A STEP Back?

Additional data released from the STEP trial raises questions about whether the vaccine may have increased the risk of HIV infection

By Kristen Jill Kresge

Clinical trials are intrinsically complex, but according to Mark Feinberg of Merck, the STEP trial may be an extraordinary case in this regard. "I've never seen more complicated data emerge from a study in any field that I've witnessed."

The public got a taste of this complexity at an open session of the HIV Vaccine Trials Network (HVTN) meeting on November 7 in Seattle. There, Merck, along with several representatives from the HVTN and the National Institute of Allergy and Infectious Diseases (NIAID), released mounds of additional data from the STEP trial, a Phase IIb test-of-concept trial of Merck's adenovirus serotype 5 (Ad5)-based vaccine candidate, known as MRKAd5. Since immunizations were stopped in this trial on September 21, investigators, biostatisticians, and clinical trial experts have spent many sleepless nights running assays, analyzing data, and interpreting the results of this pivotal study.

And the results, based on data from all 3,000 volunteers, show that even though the vaccine induced HIV-specific cellular immune responses, they were not effective at preventing HIV infection or in reducing levels of the virus in individuals who became infected despite vaccination. "The immune responses were reasonable," says Mike Robertson of Merck. "The lack of efficacy is not explained by sub-optimal immune responses."

Moreover, there is a trend toward a higher number of infections in vaccinees as the level of pre-existing antibody immunity to the Ad5 vector increases. This Ad5 immunity is due to prior exposure to the naturally-circulating strain of the cold virus, which is used in the vaccine candidate to deliver three HIV antigens: Gag, Pol, and Nef. In the STEP trial, the study's sponsors revealed in Seattle, there were 49 HIV infections overall in the vaccine group and 33 among those who received placebo as of October 17. But in individuals with the highest levels of Ad5 antibody, the imbalance was more pronounced—21 infections in vaccinees compared to 9 in placebo recipients. "This difference is clinically important for at least one subgroup, the high Ad5," says Keith Gottesdiener of Merck. "I don't really need any statistics to make a declaration that it's an important factor to take into consideration."

The explanation for this difference is still not clear. There are several possible confounding factors, including race, geographical region, age, and circumcision status of the volunteers. According to Susan Buchbinder of the University of California in San Francisco and principal investigator of the STEP trial, there is mixed data on the protective role of circumcision in men who have sex with men (MSM), the predominant population involved in this trial. In uncircumcised men there were far more HIV infections in the vaccinees while in circumcised men there was an even split of HIV infections in both vaccine and placebo recipients.

But at this stage of the analysis, the trend towards increasing rates of HIV infection among vaccinees persists even after factoring in all of the known potential confounders, says Steve Self, a biostatistician with the HVTN and the Fred Hutchinson Cancer Research Center. "The confounding factors certainly aren't the full answer," says Larry Corey of the HVTN.

Researchers are now hard at work trying to determine why the vaccine was not effective and any role it may have had in increasing susceptibility to HIV in some individuals. There is great uncertainty about the evidence that the vaccine may have enhanced the risk of HIV infection but there are some possible biological explanations for this difference, and researchers must now sort out their plausibility.

"There are going to be a lot of different hypotheses that need to be tested to try and understand what went wrong; why this wasn't efficacious and why there was a trend toward more infections with vaccine than the placebo," says Bruce Walker of Harvard Medical School in Boston, who is leading a team of scientists who will analyze the data from the STEP trial. But the devil is in the details and until the full analysis of this trial is complete, and maybe even after that, there will be many unanswerable questions. "Some of those things will take months and some may take longer than that," says Walker. "We were entering into this thinking that we will find an answer, but even that's not absolutely guaranteed."

Searching high and low

The STEP trial—also known as HVTN 502 and Merck V520-023—was co-sponsored by Merck and NIAID. It was a Phase IIb test-of-concept trial of MRKAd5, a candidate that induces cell-mediated immunity (CMI) and not antibodies against the virus. Antibody responses are how most, if not all, licensed vaccines provide protection. This study involved 3,000 healthy volunteers at high risk of HIV infection at HVTN sites in North and South America, the Caribbean, and Australia. All volunteers were scheduled to receive three shots of placebo or vaccine, which contains a mix of Ad5 vectors carrying one of three different HIV genes, gag, pol, or nef. The inserts were from HIV clade B, matching the predominant clade circulating in the areas in which the trial took place. A companion study of MRKAd5 was also conducted in South Africa (see 'Stopping a steam train').

The original plans for the STEP study only included 1,500 individuals with low levels of Ad5 antibody (less than 200 units, in which a unit is a measure of the antibody concentration required to neutralize Ad5). But after the trial began, data emerged from earlier Phase I and II trials that showed pre-existing immunity to Ad5 did not compromise HIV-specific immune responses to the degree that researchers had initially expected. In July 2005, seven months after the STEP trial began, the protocol was amended to include a second group of 1,500 volunteers who had high Ad5 antibody titers (greater than 200 units). The majority of the volunteers, 61%, in this second group were women. As a result of this modification, both primary endpoints for the trial—identifying the ability of the vaccine to prevent HIV infection or reduce viral load in volunteers who later became infected—only applied to the low Ad5 titer group. Investigators added an equivalent set of secondary endpoints relating to the group with high Ad5 antibody levels.

Immunizations in the STEP trial were halted on September 21 after the trial's independent data safety monitoring board (DSMB) reviewed the data for the first time. This interim analysis was triggered by the accrual of 30 HIV infections within the sub-group with low levels of pre-existing immunity to Ad5. The DSMB conducted what is known as a per protocol analysis of volunteers in the sub-group who had received at least two injections of MRKAd5 or placebo, who met all of the specifications of the trial protocol, and who did not become HIV infected within the first three months of the study. The DSMB concluded that based on the breakdown of infections at this time—19 in the vaccine group and 11 in placebo recipients—it was futile to continue immunizations because the vaccine was not effective. Based on the trend toward a higher number of infections in vaccine recipients, the DSMB also encouraged the continued follow-up of trial volunteers.

After futility was declared, researchers at Merck and NIAID decided to proceed "cautiously" with a post-hoc analysis of the data collected up to that point, according to Robertson, who warns that all these interpretations should be taken "with a big grain of salt," as the trial was not designed to look at different sub-groups separately. When immunizations were stopped, only one HIV infection had occurred within the 1,150 women in the trial and this volunteer received placebo, not vaccine. Since the HIV incidence was so low in female volunteers, investigators conducted all the subsequent analyses on the 1,850 male volunteers only.

The trial investigators also chose to focus on what Robertson calls the "broadest population we could look at," which is different than the per protocol analysis. This analysis, known as modified-intention-to-treat (MITT), includes volunteers who were infected with HIV during the first three months as well as those who didn't meet the exact specifications of the trial protocol.

When the individuals who became HIV infected were analyzed according to their level of Ad5 titer, researchers noticed a worrisome trend (see Table 1). In individuals with virtually no pre-existing immunity to Ad5, there were 20 infections in both the vaccine and placebo group. In the group with the lowest level of Ad5 titers, there were 8 infections amongst vaccinees and 4 in placebo recipients, while in individuals with mid-range levels of pre-existing immunity (Ad5 titers between 200 and 1,000), there were 14 vaccine recipients who became HIV infected and only 7 in the placebo group. And in volunteers with the highest Ad5 titers (concentrations of antibody greater than 1,000), there were 7 infections in the vaccine group and only 2 in those who received placebo. "As you move up [into higher levels of Ad5 antibody] there is a trend for the ratio to get higher," says Robertson.

Table 1. Number of HIV Infections According to Ad5 Antibody Titer.

View larger image

But Self says, "There is great uncertainty about some of these trends." Regardless, researchers are taking it seriously. "When looking at potential harm we have to pay close attention, even if the p value is not significant," says Buchbinder.

For many, this was an unanticipated outcome. "It was a surprise to us that there were actually more infections in vaccinees than in placebo recipients," says Robertson. "We didn't expect that," says Peggy Johnston of the Division of AIDS at NIAID.

But in 2004 Feinberg co-authored a scientific paper, in his capacity as a researcher at Emory University, issuing a potential warning to the vaccine field, saying "candidate AIDS vaccines may not simply be either efficacious or neutral; they may also have the potential to be harmful." This study indicated that vaccination of rhesus macaques with a varicella-zoster virus vaccine carrying the simian immunodeficiency virus (SIV) Env protein resulted in an increase in the number of activated CD4+ T cells upon SIV challenge (PNAS 101, 13026, 2004). This increased number of proliferating CD4+ T cells, in the absence of a balanced CD8+ T cell response, led to higher levels of SIV replication and enhanced disease in the vaccinated monkeys. A similar result was not seen in the STEP trial: even though the risk of acquiring HIV seemed to be higher amongst vaccine recipients with high Ad5 antibody titers, there was not a significant difference between the viral loads in this group compared to placebo, suggesting there wasn't increased HIV replication.

This nonhuman primate study was also done with a different virus vector, which was replication competent, unlike the Ad5 used in the STEP trial, and with different viral inserts.

Heads or tails

Despite the massive amount of data that has already been interpreted and presented on the STEP trial, there is a lot of work still to be done. One of the leading questions researchers will set out to answer is why the vaccine was not efficacious.

The results from the IFN-g ELISPOT assay show that the immune responses induced by the vaccine were similar or higher in the group with low Ad5 immunity to those seen in previous Phase I and II trials, and only somewhat different in the high Ad5 antibody titer group (see Table 2). These responses did not differ between those who became HIV infected and those who didn't.

Table 2. Percent Responders by INF-Y ELISPOT Assay at Week 8.

View Larger Image

In the group with low Ad5 immunity, 79% of volunteers mounted immune responses to at least one HIV antigen included in the vaccine, while 63% had responses to all three. Far fewer individuals, only 23%, with high Ad5 antibody titers mounted immune responses to all three HIV antigens, but 62% had immune responses to at least one. For now it is still unclear if the vaccine just didn't provide the quantity of T-cell responses necessary for protection, or if they were of the wrong quality. "What was the functionality of the immune responses that were generated?" asks Walker. "We had evidence of IFN-g production but that doesn't tell you if the cells would kill virus-infected cells, so we will obviously be looking a little bit more at the function of the immune responses." These results may also impact the future use of IFN-g ELISPOT assays for assessing the relative efficacy of vaccine candidates (see Getting it right early).

There are also many additional studies planned. Researchers will sequence the viruses that infected the volunteers to look at whether or not the HIV epitopes included in the vaccine antigens were present in the infecting virus strains. This work will be done in cooperation with Francine McCutchan of the US Military HIV Research Program and might provide valuable information about why the vaccine didn't work. It may also help elucidate whether a single vaccine candidate can provide sufficient antigenic coverage for the exceptional diversity of HIV.

There are also plans to do whole genome sequence analyses of some of the volunteers to identify any genetic components that might have enhanced susceptibility to HIV or, conversely, provided protection to placebo recipients. "Part of the problem is there are actually very few samples that are available," says Walker. "We just don't have that many so we are going to have to make decisions about what we prioritize."

The next question to tackle is whether or not the vaccine enhanced susceptibility to HIV infection. Julie McElrath of the Fred Hutchinson Cancer Research Center says this is "the question on everyone's lips." One possible explanation is that with the adenovirus there is increased activation of CD4+ T cells expressing the CCR5 coreceptor, thereby creating more target cells for HIV, says Walker. McElrath has found that individuals with high Ad5 antibody titers do have higher levels of activated, CCR5 expressing CD4+ T cells in peripheral blood. However, within this group there is no difference between vaccine and placebo recipients. "The whole thing is puzzling," Walker adds.

There is also some evidence that HIV-specific CD4+ T cells can migrate to the male genital tract—researchers have detected them in both peripheral blood and seminal fluid samples, according to Danny Casimiro of Merck. McElrath plans to further study the CD4+ T cells in both the rectum and lower male genital tract and will also investigate if the CCR5+, activated CD4+ T cells in vaccinees are more susceptible to HIV in vitro. "There is a hint of something going on here but these are very preliminary studies," she says.

Table 3. HIV Incidence Rates During STEP Trial

View larger image

Broader strokes

McElrath raised two other questions in Seattle that she and others, who are busy analyzing the STEP data, will also attempt to address: is this observation specific to Ad5? And could there be a similar issue with other viral vectors? "We have no clear clue about that but we will try to address this in the best way we can," says McElrath.

Feinberg emphasized that the way the STEP trial is handled has implications that extend far beyond this single vaccine candidate. "We have to work together. It's easy for that to sound like a platitude but it's not," he says. "If we don't do this right, the whole field will come apart at the seams and we can't allow that to happen." Merck and NIAID were openly seeking input from various investigators, advocates, and community members at the HVTN meeting, making it abundantly clear that they see issues, like the ongoing analysis of data, as a collective decision and undertaking.

Until any possible association between Ad5 immunity and increased susceptibility to HIV is ironed out, most researchers are urging caution. "Any further trials of adenoviral vectors should be done very cautiously," says Johnston. "Researchers will need to carefully consider whether enrollment of individuals with existing immunity to that adenovirus serotype can be justified from a safety perspective, at least until the STEP results are better understood." And based on the complexity of the data generated by this trial, it may be a long road. For now, most agree it is too early to close the door on CMI-inducing vaccines. "One trial does not mean the concept is not correct," says Corey.

Andreas von Bubnoff contributed reporting to this article.