Item #18: Did HIVNET 006 Lower Viral Load?

A small Phase I trial preceded HIVNET 012 that studied the safety, primarily, of nevirapine in pregnant women but also looked at efficacy. It was called HIVNET 006, and it enrolled twenty-one pregnant women for initial study. Of twenty-two infants born, four died. There were twelve “serious adverse events” reported. The study also showed that there was no lowering of viral load in the mothers who took the study drug (the industry's agreed-upon standard for interrupting maternal transmission).

Footnote 6 states there was no lowering of maternal viral load in the HIVNET 006 safety study.

The study states that “The antiviral activity of nevirapine appeared to be quite strong, resulting in a relatively consistent median 1.3 log reduction in maternal plasma HIV RNA at 1 week after a single 200mg dose in all mothers.”39

‘Viral Load’ (a measurement of the presence of a small snippet of the consensus HIV genome) was measured in only 21 women in the HIVNET 006 Phase I/II clinical trial at birth and 19 at one week after birth. The Gallo document did not mention this tiny sample size.
The reduction in viral load cited by the Gallo document was the average among the 19 women measured at 7 days after they took the nevirapine pill, several days after many of them had given birth. The viral load at birth obviously has more significance, because transmission is believed to occur during pregnancy.
At birth, in the four women whose children were later found to be HIV-positive, the viral loads were 556, 672, 291392 and 304290 copies per ml. Among the 17 whose infants were not HIV-positive, the range of viral loads was from under 400 copies/ml to 1796670. With such a huge range of values, it is obvious that average values, such as cited by the Gallo document, are meaningless.
The use of the median, rather than an average, is also interesting. The median is the central value, the one with half of the values being above, and half below. If a fifth mother had given birth to an HIV-positive child and had a viral load of 672 or below, the median of the five values would have been 672. If the mother had a value of 291392 or above, the median would have been 291392. If the value had been between 672 and 291392, the media would have been that value. Obviously the use of the median is not applicable to a small number of samples with such an enormous range (over 500 times among the mothers of HIV-positive infants and over 4000 times among the mothers of HIV-negative infants). It is quite plausible that the mean or average value did not show as large a reduction in viral load, or possibly no reduction at all. Such statistical trickery is not uncommon in papers on HIV/AIDS, unfortunately.

1. Musoke P et al. A phase I/II study of the safety and pharmacokinetics of nevirapine in HIV-1-infected pregnant Ugandan women and their neonates (HIVNET 006). AIDS. 1999 Mar 11; 13(4): 479-86.