H5N1 vaccines- how good is good enough?
A number of human clinical trials of H5N1 vaccines are in various stages of completion. Although vital tools to combat the possibilty of pandemic influenza, do we really know how to interpret the data from these trials?
Last week saw the publication of the first batch of data from human clinical trials with vaccines designed to target the contemporary H5N1 viruses. This data was presented by John Treanor and collaborators from the US National Institute of Health Vaccine Trial Efficacy Units in The New England Journal of Medicine, one of the premier medical periodicals. Their approach was as close to conventional as one can get with H5N1 vaccines (due to their extreme ability to cause disease some deviations from the conventional approach must be taken with H5N1 viruses: but that’s another issue). The vaccine was produced by Sanofi Pasteur under contract to the US government. Sanofi Pasteur is a regular in the flu vaccine manufacturing business and the H5N1 vaccine itself was made using essentially the same technology as used in yearly interpandemic flu vaccine preparations. Basically, virus was grown in large quantities in embryonated chickens eggs, the virus inactivated and split into its individual protein components using chemical treatment, then partially purified and formulated according to proprietary procedures. The clinical trial investigators then took this vaccine and administered different doses into 451 healthy volunteers aged between 18 to 64 years of age.
The purpose of this first trial was to assess the safety of the vaccine as well as to determine its optimal formulation. In flu vaccines, the active component is the amount of flu virus HA (hemagglutinin) protein. This is the protein that protrudes from the surface of the virus particle and attaches the virus to its target cells. The purpose of conventional flu vaccine approaches, such as described in the New England Journal of Medicine report, is to induce the production of antibodies that bind to the HA protein and inhibit the virus attachment to the host cell. The typical adult dose of HA required to successfully induce this immunity in an interpandemic flu vaccine is 15ìg. In their clinical trial, Treanor and colleagues examined doses ranging from 7.5 to 90ìg of H5 HA. The rationale for this approach is that it is substantially more difficult to get a good response to a vaccine in an unprimed population than it is in a primed one due to the ability of the immune system to remember foreign proteins. Unprimed describes individuals who have not “seen” the HA before (through prior vaccination or infection). After a single administration of these doses of vaccine, Treanor and team found that 28% of people vaccinated with the 90ìg dose developed what was considered a protective response as determined by laboratory measurements of antibody concentration. The results for the lower doses were correspondingly lower. This result was certainly not unexpected and was in fact encouraging. Realistic expectations were that the vaccine would need to be given twice a number of weeks apart to produce a decent level of antibody in vaccinees. Less encouraging to many, however, were the results after a second administration of vaccine. Even at the 90ìg dose, two shots resulted in only 58% of people achieving the desired response. At the standard dose of 15ìg, only 24% reached this level. Certainly the vaccine was safe with only minor side effects, but even at doses 12 times that of an interpandemic vaccine only a little over half of the participants achieved what is considered a protective response. It is certainly unrealistic to think that using this vaccination approach a typical 15ìg shot would be enough for an H5 vaccine, but 12 times? The presentation of these results has been accompanied by a number of experts and non-experts alike dismissing the vaccine as ineffective as a pandemic vaccine.
Although there is no doubt that other approaches should be aggressively pursued, there is a rather large black box that prevents an accurate appraisal of the Treanor study results. That block box is how much antibody is enough. The standard cutoff for defining a successful vaccination with conventional flu vaccine is a hemagglutination inhibition (HI) value of 40. Those that achieve a level of 40 are considered protected, those below not. The HI is an assay that measures the ability of antibodies to bind to the HA protein of flu and stop it binding to red blood cells; essentially a crude measure of the amount of antibody present in a sample. This is a very easy to run assay that is also very cheap. A HI value of 40 means that the serum from a vaccinated individual can be diluted up to 40 times and still inhibit the binding of flu to the receptors on the red blood cells. A value of 160 means that the serum can be diluted up to 160 times and still inhibit and so on. A HI titer of 40 has been considered as protective due to a number of clinical studies with human H3N2 and H1N1 viruses. Certainly in these cases 40 correlates with a decent level of protection (i.e., persons with a HI value of 40 or above will be protected). The cutoff becomes less clear when we start discussing H5 vaccines. The major problem is that we do not know what a protective level of antibody against these viruses is in humans. Additionally, we need to consider what the immediate goal of a stockpiled pandemic vaccine would be. Whereas the goals of an interpandemic vaccine are to alleviate much of the symptoms of an infection (very few vaccines can actually prevent infection), a pandemic vaccine is more targeted towards preventing severe disease and death. Under these circumstances perhaps a cut off below 40 is acceptable. Unfortunately until we actually have a H5 outbreak in humans we will never know; the ethical issues with an experimental H5 infection of humans are obvious. So what can we use to determine if 40 is overly ambitious and unnecessary for a H5 vaccine? The answer to this question is unfortunately little. The best marker we have for vaccine efficacy comes from animal models of influenza. Mice and ferrets are the most often used models. Although easier to use, mice are not great models for human influenza and ferret are superior in this regard. Ferrets are susceptible to human strains of flu and exhibit similar symptoms and immune response to humans. Correspondingly, a number of ferret trials have been conducted with H5 vaccines. In these studies, some of which have been published, others in various stages of publication, it appears that levels of antibody much lower that HI values of 40 can be protective. There is even precedent where animals had no detectable HI antibody titers but were protected from subsequent H5N1 challenge (the virus replicated, but for a shorter number of days and no severe symptoms were seen).
So what does this mean? My take is that a HI value of 40 is not necessarily needed for protection from H5N1 infection and that the results of the Treanor study are indeed very encouraging. What is very much needed, however, is the uncovering of a good “immune correlate of protection”. An immune correlate of protection is a value derived from some assay that correlates with protection from H5N1, much like a HI of 40 for H1N1 and H3N2 human flus. At the moment we do not have one, and we cannot accurately measure the success of H5N1 vaccine clinical trials. Considering the large amounts of State and industry monies going into trials of such vaccines, it is imperative that we are able to effectively judge their effectiveness. Until we are able to do this, judging the efficacy of H5N1 vaccine clinical trials in humans will remain guesswork.