| |
Fracture Prevalence Among HIV-Infected versus Non HIV-Infected Patients in a Large U.S. Healthcare System
|
| |
| |
J Clin Endocrin Metab. First published ahead of print July 1, 2008 as doi:10.1210/jc.2008-0828
Virginia A. Triant, MD, MPH, Todd T. Brown, MD, PhD, Hang Lee, PhD, and
Steven K. Grinspoon, MD
From the Brigham and Women's Hospital (BWH) and Massachusetts General Hospital (MGH) Divisions of Infectious Diseases (VAT), Johns Hopkins University Division of Endocrinology (TTB), MGH Biostatistics Center (HL), MGH Program in Nutritional Metabolism (SKG and VAT), Boston, MA 02114
Funded in part by T32 AI07387 (VAT), K23 AT002862 (TTB), K24 DK064545 (SKG).
"To our knowledge, these data are the first to compare fracture prevalence between HIV-infected and non HIV-infected patients with a large patient sample using ICD-based outcome ascertainment. The results provide strong evidence that HIV-infected patients have a higher prevalence of fractures than non HIV-infected patients, across both genders and critical fracture sites. Moreover, our data suggest that the relative difference in fracture prevalence between HIV-infected and non HIV-infected patients increases with age for both genders. As the HIV-infected population ages, reduced bone mineral density and increased fracture risk may become an even greater problem. Whether increased fractures are the sequelae of antiretroviral therapy, increased rates of traditional risk factors such as low weight among HIV-infected patients, or HIV infection - and its accompanying metabolic and inflammatory disturbances - itself remains to be determined.....Mitochondrial dysfunction has been associated with the use of nucleoside reverse transcriptase inhibitors,(23) and it is interesting to speculate that this may contribute to a "premature aging" in HIV-infected patients which may contribute to reduced BMD and increased fracture rates. This study suggests the importance of assessing bone density and minimizing factors contributing to increased fracture risk in the HIV-infected population."
Abstract
Context: Reduced bone mineral density has been demonstrated among HIV-infected patients, but fracture prevalence is unknown.
Objective: To compare fracture prevalence in HIV-infected and non HIV-infected patients.
Design: Population-based study.
Setting: A large U.S. healthcare system.
Patients: 8,525 HIV-infected and 2,208,792 non HIV-infected patients with at least one inpatient or outpatient encounter between October 1, 1996 and March 21, 2008 were compared.
Main Outcome Measure: Fracture prevalence using specific ICD-9-CM fracture codes.
Results: The overall fracture prevalence was 2.87 versus 1.77 patients with fractures per 100 persons in HIV-infected compared to non HIV-infected patients (P<0.0001). Among females, the overall fracture prevalence was 2.49 versus 1.72 per 100 persons in HIV-infected versus non HIV-infected patients (P=0.002). HIV-infected females had a higher prevalence of vertebral (0.81 vs. 0.45; P=0.01) and wrist (1.31 vs. 0.83; P=0.01) fractures per 100 persons compared to non HIV-infected females but had a similar prevalence of hip fractures (0.47 vs. 0.56; P=0.53). Among males, the fracture prevalence per 100 persons was higher in HIV-infected versus non HIV-infected patients for any fracture (3.08 vs. 1.83; P<0.0001), vertebral fractures (1.03 vs. 0.49; P<0.0001), hip fractures (0.79 vs. 0.45; P=0.001) and wrist fractures (1.46 vs. 0.99; P=0.001). Fracture prevalence was higher relative to non HIV-infected patients among African-American and Caucasian females and Caucasian males.
Conclusions: Fracture prevalence is increased in HIV-infected compared to non HIV-infected patients.
Introduction
An increased prevalence of osteoporosis and osteopenia has been reported in both men and women infected with HIV, but the mechanism and consequences of these changes are unknown. Numerous studies have evaluated bone density among patients with HIV infection (1-7) and have found the prevalence of
osteoporosis and osteopenia to be elevated in HIV-infected compared to control groups.(8-14)
A meta-analysis of studies comparing bone mineral density in HIV-infected to non HIV-infected patients demonstrated a pooled odds ratio of 3.7 for osteoporosis.(15) Although the increased prevalence of osteoporosis and osteopenia among HIV-infected patients has raised concern for increased fracture risk, few investigations have evaluated fracture rates among HIV-infected patients, and studies are largely limited to the case report level.(16) Two smaller studies compared fracture rates in HIV versus non HIV-infected
patients.(17)(18) by patient interview. To our knowledge, fracture prevalence
in HIV-infected patients has not been compared to that of control patients within a large healthcare system, using specific coded diagnoses for outcome ascertainment assessing fracture site and demographic subgroups.
Methods
Data Source and Patient Sample
Data were obtained from the Research Patient Data Registry (RPDR), a clinical care data registry capturing all data from the Partners HealthCare System, which includes two primary hospitals, Brigham and Women's Hospital (BWH) and Massachusetts General Hospital (MGH). Complex queries can be created and
submitted to the RPDR, and a set of patients matching the search criteria is returned as an aggregate de-identified summary number that is within 3 patients of the actual value, for security purposes. We queried the RPDR for all patients
with at least one encounter at BWH or MGH between October 1, 1996 and March 21, 2008, conducting separate searches for patients with and without HIV infection. Inpatient and outpatient encounters were included. Within the
HIV-stratified groups, we identified patients with vertebral, hip or wrist fracture, or any of the three sites of fracture, at any time during the study period. Patients with multiple fractures were counted only once. All searches were stratified by gender, and further stratified by age group and race using RPDR search functions.
Exposure and Outcome Ascertainment
HIV infection was identified using International Classification of Diseases, Ninth
Revision, Clinical Modification (ICD-9-CM) codes of either 042 (human immunodeficiency virus disease) or V08 (asymptomatic HIV infection status) recorded at any time. Patients with no code of 042 or V08 were presumed to be
HIV negative. Fracture outcomes were determined using ICD-9-CM codes selected on the basis of the likelihood of their being related to osteoporosis.(19) Vertebral fractures were identified using codes 805.2 (closed fracture of
dorsal [thoracic] vertebra), 805.3 (open fracture of dorsal [thoracic] vertebra), 805.4 (closed fracture of lumbar vertebra), 805.5 (open fracture of lumbar vertebra), 805.6 (closed fracture of sacrum and coccyx), 805.7 (open
fracture of sacrum and coccyx). Hip fractures were identified using codes 820.0 (transcervical fracture, closed), 820.1 (transcervical fracture, open), 820.2 (pertrochanteric fracture of femur, closed), 820.3 (pertrochanteric fracture of femur, open), 820.8 (fracture of unspecified part of neck of femur, closed), 820.9 (fracture of unspecified part of neck of femur, open). Wrist fractures were identified using codes 814.0 (closed fractures of carpal bones), 814.1 (open
fractures of carpal bones), 813.4 (fracture of lower end of radius and ulna, closed), 813.5 (fracture of lower end of radius and ulna, open).
A chart review validation of ICD-based fracture ascertainment was performed by a trained clinical research nurse on a subset of charts. The sensitivity was 100 percent for both hip and vertebral fractures, and the specificity was 84 percent for hip fractures and 79 percent for vertebral fractures.
Data Analysis
This population-based study employed database query to identify patient groups for comparison of fracture prevalence according to HIV status and gender, age, race, and site of fracture. Period prevalence between October 1, 1996 and March 21, 2008 was calculated as the number of patients with fractures per 100 persons during the given time period for each subgroup. Searches returned the number of unique patients who met search criteria. Chisquare analysis was used to calculate P values for dichotomous variables. Demographic differences between the background HIV-infected and non HIV-infected populations were
accounted for by a series of stratified analyses comparing prevalences. Statistical analysis was conducted using JMP statistical software version 5.0.1 (1989-2003, SAS Institute). The study was approved by the Partners Human Research Committee. A P value of < 0.05 was considered to indicate statistical significance.
Results
The background population who presented for care at either BWH or MGH during the study period was comprised of 8,525 HIV-infected and 2,208,792 non HIV-infected patients. The demographic characteristics of these groups stratified by gender are shown in Table 1a. A greater proportion of HIV-infected patients were male compared to the non HIV-infected patients. Within both female and male
groups, there were higher proportions of HIV-infected patients in the 30-39, 40-49, and 50-59 age groups compared to non HIV-infected patients. In gender-stratified comparisons, the HIV-infected group also had higher proportions of African-American and Hispanic and lower proportions of Caucasian patients compared to the non HIV-infected group.
Table 1b summarizes fracture prevalence comparing HIV-infected to non HIV-infected patients. Overall fracture prevalence (including any of the three sites: vertebral, hip, or wrist) was 2.87 patients with fractures per 100
persons (95% CI 2.52-3.23) in the HIV-infected group compared to 1.77 (95% CI 1.75-1.79; P <0.0001) in the non HIV-infected group. All data shown represent number of unique patients with the specified type of fracture. Similar significant differences between HIV and non HIV-infected patients were seen for vertebral
(1.01 [0.80-1.22] vs. 0.47 [0.46-0.48]; P<0.0001), hip (0.72 [0.54-0.89] vs. 0.51 [0.50- 0.52]; P=0.01, and wrist (1.38 [1.14-1.63] vs. 0.90 [0.89-0.91]; P<0.0001) fractures. Overall fracture prevalence (HIV vs. non HIV-infected) was 2.49 per 100 persons (95% CI 1.93-3.05) vs. 1.72 (95% CI 1.69-1.74; P=0.002) for women and 3.08 (95% CI 2.62-3.53) vs. 1.83 (95% CI 1.81-1.86; P<0.0001) for men.
Fracture prevalence according to anatomical site is shown in Figure 1, Panels A
and B. In gender-stratified comparisons of HIV-infected versus non HIV-infected patients, HIV-infected patients had significantly higher fracture prevalence in all categories, with the exception of hip fractures among women, where prevalence between HIV-infected and non HIV-infected patients was similar. Fracture prevalence according to gender and further stratified by age (Figure 1, Panels C and D) or race (Figure 1, Panels E and F) was also determined. Fracture prevalence was relatively higher across all age categories in each gender, though confidence intervals (Table 1b) were wide for a number of specific age groups. P values for the overall comparison among men and women, including data from all age groups, was highly significant for women (P=0.002) and men (P<0.0001). Fracture prevalence was higher for HIV-infected than non HIV-infected patients for African-American and Caucasian women and Caucasian men.
In a secondary analysis stratified by gender, followed by age group, followed by race, HIV-infected patients demonstrated a similar pattern of higher fracture prevalence compared to non HIV-infected patients in every group, with the exception of Hispanic men age 50-59 (data not shown). A series of secondary analyses were conducted limiting the study duration to three-year periods (1997-1999, 2001- 2003, and 2005-2007) to minimize the potential
time difference between HIV diagnosis and fracture, and the results were similar to the primary analysis, demonstrating higher fracture prevalence in HIV-infected patients compared to non HIV-infected patients (3.91 vs. 2.77 in 1997-1999; 3.61 vs. 2.51 in 2001-2003; and 3.25 vs. 2.09 in 2005-2007, HIV-infected vs. non HIV-infected). A significant temporal time trend in fractures was not observed comparing fractures by individual study year, from 1996 to
2008, among the HIV-infected patients. We also assessed the proportion of patients in each group with coded outpatient visits, and this proportion was more than 90 percent in both the HIVinfected
and non HIV-infected groups.
Discussion
Fracture prevalence was higher in HIV-infected patients compared to non HIV-infected patients, in this large population-based study drawing from more than eleven years of data from a large U.S. healthcare system. HIV-infected patients had a significantly higher prevalence of vertebral, hip, wrist, and combined fractures compared to non-HIV-infected patients. Within both genders, fracture prevalence was higher in HIV-infected versus non HIV-infected patients for the majority of sites assessed, across age categories, and most race categories.
The results of multiple studies have established an increased prevalence of osteoporosis and osteopenia in patients with HIV disease, and reduced bone mineral density is likely to translate to an increased risk of fracture.
For example, in non HIV-infected patients, fracture risk is thought to increase approximately two-fold for each one standard deviation reduction
in bone mineral density (BMD).(20) However, very few studies have focused on the outcome of fracture, comparing HIV-infected to non HIVinfected
patients.(17, 18) Furthermore, the studies comparing fracture rates examined relatively small cohorts of patients and used history-based outcome ascertainment.
We used a large clinical care data registry of more than 2.2 million patients, including more than 8,000 HIV-infected patients, to obtain a global estimate of fracture prevalence in HIV-infected patients compared to a background
population receiving care in the same healthcare system. Using ICD codes for outcome ascertainment avoids the potential for recall bias or inaccurate patient information obtained in history-based outcome ascertainment.
Furthermore, we selected specific ICD codes to reduce the likelihood of misclassification. Our results demonstrate modest but significant increases in fracture prevalence, among both men and women, consistent with data demonstrating moderate but significant reductions in bone density in these groups.(2, 11)
Of note, fracture prevalence appears to increase with age in both HIV-infected men and women. Comparison with a control population is critical in this regard. Fracture rates in the background population of over 2 million patients increase to a greater extent with age in women, and this increase is shown to occur as anticipated at the time of the perimenopausal transition.(21)
Among both HIV-infected men and women, agerelated increases in fracture rates appear to be greater than in the control population, suggesting that fractures may increase as the HIV-infected population lives longer and ages.
Due to database constraints, we could not investigate antiretroviral medication use, smoking, alcohol use, body mass index, socioeconomic status, or medications affecting bone metabolism such as estrogens or steroids. However, we were able to perform gender-, age- and race-stratified analyses demonstrating increased fracture rates among Caucasian males and females and African-American females, at most age ranges. We could not assess if fractures were specifically traumarelated, but the prevalence of trauma by ICD code was less than one percent for both the HIV-infected and non HIV-infected groups. Moreover, this was a population-based fracture prevalence study, and individual bone density data were not available, preventing formal multivariable regression analyses. Therefore, our data provide novel information on prevalence but do not permit determination of mechanism, and further studies will be necessary to assess specific causes. In this regard, it will be critical to investigate factors related to the menopausal transition in women and aging in men, as well as factors related to use of specific antiretroviral drugs and metabolic abnormalities associated with use of antiretroviral therapies.(10, 22) Mitochondrial dysfunction has been associated with the use of nucleoside reverse transcriptase inhibitors,(23) and it is interesting to speculate that this may contribute to a "premature aging" in HIV-infected patients which may contribute to reduced BMD and increased fracture rates.
In both HIV-infected and control populations, fracture prevalence was greater for
Caucasian patients, consistent with the known differences in bone density in African American and Caucasian patients.(24) Among women, a greater increase was seen for vertebral fractures relative to control than for hip fractures, potentially suggesting factors that affect trabecular bone. However, wrist fractures were also increased relative to controls in women, and all three categories (hip, wrist and vertebral) were increased for HIV-infected men.
To our knowledge, these data are the first to compare fracture prevalence between HIV-infected and non HIV-infected patients with a large patient sample using ICD-based outcome ascertainment. The results provide strong evidence that HIV-infected patients have a higher prevalence of fractures than non HIV-infected patients, across both genders and critical fracture sites. Moreover, our data suggest that the relative difference in fracture prevalence between HIV-infected and non HIV-infected patients increases with age for both genders. As the HIV-infected population ages, reduced bone mineral density and increased fracture risk may become an even greater problem. Whether increased fractures are the sequelae of antiretroviral therapy, increased rates of traditional risk factors such as low weight among HIV-infected patients, or HIV infection - and its accompanying metabolic and inflammatory disturbances - itself remains to be determined. This study suggests the importance of assessing bone density and minimizing factors contributing to increased fracture risk in the HIV-infected population.
|
|
| |
| |
|
|
|
