Bone Mineral Density and Fractures in Antiretroviral-Naive Persons Randomized to Receive Abacavir-Lamivudine or Tenofovir Disoproxil Fumarate-Emtricitabine Along With Efavirenz or Atazanavir-Ritonavir: AIDS Clinical Trials Group A5224s, a Substudy of ACTG A5202 - pdf attached
Download the PDF here|
The Journal of Infectious Diseases (15 June) 2011;203:1791-801
"Bone fracture was observed in 5.6% of participants."
Grace A. McComsey,1 Douglas Kitch,2 Eric S. Daar,5 Camlin Tierney,2 Nasreen C. Jahed,7 Pablo Tebas,8 Laurie Myers,9 Kathleen Melbourne,6 Belinda Ha,10 and Paul E. Sax3,4
1Departments of Pediatrics and Medicine, Rainbow Babies and Children's Hospital, Case Western Reserve University, Cleveland, Ohio; 2Harvard School of Public Health, 3Department of Medicine, Brigham and Women's Hospital, and 4Department of Medicine, Harvard Medical School, Boston, Massachusetts; 5Department of Medicine, Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, Torrance; and 6Gilead Sciences, Foster City, California; 7Social & Scientific Systems, Silver Spring, Maryland; 8Department of Medicine, University of Pennsylvania, Philadelphia; 9Frontier Science and Technology Research Foundation, Amherst, New York; and 10GlaxoSmithKline, Research Triangle Park, North Carolina
(See the editorial commentary by Yin and Overton.)
Background. Long-term effects of abacavir (ABC)-lamivudine (3TC), compared with tenofovir (TDF)-emtricitabine (FTC) with efavirenz (EFV) or atazanavir plus ritonavir (ATV/r), on bone mineral density (BMD) have not been analyzed.
Methods. A5224s was a substudy of A5202, in which HIV-infected treatment-naive participants were randomized and blinded to receive ABC-3TC or TDF-FTC with open-label EFV or ATV/r. Primary bone end points included Dual-emission X-ray absorbtiometry (DXA)-measured percent changes in spine and hip BMD at week 96. Primary analyses were intent-to-treat. Statistical tests used the factorial design and included linear regression, 2-sample t, log-rank, and Fisher's exact tests.
Results. Two hundred sixty-nine persons randomized to 4 arms of ABC-3TC or TDF-FTC with EFV or ATV/r. At baseline, 85% were male, and 47% were white non-Hispanic; the median HIV-1 RNA load was 4.6 log10 copies/mL, the median age was 38 years, the median weight was 76 kg, and the median CD4 cell count was 233 cells/μL.
At week 96, the mean percentage changes from baseline in spine and hip BMD for ABC-3TC versus TDF-FTC were -1.3% and -3.3% (P = .004) and -2.6% and -4.0% (P = .024), respectively; and for EFV versus ATV/r were -1.7% and -3.1% (P = .035) and -3.1% and -3.4% (P = .61), respectively. Bone fracture was observed in 5.6% of participants. The probability of bone fractures and time to first fracture were not different across components.
Conclusions. Compared with ABC-3TC, TDF-FTC-treated participants had significantly greater decreases in spine and hip BMD, whereas ATV/r led to more significant losses in spine, but not hip, BMD than EFV.
With the advent of potent antiretroviral therapy (ART), significant comorbidities have emerged, including osteoporosis and increased risk of fractures. Low bone mineral density (BMD) has been reported in studies of HIV-infected individuals; in a meta-analysis, the prevalence of osteoporosis was 3 times higher in HIV-infected patients than HIV-uninfected control subjects . Studies have shown that BMD decreases by 2%-6% within the first 2 years of ART initiation, regardless of the choice of therapy [2-5], with a long-term study showing that this initial decrease is not progressive . Studies reporting increased fracture rates in HIV-infected individuals are emerging [6-9].
Treatment with the nucleotide analogue reverse-transcriptase inhibitor tenofovir disoproxil fumarate (TDF) has been associated with an initial decrease in BMD . In addition, there was more bone loss in virologically suppressed persons who switched to TDF, compared with switching to the nucleoside analogue reverse-transcriptase inhibitor (NRTI) abacavir (ABC) . To date, there has been a single report of a 48-week prospective study of participants initiating their first ART with TDF-emtricitabine (FTC) or ABC-lamivudine (3TC), combined with the nonnucleoside reverse-transcriptase inhibitor (NNRTI) efavirenz (EFV) . A significantly greater decrease in spine and hip BMD was seen with TDF-FTC. To date, there has been no study comparing the effects on bone of EFV compared with those of atazanavir-ritonavir (ATV/r), a protease inhibitor (PI) combination with few metabolic effects [12, 13].
This report details changes in BMD in participants randomized to receive 1 of 4 frequently used regimens for treatment of HIV infection. As shown in prior studies, we demonstrated that ART initiation led to a large initial decrease in BMD, with ABC-3TC plus EFV being the only regimen studied that did not lead to a statistically significant decrease in spine BMD at week 96. We also found that TDF-FTC led to greater decreases in spine and hip BMD than did ABC-3TC and that ATV/r induced a significantly greater decrease in the spine BMD than did EFV. Our results are robust, because correcting for potential confounders and/or imbalances, including traditional bone risk factors and HIV disease characteristics, did not affect these results. AT analyses yielded results similar to those of ITT analyses. The incidence of fractures did not differ significantly between the regimen components.
The present study adds to a body of literature demonstrating a greater effect on reducing BMD with TDF-based therapies, compared with other regimens [11, 15]. A previous randomized clinical trial involving ART-naive participants compared TDF-FTC with ABC-3TC, both with EFV . At 48 weeks, decreases in spine and hip BMD were significantly greater with TDF-FTC. Our data are consistent with these results and extend the observation to 96 weeks and to the use of both EFV and ATV/r.
The role of PI therapy in HIV-associated osteoporosis has been debated. Our study revealed a greater decrease in BMD with ATV/r regimens, compared with EFV, but only at the spine. A trend toward greater decrease in total body BMD with another ritonavir-boosted PI (lopinavir/r), compared with EFV, was observed in another randomized trial . By contrast, other studies have not shown an effect of PI on BMD [17, 18]. Some of these discrepancies may be related to the use of whole-body DXA instead of using the more sensitive site-specific bone DXAs. Our study showed that the effect of ART varies by site, supporting the use of site-specific DXA. This site differential effect could be attributable to the trabecular nature of vertebral bone, which is more active and more subject to bone turnover and remodeling, compared with cortical (eg, hip) bone. In addition, different PIs may have differential effects on bone, analogous to their variable effect in the drug class on lipid changes.
The mechanisms involved in bone loss after initiation of ART are not well understood. TDF may affect bone through proximal tubule toxicity, resulting in phosphate wasting and increased bone turnover . EFV and PIs may affect BMD indirectly through vitamin D metabolism [20-26]. In multivariable analysis exploring the factors associated with BMD changes at 96 weeks, we found that, in addition to TDF-FTC and ATV/r each leading to greater spine BMD decrease, compared with ABC-3TC and EFV, respectively (also in hip BMD for TDF-FTC), other baseline factors were associated with BMD loss. Some of these (eg, lower BMI) are also associated with BMD decreases in the general population. Relevant HIV-specific factors that decreased BMD include higher baseline HIV-1 RNA load and lower CD4 cell count, corroborating findings of other studies showing a greater risk of osteopenia and/or osteoporosis in those with longer HIV infection duration [27-29]. These observations support the fact that HIV infection or immunologic factors linked to HIV infection play a role in bone loss after treatment initiation. Indeed, HIV proteins can increase osteoclastic activity  and promote osteoblast apoptosis [31, 32]. Furthermore, cytokines, such as IL-6 and TNF-α, may stimulate osteoclast activity [33-35].
Because most of our study participants were young (median age, 38 years), with a relatively low risk of falls, it was not surprising that we did not observe an increased rate of fractures with specific ART regimens. However, the degree of BMD loss and the between-component differences should not be perceived as clinically insignificant. Indeed, these decreases are similar in magnitude to the BMD losses sustained during the first 2 years of menopause . Furthermore, our study population was young and mostly (85%) male, a group typically spared significant loss of BMD. In the general population, the mean 2-year change in BMD in men 20-49 years of age is -0.8% at the hip and -0.3% at the lumbar spine . Even at the most vulnerable skeletal time in women (during the first 2 years of menopause), the loss of BMD accelerates, with mean annual rates of bone loss of 1.2% -1.6%. This magnitude of bone loss is equivalent to the point estimates of the mean differences shown between the ABC-3TC and TDF-FTC regimens at the hip and spine and between EFV and ATV/r at the spine, although the confidence intervals are consistent with smaller differences in the means.
Our study is notable for the observations regarding timing of BMD changes after ART initiation. Large early reductions in spine and hip BMD were observed within the first 48 weeks after ART initiation with all regimens. After the initial 48 weeks (cutoff chosen by inspection), BMD did not change or even improved slightly with some of the regimens. This is consistent with prior ART initiation studies and with longitudinal studies of treatment-experienced participants that have shown stability in BMD over time [3, 28, 38, 39].
Our study has several limitations. First, the duration of follow-up for study of bone end points was relatively short. Nevertheless, to our knowledge, our study has the longest follow-up of published prospective longitudinal studies of BMD after ART initiation. Second, the changes in the NRTI backbone of the regimen that resulted from the outcome of the DSMB review of A5202 were relatively frequent. However, our ITT results were consistent with our AT results. Other limitations are that the NNRTI-PI component was provided in an open-labeled fashion and that there was a high amount of missing data, which is not unusual for large multicentered studies. Finally, the study did not collect smoking and alcohol status, which could affect BMD.
In conclusion, we revealed that the initiation of ART leads to prompt reductions in spine and hip BMD observed within the first 48 weeks, independent of ART type. At week 96, TDF-FTC, both in the spine and hip, and ATV/r in the spine produced significantly more bone loss than did ABC-3TC- or EFV-based regimens. Studies investigating the mechanisms behind the bone loss with ART initiation are needed.
A total of 271 participants from 37 ACTG sites in the United States and Puerto Rico intended to participate in A5224s and were randomized to receive ART; of these, 2 were excluded from the analysis when found to have had an eligibility violation. Enrollment spanned from 5 October 2005 through 7 November 2007 with 69 participants randomized to receive EFV plus TDF-FTC, 70 to EFV plus ABC-3TC, 65 to ATV/r plus TDF-FTC, and 65 to ATV/r plus ABC-3TC. Baseline characteristics are summarized in Table 1. Overall, 85% of participants were male and 47% were non-Hispanic white persons. The median age was 38 years, body mass index (BMI; measured as the weight in kilograms divided by the square of the height in meters) was 24.9, CD4 cell count was 233 cells/μL, and HIV-1 RNA load was 4.62 log10 copies/mL. One hundred sixty participants (59%) enrolled had an HIV-1 RNA load <100,000 copies/mL at study screening. Overall, 3% were hepatitis B surface antigen positive, 9% was hepatitis C antibody positive, 32% reported a history of fracture, and 39% had osteopenia (t score -1 at spine or hip) at study entry. The baseline characteristics were balanced across arms.
he baseline characteristics of the A5224s participants were compared with those of the 1588 A5202 persons who did not participate in the substudy; no statistically significant differences were found for age, BMI, CD4 cell count, HIV-1 RNA load, or history of fractures. However the non-A5224s group included significantly more Hispanic persons (24% vs 16%; P = .005).
Figure 1 details the disposition of all participants. Overall, 66 (25%) of the A5224s participants prematurely discontinued the substudy, and 4 (1%) died. In addition, 31 participants (12%) discontinued, because their sites were defunded during the study. There was no statistically significant difference in time to premature study discontinuation between NRTI components (P = .13, site closure and death censored) or NNRTI-PI components (P = .86). The median time from randomization to the last clinic visit was 165 weeks.
Percentage Changes in Spine BMD.
The first coprimary analysis assessed the difference in mean percentage change in spine BMD at week 96 between ABC-3TC and TDF-FTC. Table 2 summarizes the estimated mean percentage change over time in spine and hip BMD by all regimens. Figures 2 and 3 plot the mean percentage change over time in spine and hip BMD by NRTI and NNRTI-PI components.
The estimated mean percentage change in spine BMD for all participants was -3.0% at week 48 and -2.3% at week 96. The comparison of ABC-3TC (n = 135) and TDF-FTC (n = 134) with EFV and ATV/r combined (factorial analysis) was performed, because there was no significant evidence that the treatment effect between these drugs differed at 96 weeks by the NNRTI-PI component (P = .63). Similarly, the comparison of EFV (n = 139) and ATV/r (n = 130) with ABC-3TC and TDF-FTC combined was performed.
Changes by NRTI Components: Primary Analysis.
By ITT at week 96, there was a significant decrease in mean percentage change in spine BMD for all arms except ABC-3TC plus EFV, but significantly less for ABC-3TC (estimated mean of -1.3%) than for TDF-FTC (-3.3%; difference [∼] = 2.0%; 95% confidence interval [CI], .7%-3.3%; P = .004). The AT analysis showed similar results, with the mean percentage change in ABC-3TC- and TDF-FTC-treated participants being -1.0% and -3.2%, (∼ = 2.2%; 95% CI, .6%-3.7%; P = .006). The difference between the NRTI components in the mean percentage change in spine BMD was already evident at week 48, at which point the ABC-3TC arms had an estimated mean percentage change of 1.6% (95% CI, .5%-2.8%) smaller than that in the TDF-FTC arms (P = .005).
At week 96, among participants assigned to receive EFV, there was a trend toward a greater decrease in mean percentage change in spine BMD when combined with TDF-FTC than when combined with ABC-3TC (∼, 1.7%; 95% CI, .04%-3.5%; P = .056). In ATV/r-treated arms, there was a significantly greater decrease in mean percentage change in spine BMD when combined with TDF-FTC than when combined with ABC/3TC (∼, 2.4%; 95% CI, .4%-4.4%; P = .020, by ITT).
Changes by NNRTI-PI Component: Secondary Analysis.
At week 96, by ITT analysis, the mean percentage change in spine BMD was significantly greater in those assigned to ATV/r (-3.1%) than in those in the EFV arm (-1.7%; ∼, -1.5%; 95% CI, -2.8% to -.1%; P = .035). Similar results were seen in the AT analysis. However, at 48 weeks, the mean percentage change was not significantly different between those treated with ATV/r (-3.5%) and those treated with EFV (-2.6%; ∼ = -.9%; 95% CI, -2.1% to .2%; P = .11).
Percentage Changes in Hip BMD
The second coprimary analysis involved the mean percentage change in hip BMD at week 96 between the ABC-3TC and the TDF-FTC arms. The estimated mean percentage change in hip BMD for all participants was -3.3% at both weeks 48 and 96 (Table 2). A comparison of ABC-3TC (n = 135) and TDF-FTC (n = 134) with EFV and ATV/r combined was performed, because there was no significant evidence that the treatment effect between these drugs differed at 96 weeks by the NNRTI-PI component (P = .69). Similarly, a comparison of EFV (n = 139) and ATV/r (n = 130) with ABC-3TC and TDF-FTC combined was performed.
Changes by NRTI Components: Primary Analysis.
At week 96, ITT analysis showed that the ABC-3TC arms had a significantly smaller decrease in mean percentage change in hip BMD, compared with the TDF-FTC arms (-2.6% vs -4.0%; ∼, 1.4%; 95% CI, .2%-2.5%; P = .024). The AT analysis showed similar results; at week 96, the mean percentage change in hip BMD in the the ABC-3TC arms was -2.6%, compared with -3.9% for TDF-FTC (∼, = 1.3%; 95% CI, .02%-2.6%; P = .046). The difference between the NRTI components in the mean percentage change in hip BMD was already evident at week 48, with an estimated mean change of -2.6% for ABC-3TC and -4.1% for TDF-FTC (∼, 1.5%; 95% CI, .5%-2.5%; P = .003).
For persons assigned to receive EFV, at 96 weeks, the mean percentage change in hip BMD was not statistically significantly different between the NRTI components, compared with those assigned to receive ABC-3TC; the estimated mean change was -2.5%, compared with -3.7% for those given TDF-FTC (∼, 1.2%; 95% CI, -.4% to 2.7%; P = .15). There was a trend toward a smaller decrease in mean percentage change in hip BMD for persons given ATV/r with ABC-3TC (-2.7%), compared with those given TDF-FTC (-4.3%; ∼, 1.6%; 95% CI, .2%-3.4%; P = .075).
Changes by NNRTI-PI Component: Secondary Analysis.
At week 96 and by ITT analysis, the mean percnetage change in hip BMD was not statistically significantly different between EFV and ATV/r (∼, -.3%; 95% CI, -1.5% to .9%; P = .61). Similar results were seen in the AT analysis and at week 48.
Changes in Spine and Hip BMD Adjusted for Baseline Covariates
The ITT analyses of mean percentage change from entry to week 96 of spine and hip BMD were adjusted for the following prespecified baseline covariates that could affect BMD, first individually and then jointly, with use of linear regression: NNRTI-PI (or NRTI components for the NNRTI-PI analyses), spine BMD (or hip BMD for corresponding analysis), sex, age, race/ethnicity, log10 HIV-1 RNA load, CD4 cell count, and BMI. For analyses of the NRTI component effect or the NNRTI-PI component effect, all of the adjusted models led to results similar to those of the unadjusted analyses.
Association Between Baseline Factors and Changes in BMD at 96 Weeks
Table 3 summarizes the linear regression analyses that were performed to assess the baseline factors associated with 96-week percentage change in spine and hip BMD. The covariates included in the model were the same as the ones mentioned in the previous paragraph. For spine BMD, in addition to the significant ABC-3TC and ATV/r effects, in both univariate and multivariable models, higher baseline CD4 cell count was independently associated with significant increases, and higher baseline log10 HIV-1 RNA load was independently associated with significant decreases in spine BMD at 96 weeks. For hip BMD, in addition to the significant ABC-3TC effect, in univariate and multivariable models, higher baseline BMI was independently associated with significant increases at 96 weeks.
Timing of BMD Changes: Repeated Measures Analyses.
To understand the dynamics of BMD change over time, an analysis of the slopes of changes in the early phase (0-48 weeks) and late phase (48-192 weeks) was explored in and between study components. For spine BMD, as shown in Table 4, there was a statistically significant difference between the NRTIs in the slope of BMD change during both the early and the late phase, favoring ABC-3TC. Of interest, ABC-3TC arms, but not TDF-FTC, had a significant positive spine BMD percentage change per year during the late phase. For NNRTI-PI components, there was no statistically significant difference in the slopes between NNRTI and PI arms in either phase, with both arms having decreasing BMD during the early phase and only EFV significantly increasing spine BMD in the late phase.
For hip BMD, the treatment differences and kinetics of bone loss were similar, with most of the BMD loss occurring during the first 48 weeks in both NRTI arms. During the late phase, ABC-3TC arms again showed a significant gain in hip BMD. Both EFV and ATV/r arms lost bone in the first phase, but the slope of the late phase did not reach statistical significance in either arms.
On-Study Bone Fractures
On-study bone fractures were collected in A5224s and in the A5202 parent study (n = 1857). In the substudy, 15 participants (5.6%) reported a bone fracture, all of which were a result of trauma, with 10 occurring in the EFV arms. There were no statistically significant differences in the number of fractures between the NRTIs (P = 1.00) or the NNRTI and PI study arms (P = .29). Similarly, there was no statistically significant difference in time to first bone fracture between NRTI (P = .76) or NNRTI/PI study arms (P = .27).
In the parent study-A5202, 80 participants (4.3%) reported at least one bone fracture on study (ABC-3TC plus EFV, 4.7%; ABC-3TC plus ATV/r, 3.5%; TDF-FTC plus EFV, 4.5%; and TDF-FTC plus ATV/r, 4.5%). Among these, 10 (12.7%) were atraumatic. The bone fractures were balanced across the study arms, with no statistically significant differences between the NRTI (P = .73) or the NNRTI and PI components (P = .57). No statistically significant difference in time to first bone fracture was seen between the NRTIs (P = .71) or the NNRTI and PI components (P = .49). Similarly, incidence rates were similar across arms (ABC-3TC plus EFV, 1.9 cases per 100 patient-years; ABC-3TC plus ATV/r, 1.4 cases per 100 patient-years; TDF-FTC plus EFV, 1.8 cases per 100 patient-years; and TDF-FTC plus ATV/r, 1.8 cases per 100 patient-years).