In this SMART substudy of 214 participants with mean follow-up of 2.4 years, BMD steadily declined in the group receiving continuous ART, whereas BMD remained stable or increased in the first year of intermittent, CD4 cell count-guided ART. For all three BMD measures, the between-group differences were statistically significant at year 1 and persisted through follow-up. Continuous ART may also have been associated with higher risk of fracture than intermittent ART. The unfavorable effect of continuous ART on BMD is in contrast to clinical outcomes, including progression to AIDS, myocardial infarction, and end-stage liver or renal disease, all of which were more common with intermittent ART [23
The randomized comparisons between the DC and VS groups provide evidence that ART causes BMD loss. Moreover, the pattern of BMD change is consistent with the increasing percentage of DC participants receiving ART after the first year (); BMD declined in both treatment groups during this period (). We found no evidence that uncontrolled HIV replication reduced BMD, as was previously suggested [6
]; on the contrary, BMD remained stable in the intermittent ART group during the first year, when most participants were not receiving ART.
The treatment difference and the rate of BMD decline appeared largest for spine qCT. This is consistent with studies in the general population, in which the age-related BMD decline varies by scan location and technique, and spine BMD measured by qCT declines faster than BMD by DXA of the spine or hip [34
]. Trabecular bone turns over more rapidly and, therefore, tends to show greater effects of bone-active agents.
In the VS cohort, the annual BMD decline by DXA (total hip 0.8%, spine 0.4% per year) was steeper than in healthy white men from age 35 to 55 years (hip 0.2%, spine 0.2%); BMD loss at the hip was similar to that in postmenopausal white women from 55 to 75 years (hip 0.8%, spine 0.8%) [36
]. Also, BMD declined throughout follow-up (), whereas in a study comparing tenofovir with stavudine in 600 participants over 144 weeks, BMD stabilized after the first year. Their rate of BMD loss averaged through week 144 was similar to our study, however, at 1.0% per year at the hip and 0.6% spine [22
We could not determine which antiretroviral drugs or drug classes were most responsible for the BMD declines we observed. In the analysis of the VS cohort, associations of cumulative use of specific drugs with BMD decline varied by bone site, and statistical evidence was weak. No association was confirmed in the randomized comparisons across subgroups by baseline ART regimen.
In the VS cohort, cumulative use of stavudine and zidovudine were each associated with loss in spine BMD, consistent with other studies [22
]. Cumulative use of lopinavir/r was associated with BMD loss at the spine (by qCT) and total hip. Association of lopinavir/r with BMD loss has been reported previously [21
]. Results of the cohort analysis are by design hypothesis-generating, not conclusive evidence.
The weak associations of specific drugs with BMD decline in the presence of a clear BMD difference between intermittent and continuous ART may be due to several factors. These factors are as follows:
- Sample size and follow-up time were moderate.
- Several drugs may contribute to BMD decline, and the effects of individual drugs may have been too weak to be identified in the presence of other BMD-active ART.
- In the VS cohort, 93% of follow-up time was spent on ART and follow-up time was included as covariate in the regression. Therefore, P values for individual drugs assess only the added contribution beyond the average ART effect. Such analysis tends to give weaker statistical evidence for the effect of drugs that are taken by many participants.
Traditional risk factors for low BMD such as age, sex, race, smoking, or alcohol abuse at baseline were not associated with change in BMD. It is possible that the effect of these factors on BMD is evident only with longer follow-up.
The fracture data from the parent SMART trial suggest, but certainly do not prove, that continuous ART may increase the risk of fracture as compared with intermittent ART. Fracture events were not systematically collected, but identified from grade 4 adverse event reports.
The main strengths of our study are the randomized comparisons of intermittent versus continuous ART and the standardized assessment of BMD. The study has the following limitations:
- Mean follow-up was only 2.4 years and many participants had only 1 year of follow-up due to the early termination of the substudy. The subgroup analyses in particular were underpowered.
- Fractures were only passively collected.
- The study was conducted between 2002 and 2006; some antiretroviral drugs such as TA-NRTIs and first-generation protease inhibitors are less common today in industrialized countries.
Our study did not address management of low BMD in patients receiving continuous ART. Current US Department of Health and Human Services (DHHS) guidelines suggest to consider BMD assessments in adults receiving ART [41
]. In clinical trials, the use of bone turnover markers in addition to BMD improved estimates of fracture risk in postmenopausal osteoporotic women; their diagnostic value in clinical practice is still unclear [42
]. In small, randomized trials, bisphosphonates improved BMD in HIV-infected adults [43
]; other potential agents such as parathyroid hormone have not been evaluated.
Given the current trend to initiate ART earlier in the course of HIV infection and the importance of maintaining continuous ART, our data suggest that low BMD may become more prevalent as more patients use ART for decades, possibly resulting in higher fracture rates. Studies are needed to better predict osteoporosis risk in HIV-infected persons in order to guide individualized recommendations for diagnosis and treatment of low BMD.