Prior studies examining myocardial infarction rates among HIV patients have been limited by several factors, including the absence of a reference population. The Data Collection on Adverse Events of Anti-HIV Drugs (DAD) study provided important information regarding the duration-related association of ARV therapy with AMI rates among HIV patients (11
). However, neither the DAD study nor another large study that prospectively evaluated the effect of PIs on the development of AMI (10
) included control groups for comparison in prospective analyses. Our study included comparative data and specifically assessed whether AMI rates are elevated in HIV patients compared with a non-HIV reference population. In contrast to the DAD study, the purpose of this study was not to determine the relationship of AMI rates to ARV use because detailed information on the longitudinal use of ARVs could not be captured from the database but, rather, to compare myocardial infarction rates between HIV and non-HIV patients, followed at two large U.S. hospitals.
The rate of AMI per 1000 PY was higher among HIV patients in this study than in the DAD study, but our study included older patients as well and was from a U.S. population, with potentially different AMI rates and cardiovascular risk factors than the European-based population of the DAD study. Furthermore, the background rate among the entire reference population was similar to that seen in other large U.S. databases, including the Framingham Heart Study, suggesting that this study reliably captured AMI events across the two large hospitals for which we have data. Importantly, the RR of AMI in HIV patients was approximately 2-fold increased over controls and, thus, likely to be clinically significant.
Several other studies that have included reference populations have used endpoints whose broad or composite nature may have decreased specificity. Two large studies that determined cases based on insurance claims from Medicaid and health maintenance organization data used coronary heart disease (CHD) as an endpoint (12
). When evaluated for AMI instead of CHD, one study showed only a trend toward higher rates in the HIV group (13
). A second study demonstrated increased CHD incidence among younger patients ages 25–34 using one of five ICD codes for ischemic heart disease (12
). A study of patients in the Veterans Affairs system, from an earlier time period (1993–2001), used five composite endpoints and showed decreased hospitalization rates but did not distinguish cardiovascular from cerebrovascular disease or AMI from CHD (14
). In contrast, our data using the specific ICD code for AMI as opposed to CHD, demonstrate increased AMI rates for all age groups relative to the non-HIV cohort. A study based on data from the Women’s Health Initiative showed strong agreement between hospital discharge code of myocardial infarction and validation of events by study physicians for the diagnosis of myocardial infarction (15
There was no difference in the overall unadjusted AMI rates in the comparison of HIV-infected with uninfected men. Among men, the RR became highly significant in a gender-stratified model after adjusting for age. With adjustment for all demographic and clinical variables, the RR of AMI remained highly significant in the comparison of male HIV vs. non-HIV groups. As expected, the RR was attenuated by approximately 17% when adjusting for hypertension, diabetes, and dyslipidemia. Both the unadjusted event rates and the adjusted RR were higher in the comparison of female HIV to non-HIV groups. In the gender-stratified model for women, adjustment for age increased the RR in a pattern similar to that of men. However, adjustment for hypertension, diabetes, and dyslipidemia attenuated the RR by 46% among women.
Prior studies have shown that women demonstrate increased cardiovascular disease risk indices associated with fat redistribution (16
), but only limited information is available regarding the risk of cardiac events in HIV-infected women. Two major studies excluded women due to a small number of outcomes (10
). Data from a study based on California Medicaid claims suggested an increased AMI incidence among HIV-infected women ages 25–34 (12
). In this study, we conducted both combined regression and gender-stratified models. In both male and female models, including age, race, and common cardiac risk factors, HIV status was a significant predictor of AMI, but the RR of AMI was higher in women (RR 2.98) than in men (RR 1.40), possibly because of altered inflammatory marker profiles (16
), significant changes in body composition in this group, and relative shift from gynoid to android fat distribution (7
), or other factors. Indeed, our data suggest that the rates of some common cardiac risk factors, such as diabetes, may be relatively greater in women.
Age and race were significant independent risk factors for AMI. The risk of AMI increased with age in both the HIV and non-HIV groups, and increased AMI rates were seen for all ages in the comparison of HIV and non-HIV patients. These data suggest that cardiovascular disease will likely become an increasing problem for HIV-infected patients as the population lives longer and ages. Race was also significant in the overall model as an independent predictor of AMI. AMI risk was significantly increased among African-Americans relative to Caucasians and in Hispanics relative to Caucasians. Among HIV patients, African-American race was significant as a predictor of AMI. Prior data suggest that there is an association between race and cardiovascular disease, with higher median rates of cardiovascular events (defined as AMI, angina, CHD, or stroke) compared with the national average for African-American women, and lower rates for Hispanic women and men and African-American men (17
). Racial differences between the HIV and non-HIV groups were accounted for in Poisson regression modeling, which showed a significant RR of 1.75 for AMI in the HIV group relative to the non-HIV group.
Metabolic changes are common in HIV-infected patients and likely play a role in the development of atherosclerosis, possibly in concert with other cardiac risk factors. Our study demonstrates increased rates of traditional comorbidities, including hypertension, diabetes, and dyslipidemia, seen at rates of 21.2, 11.5, and 23.3 per 100 persons in the HIV cohort, respectively. An increased rate of dyslipidemia has been well documented in smaller studies of HIV-infected populations, and may relate to effects of HIV as well as ARV therapy (18
). Among the Swiss HIV cohort, classification of hypercholesterolemia was seen in approximately one third of patients on PIs (8
). The dyslipidemia rate of 23.3 per 100 persons seen in our study was determined from a large HIV cohort not selected for PI use, providing further evidence for the widespread prevalence of dyslipidemia in this population. Dyslipidemia was an independent risk factor for AMI in separate models of HIV and non-HIV, and in the fully adjusted model of the combined groups (RR 3.03; 95% CI 2.93–3.13; P
< 0.0001), controlling for age, gender, race, hypertension, and diabetes.
In this study we also demonstrate increased rates of diabetes and hypertension in the HIV compared with the non-HIV cohort. Increased risk factors were seen in both the male and female comparisons, in gender-stratified analyses. A recent study showed a 14% prevalence rate of diabetes in HIV-infected men using ARV therapy compared with 5% in seronegative men (3
). Our data confirm and extend these data, showing an approximate 2-fold increase in diabetes rates in HIV-infected patients relative to non-HIV patients. Only limited data are available regarding hypertension rates. Another group demonstrated elevated rates of hypertension in a small study of HIV-infected patients (20
). Our data add further emphasis to the potential role played by this and other cardiac risk factors. Both diabetes and hypertension were independent risk factors for AMI in the fully adjusted model of the combined groups (RR 1.98, 95% CI 1.93–2.04, P
< 0.0001 for diabetes; and RR 1.62, 95% CI 1.57–1.66, P
< 0.0001 for hypertension). Sensitivity analyses suggested that diabetes and dyslipidemia, more than hypertension, contributed to the increase in AMI rates in the HIV group.
Relatively little is known about the risk of smoking and cardiovascular disease in HIV patients. Among non-HIV groups, numerous cohort studies demonstrate increased rates of CHD for smokers, with RRs among current smokers as opposed to never-smokers ranging from 1.3–3.5 (21
). The DAD study (11
) suggested that smoking was an independent risk for AMI in HIV patients but did not perform a comparison with control patients, controlling for smoking use. The estimated common odds ratio obtained from the analysis by smoking status in patients for whom smoking data were available in our study was comparable with that obtained using the multivariate Poisson regression model approach. However, due to database constraints, smoking data were available in only a subset of the patients and, thus, could not be added to the overall model. It is possible that increased smoking rates in the HIV group contribute to the increased rate of AMI in the HIV group. Other factors not measured, including inflammation (22
), direct effects of the virus, or ARV therapy (4
), may also contribute to the increased risk of AMI demonstrated in the HIV group.
This study investigates AMI events in both HIV and non-HIV patients followed longitudinally at two large U.S. hospitals. Furthermore, adequate women were studied to develop gender-stratified models, showing that HIV status was a relatively greater predictor of AMI among women than men. Background comparison rates for AMIs in the entire cohort of over 1 million are similar to those reported in other large outpatient databases (26
), suggesting that the database reliably captured AMI events. The relatively long and similar duration of follow-up between HIV and non-HIV groups and exclusion of all patients with a single isolated encounter, or initial encounter being an AMI, ensures that rates were determined in patients receiving longitudinal follow-up in the system, rather than patients acutely referred in for AMI assessment and treatment.
The study was limited by the constraints of the database. Overall rates of ARV use were similar to other large cohorts, but we could not investigate the contribution of ARV drugs to myocardial infarction rates in HIV-infected patients because specific data on duration of ARV use were not available through the database. Rather, the primary purpose of the study was to investigate myocardial infarction rates and major cardiac risk factors among patients with HIV disease in comparison with a large non-HIV cohort, controlling for common cardiac risk factors, age, race, and gender. Smoking data were assessed in a subanalysis among patients for whom these data were available, demonstrating that the higher AMI risk remained in HIV patients despite higher smoking rates. However, due to a lack of uniformly available data, this information was not able to be included in the overall regression model. Use of ICD coding to define AMI is standard in large epidemiological studies but is subject to misclassification bias. We performed a sensitivity analyses based on data obtained from a validation subset to determine the effect of possible misclassification bias. The hazard ratio in the sensitivity analysis, testing a 50% maximum misclassification rate, was 1.69, compared with 1.75. In addition, hypertension and dyslipidemia could only be included as dichotomous rather than quantitative traits because we were limited to the use of ICD codes in the study, and this may have limited the strength of the covariate analysis. Thus, it is not possible to know whether the differences between HIV and non-HIV after adjustment are due to HIV per se or the limited nature of covariate information in the database.
In summary, our data from a large cohort of HIV-infected patients demonstrate increased AMI rates and cardiovascular risk factors compared with a non-HIV population, particularly among women. The AMI event rate was approximately 2-fold elevated in the HIV cohort, and this difference was seen over multiple age ranges. Relative myocardial infarction rates were greater among women, in comparison with the non-HIV cohort, and this gender-specific difference may account in part for the overall differences we observed. Determining the mechanisms of increased cardiovascular disease and cardiac risk factor rates in HIV-infected women is an important area of future research. Cardiac risk modification strategies are also particularly needed and will be an important component of the long-term care of this population.