PI).
Materials and Methods: To assess this risk and to investigate its putative link with treatments, we compared the bone density of HIV+ men, who were either receiving treatment (including PI [PI+], n = 49; without PI [PI-], n = 51) or untreated (UT, n = 48). We included 81 age-matched control HIV-negative (HIV) males (age, 40 ± 8 years).
Results: BMD adjusted for age (Z-score) was lower in the HIV+ patients at the lumbar spine (HIV+: −1.08 ± 1.21, HIV-: −0.06 ± 1.26, p < 0.001) and the femoral neck (HIV+: −0.39 ± 1.05, HIV: 0.25 ± 0.87, p < 0.001). The prevalence of osteoporosis was 16% in HIV+ and 4% in HIV subjects (p < 0.01). In the HIV+ subjects, the Z-score was correlated only to body mass index (r = 0.27 at lumbar spine and 0.35 at femoral neck). Untreated HIV+ patients had a negative Z-score (−0.82 ± 1.15 for the lumbar spine), which was not different from the one of treated HIV+ patients. In the PI+ and PI- groups, the Z-score did not depend on the presence of lipodystrophy or the proportion of fat in the abdomen and legs measured by DXA. Markers of bone remodeling were measured in the 132 HIV+ and 35 HIV- subjects. Compared with controls, HIV+ patients had lower bone alkaline phosphatase and higher urinary cross-laps/Cr, which was negatively correlated with the Z-score at both the femoral neck (r = −0.22) and lumbar spine (r = −0.21). TNFα was increased in untreated compared with treated HIV+ subjects and was not correlated to the Z-score.
Conclusion: Our cross-sectional study does not show any deleterious effect of the treatment but does indicate a decrease in bone density in HIV+ patients irrespective of the treatment. This low bone density is in part related to the low body weight and is associated with increased bone resorption.
INTRODUCTION
THE INTRODUCTION OF highly active antiretroviral therapy (HAART) has dramatically modified the course of HIV infection.(1) However, long-term HAART has been associated with several metabolic complications, including hyperlipidemia,(2) abnormal fat distribution,(3) and osteoporosis in men.(4)
Osteoporosis is a common disorder in postmenopausal women, and its occurrence in middle-aged men infected with HIV is an unexpected complication of the disease or its treatment. Osteoporosis in males has been neglected during decades but has received increasing attention as the incidence of osteoporotic fractures in elderly males increases.(5) Osteoporosis in men is often secondary to detrimental environmental factors or to endocrine disease, namely alcohol abuse, glucocorticoid excess, and hypogonadism.(5,6) None of these causes can clearly account for osteoporosis in HIV-infected patients. The main complication of osteoporosis is fragility fractures, but the prevalence of osteoporosis and fractures in HIV-infected patients has not been investigated. Central to the debate is the possible link between HAART and osteoporosis: several authors have observed an association between HAART and osteoporosis(4,7); however, methodological bias, particularly the lack of control groups, could have obscured the data. Furthermore, the possible association between osteoporosis and another complication of the treatment, lipodystrophy, has also been a matter of debate.(8)
We therefore decided to investigate a cohort of HIV-infected males to determine the prevalence of osteoporosis and fractures and to identify the possible mechanisms of bone loss by evaluating markers of bone remodeling. We included in the study a group of untreated HIV-infected patients and compared the cohort to controls. Our finding show that low bone density is present in HIV-infected males before treatment and that it is neither exacerbated nor cured by the treatment.
MATERIALS AND METHODS
Patients
This was a non-interventional cross-sectional study with no individual benefit. The inclusion criteria were a documented positive HIV test, age above 20 years and under 60 years, male gender, never treated (UT) with antiretroviral drugs (ARVs) or receiving treatment for more than 18 months with the same classes of ARV including two nucleoside reverse transcriptase inhibitors (NRTIs) + one protease inhibitor (PI; PI+ group), two NRTI + 1 non-nucleoside transcriptase inhibitors (NNRTIs), or three NRTIs (PI group subdivided into PI- 2n1nn and PI- 3n), and having signed an informed consent form. The exclusion criteria were acute infection or uncontrolled chronic infection, treatment with corticosteroids, hormones, immunomodulators, cytotoxic agents, or diuretics, calcium supplementation, and any bone or rheumatic disorder. All patients were white.
A total of 148 HIV-infected patients (HIV+) were included: 48 untreated patients (UT group), 49 patients treated with PI (PI+ group), and 51 patients treated without PI (PI group: 25 in the PI2n1nn group and 26 in the PI- 3n group).
Patients completed a questionnaire about previous personal fractures and physical and nutritional habits. All fractures were reported, irrespective of site, with the date and the circumstances. Lipodystrophy was defined as the presence of peripheral loss of fatty tissue and abnormal fat distribution including one or more of the following clinical signs: breast hypertrophy, increased waist measurement, visceral abdominal fat hypertrophy, and enlargement of the dorsocervical pad "buffalo hump."
Bone density
BMD was measured at the femoral neck and at the lumbar spine (L2-L4) using a Lunar DPX-L (Lunar Corp., Madison, WI, USA). All measurements were performed with the same densitometer and by the same technician. Age-adjusted values were based on a French reference population between 20 and 89 years of age from several centers (provided by Lunar France). The data were adjusted for age and gender and expressed as a Z-score and a T-score. For the Z-score, the results were based on the observed BMD value minus the mean of normal BMD values for men of the same age, divided by the SD of this reference population. For the T-score, the results were based on the observed BMD value minus the mean of normal BMD values for men between 20 and 30 years of age, divided by the SD of this reference population. Osteopenia was defined as a T-score of between −1 and −2.5, and osteoporosis was defined as a T-score of less than −2.5 relative to this normal French population.
Whole body scans were performed to obtain the fat and lean mass. The software provided by the manufacturer included cut-off lines positioned at anatomical regions of interest. In addition to the whole body, we selected the regions of the trunk and the legs and calculated the ratio of the fat in these two regions, divided by the whole body fat.
Biochemical measurements
Testosterone was measured using a radioimmunoassay (BYK-Sangtec, Dietzenbach, Germany) (normal range, 350-1160 ng/dl). Free testosterone was calculated according to Vermeulen et al.(9) Sex hormone binding protein (SHBG), dehydroepiandrosterone (DHEA), and parathyroid hormone (PTH) were measured by automated chemiluminescent immunoassay (Diagnostic Products Corp., Los Angeles, CA, USA). Normal values were 13-71 nM, 800-5600 ng/ml, and 7-53 pg/ml, respectively. Total insulin-like growth factor (IGF)1 was measured by IRMA (Immunotech, Marseilles, France) (normal range for males of this age, 90-492 ng/ml). Leptin was measured using a radioimmunoassay (LINCO, St Charles, MO, USA) (normal range for males, 2-5.6 ng/ml). 25(OH)vitamin D was measured after extraction using a commercial RIA kit (DiaSorin, Stillwater, MN, USA) (normal range, 8-35 ng/ml). Plasma TNFα was measured using immunoradiometric assay (Biosource Europe, Brussells, Belgium). The minimum detectable concentration was 5 pg/ml. TNFα was not measured in the controls.
Bone resorption was assessed by measuring urinary type I C-telopeptide breakdown products (CTX) using an ELISA kit (Cross-laps; Osteometer, Herlev, Denmark). Calculation of the corrected Cross-laps value gave a normal range of Cross-laps/creatinine values of 100-300 μg/mmol. Bone formation was assessed by measuring both osteocalcin and bone alkaline phosphatase. Serum bone-specific alkaline phosphatase (BAP) was measured using an immunoradiometric assay (Tandem-R, Ostase) provided by Hybritech Europe S.A. (Liege, Belgium) (normal range, 7.5-16 ng/ml). Serum osteocalcin (OC) was measured using a radioimmunoassay (OSTK-PR; CIS Biointernational, Gif-sur-Yvette, France) (normal range for men, 0.9-18 ng/ml).
Statistical methods
The study involved two steps. First, to study the prevalence of osteoporosis, we compared the whole group of HIV+ patients to 81 (HIV-) male controls from 20 to 60 years of age who were employees (students, physicians, etc.) at our hospital. These subjects were volunteers and had filled out the same questionnaire as the HIV+ patients and had undergone a bone densitometry. We excluded subjects using a treatment that could induce bone disease and those who had a chronic pathology. We used the χ2 test to compare the qualitative data and Student's t-test for the quantitative data. The effect of the disease on BMD (Z-score) was also tested after adjusting for body mass index (BMI), current smoking (yes or no), and calcium intake using multivariate analysis.
Second, to assess the effect of treatment on BMD, we analyzed the four HIV+ groups PI+, PI- (PI- 2n1nn and PI- 3n) and the UT group using a one-way ANOVA. The sample size of PI+, PI-, and UT HIV+ patients were roughly equilibrated. When hypothesis of equal effect was rejected, we compared the differences, defined a priori, between treated and untreated patients, between the overall PI- group and the PI+ group and between the PI- 2n1nn and PI- 3n groups, using multiple orthogonal comparison based on Helmert contrasts. In addition, we compared each HIV+ subgroup to the HIV controls using the posthoc test of Dunnett.
The effect of the treatment on BMD was also tested after adjusting for actual BMI, the previous lowest BMI, calcium intake, smoking, age at HIV onset, how long the subject had been HIV positive, duration of the treatment, CD4 and viral load, and the interactions between these factors. We used multiple linear regression. The best fitting and most parsimonious subsets of factors were selected using the lack-of-fit method based on the likelihood ratio test.
To study the effect of the disease and treatments on biochemical parameters related to bone metabolism, we used 35 of the 81 controls and the four HIV+ groups described above. We first analyzed the difference between the five groups using a one-way ANOVA; second, when the test was significant, we used a multiple comparison between groups (Tukey test). We evaluated the relationship between the Z-score and the biochemical parameters using Pearson's correlations.
Results were expressed as mean ± SD. All tests were two-sided, and the significance level was fixed at 0.05. The statistical computations were performed with S plus 2000.(10)
RESULTS
Prevalence of osteoporosis and association with fractures
Table 1 shows that BMD, Z-score, and T-score were all significantly decreased in the patients, both at the lumbar spine and femoral neck, and the difference in BMD was 9% at both sites. The weight and BMI of the patients was lower than the controls, although BMI was only moderately decreased in the patients (23 ± 3 versus 24 ± 3; p < 0.02). Fifty-two percent of the patients and 33% of controls were current smokers. The model including disease, BMI, calcium intake, and smoking explained 21% of the variance of the Z-score (p < 0.001).After adjusting for BMI, smoking, and calcium intake, there was still a significant difference (p < 0.001) for the Z-score at the spine and femoral neck between the HIV+ patients and controls.
Sixty-six percent of the patients and 32% of the controls presented with osteopenia (−2.5 < T-score < −1) at at least one site (p < 0.001). When a T-score of ≤ −2.5 at any site was taken to be the threshold, the prevalence of osteoporosis reached 16% in this population of men (mean age, 40 years) and was significantly higher than in the 81 controls (4%; p < 0.001; Fig. 1).
Thirty-seven percent of controls reported a previous fracture. Among the HIV+ patients, 41% reported a previous fracture, including 22% who had had their first fracture after the disease had been diagnosed. All the fractures were reported as being traumatic. Six patients had a crushed vertebra, confirmed by an X-ray. It was performed because of pain after an injury that had occurred for 3/6 of the patients before the disease had been diagnosed. None of them had osteoporosis. Osteoporosis was not associated with the occurrence of fractures: 45% of the patients with fractures had osteoporosis and 39% had no osteoporosis.
BMD in HIV+ patients according to treatment
About one-half the patients were smokers (52%), and 41% had a calcium intake of less than 900 mg/day. Patients whose treatment included PI had a slightly lower calcium intake, were more frequently smokers, had a higher viral load, and had been receiving treatment for longer than those whose treatment did not include PI. Untreated patients were younger, had been HIV+ for a shorter time, and had a higher current viral load and a lower CD4 count nadir than the treated patients (Table 2).
The Z-score at the lumbar spine and femoral neck did not depend on whether the patient was receiving treatment or whether this included PI (Table 3) or the patient belonged to one of the two subgroups treated without PI (data not shown).
The overall Z-score at the lumbar spine was significantly reduced to less than zero (p < 0.001). Each subgroup of HIV+ patients had a Z-score lower (p < 0.05) than the controls. Untreated patients had a mean Z-score of −0.82 at the lumbar spine and of −0.19 at the femoral neck (Table 3). Four of 48 untreated patients had osteoporosis, which was not different from the treated patients, 18 of 100 of whom had osteoporosis (not significant).
Multivariate analysis showed that the Z-score of HIV-infected patients was significantly correlated to BMI (r = 0.27 at the lumbar spine and 0.32 at the femoral neck) and to the lowest BMI recorded since the onset of the disease (r = 0.35 at the lumbar spine and 0.39 at the femoral neck), but not to how long they had been HIV+, the treatment they received, duration of the treatment, CD4 and viral load, whether they smoked, or amount of calcium in their diet. Weight accounted for 21% of the bone density.
BMD and lipodystrophy
Lipodystrophy was present in most of the treated patients but in only one untreated patient. We also assessed lipodystrophy by measuring the amount of fat and the proportion of fat in the abdomen and the legs. Treated patients, independently of whether they were receiving PI or not, had less fat than the untreated patients. They also had a percentage of fat higher in the trunk and lower in the legs than the untreated patients (Table 4). Among the treated patients whose treatment did not include PI, patients receiving two NRTIs + one NNRTI (n = 26) had a lower fat mass (7.8 ± 3.5 kg) than patients receiving three NRTIs (n = 25, 10.1 ± 4.4 kg, p < 0.05), and they also more frequently had lipoatrophy (77% versus 48%, p < 0.04). There was no other treatment-related difference for any of the parameters describing lipodystrophy (Table 4).
In the group of HIV+ patients as a whole, lipodystrophy was not associated with osteoporosis or with Z-score. In the subgroup of treated patients, there was no influence of lipodystrophy, lipoatrophy, or hypertrophy on the Z-score. BMC was correlated to both lean mass and fat mass in the untreated patients; however, whereas the correlation between BMC and lean mass persisted, the correlation between the BMC and fat mass was not more significant in the treated patients (Table 5).
Endocrinology and bone markers
First, we compared the controls to each group of HIV+ patients (Table 6). Plasma testosterone and SHBG were higher in untreated patients than in the controls, and free testosterone was not increased in any of the patient subgroups. Vitamin D and PTH levels, as well as levels of IGF1, were in the normal ranges and did not differ in patients and controls. The one-way variance analysis among the four groups was not significant for osteocalcin. Compared with control, bone alkaline phosphatase, a marker of bone formation, was significantly lower in the untreated patients. Urinary cross-laps, a marker of bone resorption, was increased in HIV+ patients treated without PI compared with control. When comparing all the HIV+ patients to the 35 controls, urinary cross-laps was increased (0.22 ± 013 versus 0.15 ± 0.07 μg/mmol, p < 0.01) and alkaline phosphatase was decreased (9.5 ± 4.2 versus 11.9 ± 4.3 ng/ml, p < 0.01) in the patients, whereas osteocalcin was not different between the groups. There was a significant negative correlation between osteocalcin and the Z-score at the lumbar spine (r = −0.18, p < 0.04) and also between urinary cross-laps and the Z-score at both the lumbar spine (r = −0.21, p < 0.012) and the femoral neck (r = −0.22, p < 0.005). There was no correlation between alkaline phosphatase and Z-score.
Untreated HIV+ patients had higher serum TNFα values than treated patients. TNFα was not correlated to the markers of bone formation or resorption or to Z-score.
The level of leptin was not different between the HIV+ patients and the 35 controls. Among patients treated without PI, those receiving two NRTIs + one NNRTI (n = 26) had a significantly lower serum leptin level than those receiving three NRTIs (2 ± 0.9 versus 3.2 ± 1.9 ng/ml, p < 0.05).
DISCUSSION
We show in this study that there was a high prevalence of osteoporosis in the HIV+ patients. Low BMD has been reported in HIV+ patients by several groups, but our data showing a high prevalence of osteoporosis (16%) and osteopenia (66%) in all HIV+ male patients are among the highest reported thus far.(4,11) Although single case reports of fractures have already been published about HIV+ patients,(12) this is the first time that the occurrence of fragility fractures in these patients had been assessed in a cohort. In our population of young men, the fractures could not be related to osteoporosis. However, our data do not exclude the possibility that fragility fractures could occur at an older age in this population.
In the multivariate analysis, we could not see any association between bone density and nature or duration of treatment. This cross-sectional study with stratification according to treatment used a large group of untreated HIV+ patients who could be used as controls for investigating the possible implication of various treatments in the occurrence of osteopenia. This stratification was made to assess the responsibility of various treatment-related factors that have been suggested as possible etiologies for the low bone mass observed in other studies. Our study shows that BMD is reduced in HIV+ patients regardless of the treatment and quite early in the course of the disease. Several hypotheses could explain reduced bone density in HIV+ patients.
Because adipocytes and osteoblasts differentiate from a common precursor, it has been suggested that there could be some link between reduced bone density and abnormal fat repartition. Indeed, in a small group of 41 HIV+ patients, Huang et al.(8) observed that bone density was lower in patients with lipodystrophy than those without lipodystrophy and controls and that abdominal fat was a negative predictor of bone density measured by QCT. McDermott et al.(13) also observed that men receiving HAART had a higher proportion of fat in the trunk and lower bone density, both of which were related to treatment duration, but they did not detect any relationship between these two factors. However, in other studies, neither BMD nor osteoporosis was associated with fat accumulation.(4,11) We measured fat accumulation in the trunk accurately and did not find any negative relationship between bone density and the accumulation of fat in the abdomen, even when patients with lipodystrophy were selected. As in this study, the amount of fat is usually positively correlated with the BMC, but in patients treated with HAART, we showed that this had no positive or negative effect on the BMC.
Protease inhibitors have been reported to be either positively or negatively associated with bone density. In a small cross-sectional study, Tebas et al.(4) reported that the use of PI was associated with lower bone density than that found in a mixed group of patients treated without PI or not treated at all, and this was also confirmed by another study.(7) However, this was not observed in any of the subsequent cross-sectional studies.(8,13) Even in two short-term longitudinal studies, there was an increase or no decrease in BMD with time in patients receiving PI.(14,15) In our study, which included 100 treated patients, one-half of whom received PI, we could not see any treatment-related difference in bone density. Compared with controls, there was an increased level of cross-laps, a marker of bone resorption, in patients treated without PI. The same trend for this was observed in patients receiving PI and in untreated patients who also had a slight insignificant increase in Cross-laps level as already observed by other.(16) It is not certain that the increased bone resorption is attributable to the treatment, although, in vitro, some PIs can induce increased bone resorption.(17) We did not observe any decrease in either osteocalcin or bone alkaline phosphatase in these patients receiving HAART, and moreover, one study show that PI treatment is associated with an increase in osteocalcin.(16) Our data do not preclude any positive or negative action of HAART on bone, because the results of BMD were not different from the group of untreated patients, who had decreased bone density.
We could not see any biochemical endocrine change induced by treatment in the whole group or in any of the subgroups that could offer a simple explanation for the lower bone density: free testosterone was normal, and there was no patient with hypogonadism that has been shown to occur in advanced HIV disease.(18) Similarly, DHEA was, as previously reported,(19) slightly decreased in patients treated without PI but it is not likely that it can account for the decreased bone density. There was no vitamin D deficiency or increase in PTH level. Modification of the IGF system has been reported in patients infected with HIV.(20) We measured only IGF1 that has been shown to be decreased in males with idiopathic osteoporosis,(21) and we observed no changes in our HIV+ patients.
Most of the studies of BMD in HIV+ patients have included only a small proportion of untreated patients or none at all. It is in fact difficult to persuade these patients to take part in a clinical study. They also differ from treated patients in terms of age, how long they have been HIV+, and their viral load and tCD4 count. Unexpectedly, we found that they had low bone density that was similar to that of the treated patients. In accordance with previous studies,(16) this low bone density was associated with low bone formation; bone alkaline phosphatase levels are reduced in this population. That could be because of the secretion of cytokines as a result of the high level of viral replication. Pro-inflammatory cytokines have been shown to be elevated in untreated HIV+ patients.(22,23) Indeed, we also observed that TNFα serum levels were increased in untreated HIV+ patients comparatively to both treated subgroups. There is an association between cytokines and bone remodeling in several metabolic bone diseases, including postmenopausal osteoporosis.(24,25) Interleukin (IL)1 and TNF not only increased bone resorption but also decreased bone formation.(26) The role of TNF on bone remodeling in HIV infection has been suggested in a previous biochemical study based on a correlation between decreased plasma levels of osteocalcin and TNF.(15) However, in our study, we could not find any correlation between TNFα and the Z-score.
When comparing all the HIV+ patients to their controls, we observed an increase in urinary cross-laps and a decrease in alkaline phosphatase. Although these changes were of variable importance and maybe of different etiologies in the different HIV+ subgroups, these data point to an identical mechanism of the bone loss in all the HIV+ patients. Moreover, there was a negative correlation between these markers and the Z-score. The imbalance between decreased bone formation and increased bone resorption would induce bone loss.
Finally, the low bone density could be present before HIV infection or be related to any common osteoporosis risk factor. Our HIV+ patients have a slightly lower BMI than a population of uninfected patients. Indeed, we observed a relationship with the BMI and the lower BMI of these patients as previously observed in another study.(11) However, our patients did not have a wasting syndrome that is associated with markedly decreased BMD.(27) When comparing the HIV+ patients and controls, the Z-score was still lower in the HIV+ patients after adjusting for BMI. Body weight accounted for only 21% of the bone density in the patients and cannot be the only factor responsible for their low bone density.
Beside low BMI, other osteoporosis risk factors such as smoking, low physical activity, low calcium intake, and periods of immobilization could account for decreased bone density. When we adjusted Z-score for current smoking, the difference between HIV+ patients and controls persisted. Average calcium intake was around 800 mg/day (Table 2), and the patients had a normal-to-high current physical activity. However, all these parameters could have been altered in the past for significant periods of time in these patients.
In conclusion, our data show that low bone density presents early in HIV+ men and is associated with both high cytokine levels before treatment and a risk factor for common osteoporosis (low weight). However, these factors do not explain the high prevalence of osteoporosis in HIV+ men. Further studies are needed to explain why this decrease in bone density does not improve over time with treatment. Larger populations of untreated HIV+ patients followed longitudinally are required to discover whether the length of time before beginning treatment is a risk factor for osteoporosis.
