VaxGen company reveals some AidsVax data - but results change from show to show

The failure of AidsVax to prevent infection with HIV - in clinical trial results published in 2003 - triggered an intense debate about the controversial product and its manufacturer, VaxGen Inc of Brisbane, California. Mail to this website, maintained by Brian Deer, shows that existing material on a VaxGen-AidsVax index is read by significant numbers. This page seeks to further inform the discussion

Below is material from three different presentations of apparently definitive preliminary results of the AidsVax phase III US trial: (A) February 24, (B) March 19 and (C) March 31. For reasons best known to the company, the data and its format kept changing between one showing and the next, or, in the case of (C) ethnic categories are switched, making important comparisons impossible. [Puzzling time to seroconversion data for ethnic groups, also available at this site, was presented at the first event but mysteriously never resurfaced.]

(A) February 24: Presented by VaxGen CEO Dr Lance Gordon

These were the data given on the day the company admitted that the trial had failed to show overall efficacy, but claimed that AidsVax had shown astounding efficacy among US ethnic minorities. The company made clear that sophisticated statistical work had been carried out in consultation with the Centers for Disease Control. Later it emerged that UNAIDS was also involved in this statistical work.

(B) March 19: Presented by VaxGen CEO Dr Lance Gordon

At the next event - an investors conference - a different results slide was shown. This gives much less detail, but the overall percentages of vaccine efficacy are the same as on February 24 for the first three categories ("All volunteers", "white and Hispanic" and "black, Asian and other"), but in the next three categories ("black", "Asian" and "other") the efficacy has fallen slightly. With so much data from the first slide absent from the revised version, it is not easy to see why this should have happened.

(C) March 31: Presented by AidsVax inventor Dr Phillip Berman

At this event, multiple changes are made in the presentation of the data, such as in this new slide giving the overall figures. Now, in addition to the results prescribed by the trial protocol (versions of which are used in A and B)the company introduces a new "intention to treat" (ITT) category of subject, who have had only one shot, as opposed to the three shots in the "per protocol" groups. This figure (given first) raises the apparent efficacy of AidsVax from 3.2% (statistically zero), to 5.7% (still statistically zero, but sounding better all the time). The confidence interval for all volunteers is also slightly different from the first slide.

Broken down by ethnic group, the (C) March 31 presentation becomes yet more interesting. The previous binary conflation of races into BAO (black-Asian-other) and white/Hispanic has gone. Now Hispanic volunteers are moved into a new conflation HBAO (Hispanic-black-Asian-other). No compelling reason was given for this during Dr Berman's presentation, but it might tend to suggest a line of thinking that more people than envisaged by the company on February 24 might be seen as potential targets for the product. Multiple variations from previous slides are also evident in both vaccine efficacy and confidence intervals.

Discussion of the company's claims, such as exchanges on the Yahoo message board, raised the question of whether VaxGen's ethnic conflations represented any credible association, or whether they had been devised by VaxGen purely for the purpose of presenting the statistics in the most commercially-favorable light.

Time to seroconversion data can be seen on a pair of slides from the February 24 presentation.