A total of 217 HIV-infected women were enrolled in Bangkok, Thailand at two sites – Siriraj Hospital and Rajvithi Hospital [13]. At enrollment and at weeks 2, 4, 8, 16, and 24, participants had serum alanine transaminase (ALT) and aspartate transaminase (AST) measured and were evaluated clinically for signs and symptoms of hepatitis and rash. We graded adverse events including hepatotoxicity and rash according to the National Institutes of Health’s Division of AIDS guidelines as previously reported [14,15].

Cases in this analysis were women who during the first 24 weeks of treatment with nevirapine-based HAART developed 1) any grade of rash or 2) an increase of at least one grade from baseline in ALT or AST to at least grade 2. Only rash or increase in ALT or AST that the study team determined to be “possibly”, “probably”, or “definitely” related to nevirapine, considering character and timing of the reaction, was considered. Controls were selected as the next enrolled participant at the same study site who was confirmed not to have met the case definition during the first 24 weeks of treatment.

Clinical care was provided by physicians with expertise in HIV disease at each of the study hospitals. The study protocol did not direct individual treatment decisions and the clinicians used their discretion to continue nevirapine through a mild reaction, to interrupt nevirapine therapy and restart it, or to change therapy substituting either efavirenz or a protease inhibitor for nevirapine.

Blood specimens were collected in Becton Dickinson Vacutainer® tubes containing ethylenediaminetetraacetic acid as an anticoagulant and were stored at room temperature until transported on ice to the laboratory. Plasma was separated and frozen within six hours of blood collection. The time and date of specimen collection was recorded by study nurses. CD4 cell counts were measured as previously described [13]. We tested for hepatitis B surface antigen and for hepatitis C antibody with Murex Enzyme Immunoassays (Abbott Diagnostics, Dartford, UK). Stored plasma specimens collected at 2, 4, and 8 weeks of the study were assayed for nevirapine, stavudine, and lamivudine concentration.

Stavudine, lamivudine, and nevirapine concentrations in plasma specimens were determined using an LC-MS/MS method [High Performance Liquid Chromatography (HPLC, Acquity^TM Ultra Performance LC, Waters Co., Ltd. USA), Mass Spectrometer (Quattro Premier XE, Micromass Technologies, UK) and Masslynx v 4.1 software (Micromass Technologies, UK)]. The method had been developed and validated by the Department of Pharmacology, Faculty of Medicine Siriraj Hospital, Mahidol University (Bangkok, Thailand) according to the US FDA Guidance for Industry: Bioanalytical Method Validation (2001). The validation demonstrated that the method was reliable and reproducible for the intended use with respect to selectivity, sensitivity, linearity, accuracy, precision, and stability of the drugs in plasma. The assay had a linearity range of 1-8,000, 1-12,500, and 1-25,000 ng/mL for stavudine, lamivudine, and nevirapine, respectively.

Participants took versions of locally produced medications as either a fixed-dose combination of stavudine (30 or 40 mg), lamivudine (150 mg), and nevirapine (200 mg) [GPO-vir, Government Pharmaceutical Organization (GPO), Bangkok, Thailand] or as individual formulations of stavudine (30 or 40 mg) and lamivudine (150 mg) (GPO, Bangkok, Thailand). To accommodate the need for a two-week lead-in phase of nevirapine 200 mg daily, women took GPO-vir once daily in the morning and stavudine and lamivudine as individual products in the evening. After two weeks, women took GPO-vir twice daily. Adherence support and counseling was provided by the clinical staff and study nurses who met with each participant at each study visit to record adherence to medications and to provide additional counseling. Upon enrolling in the study, participants were instructed to choose a time of day for ingestion of antiretroviral drugs that would best meet their needs.

Although the time when antiretroviral medications were generally given (e.g., 9 AM and 9 PM) was known, the precise time of the most recent dose of medications was not recorded. We relied on the concentrations of stavudine and lamivudine (while not considering the corresponding nevirapine concentrations) to determine if a plasma specimen was classified as pre-dose or post-dose. Stavudine and lamivudine are both renally cleared with plasma half-lives of ~1 hour [16] and ~5-7 hours [17], respectively. There is substantial fluctuation in the plasma concentrations of stavudine and lamivudine over a 12-hour dosing interval. The typical peak and trough stavudine concentrations following twice daily doses of 30 or 40 mg are >600 ng/mL (range ~600 – 1,200) and <25 ng/mL (range, undetectable - <200) [18-21], respectively. The same values for lamivudine following twice daily doses of 150 mg are >1000 ng/mL (range ~500 – 2,000) and <200 (range, ~<50 – 500) [22-24], respectively. In contrast, there is relatively little fluctuation in plasma nevirapine concentrations over a 12-hour dosing. Nevirapine is metabolized by hepatic cytochrome P450 isoenzymes, primarily CYP3A4, and auto-induces its own metabolism so that the terminal half-life decreases from ~45 hours after the first dose to ~25-30 hours following multiple dosing [25].

Considering these pharmacokinetic data, we characterized pre-dose specimens as those that had concentrations of: 1) stavudine ≤200 ng/mL and lamivudine ≤500 ng/mL, or 2) stavudine ≤25 ng/mL and lamivudine <650 ng/mL. The latter was chosen to accommodate specimens with very low stavudine concentrations and relatively low lamivudine concentrations in order to include more data points in the pre-dose group. Post-dose specimens had concentrations of stavudine >200ng/mL and lamivudine >500 ng/mL. Specimens that did not meet either of these definitions were censored from the analyses.

We calculated standard summary statistics of the enrollment characteristics including counts, proportions, medians, and quartile ranges. Nonparametric Wilcoxon rank-sum tests were used to compare the medians of selected baseline variables between cases and controls. Chi-square or Fisher’s exact tests were used to compare categorical covariates. For comparisons of nevirapine concentrations between two groups of participants, we used Wilcoxon rank-sum tests. Women had at most one nevirapine concentration determined for analysis (either pre-dose or post-dose) at each of week 2, 4, or 8. Our primary comparison was of the plasma nevirapine concentrations at week 2, where all but one woman had a specimen available. Adverse effects present at week 2 or week 4 that resulted in substitution of another drug for nevirapine confounded statistical comparisons at week 4 and week 8 (i.e., nevirapine had been discontinued before the week 4 and week 8 visit for these women), thus we did not make statistical comparisons using those data. We present the raw values for concentrations at weeks 4 and 8 for visual inspection. We also tested for differences in nevirapine concentrations according to participants’ use of fluconazole and serum hepatitis B surface antigen status. We plotted nevirapine concentrations versus body-mass index and calculated Pearson’s correlation coefficient to determine if there was a linear relationship. All of these statistical calculations were performed using SAS software, version 9.2 (SAS Institute, Inc., Cary, North Carolina).

Written informed consent to participate in this study was obtained from each participant according to the protocol approved by the Ethical Review Committee for Research in Human Subjects of Thailand Ministry of Public Health (Rajvithi Hospital) and the Institutional Review Boards of the Siriraj Hospital and of the CDC.

We found no difference in pre-dose plasma nevirapine concentrations at week 2 comparing cases (3,528 ng/mL, n = 24) with controls (3,150, n = 30, p = 0.5) (Fig. 1). Post-dose nevirapine concentrations were higher for cases who developed rash at any time (median = 6,150 ng/mL, n = 21) versus controls (median = 4,746 ng/mL, n = 20, p = 0.02); however, when limited to cases who developed a rash at week 2, we observed no differences in either pre-dose or post-dose week 2 plasma nevirapine concentrations: median 3,270 ng/mL (n = 12, p = 0.9) and median 5,443 ng/mL (n = 9, p = 0.4), compared with the same controls, respectively. When limiting the analysis to only women who had hepatotoxicity, we found that they had similar pre-dose and post-dose plasma nevirapine concentrations at week 2 compared with controls. Women who developed both rash and hepatotoxicity likewise had similar pre-dose and post-dose nevirapine concentrations compared with controls at week 2.

Fluconazole was prescribed for 32 patients at the time of the week 2 visit, for 26 patients at week 4, and for 23 patients at week 8. There were no differences in pre-dose or post-dose nevirapine concentrations among patients taking fluconazole compared with patients not taking fluconazole at weeks 2, 4, or 8. There were also no differences in pre-dose and post-dose nevirapine concentrations at weeks 2, 4, or 8 comparing women who were positive for hepatitis B surface antigen versus negative. There were too few women who were hepatitis C antibody positive to make meaningful comparisons. There was a marginally significant relationship between body-mass index and the post-dose plasma nevirapine concentration at week 2 (Pearson’s r = -0.31, p = 0.051), but not at week 4 (r = 0.02, p = 0.90) or week 8 (r = 0.09, p = 0.64), and no significant relationship between body-mass index and pre-dose nevirapine concentrations at week 2 (r = -0.013, p = 0.92), week 4 (r = -0.06, p = 0.69) or week 8 (r = 0.20, p = 0.25) (Fig. 2).

At week 2 while on nevirapine 200 mg once daily, pre-dose nevirapine concentrations ranged from <3,000 ng/mL for 39% of women to ≥9,000 ng/mL for 2% of women. Post-dose concentrations ranged from <3,000 ng/mL for 12% of women to ≥9,000 ng/mL for 15% of women (Table 2). At week 2, cases were not more likely than controls to have a pre-dose concentration ≥6,000 ng/mL (p = 0.78 by Fishers exact test; sensitivity = 0.13, specificity = 0.93, predictive value positive = 0.60, predictive value negative = 0.57) or a post-dose nevirapine concentration ≥9,000 ng/mL (p = 0.18, sensitivity = 0.24, specificity = 0.95, predictive value positive = 0.83, predictive value negative = 0.54). At weeks 4 and 8 (while on nevirapine 200 mg twice daily), there were fewer pre-dose or post-dose nevirapine concentrations <3,000 ng/mL. Nevirapine concentrations ≥9,000 ng/mL were found at week 4 for 13% of pre-dose and 46% of post-dose observations and at week 8 for 11% of pre-dose and 13% of post-dose concentrations.

At week 2, a pre-dose nevirapine concentration <3,000 ng/mL was found for 21 (39%) of 54 women and a post-dose nevirapine concentration <3,000 ng/mL was found for 5 (12%) of 41 women (Table 2). After week 2, an additional 5 women had a pre-dose (n = 4) or post-dose (n = 1) nevirapine concentration <3,000 ng/mL for a total of 31 women who ever had a measured nevirapine concentration below the usual targeted threshold considered optimal for maintaining virologic control. Viral load was available at week 24 for 98 women in this analysis who continued on NNRTI-based ART (10 women were changed to a protease inhibitor or dropped out of the study). Viral load was <400 copies/mL for 29 (100%) of 29 women who ever had a pre-dose or post-dose nevirapine concentration <3,000 ng/mL and for 64 (93%) of 69 women who never had a pre-dose or post-dose nevirapine concentration <3,000 ng/mL (p = 0.3).