7
MORNING PLENARY DISCUSSION, DAY 2 (APRIL 5, 2005)
Moderator: Dr. Harvey Fineberg
Panel Participants:
Dr. Ferguson, Dr. Treanor, Dr. Webster
DR. FINEBERG: I invite Professor Ferguson and Dr. Treanor to join Dr. Webster on stage for questions and comments.
PARTICIPANT: Professor Ferguson, yesterday Dr. Gerberding told us that CDC planned to expand the number of quarantine stations from 8 to 30. Could you comment—based on your modeling insights—on whether you think quarantine stations are an effective component of containment?
PROF. FERGUSON: If we are not quarantining individual cases, I predict such an approach will have a fairly limited impact, because by the time cases are diagnosed, viral shedding is already occurring, although the extent depends on the characteristics of the virus. If the virus resembled human influenza, then quarantining arguably would have limited effect, because viral shedding would have declined. If the virus resembled avian H5, quarantining might have a greater effect. In any case, people who are very sick pose less risk of transmission, because they are not moving about the community. So I do not think quarantining would be the principal element of a control program.
PARTICIPANT: I have a couple of comments on the theme that history can lead us to ask certain questions. First, in the seventeenth century, when crossing the ocean took at least 6 weeks and sometimes 10 to 12 weeks, influenza made it from England to the colonies. Those were small ships. One would have thought that in a population as small as 50 and no more than 250, the virus would have burned itself out on that voyage. Information on the population of the ship that carried the disease and the exact duration of the voyage might be useful to your modeling.
The other point is that in the 1889–1890 pandemic, the third wave was the most lethal wave. In researching 1918, I found that public health officials were concerned about that. New York City was the only major city I know of that did not close its schools, but it did quarantine cases. Unlike practically everywhere else in the world, New York experienced peaks in the second and third wave, yet the killing was much more level. Philadelphia had less than one-third the population of New York City yet experienced a higher peak death toll. On a per capita basis, the death toll for Philadelphia and New York was almost identical, but the fact that the peaks were so different, and that the virus moved to the latter city so much more slowly, may be worth investigating. I can not imagine that the quarantine was effective enough to account for the lower peak death toll. Perhaps the fact that fear, prompted people to stay off the streets and normal traffic dropped significantly brought movement below a critical mass.
PROF. FERGUSON: We are very interested in looking at the latter phenomenon. I did not use my model to predict the impact of social distance measures because we regard past pandemics through the filter of intrinsic behavioral adjustments in the population. That is, people decrease their social distance spontaneously, whether by closing schools or through fear, and that was particularly true for 1918.
One of my hypotheses for why three waves occurred in some but not all locations in 1918 was that density-dependent adjustment of contact rates occurred. People did not contact each other during the pandemic, and that reduced transmission rates. The disease went away temporarily and came back multiple times. Other hypotheses to explain the waves exist, but we are trying to correlate transmission and mortality patterns with quantitative data on social distance measures such as school closures.
Our models of international travel do account for journey time and numbers of people, which is why I talked about effective epidemiological contact rates rather than absolute numbers of individuals traveling. But I agree: a 6-week journey with 50 people is a bit of a paradox.
PARTICIPANT: Is it possible that the H5 we now observe in poultry reflects better surveillance in Asia? I am referring to a paper by Ken Shortridge from 1990, which showed that about 7 percent of the rural population had H5 antibodies. If 500 million people live in rural China, this would mean that 35 million people had been infected by H5 before the 1997 outbreaks. Is the situation today truly different, or do higher infection rates simply reflect the fact that we have better ways of detecting the virus?
Dr. Webster: Surveillance in Hong Kong has certainly improved. However, if deaths had previously occurred, they would have been noticed. This virus has undoubtedly changed. Before 1997, the H5N1 viruses did not have the capacity to infect humans, as they do today. That is a learned capacity, with incremental changes then allowing the virus to kill the duck population, to transmit to cats, and so on. This virus is continuing to evolve, and we are probably creating the conditions that allow that to happen. This epidemic began just as Asian countries began to move from backyard flocks to huge chicken houses and pig-raising facilities. So no—current infection rates do not simply reflect better skills.
PARTICIPANT: Could you apply your models to try to block the annual epidemic, using antivirals, vaccines, and measures to reduce social distance, instead of waiting for a pandemic? That is, can we test the modeling and predictions in real time, rather than continue to model until we have a pandemic?
PROF. FERGUSON: I think that would be feasible, and one of my research priorities is to initiate another community study, perhaps equivalent to the Tecumseh study, where we attempt to influence transmission in isolated communities. Blocking seasonal influenza epidemics is probably infeasible, but we don't need to do that. We simply need to demonstrate the impact of control measures on transmission to be fairly sure we are exerting some influence.
Complete blocking is impractical because of the sheer weight of infectious burden on a particularly community. We are not going to treat the whole country, so the infection is constantly challenging the treated population from outside. But we certainly could use well-designed studies to refine the values of our model’s parameters in advance.
Participant: Some of the studies that Ian Stevenson and the group at CDC have done lately, with neutralization antibodies to the Duck Singapore vaccine, suggest significant
heterosubtypic immunity. My question is: what would be the effect of using an H5N1 or an H5N3 vaccine in the affected areas to induce heterosubtypic immunity, to protect against more severe disease and non-sterilizing immunity? Modeling can look at this in terms of the increased likelihood that an antiviral strategy in a population with partial immunity might affect further spread.
DR. TREANOR: Whether we can induce antibody to neutralize viruses other than the vaccine needs to be evaluated. I question whether an inactivated vaccine can do that, but finding out is critical. We could model a less-than-perfect vaccine and try to figure out how much of an effect it would have.
PROF. FERGUSON: Ira Longini's group has looked at the combination of limited-efficacy vaccines and antivirals. These produce a synergistic impact, and the overall effectiveness of the combined control policy is considerably greater than that of just one measure.
Participant: I agree with Dr. Webster that there are good veterinary vaccines and bad ones. However, I would like to comment on regulatory authority in the U.S. and the European Union, which exerts very strong control of manufacturing of vaccines. The production facilities and quality control of multinational companies such as Meriel Intervet, Lohman Animal Health, and even Biommune are very good, so the quality of the vaccines they manufacture is very high. However, one study Dr. Webster did about six years ago showed that quality control in developing countries without a strong regulatory authority is not very good, so vaccines made there show huge batch-to-batch variation. Some are very bad.
Dr. Webster also showed that veterinary vaccines, especially those for poultry, use a lower antigen content to give optimal immunity and proper protection. Vaccine studies show that the challenge dose has a huge impact on viral reduction and shedding. Thus a poor-quality vaccine used to immunize birds looks very good, but if we give the birds a challenge dose of the same vaccine, the results are not very good. Could Dr. Webster comment on how adjuvanted proteins affected the reassortment of the virus and the challenge virus dose he gave in his duck study?
DR. FINEBERG: Dr. Webster, could you also clarify the term "bad vaccine"? I understood you to mean more than poor manufacturing practice.
DR. WEBSTER: The bad vaccines I was referring to were those made in developing countries without adequate regulation. We see this problem throughout Central America and in China. We need to pay attention to unregulated companies. My message is particularly directed at people from Thailand and Vietnam, who are trying to ensure that vaccines are available when these viruses reappear. Those countries would be very wise to emphasize the quality of vaccines and ask for testing of the batches.
PARTICIPANT: There is an old adage about models: all are wrong, but some are useful. I was happy to see that Neil's model fell into the latter category—that models can help articulate many different assumptions and put them in perspective. One of the key parameters used in your model was the (R0 reproducibility number) of approximately 2, which might well be much greater than 2 if we calculate the reproductive rate based on mortality rates rather than infection rates. Can you comment on the fact that the R0 of a pandemic strain is about 2, while that of epidemic strains is much higher? I thought that influenza is one of the most highly communicative diseases, like measles and varicella.
PROF. FERGUSON: I, too, always thought of influenza having an R0 similar to that of measles and other such diseases. However, a detailed analysis of inter-pandemic flu—particularly household transmission rates—shows that the virus is not that transmissible. The inter-pandemic R0 is probably about the same as the pandemic R0.
The question is what is the R0 of influenza? The study undertaken by Mills et al., and other historical studies of transmissibility based on mortality data, show that whether we measure mortality or infection rates does not matter as long as the ratio of one to the other remains constant. I do not believe the measures are affected by the fact that we are looking at deaths rather than infections.
The one caveat regarding inter-pandemic transmission rates is that multiple strains are circulating. The accumulated reproduction number of all the strains in circulation at a particular point in time is probably somewhat higher than that of a single strain. However, it's probably no more than 3–4, which is considerably less than that of either varicella or measles.
