Selective effects of nucleoside reverse transcriptase inhibitors (NRTI) and of protease inhibitors (PI) on adipose tissue differentiation and on lipid metabolism have been described 1 2 . More recently, attention was focused on loss of bone mineral density (BMD) recorded in HIV-infected patients 3 . Prevalence of osteoporosis and osteopenia was found about twice higher in HIV-patients than in seronegative controls 4 5 . In a large population-based study, occurrence of pathological fractures was more frequent in HIV-patients 6 . Several studies have also documented low levels of vitamin D in HIV-patients, while fragments of bone collagen degradation were found elevated 7 8 9 10 . However, contrasted observations have been made regarding other markers, like parathormone or osteocalcin 7 8 . A few longitudinal studies have addressed the issue of bone loss, but patients were already treated at baseline and time of follow-up was very variable, leading to contradictory conclusions 11 12 . In order to monitor early changes in body composition and BMD, we implemented a prospective study – LIPOTRIP -, following up treatment-naive HIV-infected patients, randomized to receive either a non-nucleosidic reverse transcriptase inhibitor (NNRTI) or a PI-based regimen. In parallel, we evaluated the changes in markers of lipid and bone metabolism and looked for correlations between changes in body composition and BMD. This study enables the description of differential effects of treatments on BMD, body composition and metabolic changes.

A prospective study was set up, following up naive HIV-infected patients commencing ART, enrolled between January 2004 and April 2006. Using a random number table, they were assigned to either one of two therapeutic schemes: first group patients were treated with two NRTIs and a ritonavir-boosted PI (“PI group”) while second group patients received two NRTIs and a NNRTI (“NNRTI group”). Physicians had some recommendations for preferred regimens: ZDV-3TC + EFV in NNRTI group and ZDV-3TC + any PI in the PI group. However, at the time the study started fosamprenavir / ritonavir was extensively used.

Inclusion criteria were adult HIV-infected patients (age > 18 years) naive of ART. Patients with current or recent (< 1 month) opportunistic diseases were excluded from the study as were those with diabetes mellitus, known dyslipidaemia, alcoholism, obesity, Cushing syndrome, thyroid disease, menopause, corticosteroid treatment and anyone who could not sustain an 8-hour fasting. All patients had signed informed consent and the protocol was approved by the local Ethics Committee (“Protection Committee of people for Biomedical Research”, Toulouse 1. France). At baseline and every three months thereafter, clinical and biological data were collected, including weight and BMI, CD4-cells and other blood cell counts, HIVviral load, haemoglobin level, ionogram and creatinine concentration, biomarkers of carbohydrate, lipid and bone metabolism. Body composition was evaluated by DXA at baseline and after 9 (M9) and 21 months (M21) on treatment. Seventy three patients were screened and 70 were enrolled in the study. Complete data at baseline, 9 and 21 months were available for 35 patients, as shown in Figure 1).

The relatively small proportion of patients for whom complete data were available at 21 months can be largely explained by the constraints of laboratory tests including a glucose tolerance test and because the site where the DXA was performed was quite distant from the consultation site.

The present study had a controlled prospective design, with treatment-naive patients randomly allocated to a first ART. It clearly shows that PI administration induces a significant decrease in lumbar spine BMD. Many cross-sectional studies and few longitudinal studies have investigated alterations of body composition in HIV-infected patients under various treatments. All those studies concluded to a low BMD in infected men and women, particularly when measured at the lumbar spine 3 12 14 15 . In a large French cohort of 35–50 years-old patients, prevalence of osteoporosis was 34% in men and 8% in women, and osteopenia was diagnosed in 50–55% 5 . Differences were also observed depending on ethnicity: osteopenia was found twice more frequent in Caucasians than in Africans 9 . Finally, a low BMD may have a clinical impact as prevalence of vertebral, hip or wrist fractures was 65% higher in HIV-patients than in healthy age-paired counterparts 6 . Components of the HIV itself, like Gp 120, might be involved in impairing osteoblast differentiation while promoting their apoptosis 16 . In the present study, baseline BMD was not lower than in the general population at the same age and in the same area. These results are not consistent with those published by others and ourselves 3 8 17 18 19 . These discrepancies can be explained by a shorter duration of HIV infection in the present study as compared with previous studies. It is recognised that HIV infection itself is a risk factor for osteopenia 16 . This effect seems to be correlated with the duration of infection as reported in other chronic infections (e.g. HBV and HCV infections) 20 .

Regarding impact of different classes of ART, conflicting data have been reported. While some studies claimed that low BMD is observed irrespective of treatments 21 , others have found a strong association with use of PI 17 22 , of NRTI 12 or of Tenofovir 14 . Large meta-analyses have concluded to a 2.4-fold higher risk of osteoporosis attributable to ART 16 17 . The study by Duvivier et al. 15 including 3 groups of patients (NNRTI + PI, 2 NRTI + NNRTI, 2 NRTI + PI) also concluded to a more pronounced decrease in lumbar spine BMD in patients receiving either PI-containing regimen as compared with the NNRTI and NRTI group. Except for the presence of a third group of patients treated with PI + NNRTI, the design of this study and the characteristics of the patients and groups were fairly close to those of our study. However, the longer duration of observation in our study (90 weeks vs 48 weeks) enables to demonstrate that the more pronounced decrease of BMD in PI-treated patients occurred during the first months of therapy, as also shown in a prospective study comparing the effects on bone density of regimens containing atazanavir or efavirenz 23 . After the first 9 months of treatment, BMD remained stable, whether patients received PI or NNRTI. Moreover, we had measurements of markers of bone turn-over and body composition and found correlations between these variables and the BMD decrease. Bone loss following ART initiation seems to coincide with immune reconstitution, which causes a high bone resorption 24 .

Although the measured variations are partly dependent on sex and ethnicity, this cannot explain the differences between our groups, as the proportions of Africans and of women were identical. However, this study confirms that BMD changes were more pronounced in Europeans than in Africans.

At a variance with changes in body composition, markers of bone build-up (bone ALP, osteocalcin) and resorption (β-cross-laps) were stimulated with either therapeutic scheme. This suggests that bone turnover is activated as soon as a minimal mineral loss has occurred. Cross-sectional studies have reported elevated levels of collagen degradation fragments and of osteoblast specific proteins in HIV-infected adults and children 25 26 27 , particularly when treated with PI 8 . Although parameters of bone turn-over were not discriminators between treatment groups, changes in osteocalcin and cross-laps were significantly associated with the magnitude of osteopenia making them useful markers to follow-up treated HIV-patients.

Experiments in cultured cells may help to elucidate the molecular mechanisms underlying bone alterations upon ART. Regarding osteoclast differentiation, several antiretroviral molecules have demonstrated an impact on the RANK ligand (receptor activator of NF-КB ligand) pathway (RANKL) leading to its sustained activation 28 29 . Interestingly, leptin concentration sharply increased upon PI. Leptin, acting through its hypothalamic receptor, plays an indirect role in bone resorption, through stimulation of the sympathetic tone. Activation of the β2-adrenergic receptor in osteoblasts stimulates RANKL expression 30 . Moreover, increase in leptin upon PI may reflect development of peri-visceral, insulin-resistant, adipose tissue, as suggested here by the increased waist circumference. Hyperleptinemia, correlated with insulin resistance has been recently reported in HIV-patients under ART 31 .

ln our study, baselevels of 25-OHD₃ were below recommended values. Further on, at 21 months upon ART, a 20% decline in 25-OH D₃ was recorded (p = 0.07 for the whole population). However, clinicians have not performed supplementation before the study because 25-OH D₃ levels were not measured before the study. Supplementation was anyway an exclusion criterion. Finally, clinicians were not really aware of vitamin D deficiency at that time. Since then, a relative vitamin D deficiency (< 20 ng/ml) has been reported in a large proportion of HIV-patients, worsening after 12 months of ART 32 . In this latter study, NNRTI were incriminated in the vitamin D decline. On the other hand, several PI were shown to impair synthesis of 25-OH D₃ and 1,25-diOH-D₃ in cells 33 34 . A recent hypothesis suggests that ART, like other xenobiotics, may also favour vitamin D degradation, through the CYP24 hydroxylase 35 36 . Moreover, the decline of 25-OH D₃ might up-regulate PTH synthesis 37 . In agreement with this, hyperparathyroidism has been reported upon ritonavir 38 . Apart from its role in bone resorption, PTH stimulates renal 1α-hydroxylase and has positive effects on osteoblast marker, like ALP and osteocalcin 39 40 . Thus, we may hypothesize that a low 25-OH D₃ would trigger secondary hyperparathyroidism, which, in turn, would stimulate synthesis of 1,25-di-OH D₃ and of osteoblast differentiation factors. Up-regulation of 1,25-di-OH D3 would be somewhat less pronounced upon PI, due to possible impact of those drugs on 25-OH D3 bioactivation 33 . Elevated calcemia would then result from increased PTH and 1,25-di-OH D3. Increased PTH might thus be a sensitive marker of bone remodelling and of vitamin D defect in treated HIV-patients.

The low metabolite concentrations observed at baseline may reflect a hypercatabolic situation in HIV-untreated patients corroborating previous observations 41 . Following ART introduction, glucose, lipids and albumin increased, likely reflecting an improvement of the energetic status associated with a better immuno-virological control. Large randomised trials have demonstrated different effects of ART on metabolic markers. A recent study showed the greatest increase in lipid parameters with a treatment combining PI and NNRTI 42 . In our study, most marked increases were for TG (+ 50%) with a specific effect of PI. PI-treated patients also experienced a homogeneous gain in fat mass both in trunk and lower limbs. Interestingly, other longitudinal studies have described increases in lean and fat masses during the first weeks upon ART, followed by a subsequent loss of limb fat after 2 years, 43 44 . Thus likely, our follow-up period was too short (< 2 years) to record any significant fat redistribution, except for a homogeneous gain. In our longitudinal study, we found that the decrease in BMD was correlated with increased BMI, the latter being mostly attributable to a gain in fat mass. We speculate that PI might switch the differentiation program of mesenchymal progenitor cells towards adipocytes, at the expanse of osteoblast formation, as already hypothesized 16 .

In conclusion, despite the small size of the studied population, our observations demonstrate an early effect of ART on lipid and bone metabolisms. More specifically, PI lead to a significant gain in fat mass correlated with a sharp drop in BMD but active bone remodelling is evident with all antiretroviral treatments, associated with low vitamin D levels and hyperparathyroidism. In parallel, signs of metabolic restoration are evident. However, early increases in lean and fat mass, triglycerides, waist circumference and leptin are much more pronounced with PI.

The authors declare that they have no competing interests.

ALG and JG carried out and interpreted all biochemical assays. LM monitored the study. FML and JB conducted and interpreted all DXA. JBR and CD participated in the design of the study and performed the statistical analysis. FB, FP and MO recruited and followed almost all patients in the study. BP and EB conceived the study, and participated in its design and coordination, and drafted the manuscript. PM and BM participated in the implementation of the study, monitoring its progress and the discussion of results. All authors read and approved the final manuscript.

Author to whom reprint requests should be addressed: Eric BONNET. Hôpital Joseph Ducuing. 15, rue Varsovie. 31000 Toulouse. France.