In this study, we found a significant proportion (10%) of healthy HIV-negative MSM at risk for HIV infection had low BMD at baseline. While a number of studies using other classification criteria have shown a higher than expected prevalence of low BMD among HIV-infected MSM, including men with primary HIV infection 
, little data exist on bone mass in healthy MSM without HIV infection. Our findings suggest that some degree of the low BMD observed in HIV-infected men may pre-date HIV infection. We used ISCD classification criteria for BMD reporting in men younger than age 50. These guidelines stipulate that T-scores should not be used and a Z-score of ≤2.0 is defined as “below the expected range for age” 
; it is emphasized that in men in this age group, osteoporosis cannot be diagnosed on the basis of BMD alone. While the BMD reference ranges for men used in this study may not fully represent the current population studied, men with a Z-score≤−2.0 fall within the lowest 2.5 percentile of BMD compared with the reference population and are typically evaluated for secondary causes of low BMD if their bone density comes to medical attention. In testing 16 such individuals, we uncovered 2 cases of low vitamin D and one case of hypogonadism, suggesting the importance of pursuing a work-up for reversible causes in this population. However, the implications of low BMD for fracture risk in this cohort, particularly among younger men, are currently unknown.
Low BMD was associated with amphetamine and inhalant use in our study, while men reporting use of multivitamins and supplements containing calcium or vitamin D were less likely to have low BMD. A review of the literature shows two reports of a potential association between methamphetamine use and low BMD/altered bone metabolism 
. The association of methamphetamine and inhalant use with low BMD could be due to a direct toxic effect of these substances on bone metabolism 
, or may be confounded by another lifestyle factor that is associated with both drug use and low BMD. Experimental studies examining potential biological mechanisms for methamphetamine and inhalant-induced bone loss should be conducted, along with larger cross-sectional and cohort studies of drug-using populations to confirm and further investigate these associations. Several studies have found an independent association of amphetamine and/or inhalant use with HIV acquisition among MSM in the United States 
; this substance-using population may be more likely to benefit from PrEP if effective, but also to demonstrate low BMD at baseline.
Among men who enrolled in this PrEP trial and were included in the longitudinal analysis, TDF use resulted in a small (0.8–1.1%) but statistically significant net decrease in BMD from baseline at two anatomic sites, with the greatest loss at the femoral neck. These changes occurred within the first 12 months of tenofovir use, with no evidence of further declines in BMD at 24 months in the immediate arm, although our sample size limited power to examine differences in the treatment effect across the two time points. We also observed a higher proportion of men experiencing a >5% drop in BMD from baseline in the TDF group relative to placebo, particularly evident at the femoral neck, but this difference was not statistically significant.
Prior studies have demonstrated BMD loss with initiation of ART, regardless of regimen, with somewhat greater decreases observed with tenofovir-containing regimens 
. Our results are consistent with findings of decreased BMD associated with TDF seen in earlier randomized trials of antiretroviral-naive HIV-infected individuals. The difference in BMD decline associated with TDF compared with placebo in this study is similar in magnitude to the net BMD decline associated with TDF-containing regimens versus alternative regimens in treatment studies. In the Gilead 903 study, Gallant et al. demonstrated a net decrease in BMD of 1.2% at the lumbar spine in the TDF vs. d4T arms; bone loss in the TDF group occurred through weeks 24 and 48 in this cohort and stabilized through week 144 
. No additional bone loss was seen in a subgroup of Gilead 903 participants followed through 288 weeks 
; participants in this open-label extension received supplemental calcium and vitamin D. In the more recent ASSERT trial, Stellbrink et al. reported a net 0.8% and 1.7% decrease in BMD at the lumbar spine and total hip respectively, when comparing the tenofovir-emtricitabine vs. the abacavir-lamivudine group 
. In this European cohort, bone loss in the TDF group stabilized at week 24 at the lumbar spine but ongoing loss occurred through week 48 at the total hip.
Given the relatively short duration of follow-up in most these studies, longer term BMD data are required to better characterize the long-term effects of TDF on BMD. Current PrEP trials are testing tenofovir-based regimens in over 20,000 HIV-uninfected individuals at risk for HIV infection. Several of these trials are measuring BMD in a subset of study participants, including the iPrEx trial, a phase 3 efficacy trial of emtricitabine-tenofovir in MSM globally. Given our findings, we encourage other PrEP trials to include DEXA monitoring when logistically possible to better characterize the baseline prevalence of low BMD in different target populations for PrEP and the prevalence of risk factors for low BMD, and determine the magnitude and trajectory of BMD loss associated with ARV use for prevention. These data may help identify individuals who are at risk for low BMD or bone loss with PrEP use and guide clinical decision making on whether screening for low BMD may be warranted prior to initiation of PrEP.
The clinical significance of TDF-associated BMD loss, including whether fracture risk is increased, is currently unknown 
. In this study, we observed 6 fractures in the TDF group vs. 4 in the placebo group, although this study was not designed or powered to detect differences in fracture rates between arms. All fractures were trauma-related and assessed as unrelated to study drug. In the Gilead 903 study, 16 patients (11 in the stavudine group vs. 5 in the tenofovir group) developed fractures through 144 weeks, and almost all were related to trauma 
. However, there have been case reports of fractures during TDF therapy, in the setting of proximal renal tubule dysfunction. Additional follow-up in larger cohorts is needed to determine whether extended use of TDF increases fracture risk.
Our study is subject to some limitations. First, this study was conducted in only 1 site (San Francisco) in HIV-uninfected men, the majority of whom were white. Additional studies are being conducted in different settings and in other populations, including HIV-uninfected women, and will determine whether our findings can be generalized. Second, we had a relatively small sample size, precluding multivariable analysis of factors associated with low BMD at baseline, as well as analyses to identify any subgroups at higher risk for BMD loss during TDF PrEP use. Also, our prevalence estimate for baseline low BMD was based on a convenience sample of men screening for an HIV prevention study. Therefore, these results may not reflect the prevalence of low BMD in the larger population of MSM. Our study employed relatively short follow-up (maximum 24 months for immediate arm participants). Additional studies are needed to determine whether BMD effects of TDF are sustained or progress during longer term use, and whether these effects reverse after TDF discontinuation. Also, we did not have the opportunity to do more extensive testing for secondary causes of low BMD or to evaluate markers of bone mineral turnover to help elucidate mechanisms for TDF-associated bone loss. In the ASSERT study, increases in bone turnover markers (including osteocalcin, procollagen 1 N-terminal propeptide, bone specific alkaline phosphatase, and type 1 collagen cross-linked C telopeptide) were significantly greater in the TDF vs. comparator group 
. Future studies should incorporate testing of markers of bone resorption and formation to evaluate potential mechanisms for BMD loss associated with TDF. Finally, these analyses do not adjust for degree of exposure to study drug. Lesser drug exposure due to suboptimal pill-use may have attenuated the magnitude of effects of TDF on BMD we detected in this study. Optimally, such analyses would be adjusted using a biologic marker of long-term drug exposure. Such measures, including tenofovir concentrations in peripheral blood mononuclear cells and hair 
, are currently being explored and should be correlated with bone turnover markers and BMD outcomes in future TDF-based PrEP studies.
Our study also has several strengths. We present novel data looking at prevalence of low BMD in HIV-uninfected men at risk for HIV infection and the effects of TDF on BMD in this seronegative population. These data provide important information on skeletal health in men in the absence of HIV infection and other antiretroviral use. For the longitudinal analysis, this study utilized an intent-to-treat analysis of this randomized, placebo-controlled cohort, thus avoiding confounding by indicator in our assessments. We also achieved high levels of follow-up during the trial.
In summary, we found a significant proportion of HIV-uninfected men had low BMD at baseline. Low BMD was associated with methamphetamine and inhalant use. Similar adverse effects of TDF on BMD were seen in this cohort of HIV-uninfected MSM as seen in antiretroviral treatment studies of TDF-based regimens in HIV-infected individuals. These data suggest that low BMD may pre-date HIV infection among men at risk for acquisition of HIV, and use of tenofovir in these individuals leads to a small but statistically significant decline in BMD. The decline was not associated with an elevated fracture risk during the study.
The finding that oral FTC/TDF PrEP reduces HIV acquisition among MSM 
and the issuance of interim guidance on prescribing PrEP from the CDC to health-care providers 
will likely lead to increased PrEP use in different MSM communities. Larger controlled studies with longer follow-up are needed to assess the course of BMD loss associated with tenofovir-based PrEP regimens over the longer term, as well as the clinical significance of these findings in HIV-uninfected populations.