Two recent studies have indicated that bone mineral density (BMD) is reduced in HIV-infected adult males receiving highly active antiretroviral therapy (HAART) containing protease inhibitors (PI), raising concerns about premature osteoporosis in the long term [1,2]. In both studies, cross-sectional analysis of dual energy X-ray absorptiometry (DEXA) scans demonstrated a higher prevalence of lumbar spine osteopenia and osteoporosis in patients on PI compared with those on non-PI therapy.
In this study of male participants in the Western Australian HIV Cohort study, we sought to determine the relative independent effects of antiretroviral therapies as well as other environmental and disease related factors on BMD over time.
Comprehensive demographic, clinical and laboratory data are collected routinely on all participants of the Western Australian HIV Cohort Study which was established in 1983 (described in ). The variables relevant to this study include: age, race, weight, body mass index (BMI), duration of AIDS, Centers for Disease Control and Prevention (CDC) stage, all HIV and non-HIV related illnesses including Mycobacterium avium complex and cytomegalovirus infection, history of tobacco smoking (pack years), alcohol intake (g/day), history of antiretroviral drugs, history of prophylactic and non-HIV related medications, serial CD4 and CD8 T-cell count, plasma HIV RNA concentration (HIV Amplicor, Roche, Branchburg, New Jersey, USA, limit of detection < 400 copies/ml until November 1999; then Roche Amplicor HIV monitor Version 1.5, Ultrasensitive, limit of detection < 50 copies/ml), full blood analysis, serum biochemistry including venous lactate, bilirubin, triglyceride and alkaline phosphatase concentrations. All laboratory-derived data are electronically downloaded into the cohort study database.
A DEXA scanner (Hologic QDR-4500A machine Hologic, Inc., Waltham, Massachusetts, USA), running enhanced array whole body software (version 8.23a:3) and regional array software (version 8.26f:3), was used to perform whole body scans on participants of the Western Australian HIV Cohort Study from May 1998 to June 2000 as part of routine 6-monthly monitoring of changes in body composition. The percentage fat at the arms, legs and central abdomen was recorded. We calculated lumbar spine t-scores [(measured BMD − population mean BMD at age 30 years)/standard deviation (BMD at age 30 years)] and lumbar spine z-scores [(measured BMD − population mean BMD for age-matched subject)/standard deviation (BMD at same age)] based on population normative data for Caucasian males (Hologic 4/11/91). As lumbar spine (L1-L4) BMD measurements were derived from whole body scans rather than the localized DEXA scans used routinely in bone density monitoring, the measurements were first adjusted in line with the findings of a calibration analysis. Based on 154 participants of a rheumatoid arthritis study, the calibration analysis found that the whole body scan measurements were directly proportional to, but slightly lower than, those obtained from localized scans (R2, 0.88; proportionality factor, 1.016). Whilst the use of whole body DEXA scans may introduce additional variability in the measurements, as a result of the smaller image, no systematic biases are introduced into analyses of the data. Osteopenia and osteoporosis were defined using World Health Organization criteria (osteopenia, t-score between − 2.5 and −1; osteoporosis, z-score < −2 or t-score < −2.5). Analysis was restricted to Caucasian males, to facilitate comparison with gender- and race-specific population normative data.
Longitudinal data from 54 Caucasian males who had serial DEXA scans were analysed, representing all participants in the Western Australia HIV Cohort study who fulfilled inclusion criteria. Analysis was restricted to those who had sequential scans while on stable nelfinavir- or indinavir-containing HAART therapy, as there was insufficient data from other PI therapies (e.g. ritonavir and saquinavir) for analysis. All HAART regimens included two nucleoside reverse transcriptase inhibitors (NRTI) and a PI and were of the form: zidovudine/stavudine + lamivudine/didanosine + nelfinavir (n = 20), or zidovudine/stavudine + lamivudine/didanosine + indinavir (n = 34). All patients in the longitudinal analysis had their first DEXA scan at least 1 month after change of therapy, had no change in therapy between scans and did not have active malignancy, systemic infection, or prolonged bed rest. No patients were receiving medications likely to affect bone metabolism, specifically calcium, vitamin D, calcitriol, testosterone, bisphosphanates or thiazide diuretics. In addition, no patients were receiving simvastatin or lovastatin, therapies that may increase bone formation . In the longitudinal analysis, the mean number of days between first and last DEXA scan was 377 days (SD, 178 days). Indinavir and nelfinavir groups were well matched for age (42.92 ± 8.82 years versus 43.35 ± 10.89 years;P = 0.88), BMI prior to HAART (24.35 ± 4.11 kg/m ≤ versus 22.95 ± 2.42 kg/m ≤;P = 0.17), and average CD4 T-cell count over the time period considered (548 ± 230 × 106/l versus 557 ± 280 × 106/l;P = 0.89).
A cross-sectional analysis comparing lumbar spine z-scores in those currently treated with PI (n = 131) with those who have never had any PI exposure (n = 52) was performed. Lumbar spine z-scores and t-scores in 28 patients who were either treatment naive or had less than 1 month of any antiretroviral therapy at the time of DEXA scanning were also calculated.
Bone formation was assessed biochemically by measuring osteocalcin levels (immunoradiometric assay, Cis Elsa-osteo). We studied nine indinavir-treated and eight nelfinavir-treated patients who were antiretroviral therapy-naive prior to commencing PI treatment, and who had asymptomatic HIV infection with no history of significant opportunistic infection. Assays were performed on stored (−20°C) plasma samples, and the change in osteocalcin level from prior to commencement of therapy to 6 months following PI introduction was assessed for each individual.
For the longitudinal analysis, data obtained from DEXA scans (% leg fat, lumbar spine z-score and lumbar spine BMD) were displayed by creating individual patient plots, using HAART introduction as the time origin. Estimation and analysis of average trends in each of the plots was then undertaken by fitting linear mixed effects models, using the S-PLUS 2000 statistical package (MathSoft, Seattle, Washington, USA). Although there may be considerable measurement error associated with any single value, in these models it is assumed that the measurements of each individual fluctuate about an underlying trend line, with the individual trend lines varying randomly about the population average. Variables were tested by the application of Wald tests, providing assessment of factors that influence the rate of change over time. Factors influencing the displacement of the individual trend lines from the population average trend (i.e. factors which determine an individual's initial measurement without affecting subsequent rate of change) were also assessed.
Cross-sectional comparisons and multiple linear regression analyses were carried out using standard statistical software (SAS Statistical Package v6.12, SAS Institute, Cary, North Carolina, USA).
Estimation and analysis of average trends was undertaken by fitting linear mixed effects models as described earlier. Assessment of factors influencing the rate of change of bone density revealed an association between indinavir therapy and increasing BMD over time. In addition, analysis of factors influencing initial z-scores (displacement of the individual trend from the population trend) revealed a significant association between lower pre-HAART BMI and lower initial z-score (P = 0.003). After taking into account the effects of indinavir and BMI we found no other variables that were significantly associated with either initial z-score or change in z-score over time. These included choice of NRTI (zidovudine versus stavudine), nadir CD4 T-cell count, alcohol intake, % leg fat and average serum levels of bilirubin, lactate or triglyceride over the time period considered. BMI prior to HAART was not associated with change in z-score over time (P = 0.24).
Effect of PI therapy
The plots in Fig. 1 show the profiles of % leg fat, lumbar spine z-scores, and lumbar spine BMD (g/cm≤) for each individual, using HAART introduction as the time origin, together with the estimated population-average trend lines. We observed a significant loss of mean subcutaneous fat (expressed as % leg fat) as a result of HAART-associated lipodystrophy (P = 0.0001, indinavir;P = 0.02, nelfinavir). However, despite an association between pre-HAART BMI and lumbar spine z-score, we observed no concurrent loss in average lumbar spine z- scores and BMD over the range of DEXA scans considered. While there was little evidence of a change in average z-score over time in patients receiving nelfinavir (P = 0.92), there was a significant increase in the indinavir group z- scores over time (average increase, 0.31 z-score/year, P < 0.0001) with a significant difference between the two groups (P = 0.012).
Effect of PI therapy and BMI in cross-sectional analysis
Lumbar spine z-scores in patients who were on PI therapy at the time of first DEXA scan (n = 131; mean ± SE, − 0.969 ± 0.093) were significantly lower than those of PI-naive patients (n = 52; mean ± SE, − 0.609 ± 0.156;P = 0.043). Mean z-score was significantly lower than zero in both PI-naive (P < 0.001) and PI-treated (P < 0.001) patients, and although the prevalence of osteoporosis in the PI-treated group was higher than that in the PI-naive group (18.3% versus 7.7%), prevalence of osteopenia was comparable (38.2% and 40%, respectively) between the two groups.
Examination of the effect of BMI on BMD revealed an association between lumbar spine z- score and lowest BMI prior to first DEXA scan (P = 0.020). In a multiple linear regression analysis with adjustment for the effect of BMI, the difference between PI-naive and PI-treated patient groups lost significance (P = 0.11), whereas the BMI effect remained significant (P = 0.04).
Lumbar spine z-scores in HIV-infected patients with little or no exposure to HAART
Lumbar spine z-scores in these patients were significantly less than zero (mean ± SE, −0.58 ± 0.23;P = 0.018). Of the 28 patients who were either treatment-naive or had less than 1 month of exposure to HAART at the time of their DEXA scan, four fulfilled standard criteria for osteoporosis (14%), and nine (32%) were osteopenic.
Mean osteocalcin levels (± SD) prior to therapy in the indinavir group (15.0 ± 5.8 mg/l) and nelfinavir group (15.4 ± 2.9 mg/) were not significantly different (P = 0.85, t test), and were within the reference range for males of 11.3–34.8 mg/l. Rates of change in osteocalcin were calculated to adjust for variable time periods between the first and second tests (mean ± SD, 6.24 ± 2.9 months for the indinavir group; mean ± SD, 6.26 ± 1.2 months for the nelfinavir group). There was no significant change in osteocalcin levels over time in the nelfinavir group (P = 0.31). There was, however, a significant positive rate of change of osteocalcin in the indinavir group (mean increase, 5.01 mg/l per 6 months, P = 0.04).
In this study, we found no evidence of accelerated average bone loss in male HIV-infected patients treated with nelfinavir- or indinavir-containing HAART regimens, and propose that indinavir therapy may be associated with increased bone formation over time. The analyses showed BMI to be an independent and powerful determinant of an individual's initial z- score, and following adjustment for this effect in a multivariate cross-sectional analysis there was no independent association between PI therapy and decreased BMD. In the longitudinal analysis, BMD remained stable or increased over time despite concurrent subcutaneous fat wasting, suggesting that loss of fat mass is not predictive of loss of BMD in male patients affected by lipodystrophy.
Whereas the longitudinal analysis indicated that antiretroviral therapy was not associated with accelerated bone loss, lower pre-HAART BMI was found to correlate with decreased BMD in both the longitudinal and cross-sectional components of the study. This association between body mass and bone density is consistent with data obtained from the general population, where studies of body composition in young adult males [5–7] and elderly males  has shown a consistent correlation between body mass and bone density that is determined by lean body mass rather than by fat mass. In pre-menopausal and post-menopausal women, however, bone density appears to be influenced by fat mass as well as by lean body mass [7,8].
In the context of HIV infection, studies of body composition in HIV infected individuals in the pre-HAART era have demonstrated that adverse clinical outcomes associated with uncontrolled HIV infection lead to malnutrition with prominent depletion of lean body mass . By contrast, use of HAART appears to have a beneficial effect on lean body mass despite a propensity to cause concurrent loss of fat mass as a result of the HAART-associated lipodystrophy syndrome . It is likely, given the strong association between pre-HAART BMI and initial lumbar spine BMD in this study, and the maintenance of BMD in both PI treatment groups despite subcutaneous fat wasting, that the effect of BMI in this analysis reflects the influence of lean body mass. The dominant effect of pre-HAART BMI in determining BMD in the cross-sectional analysis appears to account for PI versus non-PI differences, and suggests the possibility that PI may be used in those with more advanced HIV disease in preference to non-nucleoside analogue reverse transcriptase inhibitors or triple NRTI regimens. Thus, it is possible that PI therapy acts as a surrogate marker of other osteoporosis risk factors in univariate, cross-sectional analysis.
This study found increased average lumbar spine BMD and osteocalcin levels over time were associated with indinavir use, providing evidence of increased bone formation induced by this therapy. It is important, however, to recall that this effect was observed for the average of the study population and that the magnitude of this effect cannot be extrapolated to an individual case. An explanation for such an effect has been provided by in vitro studies of mesenchymal stem cell differentiation, which were originally designed to examine the pathogenesis of the lipodystrophy syndrome associated with HAART. Lenhard and colleagues  demonstrated that indinavir (but not amprenavir, nelfinavir, ritonavir or saquinavir) inhibits the differentiation of stem cells (C3H10T1/2) into adipocytes, in a manner dependent on the presence of vitamin A1 acid (all-trans-retinoic acid, or ATRA). This was associated with increased expression of the osteoblast-specific marker alkaline phosphatase, as well as the induction of morhphological changes in stem cells typical of osteogenic rather than adipogenic differentiation. The effect appears to be mediated by increased retinoic acid receptor (RAR) signalling, and could be inhibited by a specific RAR antagonist (AGN193109). Thus indinavir may promote osteogenic differentiation of mesenchymal stem cells by stimulating retinoid signalling. In this study, the increase in osteocalcin concentration associated with indinavir therapy also supports an increase in osteoblastic activity rather than suppression of bone resorption as the cause of the increase in bone density.
These results are in keeping with recently developed models of osteoporosis pathogenesis. Following the observation that osteoblasts and adipocytes share a common progenitor arising from bone marrow [12–14], and that loss of osteogenic cells in osteoporotic bone is accompanied by increased adipocytes [14–16], the ‘lipid hypothesis of osteoporosis’ has been developed . The basic tenet of this hypothesis is that the loss of BMD that characterizes osteoporosis reflects a shift in phenotype of marrow-derived stem cells, where adipogenesis increases at the expense of osteogenic differentiation. The cellular signalling mechanisms that underlie the choice of differentiation pathway involve, in part, two families of nuclear transcription factors. Increased levels of retinoids stimulate osteogenic differentiation through the activation of RAR . Opposing this activity are factors that stimulate adipogenesis by activating the peroxisome proliferator activated receptor (PPAR) family of transcription factors (PPAR-γ and PPAR-α), such as free fatty acids , low density lipoprotein oxidation products , and anti-diabetic thiazolidinediones . The balance between these factors appears to determine the fate of mesenchymal precursors. These in vitro studies indicate that RAR signalling is increased in the presence of indinavir, resulting in increased osteogenic differentiation.
When considering the potential benefits of indinavir therapy on bone metabolism, caution must be exercised in associating increased lumbar spine BMD with improved bone strength and decreased fracture risk. While this therapy appears to enhance osteoblast differentiation and bone formation, it must be borne in mind that the only currently available therapeutic agent that specifically stimulates bone formation – fluoride – increases bone mass without necessarily increasing bone strength or reducing the risk of subsequent bone fractures . In this context, bone strength appears to reflect not only bone mass but also its architectural properties, which are determined by factors that influence mechanical load on bone such as physical activity and muscle strength .
More longitudinal data is required to confirm the findings of this study. Ideally, future studies should be based on data obtained from localized DEXA scans where there is greater measurement precision, so that changes in bone density may be assessed with greater accuracy. The average coefficient of variation that was calculated in this study (4.4%) is greater than is seen with localized DEXA scanning in our department (2%). This variability does not introduce systematic bias that would invalidate the conclusions of the study in terms of average trends, but it does reduce the ability to ascribe changes in bone density as measured in individuals to interventions (such as indinavir) rather than to measurement imprecision. We were unable to investigate the influence of other PI on bone density, and given the differences described between indinavir and nelfinavir it is important to emphasise that these results may not be generalized for PI that were not studied. It is notable in this context that there was a relatively high use of ritonavir (24%) and saquinavir (22%), in addition to indinavir (27%) and nelfinavir (27%), in the group of patients studied by Tebas et al., in which PI use was associated with decreased lumbar spine BMD. Longitudinal studies of BMD in patients receiving ritonavir- and saquinavir-based HAART are therefore warranted. We also propose that the factors that influence BMD in males and females may be significantly different, particularly in relation to the association between fat mass and bone density. While the overall loss of subcutaneous fat mass did not coincide with a corresponding loss of bone density over the time period studied, the same may not be true for women where fat mass is known to be a determinant of bone density .
While it has been proposed on the basis of cross-sectional studies that long-term use of HAART may accelerate bone density loss and induce premature osteoporosis, the prospective longitudinal data presented in this study indicate that indinavir- and nelfinavir-based HAART, at least, is not associated with bone loss in HIV-infected men. Moreover, the findings add to the increasing evidence that different PI differ in their metabolic effects, and that the contributions of individual PI should be considered in studies examining the long-term complications of antiretroviral therapy.
1. Tebas P, Powderly WG, Claxton S. et al
. Accelerated bone mineral loss in HIV-infected patients receiving potent antiretroviral therapy. AIDS 2000, 14: F63–F67.
2. Hoy J, Hudson J, Law M, et al.Osteopenia in a randomised, multicenter study of protease inhibitor (PI) substitution in patients with the lipodystrophy syndrome and well-controlled HIV viremia.Seventh Conference on Retroviruses and Opportunistic Infections.
San Francisco, January–February 2000 [abstract 59].
3. Mallal S. The Western Australian HIV Cohort Study, Perth, Australia. J Acquir Immune Defic Syndr Hum Retrovirol 1998, 17 (Suppl 1): S23–S27.
4. Mundy G, Garrett R, Harris S. et al
. Stimulation of bone formation in vitro
and in rodents by statins. Science 1999, 286: 1946–1949.
5. Ellis KJ. Body composition of a young, multiethnic, male population. Am J Clin Nutr 1997, 66: 1323–1331.
6. Kroger H, Laitinen K. Bone mineral density measured by dual-energy X-ray absorptiometry in normal men. Eur J Clin Invest 1992, 22: 454–460.
7. Reid IA, Plank LD, Evans MC. Fat mass is an important determinant of whole body bone density in premenopausal women but not in men. J Clin Endocrinol Metab 1992, 75: 779–782.
8. Baumgartner RN, Stauber PM, Koehler KM, Romero L, Garry PJ. Associations of fat and muscle masses with bone mineral in elderly men and women. Am J Clin Nutr 1996, 63: 365–372.
9. Kotler DP. Body composition studies in HIV-infected individuals. Ann N Y Acad Sci 2000, 904: 546–552.
10. Pernerstorfer-Schoen H, Schindler K, Parschalk B. et al
. Beneficial effects of protease inhibitors on body composition and energy expenditure: a comparison between HIV-infected and AIDS patients. AIDS 1999, 13: 2389–2396.
11. Lenhard JM, Weiel JE, Paulik MA, Furfine ES. Stimulation of vitamin A1 acid signalling by the HIV protease inhibitor indinavir. Biochem Pharmacol 2000, 59: 1063–1068.
12. Bruder SP, Fink DJ, Caplan AI. Mesenchymal stem cells in bone development, bone repair, and skeletal regeneration therapy. J Cell Biochem 1994, 56: 283–294.
13. Nuttall ME, Patton AJ, Olivera DL, Nadeau DP, Gowen M. Human trabecular bone cells are able to express both osteoblastic and adipocytic phenotype: implications for osteopenic disorders. J Bone Miner Res 1998, 13: 371–382.
14. Parhami F, Jackson SM, Tintut Y. et al
. Atherogenic diet and minimally oxidised low density lipoprotein inhibit osteogenic and promote adipogenic differentiation of marrow stromal cells. J Bone Miner Res 1999, 14: 2067–2078.
15. Meunier P, Aaron J, Edouard C, Vignon G. Osteoporosis and the replacement of cell populations of the marrow by adipose tissue: A quantitative study of 84 iliac bone biopsies. Clin Orthop 1971, 80: 147–154.
16. Majors AK, Boehm CA, Nitto H, Midura RJ, Muschler GF. Characterization of human bone marrow stromal cells with respect to osteoblastic differentiation. J Orthop Res 1997, 15: 546–557.
17. Diascro D Jr, Vogel RL, Johnson TE. et al
. High fatty acid content in rabbit serum is responsible for the differentiation of osteoblasts into adipocyte-like cells. J Bone Miner Res 1998, 13: 96–106.
18. Paulik M, Lenhard JM. Thiazolidinediones inhibit alkaline phosphatase activity while increasing expression of uncoupling protein, deiodinase, and increasing mitochondrial mass in C3H10T1/2 cells. Cell Tissue Res 1997, 290: 79–87.
19. Haguenauer D, Welch V, Shea B, Tugwell P, Wells G. Fluoride for treating osteoporosis (Cochrane Review). Cochrane Database Syst Rev 2000, 4: CD002825.CD002825.
20. Frost HM. Obesity, and bone strength and ‘mass': a tutorial based on insights from a new paradigm. Bone 1997, 21: 211–214.