March 3, 2010
Below is the transcript of a press conference held at CROI 2010 on Feb. 18, in which Nelson Michael, M.D., Ph.D., of the Walter Reed Army Institute of Research, and Morgane Rolland, Ph.D., of the University of Washington-Seattle, discuss the outcomes of the Ad5 and RV144 ("Thai") HIV vaccine trials, and how these results might impact future vaccine trial structures. This discussion was moderated by Richard Koup, M.D.
Richard Koup: I think we could call this session "A Tale of Two Vaccines." As you are all aware, over the last five years there have been two major vaccine products that have been tested in efficacy trials. Approximately two years ago, one of those vaccine trials was stopped due to lack of efficacy. That was actually two trials of the Merck Ad5 vaccine, which was a pure T cell-based vaccine, and was supposed to limit viral load in those who became infected.
More recently, just last fall, we completed a trial in Thailand of RV144, which was a vaccine with two components: an ALVAC vector, followed by a gp120 boost. That vaccine actually showed partial efficacy, about 31% protection against infection.
Moving forward, figuring out why one vaccine failed and why the other vaccine succeeded is a very important basic scientific question that needs to be addressed in the very near future. Today we heard the beginnings of results of the analyses of those trials that start to address why one vaccine may have protected, and one vaccine may have failed.
We heard from Nelson Michael that there were good antibody responses in the RV144 trial, that efficacy looked to be stronger earlier rather than later and that the antibody responses dropped concomitant with the potential loss of efficacy, although those results did not reach statistical significance.1 We also heard from Morgane Rolland that, looking at the T cell responses in the STEP trials, there is evidence that there was an impact on either the transmitted virus or the evolution of that virus after infection that was induced by the Merck vaccine.2
What I'd like to do now is turn the microphones over for brief statements by both of the investigators, and then we'll open it up for questions. We'll start with Nelson.
Nelson Michael: Thanks, Rick. Actually, you gave a very nice summary of the results that we presented. I would only add one additional comment about the data, which is that in Paris, and in our publication in The New England Journal of Medicine, we described that in this low-incident population, largely of people at heterosexual risk for HIV infection in Thailand, we looked at baseline risk and identified individuals that were at relatively higher or lower risk for HIV infection. That analysis did not achieve statistical significance, in terms of what their risk category was, and whether or not there was a difference in the protective efficacy -- even though the numbers looked like they were headed in that direction. In the individuals who had the highest incidence, the estimate of efficacy was 3%, versus 31%, but the confidence intervals were very, very large.
What I presented today is that we looked at individuals at every six-month time point when they came in, and put individuals in two categories: those that never described a higher risk behavior and those that, even if they described a single episode of having a higher risk category, would be in a different bin. If you compared those two groups, there was a lower efficacy in the group of individuals that described a higher risk at any point in the study. That p value was .008.
The caveat to that is that that is a post hoc analysis. In a clinical study, anything that you dream up to study after the data are unblinded, you should take with a pretty big grain of salt, because the chances that those kinds of associations are influenced by statistical variation go up. So even a p value of that magnitude is one that should still be interpreted with caution.
Secondly, that doesn't necessarily indicate that this vaccine would not have worked in individuals in a higher incidence population. It's possible that what we think is an early, and nondurable, effect of this vaccine may be confused with the fact that, in the higher-risk individuals, they may simply have been having risk exposures outside of a protective window that occurred in the first part of the study.
It's an exploratory observation, but it's one that we think is going to be important as the debate goes forward, as to what we do next -- not only with these products, but what we do next with other products, in terms of next-generation efficacy studies.
There are discussions in the field as to whether or not we should be doing studies in very, very high-incidence populations, especially those in sub-Saharan Africa. Essentially, should we go to where the epidemic is raging and do studies in parts of the world and in populations where you have very, very high transmission rates? Ultimately, if one were to deploy a globally effective vaccine, it would be logical that those would be the first groups that you would want to provide a public health benefit like a vaccine to.
On the other hand, the other part of the discussion is that if what you learned from RV144 is that this combination of vaccines worked in a Thai population at relatively low incidence, and worked with a circulating type of the virus that you see in Thailand, should you perhaps use that as an extension? Have at least one arm of the next study have those attributes with it? So that's part of the debate and dialogue in the field right now. As a consequence of what we're trying to learn from RV144, I don't think we can come to the conclusion that this vaccine would only have worked in a low-incident population.
Even though our newer data may lead some people to believe that's what we're saying, because of other uncertainties about the duration of the benefit, I think that those two issues may be confused. As I look at these data, I think it provides a rationale for us to have equipoise on that question. My own personal bias would be that I think it makes sense for us to test next-generation vaccines, or even this combination again, in a study design where you have two arms: a relatively higher incidence arm, and a relatively lower incident, heterosexual population. I think that would be a reasonable accommodation of what we actually know, and what we can infer from the data.
Morgane Rolland: What we did is look at the HIV sequences from individuals, volunteers in the STEP vaccine trial who became infected after receiving either the vaccine or the placebo. We had sequences from 40 vaccinees, and 28 placebo recipients. When we compared those sequences, we found that the sequences from the vaccinees were genetically different from the ones in the placebo.
Since this was a CTL-based vaccine, we looked specifically at CTL epitopes and measured distances between CTL epitopes that we identified in the breakthrough viruses. We compared those to the corresponding epitopes in the Merck vaccine insert. The distances were larger among vaccinees than among placebo recipients.
This shows that there are more mutations in sequences from vaccinees than in sequences from placebo recipients, when you compare them to the vaccine insert. And we found that this signal was only seen in epitopes coming from the proteins that were included in the vaccines: gag, pol and nef. When we compared the other proteins of the virus, there was no difference between vaccine and placebo recipient.
This suggests that either the vaccine blocked the outgrowth of specific HIV-1 variants that were the ones that were the most similar to the vaccine insert sequence, or that the CTL responses induced by the vaccine led to further evolution, or faster, or more, mutations in the sequences from the infected individuals. Those were sequences from the earliest time point in infection. We are going to do further studies to see over time.
Reporter #1: To ask Dr. Michael: If you just looked at the people who had some evidence of risk, then it was clear that the vaccine did not work? Was that it?
Nelson Michael: Well, first, let me point out that individuals that described continuous low risk were clearly having risky behaviors. We saw infections in individuals that described that they never had high-risk behaviors. That's the first point to capture.
It is true that, in individuals who at any point in the study said that they had a high-risk behavior, that category of individual was less likely to have a protective effect than individuals who always describe a lower-risk lifestyle. I think the critical point is that that may very well be confounded by the fact that the vaccine appears to have had a transient effect, so that it may not imply that this vaccine would not have worked in higher-incident populations.
Really, the key point is this: We can't confuse incidence, the rate of disease transmission, with risk. I think that's part of the debate we're having right now, collectively, as vaccine development groups, as to what those results may really mean as we go and design next studies. Just be careful with that interpretation. I think that risk and transmission and durability of vaccine effect may all be intertwined. Our ability to tease those out with this study is going to be extremely challenging. I think it is likely going to require additional science and additional studies for us to de-convolute those points.
Reporter #2: This is for Dr. Michael. Are there new candidates for correlates of immunity that have emerged from this study that were not suspected before?
Nelson Michael: That's a great question. I think the simple answer is, within the next two weeks we're going to begin doing the wet work. The short answer to your question is, I don't know yet. And a longer answer will be, probably about one year from now is when we'll be in a position to answer your question.
I described in the latter part of my talk how it's really been, I called it, "all hands on deck." It's a very, very large number of individuals in the field, the majority of whom were not involved in the study. We went out and are trying to collaborate with the best and brightest in the field, writ large, to ask precisely the questions that you have enumerated.
Over the next six months, we'll be doing exploratory studies, trying to understand which of the handful of potential assays we'll use in the ultimate, what we call a case control, study design, which will be the true test of looking for a correlate. We'll initiate those studies in the summer. We should have all those data in by roughly the holiday period of 2010, and hope to be able to publicly discuss those data sometime early in 2011.
Reporter #3: Just one more follow-up question on RV144. Did you do any work to tease out, within the people who reported having any higher-risk behavior, whether that was IDU versus just heterosexual risk?
Nelson Michael: As I described in our presentation, only a very small number of individuals had what we would consider traditional risk behaviors of that type. There were only two individuals that reported commercial sex work, and another 12 individuals who had same-gender sex. We had no individuals at baseline that described injection drug use.
The problem isn't using substances of abuse, it's that if you actually asked, "Did you ever expose yourself to needles in any fashion?" the answer is yes. There were some individuals that had behaviors that were not associated with, as an example, the use of something like heroin, but were behaviors like tattooing, that could be associated with needle use.
What we tried to do -- remember, this was not a part of the study that was powered to ask this question -- we said, can we look at when individuals became infected during the period of time of the study, and look at their proximate risk? Because every six months they are coming in, and we're asking the risk behavior questions. So can we say, "Aha! Here's an individual that described risk, and it looks like there's a blip in infection?"
The answer is that we don't have the power to answer that question. We really don't know. So, yes: There were no individuals at baseline that had injection drug use behaviors, but there were sporadic instances of people being exposed to needles in other fashions. I don't think we're going to be able to rationally come to an answer as to whether or not those other behaviors were associated with transmission.
Just to reiterate: This was really a study of lower-incident heterosexuals in Thailand. And so, in that sense, this was a study not done in traditional high-risk groups, like individuals that are commercial sex workers, or who have same-gender sex.
Reporter #4: I've got a question for Dr. Rolland. Does your study at all address Ad5 seropositivity? Can you look at sequence to see anything? And is there any clarification at this point as to whether circumcision was the determining factor? Can you clarify it at all?
Morgane Rolland: Our own study did not look at Ad5 seropositivity. Ad5 seropositivity is not associated with the higher risk of infection in the largest data that Susan Buchbinder has presented. The significance of the Ad5 was only a .08 significance in the first study.
Since a larger number of cases have been accumulating, you would have to double-check, but this has dropped to .2 now. So the significance with Ad5 seropositivity is gone.
By looking at the genetic influence of the sequences, we are really not looking at acquisition, but more at what's happening afterwards with the HIV-1 sequences. But there was no difference in circumcision. We had the same numbers in the placebo and vaccinees in that.
Reporter #5: One more question. If there were to be a follow-up study with a low-incidence, heterosexual population, where would be places that could be conducted?
Nelson Michael: In our own network, we have sites in East Africa and in Thailand. We actually did a study just about an hour north of Kampala, Uganda, looking for a high-risk cohort. All of us really have conceived of that as the right place to work, for every reason under the sun. It's where the epidemic is raging. That's where you can get answers quickly. That's where you can get answers efficiently, in terms of cost.
And so, we went to Kayunga and we did a formal incidence study. We enrolled several thousand individuals and had them come back every six months -- you know, the old-fashioned way. We were excited to tell the government of Uganda that the incidence in that population was quite low, 0.8, about five- to tenfold lower than the data that you've seen from Glenda Gray, presenting from heterosexual cohorts in the Republic of South Africa.
But for our initial review of that data, we said that's disappointing for us as scientists, because we probably can't work there, in terms of efficiently doing vaccine studies. Now we're rethinking that. So it is possible, certainly, for you to go to what we call "community risk populations" all over Africa, and not necessarily concentrate on the highest transmitting, or the highest incident, population.
Cohorts of a similar ilk to those that we have identified in Southeast Asia, like Thailand, certainly exist in Africa. But there's a debate in the field right now about taking what seemed to be an obvious operational step. What you learned in RV144: you change as few variables as possible. One of those variables might be the level of transmission risk, relatively high versus relatively low.
It's critical to not think that you can only do studies of the type of RV144 only in a place like Thailand. You can do them elsewhere including sub-Saharan Africa. You just have to use good epidemiology to choose where you would do it.
Reporter #6: This is a question for Dr. Michael. What's your budget for this year? How much do you have to conduct the trial again? Do you have enough money to do it, or would you have to get funds from elsewhere?
Nelson Michael: Our budget is complex. I can break it down pretty quickly. The U.S. Army provides my program roughly $30 million a year to do vaccine research. That's hardwired into the president's budget. Then we have a substantial portfolio of grants from the NIAID [U.S. National Institute of Allergy and Infectious Diseases] that runs roughly about another $20 million, but those are constrained to certain projects, so I don't have discretion, in terms of taking those dollars and refocusing them on something else.
Then, a very, very large commitment from the State Department. We have a very large PEPFAR [U.S. President's Emergency Plan for AIDS Relief] program, over 70,000 Africans under antiretroviral therapy in four countries in Africa. And that PEPFAR piece is roughly $70 million. Obviously, I can't touch those dollars, either.
So the straight up answer to your question is: The Army can probably bring roughly $8 million a year to any further study that we would ever do. Just to give you some context of what $8 million means: The most expensive year for RV144 was $17 million. So, if you were to do a study that cost roughly that much, John, the Army could contribute roughly 50% of that. We would need to look to traditional funders like NIAID and the Gates and, potentially, others.
But that's a critical point. Because what the audience should take back from that is, MHRP is not only not willing, but also not able to, if you will, go rogue, and do a next study by itself. We did RV144 with approximately 75% of those funds coming from NIAID. There was a powerful synergy to work in a collaborative way, and we would definitely not look to any follow-on study in any different philosophy; we would do it collaboratively.
This transcript has been lightly edited for clarity.