The leading example of a biotech company developing an HIV preventive vaccine is VaxGen with its AIDSVAX products. VaxGen is a company spun-off from Genentech (a biotech company now largely owned by Roche) to develop Genentech's gp120 recombinant vaccine after the US government refusal to support Phase III trials of a prototype in 1994. VaxGen succeeded in raising private funds to run Phase III trials in the USA, Canada, the Netherlands, Puerto Rico and Thailand. It is greatly to their credit that they completed the trials despite many expert opinions to the effect that it would be impossible to recruit or retain volunteers. In fact, trial volunteer retention was higher than predicted in all settings and higher in Thailand than in North America.

The first VaxGen trial

The first VaxGen trial recruited 5,417 volunteers at risk of HIV infection from sexual transmission, around 90% of whom were gay men. One third of the volunteers were randomised in a double-blind trial to receive placebo injections and two thirds a ‘bivalent’ vaccine based on a genetically engineered version of the gp120 surface protein from two different HIV-1 subtype B isolates (Harro).

The protocol for the trial involved seven injections over thirty months (0, 1, 6, 12, 18, 24 and 30 months) with follow-up visits for blood tests two weeks after each injection and six months after the last injection. All volunteers were given prevention advice and counselling.

All volunteers who became HIV-positive during the study are due to be followed up every four months for 24 months. There is no restriction on the treatments HIV-positive volunteers may receive, although treatments are not provided directly as part of the study. Social harms experienced by volunteers, including self-reported risk behaviour, were closely monitored (Jermano).

5,009 volunteers received at least three of the seven scheduled injections and were included in 'on treatment' analysis which did not, in the event, differ substantially from the 'intent to treat analysis' which included all volunteers for whom an outcome was known.

VaxGen's results: no overall protection

The results of the first phase III HIV vaccine study released in February 2003 showed that Vaxgen’s AIDSVAX offered no significant protection against HIV infection in the study population as a whole.

The annual study infection rate 2.7% which was not significantly different between placebo and vaccine recipients. Among all volunteers the level of apparent 'protection' was just 3.8% (p-value = 0.76; confidence interval: -23% to 24%). In other words, it was 95% certain that the upper limit for protection was below 25%. The trial had been designed in the hope that if the lower limit was above 30%, the vaccine could still be licensed for use in the USA.

A vaccine for minorities?

Despite the apparently clear outcome of the trial, that AIDSVAX did not protect the volunteers who received it any better than those who received placebo, VaxGen chose to highlight apparent findings of protection among non-white volunteers. On closer examination, this turned out to depend entirely on a difference in infection rates among black volunteers.

Among those black volunteers, observed protection (the amount by which the infection rate was reduced in the vaccinated group) was 78%. The p-value was < 0.02; confidence interval: 29% to 93%.

The main problem is that the number of volunteers categorised as 'black' was only 314. VaxGen's analysis was based on just 13 infections among black volunteers, four in the vaccine group and nine in the placebo group. Of course, since twice as many people received vaccine as received placebo, the reduction from the 'expected rate' based on the infection rate among black volunteers was from 18 to 4. If such a difference had been seen in a freestanding trial, then VaxGen would be justified in using the statistics they did, to conclude that 'There is less than a 2% possibility that the observed difference in infection rates among black volunteers could have occurred by chance.'

Unfortunately, this claim is highly questionable and has been vigorously challenged by a number of statisticians and vaccine researchers. The problem is, that if enough 'subgroup analyses' were performed on entirely arbitrary groups of volunteers, there would eventually be groups in which the distribution of infected volunteers is skewed in favour of the vaccine. For this reason, such analyses need to be treated with extreme caution.

The black populations recruited into the trial are genetically diverse; 'black' in the USA is a social category not strictly a genetic one. Many African-Americans have caucasian as well as African ancestors. The idea that this difference could make the difference between a vaccine working and not working is has no clear precedent. Although VaxGen has searched the literature and found a few papers which do point to ethnic differences in immune responses, these do not include a difference in vaccine response between US black and caucasian populations.

The proper use of subgroup analyses is normally, to give leads to follow up with further studies. In this case, the leads don't seem to go very far. All black volunteers who became HIV positive during the study were infected with subtype B viruses, so there is no evidence for cross-subtype protection, as might be hoped for before such a vaccine were offered in African clinical trials.

The second and future trials of AIDSVAX

The second VaxGen Phase III trial commenced in Thailand during 1999, using a gp120 vaccine based on ‘subtype E’ (now classified as a ‘circulating recombinant form’) and subtype B isolates of the virus. This recruited 2,500 volunteers from drug treatment centres around Bangkok, to evaluate protection from direct blood exposure, with an equal number of placebo recipients and vaccine recipients. (For a description of the trial volunteers, see Vanichseni, and on social harm monitoring, Pitisuttithum 2000a.)

When the results of this trial are reported, towards the end of 2003, it is likely that they will either revive or finally bury the AIDSVAX project. In the event that they do show some evidence of protection, there will be a fresh impetus to go ahead with the planned trial of canarypox and AIDSVAX in prime-boost combination. If not, then it will surely be difficult to proceed with further studies in the USA or elsewhere, even among beleaguered minority communities that definitely need better means of HIV prevention.