In their recent meta-analysis, Brown and Qaqish  examined the prevalence of low bone mineral density (BMD) in HIV-infected and uninfected individuals, and the risk of low BMD associated with antiretroviral drugs and protease inhibitors. Existing literature regarding the relationship between BMD and both HIV and various antiretroviral agents consists of relatively small studies with inconsistent results, making meta-analysis appropriate. In the current meta-analysis, the authors reported a 6.4-fold increase in the pooled odds of low BMD among adults with HIV infection, compared with uninfected adults. We believe that there are important aspects of this meta-analysis that should be noted.
Numerous medical and lifestyle factors that are highly prevalent among HIV-infected individuals are associated with low BMD, including low body weight, physical inactivity, decreased intake of calcium and vitamin D, cigarette smoking, heavy alcohol use, and opiate exposure. Female sex, older age, white race and estrogen deficiency are also powerful predictors of low BMD. Moreover, studies of BMD that recruit participants from osteoporosis referral sites have limited generalizability. The interpretation of any meta-analysis wholly depends on the generalizability and quality of the original studies, and an assessment of study quality is usual when reporting meta-analyses of observational studies . Brown and Qaqish , however, did not note the exclusion of studies of poor quality, nor comment on the quality of the studies included. All the studies included were cross-sectional and lacked adequate comparison groups, and all of the extracted data examining the effect of HIV on BMD were unadjusted. The lack of adjustment for potentially confounding variables is a major limitation of this meta-analysis, and has important implications for interpreting its findings.
Several studies have suggested that adjustment for potential confounders, such as race, opiate exposure, circulating estrogen, or body mass index (BMI), alters the observed association between HIV and BMD. For example, we recently compared BMD in 263 middle-aged HIV-infected women and 232 uninfected simultaneously recruited women with similar high-risk drug use or sexual behaviors . Although HIV infection was independently associated with low BMD, the differences between the groups were modest. In race-stratified multivariate analyses, HIV was not independently associated with low BMD in black women, suggesting that black race may lessen the effect of HIV on BMD. Methadone maintenance therapy was also independently associated with reduced BMD. An accompanying editorial suggested that estrogen deficiency might influence the effect of HIV on BMD . We also recently completed an analysis of older men with or at risk of HIV infection, which demonstrated that the effect of HIV on BMD was modest after adjustment for other risk factors . As in women, methadone maintenance therapy was independently associated with reduced BMD in that analysis. In addition, HIV was not independently associated with low BMD in men who were overweight or obese, suggesting that high BMI may also lessen the effect of HIV on BMD. The meta-analysis by Brown and Qaqish  did not adjust for any of these potential confounders.
These limitations are equally applicable to the elevated pooled odds ratios in those receiving antiretroviral therapy compared with those not on antiretroviral therapy, and in those taking a protease inhibitor versus those not treated with a protease inhibitor, reported in their meta-analysis. Despite imbalances in BMI, alcohol use, and physical activity between study subjects in the antiretroviral treated and untreated groups, the pooled odds ratios did not adjust for those potential confounders. In their analysis of the effect of treatment with a protease inhibitor on BMD, the authors noted sex imbalances in several studies. Whereas the authors separately pooled adjusted odds ratios for three studies and concluded that the crude and adjusted odds ratios were similar, the adjustment was minimal; two studies adjusted for sex, CD4 cell count nadir, and a history of AIDS, and the third adjusted for age, BMI, duration of antiretroviral therapy, antiretroviral drugs, and a history of AIDS. Well-controlled studies have found no independent effect of either antiretroviral therapy or protease inhibitors [3,5].
In conclusion, although HIV has an independent effect on bone, the lack of adjustment for potentially confounding variables calls into question the magnitude of the pooled odds reported in the meta-analysis of Brown and Qaqish .
1. Brown TT, Qaqish RB. Antiretroviral therapy and the prevalence of osteopenia and osteoporosis: a meta-analytic review. AIDS 2006; 20:2165–2174.
2. Struop DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al
. Meta-analysis of observational studies in epidemiology: a proposal for reporting. JAMA 2000; 283:2008–2012.
3. Arnsten JH, Freeman R, Howard AA, Floris-Moore M, Santoro N, Schoenbaum EE. HIV infection and bone mineral density in middle-aged women. Clin Infect Dis 2006; 42:1014–1020.
4. Yin MT, Glesby MJ. Low bone mineral density, HIV infection, and women: fracture or fiction? Clin Infect Dis 2006; 42:1021–1023.
5. Arnsten JH, Freeman R, Howard AA, Floris-Moore M, Lo Y, Klein RS. Decreased bone mineral density and increased fracture risk in aging men with or at risk for HIV infection.AIDS
: in press.