HIV infection and antiretroviral therapies have detrimental effects on bone metabolism, but data on their impact on fracture risk are controversial. We conducted a population-based cohort study to explore the association between clinical diagnosis of HIV infection and hip and major osteoporotic fracture risk. Data were obtained from the SIDIAPQ database, which contains clinical information for >2 million patients in Catalonia, Spain (30% of the population). We screened the database to identify participants with a clinical diagnosis of HIV infection, and ascertained incident hip and osteoporotic major fractures in the population aged 40 years or older in 2007 to 2009. In addition, data on incident fractures involving hospital admission were obtained from the Hospital Admissions database. Cox regression models were used to estimate hazard ratios (HRs) for the HIV-infected versus uninfected participants. Models were adjusted for age, sex, body mass index, smoking status, alcohol drinking, oral glucocorticoid use, and comorbid conditions (Charlson index). Among 1,118,156 eligible participants, we identified 2489 (0.22%) subjects with a diagnosis of HIV/AIDS. Age- and sex-adjusted HR for HIV/AIDS were 6.2 (95% confidence interval [CI] 3.5–10.9; p < 0.001) and 2.7 (2.01–3.5; p < 0.001) for hip and major fractures, respectively; this remained significant after adjustment for all mentioned potential confounders: HR 4.7 (2.4–9.5; p < 0.001) and 1.8 (1.2–2.5; p = 0.002). After stratifying by age, the association between HIV infection and major fractures was attenuated for those aged <59 years (adjusted HR 1.35 [0.88–2.07], p = 0.17) but appeared stronger in older patients (adjusted HR 2.11 [1.05–4.22], p = 0.035). We report a strong association between HIV infection and hip fracture incidence, with an almost fivefold increased risk in the HIV infected, independent of sex, age, smoking, alcohol drinking, and comorbidities. Similarly, we demonstrate a 75% higher risk of all clinical fractures and a 60% increase in risk of non-hip clinical fractures among patients with a diagnosis of HIV infection.
Because high-activity antiretroviral therapy (HAART) for HIV infection allows patients to live longer, many are being confronted with additional health challenges related to aging. Morbidities that were not classically considered to be HIV related are now found associated with ongoing HIV replication, chronic immune activation, and also with long-term HAART.[1, 2] Although potentially severe, osteoporosis and fractures historically have been neglected. Despite the previously established effects of HIV on bone metabolism, the impact of HIV-related bone disease on fracture risk remains uncertain. Numerous studies have found that HIV-infected patients have lower bone mineral density (BMD) and higher bone loss rates compared with the general population,[3, 4] but studies analyzing whether low bone density actually leads to greater incidence of fractures in HIV-infected patients have been inconclusive.[4-6] In addition, Collin and colleagues found that the incidence rate of first fractures in HIV-infected patients was in the same range as that reported in the general European population for the same age group. We used a large population-based primary care database to explore the association between HIV infection and the risk of hip, non-hip, and all clinical fractures.
Materials and Methods
This study is a population-based retrospective cohort study.
The Spanish public health-care system covers the practical totality of the population. General practitioners (GPs) play an essential role, being responsible for primary health care, long-term prescriptions, and specialist and hospital referrals. The data in this study were obtained from the SIDIAPQ (Sistema d‘Informació per al Desenvolupament de l‘Investigació en Atenció Primària-Q) database. SIDIAP is composed of electronic medical records of a sample of patients attending GPs in Catalonia (northeast Spain), covering a population of about 5 million patients (80% of the total population) with a total of 3414 participating GPs. Only data registered by those GPs with the highest scores in coding quality within SIDIAP are included in SIDIAPQ, a higher-quality version of the SIDIAP database, including information on a representative sample of 1.9 million participants (30% of the population of Catalonia). SIDIAPQ contains the clinical and referral events registered by primary care health professionals in electronic medical records, comprehensive demographic information, prescription and corresponding pharmacy invoicing data, specialist referrals, primary care laboratory test results, hospital admissions, and their major outcomes. Health professionals gather this information using ICD-10 codes and use structured spreadsheets designed for the collection of variables such as body mass index, smoking, alcohol drinking, and blood pressure. Encoding personal identifiers ensures the confidentiality of the information in SIDIAPQ.
All patients aged ≥40 years in the database in the period 2007 to 2009 were eligible for this study (N = 1,118,587). Participants with a clinical diagnosis of HIV infection were identified amongst these using ICD-10 codes (B20, B22, and B24).
Ascertainment of fractures
Clinical fractures registered in the study period (January 1, 2007 to December 31, 2009) in SIDIAPQ were identified using medical codes for a list of sites of fracture, which are based on the ICD-10 classification. Fracture sites considered for these analyses were those defined by Center and Eisman as major fractures (hip, clinical spine, pelvis, tibia, multiple rib, and proximal humerus) and the most prevalent minor osteoporotic fracture in our data (wrist/forearm). Fracture coding has been validated in SIDIAP using both prospective cohort and hospital admission data as a reference: Hip, clinical spine, and wrist/forearm fracture coding have been shown to be highly specific (99%, 99%, and 98%, respectively) in SIDIAP.
Cox proportional hazard regression models were used to estimate multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CI) for the HIV-infected versus uninfected participants. Similar models were fitted for any clinical fracture, hip fracture, and non-hip fracture. Age- and sex-adjusted and multivariate HRs are reported. The latter were adjusted for the following potential confounders: age, sex, body mass index (BMI), smoking status, alcohol drinking, oral glucocorticoid use, and comorbid conditions, as listed in the Charlson index (type 2 diabetes mellitus, diabetic complications, chronic obstructive pulmonary disease, heart failure, myocardial infarction, peripheral vascular disease, cardiovascular disease, chronic renal failure, liver disease, rheumatoid arthritis, paraplegia, gastro/duodenal ulcer, dementia, malignancies, and metastatic neoplasm). We then replicated these analyses to look at the effect of HIV on major fracture risk stratified by median age (59 years).
Missing values for smoking status and alcohol drinking were accounted for by addition of a missing category. The validity of the proportional hazards assumption was verified using the Schöenfeld's residuals formal test. All model fitting was carried out using Stata for Mac version 12 (StataCorp, College Station, TX, USA).
We identified 1,118,156 people aged 40 years or older in SIDIAPQ in 2007 to 2009. Of these, 2489 (0.22%) were either prevalent or incident cases of HIV infection in this same period. HIV-infected and uninfected participants were followed up for a median (interquartile range) of 2.997 (0.91) and 2.997 (0.001) years, respectively. During the study period, 41,907 (3.75%) patients died (178 [7.2%] HIV-infected and 41,729 [3.7%] in the HIV-free population) and 26,126 (2.34%) were lost to follow-up (92 [3.7%] and 26,034 [2.3%] among HIV and non-HIV subpopulations, respectively). When compared with the general population in SIDIAP, HIV-infected participants were younger (mean [standard deviation] 50.0 [7.6] versus 61.3 [14.2]; p < 0.001), thinner (BMI 24.5 [4.4] versus 28.4 [4.9]; p < 0.001), more likely to be male (75.3% versus 47.8%; p < 0.001), current smokers (53.3% versus 18.9%; p < 0.001), severe alcohol drinkers (2.7% versus 1.8%; p < 0.001), and to suffer from mild (34.3% versus 2.3%; p < 0.001) and severe liver disease (0.4% versus 0.1%; p < 0.001) and malignancies (3.8% versus 2.9%; p < 0.001).
During the study period, 49 and 24,408 clinical fractures (12 and 7299 hip fractures) were observed in the HIV-infected and uninfected patients, respectively. Corresponding unadjusted fracture incidence rates were 8.03/1000 patient-years (95% CI 6.07–10.62) and 7.93/1000 (7.83–8.03). Age-specific fracture incidence rates in the HIV-infected versus the disease-free participants for all clinical fractures have been plotted (Fig. 1). Hip fracture incidence rates were 2.03 (1.15–3.57) for the HIV-infected and 2.37 (2.31–2.42) for HIV-free participants.
Age- and sex-adjusted HRs for all clinical, non-hip, and hip fractures for the HIV-infected patients were 2.67 (2.01–3.53; p < 0.001), 2.39 (1.76–3.25; p < 0.001), and 6.16 (3.49–10.86; p < 0.001), respectively. Fracture risk remained increased for the HIV-infected group even after adjustment for potential confounders including BMI, smoking, alcohol drinking, oral corticosteroid use, and history of comorbid conditions: HRs were 1.75 (1.24–2.48; p = 0.002), 1.63 (1.12–2.37; p = 0.010), and 4.72 (2.35–9.47; p < 0.001) (Table 1). After stratifying by age, the association between HIV infection and major fractures was no longer significant for those aged <59 years (adjusted HR 1.35 [0.88–2.07]; p = 0.17) but appeared stronger in older patients (adjusted HR 2.11 [1.05–4.22]; p = 0.035).
|No. of fractures||Fracture IR/1000 py (95% CI)||Age- and sex-adjusted HR (95% CI); p value||Multivariate-adjusted HRa (95% CI); p value|
|HIV uninfected||7299||2.37 (2.31–2.42)||Ref.||Ref.|
|HIV infected||12||2.03 (1.15–3.57)||6.16 (3.49–10.86); p < 0.001||4.72 (2.35–9.47); p < 0.001|
|HIV uninfected||17,839||5.78 (0.70–0.87)||Ref.||Ref.|
|HIV infected||41||6.70 (4.93–9.10)||2.39 (1.76–3.25); p < 0.001||1.63 (1.12–2.37); p = 0.010|
|All clinical fractures|
|HIV uninfected||24,408||7.93 (7.83–8.03)||Ref.||Ref.|
|HIV infected||49||8.03 (6.07–10.62)||2.67 (2.01–3.53); p < 0.001||1.75 (1.24–2.48); p = 0.002|
We report a very strong association between HIV infection and hip fracture occurrence, with an almost fivefold increased risk in the HIV-infected patients when compared with uninfected participants, independentely of sex, age, smoking, alcohol drinking, and comorbidities. Similarly, we demonstrate a 75% higher risk of all clinical fractures and a 60% increase in risk of non-hip clinical fractures among patients with a diagnosis of HIV infection. These effects were also independent of potential confounders. However, stratified analyses showed that the increase in risk of non-hip major fractures was only significant in older patients (aged 59 years or older), among whom HIV infection appeared related to a more than double risk of fracture.
There is limited prior evidence assessing the relationship between HIV infection and the risk of fragility fractures. Our findings reinforce other studies that have found a positive correlation between HIV infection and fractures.[10, 11] Triant and colleagues support the hypothesis that HIV infection is associated with an elevated risk of fracture in all fracture sites. However, they did not find differences in hip fractures in women. Arnsten and colleagues found that HIV infection is independently associated with reduced BMD in a relatively aged cohort of men, and showed that lower BMD was associated with increased fracture risk. Other population-based cohort studies have been carried out to date,[10, 11, 13] which also showed an increased fracture risk in HIV-infected patients. However, some of these studies were methodologically different from ours, like the study by Young and colleagues, which reported age and sex indirectly standardized fracture rates from the HOPS cohort. Nevertheless, similarly to our study, there was a higher risk of incident fractures in HIV-infected patients within this cohort, even when adjusting by age, sex, substance abuse, and comorbidities. Conversely, other studies have found no association between HIV infection and fractures,[5, 14] although reduced sample size and restrictive inclusion criteria might limit the validity of these findings.
Several potential explanations for the association between HIV infection and fragility fractures have been proposed, including a lower bone mass in these patients: A systematic review of 12 cross-sectional studies in HIV-infected adults found that the probability of osteopenia and osteoporosis was 6.4 and 3.7 times higher in HIV-infected patients, respectively. Pro-inflammatory effects of HIV, including release of cytokines (interleukins 1,6 and tumor necrosis factor) and HAART side effects on bone metabolism shown in clinical trials[2, 10, 17-19] have been proposed as the causal pathway for the reduced bone mass observed in the HIV-infected group. A recent study has reported that the cumulative use of HAART treatments appears independently associated with an increased risk of osteoporotic fracture. In addition, different lifestyle and hormonal factors that are prevalent among HIV-infected persons could partly account for the increased fracture risk in these patients. Such factors include physical inactivity, decreased intake of calcium and vitamin D, cigarette smoking, alcohol use, opiate use, low testosterone levels, and hepatitis B or C coinfection.[5, 21] According to our data, HIV patients have higher prevalence of alcohol consumption, smoking, and viral hepatitis coinfection than the general population.
The main limitation of our study is the lack of individual validation of each one of the fractures observed. However, coding of hip fractures (and other clinical fractures) has been recently validated in the SIDIAP database and shown to be highly specific when compared with prospective cohort studies and to the official national hospital admission database. In contrast, fracture coding in SIDIAP has low sensitivity when compared with conventional cohort studies (between 50% and 70%, depending on fracture site). Nevertheless, SIDIAP data were completed with hospital-based diagnoses to minimize misclassification in this study. If there was still some degree of under-register in our data, this is likely to be at random and would hence only drive the risk estimates toward the unity. Another limitation of these data is the lack of detailed information on HIV infection (virus load, disease stage, etc.) as well as on antiretroviral therapies used, which are given for free to HIV patients in hospital settings in Spain and hence do not appear in pharmacy invoice databases. Finally, the low number of HIV-infected patients included among the elderly suggests that the age-stratified results should be interpreted with caution and need replication in external cohorts.
Strengths of our study are the high representativeness and generalizability of the data used: SIDIAPQ covers a representative sample of more than 30% of the total population, and these data are gathered in actual practice conditions. In addition, loss to follow-up is low (<2.5%) when compared with other cohort studies, which limits the possibility of loss to follow-up bias.
In conclusion, we have shown a strong association between HIV infection and hip fracture risk, with more than fivefold higher risk when compared with the general population. Similarly, HIV-infected patients are at 75% higher risk of major osteoporotic fractures. This is independent of classical fracture risk factors as well as of comorbidities. The excess risk of non-hip major fractures appears to be highest in older patients.
All authors state that they have no conflicts of interest.
The Internal Medicine Department and the URFOA IMIM receive support from the RETICEF (Red Temática de Investigación Cooperativa en Envejecimiento y Fragilidad, Instituto Carlos III, Government of Spain). We would like to thank the collaboration of the “Servei Català de la Salut (CatSalut)”, and in particular, Mrs Montserrat Bustins, by providing us with data from the CMBD-AH. We finally want to thank all the health professionals involved in registering data in computerized medical records for SIDIAP.
Authors’ roles: Study Design: DPA, RGF, XN, PV, and ADP. Study conduct: FFA, DPA, RGF, PV, CCA, ASC, HK, and ADP. Data management: FFA. Data interpretation: RGF, PV, CC, HK, FFA, ASC, XN, DPA, and ADP. Drafting manuscript: RGF, DPA, HK, and ADP. Revising manuscript content: All the authors. Approving final version of manuscript: All the authors.