Cardiovascular disease (CVD) has been associated with HIV infection. However, it is uncertain whether increased CVD rates are associated with HIV-related factors (e.g., HIV-infection or highly active antiretroviral therapy (HAART) may worsen dyslipidemia) or reflect differences in the prevalence of underlying risk factors for CVD. Furthermore, the association between initiation and duration of HAART exposure and CVD risk, including which specific drugs within the HAART classes may contribute to the increased risk, is unknown. The primary objectives of the study are therefore:

1. To estimate the absolute and relative incidence rate (IR) of CVD claims-based diagnoses among a cohort of adult patients from a large managed care population with a claims diagnosis of HIV, AIDS, or AIDS-related complex (ARC) during periods of exposure to:

• Any HAART compared to no HAART exposure
• HAART class [i.e., NRTIs, NNRTIs, PIs, and Other (i.e., fusion inhibitors)] compared to no HAART class exposure
• Specific NRTI medications compared to no specific NRTI exposure

The scientific approach is to conduct a retrospective cohort study of patients with a claims diagnosis of HIV, AIDS, or ARC (simply referred to as HIV infection) among commercially insured people in the U.S. to estimate the IR of acute MI (AMI), AMI including coronary revascularization (CR) procedures, and CVD, comparing exposure to HAART treatments. We will identify a primary cohort of patients with an initial claims diagnosis of HIV infection occurring after at least 6 months of continuous enrollment in the Ingenix National Health Informatics (NHI) or Ingenix Impact National Managed Care (Impact) Databases between 01 January 1998 and 31 December 2007. For each eligible primary cohort member, we will identify all HAART dispensings during the baseline and follow-up periods and categorize person-time of exposure by any HAART, HAART class, and specific NRTI medications. We will also specifically evaluate the risk associated with fosamprenavir and amprenavir use. Study outcome events, including AMI, AMI including CR procedures, and CVD, will be identified from the medical claims data for all patients during the follow-up period.

For the primary analysis, we will estimate the absolute IR and 95% confidence intervals (CIs) of each study outcome event during periods of recent, past, non-use, and cumulative duration (non-use, < 1 year, 1 - 2 years, and > 2 years) of each HAART exposure classification. We will also estimate the rate difference (RD) and 95% CI of each study outcome event during periods of recent, past, and cumulative duration (< 1 year, 1 - 2 years, and > 2 years) of each HAART exposure classification relative to non-use. We will use Poisson regression models to estimate the multivariable adjusted rate ratio (RR) and 95% CIs of each study outcome event during periods of recent, past, and cumulative duration (non-use, < 1 year, 1 - 2 years, and > 2 years) of HAART use relative to non-use for each HAART exposure classification, adjusting for confounding variables derived from the claims history of each person during their 6-month baseline period to account for channeling bias and during the follow-up period to account for changing risk factors for CVD. To further account for confounding of indication by abacavir use, analyses will be stratified by calendar year 2004, during which time there was a significant change in the primary indication for abacavir use in HIV. Analyses will be repeated as a secondary analysis on the subgroup of patients in the NHI Database with any outpatient laboratory test result; these analyses will account for values of cholesterol, cluster of differentiation 4 (CD4) count, and viral load [HIV-1 ribonucleic acid (RNA)] laboratory tests.

Among the primary cohort, we will identify a subcohort of patients with an initial claims diagnosis of HIV infection occurring after at least 6 months of continuous enrollment and without HAART medication in the 6-month baseline period prior to the cohort entry date. We will repeat the primary and secondary analyses among this naïve subcohort of HAART initiators.

ViiV Healthcare is the new sponsor of this study, and GlaxoSmithKline is in the process of updating systems to reflect the change in sponsorship.

Inclusion Criteria:

• All patients aged 18 years or older with an initial claims diagnosis of HIV infection associated with a physician visit occurring after at least 6 months (183 days) of continuous enrollment between 01 January 1998 and 31 December 2007 will be included in the primary cohort. Patients will be required to have commercial health insurance coverage and complete medical and pharmacy benefits. The cohort entry date for each patient will be limited to the first claim associated with a HIV infection diagnosis that meets the 6-month continuous enrollment criteria. Patients with a HIV infection diagnosis claim who do not have at least one continuously enrolled segment that is equal to or greater than 6 months prior to the cohort entry date will be excluded. Six months of continuous enrollment prior to the initial claim associated with a HIV infection diagnosis will be required to guarantee the availability of a 6-month baseline period for ascertaining exposure to HAART prior to cohort entry and covariates of interest for the Poisson regression models. Patients may enter the primary cohort already exposed to HAART and with a HIV infection diagnosis claim in the 6-month baseline period.

Exclusion Criteria:

-

Source population is from the proprietary Ingenix NHI Database, a large, geographically diverse population of U.S. health insurance plan enrollees. Records are organized into a medical file containing claims from providers and facilities, a pharmacy file with outpatient pharmacy dispensing records, and an enrollment file that provides demographic data and dates of insurance eligibility for persons in the database. Laboratory test results are also available for a fraction of the population. In addition, i3 Drug Safety will complement the data from the NHI Database with data from the Ingenix National Managed Care (Impact) Database, a fully de-identified, HIPAA compliant database from health insurance claims and enrollment data of more than 30 health plans separate from the plan affiliated with the NHI Database. Patients in each data source have a common unique identifier; therefore, it is possible to construct a combined population without concern that some patients may be duplicated.