From June 1998 through October 1999, VaxGen, Inc. (Brisbane, CA) enrolled 4697 HIV-seronegative MSM into a 36-month, randomized (2∶1 vaccine∶placebo), double-blind, placebo-controlled phase III HIV-1 vaccine efficacy trial at 56 clinical trial sites in 47 cities in the United States. Enrolled MSM were healthy, 18–60 years of age, and reported having anal sex with ≥1 HIV positive or unknown serostatus male partner during the past year. Men in a monogamous relationship for ≥1 year with a HIV seronegative male partner or who reported injection drug use (IDU) in the preceding three years were excluded from enrollment. Recruitment strategies included advertising, outreach, and referrals from other cohort studies, from HIV-infected partners, and current trial participants.

HIV counseling and testing was conducted at enrollment, semiannually, and at participant request. Participants diagnosed with HIV infection during the trial were referred for medical care and social services. Nucleic acid testing (NAT) detected 13 HIV infections among MSM at enrollment; these men were discontinued from the trial resulting in 4684 MSM eligible for subsequent incidence analyses. Details of the trial baseline methods, study procedures, and efficacy results have been published elsewhere [17], [18].

The vaccine study protocol was approved by the institutional review boards (IRB) of Saint Louis University, John Hopkins University School of Medicine, Fenway Community Health Center, Philadelphia Fight, AIDS Research Alliance, Louisiana State University Medical Center, University of California, Irvine, University of California, San Francisco, University of Washington, Hennepin County Medical Center, Mount Sinai Medical Center, the University of New Mexico, the University Medical Center of Southern Nevada, Abbot Northwestern Hospital, New York Blood Center, New York Medical Center, Howard Brown Health Center, Ohio State University, the University of Texas, Galveston, the University of California, Davis, Community AIDS Resource, Inc., University of Hawaii, the AIDS Research Consortium of Atlanta, Erie County Medical Center, Santa Clara County Public Health Department, Albany Medical College, New Jersey Community Research Initiative, Duval County Health Department, University Hospitals of Cleveland, University of Alabama at Birmingham, Community Hospitals Indianapolis, Wisconsin AIDS Research Consortium, University of Pittsburgh, Miriam Hospital, Nalle Clinic, and Memorial Hospital of Rhode Island, the Colorado Multiple IRB, the Western IRB, and the IRB of the Centers for Disease Control and Prevention (CDC). All participants provided written informed consent for trial participation.

At enrollment, a standardized interviewer-administered questionnaire was used to collect information on demographics, medical history, and reasons for trial participation. Data on STI diagnoses, concomitant medications, and sexual risk behaviors (i.e., unprotected receptive or insertive anal intercourse [UAI]) with partners by HIV-serostatus (negative, positive, and unknown) in the previous 6 months were collected at enrollment and 6 month intervals. Participants were also asked if they had injected drugs during the past six months, or if they had used marijuana/hash, poppers, crack or cocaine, amphetamines (e.g., crystal or speed), tranquilizers/sedatives (e.g., valium, Xanax, Klonopin, barbiturates, Quaaludes), hallucinogens (e.g., PCP, Special K, angel dust, acid, LSD, mushrooms, ecstasy), heroin, or drugs to enhance sexual performance (e.g., Viagra, Yohimbine). Two questionnaire items designed to assess beliefs about HIV vaccine trial participation that may increase HIV risk behavior, such as perceiving a high degree of protection against HIV if the vaccine was shown to be efficacious and perceiving vaccine treatment assignment were asked at enrollment (the first item) and at 12, 24, and 36 months (the second item).

Specimens were tested using an enzyme immunoassay (EIA) at a central laboratory; EIA reactive specimens were confirmed by Western Blot. If HIV seropositivity was detected, a specimen available at the time of the last negative EIA was tested by NAT (Procleix HIV-1 Discriminatory Assay, Gen-Probe, Inc., San Diego, CA) to determine viremia at an earlier visit.

Incidence density was calculated by baseline characteristics based on person-years (py) of follow-up for HIV seronegative MSM who received at least one additional HIV test during the trial. The Breslow-Day test was used to assess heterogeneity in HIV incidence over time. Date of HIV infection was defined as the midpoint between the last negative and the first positive EIA or date of first positive NAT. A Cox proportional hazards model identified baseline variables predictive of loss-to-follow up. Baseline variables were evaluated to examine differences in reported risk behaviors between young (18–30 years of age) and older MSM using the Mantel-Haenszel [chi]² test. Potentially predictive (p<.20) characteristics for HIV seroconversion identified in univariate Cox proportional hazards models with time-varying covariates updated every 6 months and forced inclusion of treatment arm assignment were assessed in a multivariate Cox proportional hazards model including partner serostatus, numbers of sexual partners, UAI, recreational drug use, age, and STI. Analyses were conducted using SAS (version 8.0; SAS Institute, Cary, NC).

From June 1998 through October 1999, 4684 HIV-1 seronegative male trial participants enrolled at 56 clinical trial sites in 47 U.S. cities, encompassing 27 states, Washington, DC, and Puerto Rico. The median number of MSM enrolled per site was 75 (range, 5–303). Retention at 6, 12, 18, 24, 30, and 36 months was 96%, 92%, 89%, 87%, 85% and 82%, respectively. Loss-to-follow-up was independently associated with the following baseline characteristics: age 18–30 years (p<.0001), African American (p<.01) and Hispanic (p<.01) race, tranquilizer (p<.0001) and crack (p<.01) use. Retention was associated with popper (p<.01) and sexual performance enhancing drug (p<.05) use and completing college (p<.0001).

The baseline demographic and sexual risk behavioral characteristics of the U.S. MSM are presented in Table 1. The median age was 35 years, range 18–62; 25% were 18–30 years of age. Men were predominantly of white race/ethnicity, and most reported attaining at least one college degree. The majority of MSM were enrolled at clinical sites in the southern or western United States. The median number of male sex partners in the 6 months before enrollment was 5 (range 0–960 partners); 32% reported having ≥10 male partners. In the previous six months, 10% reported having a STI, and over half reported UAI. IDU was minimal (0.23%), however 61% reported non-injection recreational drug use (e.g., amphetamines, tranquilizers, poppers, cocaine, hallucinogens, and sexual performance enhancing drugs), and 87% reported alcohol use.

Young men (≤30 years of age) were less likely than older men to report a male HIV-infected partner at enrollment, but were more likely to report any receptive UAI 41% vs. 35% (p = <.001). They were also less likely than older men to report use of poppers or sexual performance enhancing drugs, but more likely to report using marijuana, amphetamines, cocaine, and hallucinogens. Furthermore, over the course of the study, amphetamine use significantly increased among young men (from 11% at baseline to 16% at 36 months; p = <.01). The use of other recreational drugs did not differ by age or time.

A total of 338 HIV infections were detected among the 4684 MSM during the trial. Fifty-one sites reported ≥1 case of HIV infection. Participants who became infected with HIV were predominantly white (85%); 60% reported a college degree. The median age at the time of infection was 36 years (range, 20–62 years), and the median time to infection after study enrollment was 16 months (range, .6–35 months).

Overall, HIV seroincidence was 2.7/100 py (95% CI, 2.5, 3.0) and was relatively consistent across study year (1998–99, 78 cases, 2.5/100 py, 2000, 137 cases, 3.2/100 py, 2001, 101 cases, 2.7/100 py, and 2002, 22 cases, 1.8/100 py [p = .05]). Incidence did not vary by geographic region or race/ethnicity (Table 1), but was substantially higher among young MSM (3.7/100 py; 95% CI 3.0, 4.5). Twenty-three cities had HIV seroincidence rates above 2.7/100 py; twelve cities exceeded 3.5/100 py (Figure 1). Although, HIV incidence was expectedly high among MSM from New York City and Denver, it was also high among MSM from Atlanta, Houston, Jacksonville, and Phoenix. HIV seroincidence was highest among MSM who reported receptive UAI with HIV positive or unknown partners at baseline, followed closely by any receptive UAI, and then any UAI with HIV positive or unknown male partners. Elevated HIV seroincidence rates were noted among men who reported baseline recreational drug use, especially amphetamines (5.0/100 py; 95% CI 3.8, 6.4) and crack (5.2/100 py; 95% CI 2.7, 8.8).

In univariate analysis, young men, men who reported a STI, or recreational drug use were more likely to seroconvert (Table 2). Variables, such as education level, geographic region, and race/ethnicity, and treatment arm assignment (HR = 1.01; 95% CI 0.78, 1.31) were not significant. Sexual risk behaviors predictive of HIV infection included: multiple male sex partners, reporting ≥1 HIV positive male sex partners and reporting UAI with HIV positive or unknown male partners. Perceived vaccine efficacy and perceived treatment assignment were not predictive of HIV seroconversion, p = .47, and p = .62, respectively.

In the multivariate time-varying model, significant predictors of HIV seroconversion included younger age, reporting greater numbers of male sex partners, having at least one HIV-infected male sex partner, UAI with HIV positive or unknown male partners during the preceding 6 months, and amphetamine and popper use (Table 2). The following variables were included in the model but were not significant: date of study entry, race, education level, geographic region, and treatment arm assignment (aHR = 1.04; 95% CI 0.80, 1.36).