Aug 3, 2005 (CIDRAP News) – Researchers relying on mathematical modeling claim that early containment of pandemic influenza, an eventuality widely accepted as not an "if" but a "when," may well be attainable through targeted public health strategies. A leading public health expert, however, cautions against viewing these study results as fodder for relaxed planning efforts, stressing the myriad variables that would come into play.
In a report published online today in Nature, Neil M. Ferguson of Imperial College London and international colleagues outline a simulation of flu transmission among a modelled population of 85 million people in Thailand and environs, chosen because of the availability of relevant demographic and other data. Ira M. Longini, Jr, of Emory University and colleagues from that institution as well as from Johns Hopkins and the Thai Ministry of Public Health, describe in a Science report also released today a model of flu transmission based on a largely rural population of 500,000 in Southeast Asia.
Both groups of researchers examined various intervention strategies aimed at containment of an emerging pandemic strain. The avian influenza A H5N1 strain now circulating in birds in Asia, which has caused more than 100 human cases and more than 50 deaths, is so far not efficiently transmissible from human to human. When the virus gains this ability, it is widely expected to launch the next flu pandemic.
The Thailand model (Ferguson group) based calculations on human viral reproduction, or transmissibility (RO), rates ranging from about 1.0 to 2.0. These rates refer to the average number of secondary cases of disease generated by a typical primary case in a susceptible population; an RO rate of 1.0 would thus indicate no transmission. They set the generation time (Tg), meaning the average interval between infection of an individual and infection of contacts, at 2.6 days. This Tg factor was arrived at on the basis of analysis of past estimates of transmissibility of respiratory diseases and is less than the approximately 4 days assumed in most past modeling studies, say the authors. A predicted attack rate of 50% to 60% derived from these factors is consistent with the first two waves of past flu pandemics, according to the researchers. The model also assumed that 50% of infections were clinically recognizable.
Using these variables and beginning from the scenario of a single infection in a rural (low population density) area, the authors found that the diseases remained mainly local for the first 30 days but then spread nationally between days 60 and 90. The critical period for containment is obviously before the second time period, as they point out.
Containment strategies in this study were antiviral prophylaxis using oseltamivir, social distancing (closing schools and workplaces), and quarantining (restricting movement into and out of the affected areas).
The researchers found that blanket prophylaxis could contain a pandemic strain of influenza with high transmissibility (RO >3.6). This approach is obviously unrealistic because of the drug quantities that would be required. Thus targeted prophylactic approaches were modeled.
Social targeting, meaning administering the drug to contacts in an infected person's household, school, and/or workplace was effective only at very low transmissibility rates (RO=1.25) so would be unlikely to be effective if used alone.
Geographic targeting, meaning prophylaxis of the whole population within a set radius of a newly identified case, was capable of containment at an RO of 1.5 assuming a 2-day delay from case onset and a 5-km ring. About 2 million antiviral courses would be required. The authors considered this method effective only if it were untaken early and if a major urban center was not involved early on.
A drug-sparing targeted approach, in which only a set number of people nearest an index case within a set radius are given antivirals, was found to have increased effectiveness over just geograhic targeting, and the method would required less drug usage.
In studying the effects of social distancing, the authors made the conservative assumption that household and random contact rates would increase 100% and 50%, respectively, for infected individuals and so would affect attack rates in those settings. Results showed that this approach added to a drug-sparing prophylaxis policy would have a greater than 90% chance of containing a virus with an RO of 1.7.
Quarantine was found to increase the effectiveness of other strategies even if only 80% of movement into and out of the zones occurred (90% containment at RO =1.8 when added to geographic targeting of antiviral prophylaxis and similar containment at RO =1.9 when social distancing is also added).
These authors point out a number of factors that would affect practical application of containment approaches they found to be effective. Among them are the applicability of the assumptions, such as transmissibility, used in the model to the characteristics of an actual pandemic virus strain; the effectiveness of surveillance, since success will depend on early identification of the first cluster of cases; effective delivery of treatment to targeted groups, which requires an infrastructure not present in many areas; and antiviral resistance, which could arise with widespread drug use.
Among the broad conclusions drawn by this group, who claim that their study is "perhaps the largest-scale detailed epidemic microsimulation yet developed" are that a stockpile of at least 3 million drug courses would be needed to contain a pandemic flu outbreak and that containment is unlikely if a new flu strain has an RO of greater than 1.8. The use of multiple approaches to containment is key, they stress.
The study published in Science (Longini group) used a discrete-time, stochastic influenza simulation model for a geographically distributed population of 500,000 in rural Southeast Asia. Like the study discussed above, it examined the effects of targeted antiviral prophylaxis and quarantine on containment of an emerging viral strain, and it added another approach as well: preexposure vaccination with a low-efficacy vaccine, which is all that would be available early on in a pandemic situation.
In the model, a target overall attack rate of 33% was used, corresponding to an RO of 1.4. This is similar to the rate in the first wave of the Asian (1958) and Hong Kong (1968) pandemics, the authors say. They made calculations based on delays of 7, 14, and 21 days between infection and implementation of interventions.
All of the intervention were found to work well if the viral strain had a very low rate of transmissibility (RO =1.1). Social targeting and geographic targeting of antiviral prophylaxis that reached 80% and 90% of the targeted population, respectively, were each effective if RO was 1.4 or lower, but neither was sufficiently effective at an RO of 1.7 or greater. When preexposure vaccination was added, containment of strains having an RO of 1.7 was possible, even when only 50% of the targeted population was vaccinated; it could be effective at an RO as high as 2.1 with higher vaccination. If quarantine were also added, strains with an RO of up to 2.4 could be contained, the researchers found.
These authors conclude on the basis of their findings that the current World Health Organization (WHO) stockpile of antivirals, which amounts to 120,000 courses, "could be sufficient to contain a pandemic if the stockpile were deployed at the source of the emerging strain within two to three weeks of detection," and that "up to one million courses could be needed to deal with multiple outbreak foci."
"I want these strategies to work," infectious disease expert Michael Osterholm, PhD, MPH, told CIDRAP News. "But in all my years in public health, I have yet to see mathematical models that have driven public health actions in meaningful ways." Osterholm used HIV and bovine spongiform encephalopathy (BSE) as examples of diseases for which there have been what he calls a "pandemic of modeling studies."
"My concern is that papers like these suggest more direction for planning than is warranted and may placate policymakers who believe the planning puzzle has clear solutions. . . . The issue of antiviral treatment, for example, has to be looked at against the whole system of disease occurrence and transmission. How well can we detect the disease when it starts occurring? How can we make sure travelers who appear healthy aren't unknowingly spreading the virus?" Osterholm, who is director of the Center for Infectious Disease Research and Policy at the University of Minnesota, publisher of this Web site, used as illustration the example of SARS' fast jump from the Far East to Canada in 2003.
Osterholm also made the point that since a flu pandemic will very likely be caused by a mutation of the H5N1 virus currently spreading among birds in Asia, we will be facing a "reloading" problem at the source—that since birds are a reservoir that is constantly replenished, "We are dealing with a moving target, not a static population like humans. . . . Culling [the birds] won't work. It's like throwing fresh wood on a fire."
The two studies published today were funded by the National Institutes of Health (NIH).
Ferguson NM, Cummings D, Cauchemez S, et al. Strategies for containing an emerging influenza pandemic in SE Asia. Nature 2005 (published online Aug 3) [Abstract]
Longini IM, Nizam A, Xu S, et al. Containing pandemic influenza at the source. Science 2005 (published online Aug 3) [Abstract]
Aug 3 NIH press release
Aug 3 WHO statement
CIDRAP overview of pandemic influenza