Thai HIV Vaccine Trial results presented and published
Richard Jefferys, TAG
In tandem with the presentation of the data that took place at the AIDS Vaccine 2009 conference in Paris, the results of the RV144 trial were published online in the New England Journal of Medicine.
Access to the paper and the accompanying editorial is free of charge. Three different analyses of the results are presented in sequence: the intent-to-treat analysis (ITT), which includes everyone enrolled and randomised to receive vaccine or placebo, a per protocol (PP) analysis limited to everyone who received all immunisations on schedule, and finally a modified ITT (mITT) analysis that excludes seven individuals who reportedly turned out to be HIV-infected at the time of their first immunisation.
- ITT: Total n=16,402. Cases of HIV infection: 76 placebo, 56 vaccine. Efficacy: 26.4% (95% confidence interval [CI], 4.0 to 47.9; p=0.08)
- PP: Total n= 12,452. Cases of HIV infection: 50 placebo, 36 vaccine. Efficacy: 26.2% (95% CI, 13.3 to 51.9; p=0.16)
- mITT: Total n=16,395. Cases of HIV infection: 74 placebo, 51 vaccine. Efficacy 31.2% (95% CI, 1.1 to 51.2; p=0.04)
The data suggests the possibility of a marginal protective effect, almost entirely concentrated during the first year of the study. Kaplan-Meier plots of the infection rate over time show a divergence initially, but the rates in the vaccine and placebo groups are superimposable from week 52 onwards. Subgroup data are also reported, but the statistics are uncorrected for multiple analyses and should be interpreted with great caution. With this caveat, there is a hint that the difference between the vaccine and placebo was greatest among those at lowest risk of HIV exposure. The age group breakdown also indicates the difference between vaccine and placebo groups was concentrated in the 20-25 age group; there is no difference in the number of infections between the groups among those under 20, and very little difference among those over 26. Among participants aged 20-25, there were 20 infections in the vaccine group and 40 in placebo.
In terms of the viral load outcomes in people who acquired infection, the ITT analysis shows a trend in the wrong direction; viral load was higher on average among vaccine recipients (4.36 log vs. 4.21 log, p=0.09). However this trend disappears in both the PP and mITT analyses. There were no differences in post-infection CD4 T cell counts in any of the analyses.
As to why the only statistically significant result is reported last in the published paper (in contrast to the September 24 press announcement, in which the mITT was the only result given), it appears that the RV144 protocol specified that the primary analysis would be ITT. The paper states that the mITT analysis was used as the primary analysis for the interim efficacy evaluation (which was conducted by the Data Safety Monitoring Board in July of 2007) and then, five months before the study was unblinded, a decision was made to make the mITT the primary analysis. Reading between the lines, perhaps the reviewers of the manuscript were not satisfied that this late adoption of the mITT as the primary analysis justified listing the result first in the paper. It is currently unclear why the RV144 protocol did not specify the mITT as the primary efficacy analysis from the start. As the import of the trial results are mulled by the larger community, it will be important to gain some clarity as to exactly how these events played out.
So far, in the limited time observers have had to digest the data, the main issues that are being discussed are the suggestion of a transient, time-limited effect and what might explain it (vaccines generally work by the induction of immunological memory, which is typically long-lived) and the hint that vaccine-mediated protection might be easier to achieve in individuals with less frequent HIV exposure compared to those at high risk.
Regrettably, the release of the data today does not change the fact that it was an appalling and woefully short-sighted decision to only release the mITT analysis to the press on September 24. On the conference call hosted by AVAC that took place that day with investigator Merlin Robb and Peggy Johnston from NIAID, Robb explicitly stated that only 16,395 people had been enrolled into the trial. Not only was this not true, but it turns out that vaccine/placebo distribution of the 7 people excluded from the mITT was crucial to the attainment of statistical significance: five of these individuals were in the vaccine group and two in placebo. By cherry-picking the mITT to announce, the RV144 investigators have created suspicion and uncertainty in a field that they well know is already plagued by controversy. Their decision will only serve to complicate efforts to glean useful information from the trial data.
Source: www.tagbasicscienceproject.typepad.com (20 Oct 2009)
The webcast of the press conference about the trial results is now available online (scroll down to the bottom of the page to the Tuesday, 20 October press conference link).
- Supachai Rerks-Ngarm et al. Vaccination with ALVAC and AIDSVAX to prevent HIV-1 infection in Thailand. 20 October 2009 (10.1056/ NEJMoa0908492).
- Dolin R. HIV vaccine trial results – an opening for further research. NEJM Editorial. 20 October 2009 (10.1056/NEJMe0909972).
Earlier articles detailing the unfolding controversies around this study and the early press release focusing on a positive trial result are covered in a number of articles from the TAG basic science web log.
– Marginal HIV vaccine trial result raises hopes, eyebrows. (25 Sep 2009).
– Thai HIV vaccine trial: additional history & links. (28 Sep 2009).
– Deconstructing the Thai trial vaccines. (29 Sep 2009).
– Did the world get a fair glimpse of the Thai vaccine trial data? (05 Oct 2009).
– Thai HIV vaccine trial update: uncertainty reigns. (15 Oct 2009).