Item #21: HIVNET 012: Phase II or Phase III?

HIVNET 012, according to its original 1997 protocol, was intended to be a four-arm, Phase III, randomized, placebo-controlled trial.6 Its sole sponsor was listed as the National Institute of Allergy and Infectious Diseases (NIAID), though one of the investigators was a Boehringer employee. The “sample size” was to be 1,500 HIV-1 infected Ugandan women more than thirty-two weeks pregnant. The four arms they would be divided into were 1) A single dose of 200mg nevirapine at onset of labor and a single 2mg dose to the infant forty-eight to seventy-two hours post-delivery, and 2) a corresponding placebo group; 3) 600mg of AZT at onset of labor and 300mg until delivery, with a 4mg AZT dose for the infant lasting seven days after birth, and 4) a corresponding placebo group. There were to he 500 women in each “active agent” arm and 250 in each placebo arm. The study was to last eighteen months, and its “primary endpoints” were to see how these two regimens would affect rates of HIV transmission from mother to child, and to examine the “proportion of infants who are alive and free of HIV at 18 months of age.” Another primary objective was to test the “safety/tolerance” of nevirapine and AZT. HIVNET's architects estimated that more than 4,200 HIV-positive pregnant women would deliver at Mulago hospital each year, allowing them to enroll eighty to eighty-five women per month. Consent forms were to be signed by either the mother or a guardian, by signature or “mark.” One of the exclusion criteria was “participation during current pregnancy in any other therapeutic or vaccine perinatal trial.”
Although HIVNET was designed to be a randomized, placebo-controlled, double-blind, Phase III trial of 1,500 mother/infant pairs, it wound up being a no-placebo, neither double- nor even single-blind Phase II trial of 626 mother/infant pairs. Virtually all of the parameters outlined for HIVNET 012 were eventually shifted, amended, or done away with altogether, beginning with perhaps the most important—the placebo controls. By a “Letter of Amendment” dated March 9, 1998, the placebo-control arms of HIVNET were eliminated.
The study as reconstituted thus amounted to a simple comparison of AZT and nevirapine.

Farber claims that HIVNET 012 was supposed to be a phase III trial but wound up being a phase II trial.

Farber appears not to know the difference between a phase II and phase III trial, because HIVNET 012 was a randomized phase III trial. It was not double-blind, because the drug administration procedures were so different in each of the two arms. While phase III trials are ideally double-blind, this is not an indispensable requirement. Frequently drugs are tested using an “open-label” procedure.

The most important question is not whether HIVNET 012 was a Phase II or Phase III clinical trial, but whether it was a valid and useful clinical trial at all.
Farber describes the many problems with this trial, focussing on the lack of a placebo, which removed any ability to determine the adverse effects of the two study drugs. Without this it is not possible to claim that either intervention is safer or more effective than doing nothing.
Double-blinding (preventing both researchers and patients from knowing which intervention is being used for an individual) is important to remove the bias, often unconscious, towards showing that the new drug worked better than the old (or, when a placebo is included, showing that the intervention was better than doing nothing).
Perhaps the biggest problems with this trial were the ‘book-keeping’ errors, the loss of records of adverse events and other critical information. Without accurate records no clinical trial is meaningful. And losing records that could show that the intervention is unsafe will bias the results in favor of the therapy.