As mentioned in the previous post, a large amount of information about the Merck HIV vaccine trial was shared yesterday morning at the HVTN conference. Here, in bullet point form, are some more of the potentially important nuggets of information related to yesterday’s presentations:
- Since the September 18 DSMB review, an additional 8 infections have been confirmed in the first 1,500-person STEP cohort (which enrolled people with anti-Ad5 antibody titers less than 1:200).These infections were evenly split between placebo and vaccine recipients, bringing the totals in this cohort to 28 infections among vaccine recipients and 24 among placebo recipients, a non-significant difference.
- HVTN statistician Steve Self explained that, in strict statistical terms, even when the larger difference in infections between vaccine and placebo recipients in the second STEP cohort is factored in, the fact that multiple, unplanned subanalyses are involved means that the difference represents a very strong trend but may not quite reach significance.
- The association between higher levels of anti-Ad5 antibodies and increased susceptibility to HIV infection makes it unlikely that the HIV-specific T cell responses induced by the vaccine were responsible for the observed effect, because HIV-specific T cell responses were of higher magnitude in individuals with low levels of anti-Ad5 antibodies. It also suggests that generalized immune activation induced by the vaccine is not a likely culprit because, again, levels of generalized immune activation would be predicted to be higher in individuals with low levels of anti-Ad5 antibodies. Nevertheless, both of these possibilities will be looked at in detail as part of the plan to evaluate any potential biological mechanism that could explain the study results.
- One potential explanation for the vaccine’s failure to significantly reduce viral load levels in infected participants may be that the vaccine-induced HIV-specific T cell responses were very narrowly targeted. Juliana McElrath showed that recipients developed a T cell response to an average of only one epitope per protein in the vaccine (one epitope from Gag, one from Pol, and one from Nef). Planned analyses will look at the question of whether the epitope-specific T cells could recognize infecting HIV strains and, if so, whether there is evidence that the virus mutated to escape these T cell responses.
- Additional planned immunological and virological analyses include studies of the phenotype (effector memory vs. central memory), functional profiles (e.g. production of various cytokines and proliferative capacity), gene profiles and in vitro antiretroviral efficacy of vaccine-induced T cell responses. A variety of viral sequencing studies in collaboration with James Mullins and Frances McCutchan will also characterize infecting HIV strains and look for evidence of transmission clusters among participants (a phenomenon that was seen in a prior HIV vaccine efficacy trial).
- Among the men in the second STEP cohort who became infected, average set point viral load levels were 19,070 in the 21 vaccine recipients and 89,810 in the 9 placebo recipients. Although this may seem like a large difference, it’s driven by a single individual in the placebo group with a very high viral load and is not significant. These data do, however, offer some reassurance that even if a vaccine-related factor did enhance susceptibility to HIV infection, it did not lead to enhancement of HIV replication among the infected vaccine recipients.
- Susan Buchbinder outlined the three possible non-biological explanations for the STEP results: baseline differences among vaccine and placebo recipients, differential risk behaviors between these groups and simple chance. There were a number of significant baseline differences between the men in the first and second STEP cohorts: more participants in the second cohort were from sites outside the US, of non-Caucasian ethnicity, less than 30 years old, and uncircumcised. However, there were no significant baseline differences between vaccine and placebo recipients in either cohort. Likewise, analyses conducted to date have not found any significant difference in risk behaviors between the vaccine and placebo groups. Buchbinder noted that the small size of the overall study combined with these analyses being restricted to men means that these “post hoc” (unplanned) subset analyses come with very wide confidence intervals, indicating a great deal of uncertainty regarding the robustness of the findings. Given this fact, chance remains a possible explanation of the STEP results, but Buchbinder stressed that the serious implications of the Ad5 vector having an enhancing effect means that it can only be considered as an explanation of exclusion; i.e. all other possibilities must be considered and investigated first.
- At the current time, any STEP participant can ask to be unblinded (told whether they received vaccine or placebo) and informed of their anti-Ad5 antibody titer. Peter Gilbert, a statistician from the HVTN, made the case that maintaining the blind could help show whether any enhancement of susceptibility persisted over time – important information for the long term safety of any recipient of an Ad5 vaccine. Unblinding the study, he argued, could introduce confounding factors that would preclude this analysis. However, other meeting attendees, including Mitchell Warren from AVAC, argued that the blind cannot be ethically maintained if informing participants who received the vaccine might cause them to further reduce their risk of HIV infection. A decision about unblinding the study is expected in a matter of days.
- In a presentation on the unblinding of the Phambili trial, Glenda Gray revealed that the Phambili DSMB met twice – once on September 18, after which they requested additional information about the STEP trial results, and then again on October 8 when they recommended stopping and unblinding the Phambili trial. This appears to contradict recent statements by representatives from the HVTN and Merck that the Phambili DSMB did not review more data than the STEP DSMB.
While the last few weeks have been a nightmare of confusion and uncertainty regarding the STEP and Phambili trials (it is still unclear when the evidence of enhancement among the second STEP cohort came to light), further criticism of the communications strategy employed by the trial sponsors would be redundant and misplaced given the shocking and complex nature of the data. None of the many prior studies conducted with the Ad5 vaccine – in people and in a long series of animal model experiments - have suggested that the presence of anti-Ad5 antibodies could cause the vaccine to increase susceptibility to HIV infection.
Amidst the uncertainty, one thing was very clear from yesterday’s meeting: NIAID, HVTN’s leadership, investigators and collaborators from Merck and elsewhere deserve extraordinary credit for the work they are conducting and planning to figure out what happened in the STEP trial. Both Susan Buchbinder’s and Juliana McElrath’s presentations offered detailed, thoughtful and comprehensive approaches to resolving the questions raised by the study data. The willingness of HVTN and Merck to support outside investigators who might be able to contribute to the effort by suggesting and conducting additional analyses is also to be saluted.
As you’d expect, many more news outlets have now picked up the story and, in general, the coverage is remarkably accurate and well-informed. In hindsight, perhaps this was the trial sponsors justification for initially being economical with the details regarding the data which informed the Phambili DSMB’s decision. Given what had occurred, this may have been the best of a bunch of unpleasant options as to how best to communicate the study findings to the wider public.