Effect of Nevirapine Toxicity on Choice of Perinatal HIV Prevention Strategies

Abstract

Objectives. This study evaluated the validity of concerns about the toxicity of nevirapine (NVP) that have delayed its implementation as a perinatal HIV prevention strategy.

Methods. A decision analysis model compared 3 strategies: single-dose NVP, short-course zidovudine (ZDV), and no intervention.

Results. NVP would prevent more deaths than ZDV and no intervention as long as the rate of NVP toxicity did not exceed, respectively, 9 times that observed in the earlier NVP clinical trial and 42 times that observed in the clinical trial. NVP would be economically preferable to ZDV as long as the rate of toxicity did not exceed 22 times that observed in the clinical trial.

Conclusions. Field implementation of NVP should not be delayed by concerns about its toxicity. (Am J Public Health. 2002;92:365–366)

A single dose of nevirapine (NVP) given to an HIV-infected mother at labor onset, followed by a dose given to the infant after delivery, significantly reduces perinatal HIV transmission.1 However, concerns have been raised over possible toxicity arising from widespread use of the drug.2 Given NVP's high degree of efficacy and its favorable cost-effectiveness relative to interventions involving zidovudine (ZDV),3 we sought to characterize, in this study, the magnitude of toxicity that would be required for its known benefits to be outweighed by any as yet unrealized detrimental effects.

METHODS

We used decision analysis to evaluate the potential effects of NVP toxicity on the clinical effectiveness and cost-effectiveness of perinatal HIV prevention strategies. The general structure of the decision model and its baseline estimates have been described previously.4 In most of the developing world, interventions such as formula supplementation, combination antiretroviral therapy, and elective cesarean delivery are not sustainable options; thus, decision makers must choose between providing no therapy and providing a short-course antiretroviral regimen (either NVP or ZDV).1,5,6 Our model compared the following 3 strategies: (1) single-dose maternal–infant NVP, (2) short-course maternal–infant ZDV (oral maternal ZDV from 36 weeks' gestation), and (3) no intervention.

Costs are expressed in year 2000 US dollars. Future costs and effects were discounted at a rate of 3% per year. The perspective used was that of a district health project. We assumed that decision makers would be willing to pay $50 per disability-adjusted life-year saved to achieve health gains in resource-poor nations.7

The following model estimates (well justified elsewhere3,4) were used: HIV transmission rate without treatment, 25%; reduction in transmission with NVP, 47% (vs 37% with ZDV); HIV prevalence, 30%; discounted lifetime cost of infant HIV infection, $300; cost of voluntary counseling and testing, $7.60; cost of counseling for mass therapy, $0.45; cost of maternal–infant single-dose NVP, $4; cost of maternal–infant ZDV, $60; discounted life expectancy without HIV, 20 years; and discounted disability-adjusted life-years lost with HIV, 15.

The HIVNET clinical trial of NVP1 reported 2 adverse events that were “possibly, but unlikely to be, related to” the drug: a nonmacerated stillbirth to a woman who had taken NVP 3.5 hours before delivery and a transient episode of respiratory distress in an infant born with meconium staining. Thus, the overall rate of toxicity potentially related to NVP was 0.645% (2/310). We selected the upper 95% confidence interval (2.3%) as our baseline estimate for this rate and assumed that half of all toxic events would result in stillbirths (estimated to incur no additional cost and to save no disability-adjusted life-years) and half would result in transient respiratory distress (estimated to be one third as costly as HIV infection to involove no decrement in disability-adjusted life-years). ZDV therapy was assumed to produce no toxic sequelae.8 By adopting this liberal estimate of NVP toxicity and conservative estimate of ZDV toxicity, we both biased the model against the choice of NVP and accounted (in terms of lost disability-adjusted life-years) for the unlikely possibility of nonimmediate adverse infant effects arising from the single-dose NVP regimen.

RESULTS

Results were calculated for a hypothetical cohort of 10 000 women enrolled under a targeted strategy (HIV testing and treatment of seropositive women) We found that provision of NVP would cost less, produce more gains in disability-adjusted life-years, and prevent more deaths than the alternative strategies (Table 1 ▶). From a clinical effectiveness standpoint, NVP would prevent more deaths than short-course ZDV and no intervention as long as the actual rate of NVP-related toxicity did not exceed, respectively, 9 times that observed in the HIVNET clinical trial1 (absolute rate: ≤6%) and 42 times that observed in the clinical trial (absolute rate: ≤28%). From an economic standpoint, NVP would be preferred as long as the actual rate of NVPrelated toxicity did not exceed 22 times that observed in the clinical trial (absolute rate: ≤14%).

Table 1—

Costs and Effects of Provision of 4 Perinatal HIV Prevention Strategies Under Assumptions of High Nevirapine (NVP) Toxicity

Strategy	Cost, $	HIV Cases Prevented, No.a	NVP-Related Deaths, No.b	Deaths Prevented, No.	DALYs Saved, No.
Mass NVR	166 000	353	149	204	1650
Targeted NVP	198 000	353	30	323	4695
No intervention	225 000	0	0	0	0
Targeted zidovudine	386 000	278	0	278	4170

Note. Data were applied to a hypothetical cohort of 10 000 pregnant women at 30% HIV prevalence. DALYs = disability-adjusted life-years.

^aCases prevented relative to no intervention.

^bDeaths in uninfected infants.

In a separate analysis, we considered drug provision under a mass strategy (treatment of all gravidas without testing)3,4 and found that perinatal NVP would cost less and generate more health benefits than no intervention. Provision of ZDV would be considerably more expensive and would produce more gains in disability-adjusted life-years (because NVP was modeled to be highly toxic). However, ZDV's additional efficacy would be achieved at a cost of $228 per disabilityadjusted life-year saved, a prohibitively high incremental sum in most developing-world settings. Assuming that decision makers are willing to pay $50 per disability-adjusted life-year, NVP would be the economically preferred strategy as long as the actual rate of NVP-related toxicity did not exceed 8 times that observed in the HIVNET clinical trial (absolute rate: ≤5%). When the estimate of HIV prevalence was reduced to 15%, mass NVP remained preferable as long as the modeled rate of NVP-related toxicity did not exceed 3.3 times that observed in the clinical trial (absolute rate: ≤2.1%).

DISCUSSION

This analysis demonstrates that, even with a number of underlying assumptions biased against its use, NVP is preferable to ZDV, as well as to no intervention, from both a clinical effectiveness standpoint and a costeffectiveness standpoint. The rate of NVP-attributable toxicity would have to exceed the toxicity observed in the HIVNET clinical trial by many times to overturn this conclusion.

Another concern surrounding the field implementation of NVP is the potential for emergence of NVP-resistant viral strains.9 Although a complete discussion of this issue is beyond the scope of the present article, a brief response to resistance concerns can be constructed with the findings of our toxicity model. Because NVP emerges as the preferred perinatal HIV prevention strategy under our assumptions of extremely frequent and lethal (50%) toxicity, it is exceedingly unlikely that selection of a transient reverse transcriptase mutation in approximately 20% of exposed women would produce sufficient harm to outweigh the known and immediate life-saving benefits of NVP. This is especially true in that such mutations (1) seem to revert within 6 to 12 months and (2) would be of clinical importance only among women with the means to receive NVP as part of a chronic combination regimen (a very small minority in most African settings). We conclude that field implementation of NVP as a perinatal HIV prevention strategy should not be delayed by concerns about its toxicity.