Clinical and immune responses were compared among women with and without a history of NVP use who were currently receiving NNRTI-containing ART across 26 public ART clinics in Zambia. This setting is well suited for such an evaluation because of robust programs for both PMTCT [3
] and ART provision [18
]. At the start of 2007, 52 919 individual doses of NVP had been distributed to HIV-infected pregnant women, while 26 590 women had initiated ART.
Across all ART sites, clinical flow, patient care, and medical documentation have been standardized in accordance with the Zambian national guidelines for adult HIV treatment [19
], which closely follow those outlined by the World Health Organization (WHO) [20
]. In the first months of the program, ART was initiated in individuals with either CD4 cell count <200 cells/μl or WHO stage 3 or 4. Since 2005, the Zambian guidelines have been revised to exclude ART eligibility for patients in WHO stage 3 with CD4 cell count >350 cells/μl. First-line drug regimens in Zambia comprise two nucleoside reverse transcriptase inhibitors (lamivudine with either zidovudine or stavudine) and one NNRTI (NVP or efavirenz).
In our setting, routine testing for HIV viral load or viral drug resistance is not available as part of treatment monitoring. Instead, patients are switched to second-line drug regimens based on clinical and immunological (CD4 cell) criteria, such as the development of new opportunistic infections or the return of CD4 cell counts to (or below) pretreatment levels. Since several months are required for immunological reconstitution, a patient must be adherent to ART for a minimum of 3 months before a second-line regimen is considered [20
At each site, patient information is collected on a comprehensive set of clinical care forms based on national treatment guidelines. Each patient visit generates a paper form, and selected data from that form are entered by clerical staff at the clinical facility. Through these procedures, key individual indicators are collected in a large database for monitoring and reporting, including demographic characteristics, medical history, laboratory results, pharmacy dispensations, and vital status [21
]. Previous exposure to single-dose NVP is ascertained through patient’s ART history, collected at the enrollment visit. In May 2005, the Lusaka District clinical care forms were revised to elicit more detailed history of ART, including date and number of NVP exposures for PMTCT. New patients are asked these questions at time of enrollment. For those already enrolled into care, this information was obtained at the next clinical visit following introduction of the new forms.
For the present analysis, patients were categorized as “active” if they attended their last scheduled clinic and/or pharmacy visit within 30 days of their appointment date. “Dead” patients were so classified only after confirmation from family members, friends, or clinic staff. “Inactive” patients had formally withdrawn from the HIV treatment program. “Late” patients were more than 30 days overdue for their last scheduled visit and could not be located via community health worker contact tracing. Censorship of patient outcomes began at the last patient contact for late patients. For inactive or dead patients, it was based on the date of report or date of death.
Outcome measures were mean CD4 cell count change, survival, and treatment failure. Patients were classified as having failed their first-line ART regimen if they met any of the following criteria: (1) worsening WHO stage after 3 months on ART, (2) fall in CD4 cell count to below 95% of the pretreatment level after at least 3 months on therapy, or (3) death. Comparisons were made between women previously exposed and unexposed to single-dose NVP. In order to determine what effect the timing of NVP exposure may have on patient outcomes, the exposed population was also stratified according to the interval between drug ingestion and ART initiation. The categories of “recent” (i.e., <6 months between NVP ingestion and ART initiation) and “remote” (i.e., ≥6 months) NVP exposure were used based on conventions established by previous work [17
To determine the optimal study population, women with known NVP status (i.e., exposed or unexposed) were grouped as: (1) those with NVP exposure status determined at time of enrolment, and (2) those with NVP exposure status missing at time of enrollment but determined at some point after initiating NNRTI-based ART. To assess similarity between these groups, the proportion of NVP-exposed women within each was compared. There was a higher proportion of women with NVP exposure among those whose status was determined after ART initiation than among those whose status was recorded at enrollment (40% versus 11%, P <0.0001). This was thought to result from ascertainment bias. Among women who had already started ART and returned for clinic visits later in care, it appeared that health-care workers were less likely to document status information if they denied previous NVP use. In an effort to minimize selection biases, the primary analysis was limited to women with NVP exposure status determined at time of enrollment (the “primary evaluation cohort”). The one exception was a supplementary comparison of recent and remote NVP exposure, where all women with known NVP exposure were included in the analysis. It was reasoned that differences in exposure ascertainment were unlikely to influence comparisons within the NVP exposed group itself. Secondary analyses were performed on the larger cohort of women.
To compare various demographic and medical characteristics among members of our study population, categorical variables were analyzed using either Pearson’s χ2
test or Fisher’s exact test. Two-tailed t-tests were used for normally distributed continuous variables, while Wilcoxon rank-sum tests compared parameter medians when distribution was not believed to be normal. Log-rank tests were used to evaluate differences in survival and treatment failure using Kaplan–Meier analysis. Cox proportional hazards models were used to assess the risk of death and treatment failure [22
]. These multivariate analyses were performed with covariates significant at a P
≤0.10 level in bivariate analysis comparing NVP-exposed and NVP-unexposed women. All analyses were performed using SAS version 9.1 (SAS Institute, Cary, North Carolina, USA). This analysis, using routinely collected clinical data, was deemed exempt [23
] from human subjects review by the Institutional Review Boards of the University of Zambia (Lusaka, Zambia), the University of Alabama at Birmingham (Birmingham, Alabama, USA), and the US Centers for Disease Control and Prevention (Atlanta, Georgia, USA).