Analysis of Site Heterogeneity and HIV Outcomes Across Rural and Urban Study Sites in Phidisa II - A Multi-site Randomized Controlled Antiretroviral Treatment Trial in a South African Military Cohort.

BACKGROUND: Clinical trials frequently enroll subjects from different study sites. Few such studies provide analysis by individual site. Between 2004-2007, the South African Military Health System (SAMHS) established 6 research sites (3 urban, 3 rural) to build capacity for clinical research and HIV care. We explore differences in clinical, virologic and CD4 outcomes by site in the context of a randomized controlled trial.

METHODS: Phidisa-II is the first randomised controlled trial conducted in the South African military setting, which compared 4 antiretroviral regimens in treatment-naïve advanced HIV subjects. Primary study outcome was first AIDS event or death. Kaplan-Meier curves for AIDS events and mortality were compared across sites. Hazard rates were adjusted for baseline risk factors to assess the independent effect of site. Secondary outcomes of CD4 count and viral responses are also compared across study sites.

RESULTS: 1,771 subjects [average age=35.4 ± 5.5 years old, 68% male, with median CD4 count=105 (IQR 41, 157) cells/mm3 and HIV RNA=144,000 (IQR 53,900-305,000)copies/mL] enrolled in 3 urban and 3 rural sites. Sites varied considerably in resources and diagnostic capacities. After adjusting for baseline characteristics, study site was found to be a factor significantly associated with mortality (p=0.008), with Urban 2 and Rural 2 sites had the lowest mortality. Site was also associated with the adjusted hazard for AIDS events (p=0.038). At 24 months, CD4 count was similar across sites, but HIV suppression rate varied considerably (range 40-70%).

CONCLUSION: Site heterogeneity was found in primary clinical outcomes of mortality and AIDS event rates, but there were no clear patterns for differences between the rural versus urban sites. Site differences were also found in the proportion of confirmed AIDS events. Factors within study sites that may have contributed to poorer outcomes need further investigation.