In his Correspondence letter “HIV drug remains unproven without placebo trial” (Nature 434, 137; 2005), Valendar Turner asks on what basis it can be claimed that the single-dose nevirapine regimen is effective in reducing mother-to-infant HIV transmission.

He points out that the placebo arm was dropped in the HIVNET 012 trial (of which we were team members) and notes the high variability in reported historical HIV transmission rates.

But he ignores the valid comparison made by the HIVNET 012 trial — whose conclusions were recently accepted by a US Institute of Medicine panel — between nevirapine and another drug, zidovudine, which had itself been tested against a placebo in other studies starting at least four weeks before labour.

The HIVNET 012 trial in Uganda originally placed HIV-infected pregnant women and newborns in three groups, randomly selected, to receive either the nevirapine regimen (a single dose of nevirapine to the mother during labour and a single dose to the baby shortly after birth), or a very short course of zidovudine from the onset of labour through the first week of the baby's life, or a placebo.

As noted by Turner, the placebo arm of HIVNET 012 was dropped shortly after the study began, when a Thai study showed that a short course of zidovudine, given for the last four weeks of pregnancy and during labour, reduced transmission by 50% compared with placebo (N. Shaffer et al. Lancet 353, 773–780; 1999).

Women on HIVNET 012 continued to be enrolled in the two active drug arms, following recommendations — made jointly by the US Centers for Disease Control and Prevention, the National Institutes of Health, the French National Agency for AIDS Research, and the Joint United Nations Programme on HIV/AIDS — to drop the placebo arm from all ongoing perinatal HIV trials.

Once the placebo arm was dropped, the redesigned trial compared the mother-to-infant transmission rates for nevirapine and on zidovudine within the same trial, rather than making a comparison with historical transmission rates, as Turner implies.

When one compares two drugs, if the test and control drugs are found to be similar in efficacy, it is not possible to know whether the two drugs are equally effective or equally ineffective — unless there is clear evidence that the control drug is effective. This is why use of placebos is advocated in settings where the active control has not been established to have efficacy that is clinically and statistically significant.

But when an experimental drug is found to be superior to a control that itself is not harmful (thus replacing a placebo), the effectiveness of the experimental drug is thereby established.

In the HIVNET 012 trial, a very short course of zidovudine was used as the active control. Many studies had already reported that longer zidovudine regimens were more effective than placebos.

Of the HIVNET 012 patients who gave birth before discontinuation of the placebo arm, the transmission rate at 6–8 weeks of age was 37% in the placebo group, compared with 28% in the zidovudine group and 7% in the nevirapine group. (The 6–8-week transmission rates at the end of the trial were 20% for zidovudine and 11.8% for nevirapine.)

Most people would conclude that zidovudine does not increase the risk of transmission. Hence, our finding that single-dose nevirapine was significantly more effective in preventing HIV transmission than a very short course of zidovudine justifies the conclusion that the HIVNET 012 nevirapine regimen is more effective than nothing in preventing mother-to-infant HIV transmission.