Secondary Logo

Journal Logo

Clinical Science

Impact of Antiretroviral Drugs on Fracture Risk in HIV-Infected Individuals: A Case–Control Study Nested Within the French Hospital Database on HIV (FHDH-ANRS CO4)

Costagliola, Dominique PhD*; Potard, Valérie MSc*,†; Lang, Sylvie PhD*,†,‡; Abgrall, Sophie MD*,§; Duvivier, Claudine MD; Fischer, Hugues MSc; Joly, Véronique MD#; Lacombe, Jean-Marc MSc*,†; Valantin, Marc-Antoine MD*,**; Mary-Krause, Murielle PhD*; Rozenberg, Sylvie MD**, on behalf of FHDH ANRS CO4

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: February 1, 2019 - Volume 80 - Issue 2 - p 214-223
doi: 10.1097/QAI.0000000000001903



People living with HIV have a lower bone mineral density (BMD)1 and a higher incidence of fractures than the general population of the same age and sex.2–7 Data from randomized trials in antiretroviral (ARV) treatment-naive patients show that BMD declines after ARV initiation,8,9 especially with tenofovir disoproxil fumarate (TDF) compared with other nucleoside reverse transcriptase inhibitors (NRTIs) both at the lumbar spine and the hip.10,11 The decline was also greater with protease inhibitors (PIs) than with non-NRTIs (NNRTIs) at the lumbar spine.10,12 The data concerning PIs might be artifactual being linked to an increase in visceral fat. A significant gain in fat mass correlating with a sharp drop in BMD has been observed during PI exposure.13 Simulated increases in body fat reduced the mean dual-energy X-ray absorptiometry spine BMD but did not affect the mean dual-energy X-ray absorptiometry hip BMD, a finding that might explain the trial results.14 In trials of pre-exposure prophylaxis, in HIV-uninfected individuals, TDF was associated with a small decline in BMD.15,16 In the SMART trial, ARV discontinuation was followed by an increase in BMD.17

Ten studies (see Table 1, Supplemental Digital Content, have examined links between specific ARV and the risk of fracture.5,7,18–25 They variously considered fractures, fractures at osteoporotic sites (or both), or low-energy fractures. They also differed in the confounders they took into account. In particular, few considered classic risk factors for osteoporotic fractures such as the body mass index (BMI), ethnic origin, current smoking status, daily alcohol consumption exceeding 2 units, a family history of hip fracture, systemic glucocorticoids, rheumatoid arthritis, and menopausal status.26 Few of these studies showed an association between an increased risk of fracture and exposure to TDF or to PIs (1/7 and 1/6, respectively), or of fractures at osteoporotic sites (2/6 and 2/3, respectively). In one of the study5 showing a statistical association between ongoing PI exposure and an increased risk of fracture, the authors considered that the link was not causal. The purpose of this study was to assess the possible impact of ARV on the risk of low-energy fractures at potential osteoporotic sites focusing on PIs and TDF.


The French Hospital Database on HIV (FHDH-ANRS CO4)

The FHDH is a hospital-based open multicenter cohort in which inclusions have been ongoing since 1989.27 Individuals are eligible if they have documented HIV-1 or HIV-2 infection and give their written informed consent to participate. Data are collected prospectively by trained research assistants on standardized forms, which include demographic characteristics, the date and type of clinical events recorded according to the International Disease Classification ART, and biological markers. The FHDH project was approved by the French data protection authority (Commission National de l'Informatique et des Libertés on November 27, 1991, Journal Officiel, January 17, 1992).

Study Design

The protocol, including a detailed statistical analysis plan, was written at the time of submission of the project to the French medicines agency (Agence Nationale de Sécurité du médicament et des produits de santé (ANSM)) and to France REcherche Nord & Sud Sida-hiv Hépatites (ANRS).

We conducted a case–control study nested in the FHDH and focusing on HIV-1–infected individuals who were enrolled in the FHDH while ARV-naive. We chose this approach because of the time-varying nature of ARV use, the large size of the cohort, and the long duration of follow-up.28 Moreover, compared with a full cohort approach with a survival analysis using time-dependent variables, a nested case–control analysis provides estimates of odds ratios (ORs) from conditional logistic regression models that are unbiased estimates of relative risk.29

Case Definition

Cases were individuals enrolled in the FHDH while ARV-naive and who had a first prospectively reported low-energy fracture at a potential site of osteoporotic fracture between January 2000 and December 2010. Individuals who were selected as controls but for whom a fracture report was found in their medical records when extracting the data were also considered as cases. The potential osteoporotic sites were the vertebrae, hip, wrist, upper humerus, lower femur, upper tibia, and simultaneous fracture of 3 or more ribs. A low-energy fracture was defined as a fracture sustained after mild trauma such as a fall from standing height.

Selection of Controls

Controls were individuals enrolled in the FHDH while ARV-naive and with no history of fracture. They were randomly selected with the incidence density sampling method,30 after individual matching for sex, age (±3 years), period of HIV diagnosis (<1997/≥1997), being under follow-up at the date of fracture in the corresponding case (±3 months), and, if possible, the clinical center.

Potential Confounders

It was important to take into account preexisting risk factors for fractures that might have influenced the choice of ARV during the study period. The following traditional risk factors were explored: geographic origin (sub-Saharan Africa/other), BMI [underweight (<18.5)/normal (18.5 ≤ BMI ≤ 25)/overweight (>25)], smoking status (no/past/current), alcohol consumption above 2 glasses per day (no/yes), family history of hip fracture (no/yes), prior exposure to systemic glucocorticoids (no/yes), and menopausal status for women (no/yes). We also studied the potential effect of the following HIV-related variables on the risk of fracture: period of enrollment in the FHDH (≤1996/1997–2001/2002–2010), transmission group [men who have sex with men (MSM)/other], period of ARV initiation (naive/<1997/1997–2001/2002–2010), AIDS status, CD4 T-cell nadir, and anti–hepatitis C virus antibody status (HCV Ab−/HCV Ab+) before the date of fracture in the cases. In addition, we considered the CD4 T-cell count, CD8 T-cell count, CD4 to CD8 T-cell ratio, and viral load (VL) measured within 6 months of the date of fracture in the cases. Finally, we collected available data on osteoporosis, which could be on the causal pathway between ARV exposure and the risk of fracture, and chronic kidney disease [dialysis or estimated glomerular filtration rate using the CKD-EPI equation from creatinine level <60 mL/min/1.73 m2 (no/yes)], which may lie on the causal pathway between TDF exposure and the risk of fracture. These last 2 variables were only to be used in sensitivity analyses if a significant association was found between exposure to any ARV and the risk of fracture.

Data Collection

The date of fracture diagnosis, sex, age, geographic origin, and HIV parameters were extracted from the FHDH and validated in the medical records by trained medical assistants using a predefined case report form. We also collected the site and circumstances of the fracture, and potential confounders described above from the medical records. In addition, we used self-administered questionnaire filled by the study participants to collect the date(s), site(s), and circumstance(s) of antecedent if any and referent fracture(s), smoking status, alcohol consumption above 2 units per day, a family history of hip fracture, systemic glucocorticoids, menopausal status, known osteoporosis, and dialysis. When data were recorded both in the medical records and in the questionnaire, we used the data from the questionnaire. When a value was missing both from the medical records and from the questionnaire, we used the value recorded in the FHDH if available.

Statistical Analysis

Conditional logistic regression models were used to quantify the relation between exposure to individual ARV drugs and the risk of fracture. Exposure to each drug was considered as the cumulative duration of exposure (model 1) or as “ever exposed” (yes/no) (model 2). In a third model, the exposure variable for each ARV drug was chosen according to the lowest values of Akaike's information criterion (AIC) in univariable models of the risk of fracture. In sensitivity analyses, exposure to each ARV drug was modeled by a 3-category variable: never exposed, exposed for less than 2 years, and exposed for 2 years or more. An ARV for which of less than 10% of controls were exposed was included in the models, but the results are not reported. We also explored exposure to all PIs [darunavir (DRV), atazanavir (ATV), fosamprenavir/amprenavir (FPV/AMP), indinavir (IDV), lopinavir (LPV), nelfinavir (NFV), ritonavir, and saquinavir (SQV)] and in separate models, exposure to first-generation PIs (all PIs except ATV and DRV).

Characteristics of the cases and controls were compared by using univariable conditional logistic regression. Known risk factors for fracture, except for menopausal status, were included in the multivariable models if they were present in at least 5% of individuals. Regarding factors related to HIV infection, these were included in the multivariable models if in the corresponding univariable conditional logistic regression model, the P value was below 0.10. For VL, CD4, the CD4 nadir, CD8, and CD4/CD8, the choice between continuous and categorical classification was based on the lowest AIC value in the corresponding univariable conditional logistic regression model.

All values missing for fewer than 50% of individuals were replaced by using a multiple imputation method, missing values being randomly sampled from their predicted distributions.29 Ten sets of imputations were used to create 10 complete data sets. All 10 data sets were analyzed and combined with Rubin's rules. SAS software (v9.4; SAS Institute Inc, Cary, NC) was used for all statistical analyses.


Baseline Characteristics of Participants

The study flow chart is shown in Figure 1. Among the 861 reviewed fractures, 261 low-energy fractures at a potential osteoporotic site were validated, and 254 of the patients concerned were matched with at least one control (376 controls). With a total of 254 cases and 376 controls, the ORs that could be detected with 80% power and a 5% type-1 error were above 1.6–2.0 for exposure in the control group ranging from 50% to 10%. There were 285 self-administered questionnaires completed of which 101 (41%) for the cases and 184 (49%) for the controls. Two hundred (78.7%) of the 254 cases had had only one fracture, 37 (14.6%) 2 fractures, 11 (4.3%) 3 fractures, and 6 (2.4%) 4 fractures. There were 53 spine, 69 hip, 51 wrist, 30 humerus upper end, 11 femur lower end, 14 tibia upper end, 6 simultaneous 3 rib, and 20 other fractures. The median year of fracture diagnosis was 2007 [interquartile range (IQR), 2004–2009]. Characteristics of the cases and controls are shown in Table 1. The rate of missing values was lower than 50% for all potential confounders: 15% for BMI, 7% for smoking status, 14% for alcohol consumption, 38% for a family history of hip fracture, 16% for prior systemic glucocorticoid exposure, 13% for chronic kidney disease, 1% for the CD4 nadir, 1% for CD4, 6% for CD8, and 2% for VL. In the case population, median age was 49 years, 67% of patients were men, and 71% were diagnosed with HIV infection before 1997. Their median CD4 cell count was 436/mm3 (IQR, 293–592), their nadir CD4 cell count was 172/mm3 (IQR, 75–298), and 65% of them had VL <50 copies per milliliter. The corresponding values in controls were not significantly different. The proportion of individuals with AIDS was higher among the cases than the controls (31% versus 20%). Regarding classic fracture risk factors, the cases were less likely to be of sub-Saharan origin and more likely to have low BMI, alcohol consumption ≥2 glasses/d, and to have been exposed to systemic glucocorticoids. Unsurprisingly, osteoporosis had been diagnosed in more cases than controls.

Flow chart.
Characteristics of Participants
Characteristics of Participants

The proportions of ARV-exposed cases and controls and the mean duration of exposure to individual ARVs are shown in Table 2. At the date of fracture diagnosis, 49% of cases had been exposed to TDF and 82% to PIs, for 2.5 and 4.3 years, respectively.

ARV Exposure
ARV Exposure

Less than 10% of controls had been exposed to DRV, T20, raltegravir, or maraviroc. In model 3, ARV exposure was modeled as follows: ever exposed (yes/no) to any ARV except efavirenz (EFV), ATV, DRV, FPV/APV, and emtricitabine (FTC), the latter 5 drugs being modeled with their cumulative duration of exposure. In addition to ARV exposure, the multivariable models were adjusted for the period of FHDH enrollment, the HIV transmission group, prior AIDS-defining events, geographic origin, BMI, smoking status, alcohol consumption, and prior systemic glucocorticoid exposure.

Impact of ARV Exposure on the Risk of Fracture

Univariable models of ARV exposure are shown in Table 2. Multivariable models 1, 2, and 3 with adjustment for all ARV exposure and for all ARV exposure plus confounders are shown in Table 3. Models 1 and 3 had lower AIC values than model 2, whether they were adjusted for all ARVs alone or for ARVs and confounders.

Multivariable Models of ARV Exposure

Impact of TDF and PIs

In both the univariable and multivariable models, and regardless of how ARV exposure was modeled, no association was found between TDF and the risk of fracture: the OR was 1.21 [95% confidence interval (CI): 0.61 to 2.39] in the ever-exposed model adjusted for ARV plus confounders, and similar results were obtained for cumulative exposure (OR: 1.04, 95% CI: 0.86 to 1.27). Results were not changed after including existence of chronic kidney disease in the models (data not shown).

ATV was associated with an increased risk of fracture in univariable models using both cumulative exposure (OR: 1.32, 95% CI: 1.08 to 1.62) and ever exposure (OR: 1.59, 95% CI: 1.01 to 2.51). After accounting for other ARVs and confounding factors, we found no significant association in model 2 (OR: 1.89, 95% CI: 0.96 to 3.72), whereas the association was significant in model 1 (OR: 1.52, 95% CI: 1.06 to 2.17) and in model 3 (OR: 1.49, 95% CI: 1.04 to 2.13). In sensitivity analyses (see Table 2, Supplemental Digital Content,, ATV exposure for more than 2 years was associated with an increased risk of fracture in the univariable model (OR: 2.40, 95% CI: 1.07 to 4.15) but not after accounting for other ARVs and confounding factors (OR: 2.37, 95% CI: 0.78 to 7.23). We also checked whether there was an interaction between TDF and ATV and found no significant interaction (P value = 0.56 in model 2 adjusted for ARV and confounders). Other PIs showed no significant association with the risk of fracture in any of the 3 multivariable models. Finally, there was no significant association between the risk of fracture and exposure to either all PIs or only to all first-generation PIs.

Impact of NRTI Exposure

Zidovudine (ZDV) was associated with an increased risk of fracture in the univariable model using ever exposure (OR: 1.58, 95% CI: 1.04 to 2.41) but not cumulative exposure (OR: 1.01, 95% CI: 0.96 to 1.06). After accounting for other ARVs and confounding factors, ZDV was not associated with an increased risk of fracture in the 3 models. Zalcitabine (DDC) was not associated with the risk of fracture in univariable models using either cumulative or ever exposure. After adjustment for ARV and confounders, cumulative exposure to DDC was associated with a lower risk of fracture in model 1 (OR: 0.66, 95% CI: 0.45 to 0.97), contrary to ever exposure to DDC. In adjusted models of sensitivity analyses, DDC exposure for more than 2 years was associated with a lower fracture risk (OR: 0.19, 95% CI: 0.05 to 0.79), whereas DDC exposure for less than 2 years was not associated with the risk of fracture. However, only 6% of controls and 2% of cases had been exposed to DDC for more than 2 years. No other association was found between NRTIs and the risk of fracture in any of the 3 models.

Impact of NNRTI Exposure

Univariable models showed no association between cumulative exposure and ever exposure to EFV. After accounting for ARV and confounding factors, cumulative EFV exposure was associated with a lower risk of fracture in models 1 and 3, with respective ORs of 0.81 (0.69–0.96) and 0.82 (0.70–0.96) per year of exposure. In sensitivity analyses, exposure to EFV for either less than 2 years or more than 2 years was not associated with the risk of fracture. The proportion of individuals exposed to EFV for more than 2 years was small (15%).

Univariable models showed no association between cumulative exposure to nevirapine (NVP), whereas ever exposure to NVP was associated with an increased risk of fracture, with an OR of 1.93 (1.32–2.83). After accounting for other ARVs and confounding factors, NVP exposure was associated with a higher risk of fracture only in model 2.


We found no robust association between the risk of fracture in HIV-infected patients and exposure to ARV drugs including exposure to TDF and PIs. Some drugs, particularly EFV associated with a lower risk and ATV associated with a higher risk, were significantly associated with the fracture risk in some of the models we constructed; however, sensitivity analyses showed that these associations were not robust.

We used several approaches to minimize biases including the choice of a case–control design nested within the FHDH cohort, use of controls matched for age and sex, which are 2 important risk factors for fracture,26 as well as the period of HIV diagnosis (which influences exposure to the different ARVs), and adjustment for both HIV-related parameters and classic fracture risk factors. These classic risk factors were associated with the risk of fracture in our study, supporting the reliability of our data. HCV coinfection was not taken into account in the multivariable models because its prevalence was low (6%). Adding the variable in the models did not change our results. Several published studies have shown an association between HCV coinfection and the risk of fracture in populations with a higher prevalence of HCV infection. We recognize that our study does not provide information on the most recent drugs such as DRV, rilpivirine, raltegravir, dolutegravir, or elvitegravir.

Only one previous study has linked TDF to an increased risk of all fractures.23 The difference between the 2 analyses reported in this article is surprising. If the impact of TDF on the risk of fracture is mediated by its effect on BMD, the hazard ratio should be larger when the analysis is restricted to osteoporotic fractures rather than all fractures. And, even if the exact hazard ratio value for osteoporotic fractures was not reported, it does not appear to be the case when looking at the figures of this article. This does not support a causal link of the reported association. In addition, one study19 concluded that there was a modestly increased osteoporotic fracture risk associated with a cumulative TDF exposure; however, the association was only significant when restricting the analysis to participants enrolled after 1996. In the ACTG study A5202,10 despite a significant impact of TDF on BMD compared with abacavir, there was no significant difference in the incidence of fractures. Finally, a study based on the proportional reporting ratios of fracture in the EudraVigilance database between 2001 and 2016 did not show a disproportionality for all fractures in patients treated with TDF, whereas the disproportionality observed for osteoporotic fractures was based on 13 cases only.25 This may well be explained by difference in reporting rates in EudraVigilance when TDF is part of an ARV regimen. All other studies (see Table 1, Supplemental Digital Content, did not report an association between exposure to TDF and the risk of fractures. Exposure to TDF may affect BMD but not necessarily the risk of fractures, as the decline in BMD occurs mainly during the first 2 years of treatment. Recent studies have shown no effect of TDF on BMD except during treatment initiation.31,32 In addition, only 10%–44% of the risk of fractures can be attributed to low BMD.33 In our study, the selected fractures occurred years after ARV initiation and only 9.0% of case patients started ARV with a TDF-containing combination.

Only 3 studies have shown an increased risk of fracture with some PI exposure,5,19,20 but it was not necessarily causal in one study5 and consisted only of a moderate, nonsignificant increase in the risk of osteoporotic fracture in the study by Bedimo et al,19 where only exposure to LPV was significantly associated with the risk of fracture. In the study by Mundy et al,20 although exposure to all PIs was not associated with an increased risk of fracture; exposure to DRV and to SQV were associated with an increased risk of fracture. As in our study, PIs were not found to be associated with the fracture risk in the remaining studies.7,22–25 This is not unexpected, as the observed decline in spine BMD after PI exposure10,12 was recently suggested to be a measurement artifact: BMD measurement at lumbar spine is less accurate after weight gain.14


We found no evidence of an excess risk of fracture after exposure to TDF or to PIs. It would be interesting to study this risk during a more recent period to include more individuals initiating ARV with a TDF. These results have important implications for the use of tenofovir alafenamide versus generic TDF.


The authors are grateful to all FHDH participants and research assistants, without whom this work would not have been possible.


1. Brown TT, Qaqish RB. Antiretroviral therapy and the prevalence of osteopenia and osteoporosis: a meta-analytic review. AIDS. 2006;20:2165–2174.
2. Triant VA, Brown TT, Lee H, et al. Fracture prevalence among human immunodeficiency virus (HIV)-infected versus non-HIV-infected patients in a large U.S. healthcare system. J Clin Endocrinol Metab. 2008;93:3499–3504.
3. Güerri-Fernandez R, Vestergaard P, Carbonell C, et al. HIV infection is strongly associated with hip fracture risk, independently of age, gender, and comorbidities: a population-based cohort study. J Bone Miner Res. 2013;28:1259–1263.
4. Prieto-Alhambra D, Güerri-Fernández R, De Vries F, et al. HIV infection and its association with an excess risk of clinical fractures: a nationwide case–control study. J Acquir Immune Defic Syndr. 2014;66:90–95.
5. Womack JA, Goulet JL, Gibert C, et al. Increased risk of fragility fractures among HIV-infected compared to uninfected male veterans. PLoS One. 2011;6:e17217.
6. Young B, Dao CN, Buchacz K, et al. Increased rates of bone fracture among HIV-infected persons in the HIV Outpatient Study (HOPS) compared with the US general population, 2000-2006. Clin Infect Dis. 2011;52:1061–1068.
7. Sharma A, Shi Q, Hoover DR, et al. Increased fracture incidence in middle-aged HIV-infected and uninfected women: updated results from the Women's Interagency HIV Study. J Acquir Immune Defic Syndr. 2015;70:54–61.
8. Grund B, Peng G, Gibert CL, et al. Continuous antiretroviral therapy decreases bone mineral density. AIDS. 2009;23:1519–1529.
9. Hansen AB, Obel N, Nielsen H, et al. Bone mineral density changes in protease inhibitor-sparing vs. nucleoside reverse transcriptase inhibitor-sparing highly active antiretroviral therapy: data from randomized trial. HIV Med. 2011;12:157–165.
10. McComsey GA, Kitch D, Daar ES, et al. Bone mineral density and fracture in antiretroviral-naïve 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. J Infect Dis. 2011;203:1791–1801.
11. Stellbrink HJ, Orkin C, Arribas JR, et al. Comparison of changes in bone density and turnover with abacavir-lamivudine versus tenofovir-emtricitabine in HIV-infected adults: 48-week results from the ASSERT study. Clin Infect Dis. 2010;51:963–972.
12. Duvivier C, Kolta S, Assoumou L, et al. Greater decrease in bone mineral density with protease inhibitor regimens compared with non nucleoside reverse transcriptase inhibitor regimens in HIV-1 infected naïve patients. AIDS. 2009;23:817–824.
13. Bonnet E, Ruidavets JB, Genoux A, et al. Early loss of bone mineral density is correlated with a gain of fat mass in patients starting a protease inhibitor containing regimen: the prospective Lipotrip study. BMC Infect Dis. 2013;13:293.
14. Yu EW, Thomas BJ, Brown JK, et al. Simulated increases in body fat and errors in bone mineral density measurements by DXA and QCT. J Bone Miner Res. 2012;27:119–124.
15. Liu AY, Vittinghoff E, Sellmeyer DE, et al. Bone mineral density in HIV-negative men participating in a tenofovir preexposure prophylaxis randomized clinical trial in San Francisco. PLoS One. 2011;6:e23688.
16. Mulligan K, Glidden DV, Anderson PE, et al. Effects of emtricitabine/tenofovir on bone mineral density in HIV-negative persons in a randomized, double-blind, placebo-controlled trial. Clin Infect Dis. 2015;61:572–580.
17. Hoy J, Grund B, Roediger M, et al. Interruption or deferral of antiretroviral therapy reduces markers of bone turnover compared with continuous therapy: the SMART Body Composition substudy. J Bone Miner Res. 2013;28:1264–1274.
18. Hansen AB, Gerstoft J, Kronborg G, et al. Incidence of low and high-energy fractures in persons with and without HIV infection: a Danish population-based cohort study. AIDS. 2012;26:285–293.
19. Bedimo R, Maalouf NM, Zhang S, et al. Osteoporotic fracture risk associated with cumulative exposure to tenofovir and other antiretroviral agents. AIDS. 2012;26:825–831.
20. Mundy LM, Youk AO, McComsey GA, et al. Overall benefit of antiretroviral treatment on the risk of fracture in HIV: nested case-control analysis in a health-insured population. AIDS. 2012;26:1073–1082.
21. Gedmintas L, Wright EA, Dong Y, et al. Factors associated with fractures in HIV-infected persons: which factors matter? Osteoporos Int. 2017;28:239–244.
22. Yin MT, Kendall MA, Wu X, et al. Fracture after antiretroviral initiation. AIDS. 2012;26:2175–2184.
23. Borges AH, Hoy J, Florence E, et al. Antiretrovirals, fractures and osteonecrosis in a large international HIV cohort. Clin Infect Dis. 2017;64:1413–1421.
24. Gonciulea A, Wang R, Althoff KN, et al. An increased rate of fracture occurs a decade earlier in HIV+ compared with HIV- men. AIDS. 2017;31:1435–1443.
25. Garcia MG, Larrinaga-Torrontegui U, Eduardo Martinez M, et al. Fractures related to tenofovir: a case/noncase study in the European pharmacovigilance database. Am J Ther. 2018 [in press].
26. Kanis JA, Johnell O, Oden A, et al. FRAXTM and the assessment of fracture probability in men and women from the UK. Osteoporos Int. 2008;19:385–397.
27. Mary-Krause M, Grabar S, Lièvre L, et al. Cohort Profile: French hospital database on HIV (FHDH-ANRS CO4). Int J Epidemiol. 2014;43:1425–1436.
28. Lang S, Mary-Krause M, Simon A, et al. HIV replication and immune status are independent predictors of the risk of myocardial infarction in HIV-infected individuals. Clin Infect Dis. 2012;55:600–607.
29. Suissa S. Novel approaches to pharmacoepidemiology study design and statistical analysis. In: Strom B, ed. Pharmacoepidemiology. 4th ed: Hoboken, NJ: Wiley; 2005:811–829.
30. Richardson DB. An incidence density sampling program for nested case-control analyses. Occup Environ Med. 2004;61:e59.
31. Tinago W, Cotter AG, Sabin CA, et al. Predictors of longitudinal change in bone mineral density in a cohort of HIV-positive and negative patients. AIDS. 2017;31:643–652.
32. Grant PM, Kitch D, McComsey GA, et al. Long-term bone mineral density changes in antiretroviral-treated HIV-infected individuals. J Infect Dis. 2016;214:607–611.
33. Stone KL, Seeley DG, Lui LY, et al.; for the study of Osteoporotic Fractures Research Group. BMD at multiple sites and risk of fractures of multiples types: long term results from the Study of Osteoporotic Fractures. J Bone Miner Res. 2003;18:1947–1954.

fracture; bone mineral density; antiretroviral drugs; tenofovir; protease inhibitors

Supplemental Digital Content

Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.