The Veterans Aging Cohort Study (VACS) is an ongoing prospective cohort study involving HIV-infected and HIV-uninfected veterans receiving care at VHA clinics throughout the United States. The overarching aim of VACS is to study the role of alcohol consumption and comorbid medical and psychiatric disease on clinical outcomes in HIV infection.19
VACS participants were recruited from infectious disease (HIV-infected) and general medicine (HIV-uninfected) clinics at eight sites (Atlanta, GA; Baltimore, MD; Bronx, NY, Houston, TX; Los Angeles, CA; New York City, NY; Pittsburgh, PA and Washington, DC).19
Overall, 58% of HIV-infected men at the eight sites were enrolled, with only 9% of those approached refusing to participate.19
HIV uninfected controls were targeted to match the demographics of the HIV-infected participants on 5-year age blocks, race and gender. Subjects completed a comprehensive baseline survey at enrollment and then at one year follow-up intervals. Further descriptions of the VACS sample and methodology are available online (www.vacohort.org
The study sample for these analyses includes the subset of male VACS participants who completed a follow-up survey between September 2005 and January 2007 and who reported any sexual activity in the past year (n
2,787). Data were obtained from three linked sources: participant surveys, electronic medical records, and pharmacy data that are collected nationally through the Pharmacy Benefits Management (PBM) program (Hines, IL). The PBM program includes all outpatient prescriptions funded through the VHA healthcare system and are likely to be representative of the use of prescribed EDD in this population.
Measures EDD use was defined as two or more prescriptions for sildenafil citrate, tadalafil, or vardenafil HCL, documented in the PBM database in the year prior to and up to the follow-up survey date. Thus, men who had received only one EDD prescription were not considered to be EDD users, because such men may have never used the medication, or simply tried it once. Although men were not asked about non-prescribed use of EDD at this assessment, a previous assessment in the same cohort demonstrated substantial agreement between self-reported EDD use and the PMB database, with fewer than 5% of men reporting EDD use exclusively from non-VA sources (unpublished data).
Risky sexual behavior was defined as “unprotected sex with a partner of serodiscordant or unknown HIV status”. Specifically, men were asked, “During the past 12 months, did you ever, even once, have unprotected vaginal or anal sex (sex without a condom) with any of the following types of partners? Any partner who was HIV positive; Any partner who was HIV negative; Any partner whose HIV status was unknown.” The following were the response options: Yes (unprotected sex at least once); or No (always used a condom). Those who reported unprotected sex with anyone of serodiscordant or unknown HIV status were classified as having risky sexual behavior.
STDs were identified by self report. Participants were asked, “In the past 12 months, have you been diagnosed with any of the following sexually transmitted diseases?” Persons were classified as having had a self-reported STD if they reported having genital warts, chlamydia, gonorrhea, syphilis, trichomonas, chancroid, or herpes.
Participant demographic characteristics, including gender, race/ethnicity, and marital status/living with partner were measured by self-report on the VACS baseline survey. Measures of alcohol and drug use and depression were included on the VACS follow-up surveys. Hazardous drinking (drinking associated with possible harm) was defined as a score of eight or more on the Alcohol Use Disorders Identification Test (AUDIT).20
Participants were asked to report the frequency of use of marijuana, cocaine, stimulants, and heroin; current use for these drugs was defined as “at least monthly.” Non-prescribed use of pain medications, defined as use of pain medications that were not prescribed, was based on self-reported use in the past year from a list of 20 specific narcotic pain medications. Depression, measured by the Patient Health Questionnaire (PHQ-9) at each follow-up assessment, was classified as present if the PHQ-9 score was 10 or more.21
Comorbid medical conditions, including diagnoses of hypertension, diabetes, and coronary artery disease (CAD), were determined using International Classification of Diseases, Ninth Revision (ICD-9) codes from the electronic medical record. Persons were considered to have a comorbid diagnosis if at least one inpatient or two outpatient ICD-9 code diagnoses were recorded between one-year prior to and six months after the survey date. Persons were classified as having sex with men if they reported having “sex with males” or “sex with males and females” in the past year. Further details on variables and surveys can be found at www.vacohort.org
Demographic and behavioral characteristics were described and compared by HIV status using chi-square tests and t-tests, as appropriate. We used chi-square tests to determine whether EDD receipt varied by demographic and descriptive variables and to determine the bivariate associations of EDD receipt to risky sex and STDs.
Multivariable logistic regression models were used to identify independent factors associated with EDD receipt, risky sexual behavior, and STDs. Logistic regression models predicting EDD receipt were adjusted for demographic variables (age, race/ethnicity, married/living with partner, sex with males in past 12 months), alcohol and drug use, and comorbidities identified as potential confounders (depression, diabetes, hypertension, and coronary artery disease). The logistic regression models predicting risky sexual behavior and STD included the same variables plus EDD receipt. There was significant collinearity between the sex with males and HIV status variables; therefore, we ran multivariate models stratified by HIV status. For each set of models, we first ran the multivariate logistic regression models including the variables described above. We then reran the models excluding variables with p
0.2 in both HIV-infected and uninfected models, and we present these models in the results. We also examined the relationship of EDD to risky sexual behavior or STDs in the following subgroups: those with hazardous drinking, cocaine use, and sex with men. All analyses were conducted using Stata 10.0 (College Station, TX).
Multiple imputation was used to address missing data.22–24
For each of the following covariates, 1% or less were missing a response: married or living with partner, had sex with males, hazardous alcohol use, and depression. Pain medication use information was missing for 2%, and marijuana and cocaine use information was missing for 4%. Multiple imputation was conducted using the Stata v10.0 (Stata Corporation, College Station, Texas) ice command.22–24
The imputation model included EDD receipt, risky sexual behavior, and all covariates included in the initial models (listed above), undertook ten switching procedures, and generated five datasets. The Stata v10.0 mim command was used to combine the results of the analyses from the multiply imputed data sets.23
Analyses were also conducted on the complete case dataset (using categories to define missing data). Results from analyses of imputed data did not differ substantively from those using complete cases; therefore, the results from the imputation models are included here.