Background:  Nevirapine (NVP) is used by >60% of patients on first-line regimen in resource-limited settings. One of the primary concerns with NVP is associated severe hepatoxicity (HT). International guidelines recommend frequent screening of alanine aminotransferase (ALT) serum levels. This is often not feasible in resource-limited settings. We describe the incidence, determinants, and associated mortality of NVP-associated HT in a primary care ART program (Khayelitsha, South Africa).

Methods:  Prospective cohort study of all treatment naïve adults initiating NVP-based ART in Khayelitsha from November 2002 until December 2006. We used Kaplan-Meier analyses to estimate time to HT and mortality, and Cox proportional hazards regression to model determinants of HT and mortality. Logistic regression was used to determine factors associated with HT at baseline. HT was defined as ALT above 5 times the upper limit of normal.

Results:  We initiated 1832 adults on NVP-based ART during the study period, of whom 382 (21.1%) were male. The median baseline CD4 count was 112 cells/µL. South African guidelines recommend ALT measurement at 4, 8, 12, and 24 weeks. Only 7.6% of patients had all their measurements done according to protocol. A baseline CD4 count <50 cells/µL was associated with baseline HT (OR 16.5, 95%CI 1.8 to 151.3). Of the total, 35 (1.9%) patients developed HT on ART:  15 (42.9%) in the first 4 weeks, 21 (60%) by 8 weeks, and 26 (74.3%) by 12 weeks. Of the 12 who inadvertently continued NVP, all resolved their HT. The median time to HT was 56 days (IQR 28 to 121). The incidence of HT in patients with all measurements done was 6.1% during the first 6 months. Of the 3 (0.16%) patients who died, all had discontinued NVP and had been restarted on efavirenz.

Conclusions:  The overall incidence of HT in this cohort was lower than reported in clinical trials. HT at baseline and female gender were not found to be associated with HT, contrary to data reported by clinical trials. A third of HT resolved despite continuing NVP. ALT screening according to protocol was only accomplished in a very small proportion of patients. This highlights the difficulties in implementing serial ALT monitoring in a high-burden resource-limited setting. In this context, ALT measurements are usually prompted by clinical symptoms. Further research into the sensitivity of symptoms as opposed to ALT screening is needed to determine whether this intervention should be recommended in poor-resource settings.