Revisiting the swine flu puzzle of 1976

On 4 February 1976, a soldier died at Fort Dix army base from acute respiratory disease. Analysis of tracheal swabs from this soldier showed that he was infected with a novel H1N1 influenza similar to those circulating in swine. The new virus, dubbed A/New Jersey/76, was of concern since H1N1 strains of influenza had not circulated in the human population since the 1957-1958 pandemic. Since Fort Dix was an infantry training facility, the population was younger and nearly all inhabitants were immunologically naive to any H1N1 influenza strain. Subsequent investigations revealed that A/New Jersey/76 had circulated widely in the trainees at Fort Dix between 5 January and 14 February, by which time the virus had apparently gone extinct. (Lessler et al., J. R. Soc. Interface [cites omitted].

Fort Dix was a training center, organized into 7 companies. A company had 4 platoons, each of 50 new recruits. There was very little contact between platoons. The recruits had just returned from winter break when an outbreak of respiratory disease occurred about a week after. The last evidence of transmission occurred about 33 days after the first case, so this was an explosive outbreak in an isolated and contained population of young men, in close contact, that lasted a month. No case outside of Fort Dix was ever identified, although the outbreak wasn't recognized until later and thus there was no quarantine or isolation. It just never went anywhere. Why?

Using data from 1976 and modern analytic methods, Lessler et al. estimated two parameters of epidemiologic interest, the basic reproductive number, R₀, and the serial interval. R₀ is the average number of new cases produced by introducing an infected index case into a completely susceptible population. The same number for seasonal flu isn't really R₀ because many people have some immunity previous years. The average number of new cases in that kind of population isn't called R₀ but "effective R." But these new recruits were mostly too young to have acquired immunity the last time H1N1 was around (1957) so the estimate in this case was considered reasonable for an R₀. The other number, the serial interval, is the average time from one case producing another infected case. The explosiveness of an outbreak is not just related to R₀ but the combination of R₀ and a short serial interval. We know that explosive outbreaks can occur with rather modest R₀'s combined with short serial intervals. However it is R₀ that tells us if the virus will spread or not. If R₀ is less than one, it is destined to extinction.

But even a virus in a susceptible population that has R₀'s above 1.0 will burn out in closed communities. A virus infecting recruits in a more or less isolated setting of an Army base can run out of hosts. As long as the infected soldiers didn't come in contact with new susceptibles inside or outside the base, the outbreak was destined to burn out.

But of course they weren't completely isolated. There was contact with the outside world. So why didn't it spread? It certainly spread quickly within Fort Dix, so why not outside it? The answer Lessler et al. give in their paper (and I find it plausible) is that this virus just wasn't very fit. The R₀ they estimated was extremely low, only about 1.2, which is lower than any other R₀ estimate for a seasonal or pandemic virus in the literature [NB: if you are wondering about confidence intervals, these folks are from Hopkins where likelihood methods are favored, so they talk about "supported regions": I won't bore you with the details of internecine warfare amongst biostatisticians except to say it is surprisingly fierce. At any rate, the supported region in this case runs from 1.1 to 1.4, so even the high end is relatively low]. Now estimating R₀ for seasonal or previous pandemic flu is subject to many uncertainties, but they are virtually all above 1.5, often above 2.0 or 3.0. So we are talking about a virus that is a weakling in the R₀ area. The estimated serial interval was also fairly short: 1.9 days, so even a modest R₀ could have explosive spread.

So let's return to our puzzle. We have a virus that spread quickly among Army recruits but not at all in the general population. The Lessler et al. explanation is that the recruits in basic combat training were such a fertile ground that even a puny virus that transmitted poorly could succeed. The recruits were under stress, in close contact and immunologically naive to the virus. When the virus used up its "food" (the recruits), it died out. In the outside world the conditions were different. Even without considering the special living circumstance of the recruits, the general population still may have had herd immunity to H1N1. "Herd immunity" is the effect of having a proportion of people in the community who are immune, that is, not susceptible. For each effective contact between people in the community there is a probability that the virus will transmit. If the contact is with an immune person, however, the probability of transmission to that person is zero. This effectively reduces R₀, and if enough people in the community are immune (because of previous infection or having been vaccinated), R₀ is effectively reduced below 1.0. What proportion of the population do you need for effective herd (i.e., population) immunity? That depends on R₀, but for low values it is surprisingly small. For an R₀ = 1.2, you only need 17% (one in six) to be not susceptible; for R₀ = 1.1 it only needs to be 9%. Moreover there is much less contact than between recruit to recruit, reducing effective R₀ even more. This virus just didn't have the muscle power to make it in the real world.

So how often does this happen? We don't know. This situation is very unusual in that fairly detailed epidemiologic data were available to make these estimates. That almost never happens. The paper was written before the current swine flu outbreak and its point of reference (in 2007) was general concern with H5N1, which transmits even more poorly than the 1976 swine flu. But the authors did have some sobering cautions. The data did indicate that even with such a poorly transmitting virus, in the special setting of his military encampment it was able to transmit through an estimated chain of six links:

The Fort Dix experience illustrates that the particular circumstances of emergence of a virus can profoundly affect its subsequent course. The circulation of A/New Jersey/76 within the population of military recruits, despite its apparent low fitness, raises concerns that similarly compact populations may serve as bridges for emerging viruses, providing viruses not yet capable of spreading efficiently in the general population time to adapt. The detection of A/New Jersey/76 and its failure to spread beyond the confines of Fort Dix provides hope that with improvements in surveillance and control techniques, future outbreaks of novel influenzas can be contained before they cause a pandemic.

It's now clear that for an R₀ more like a seasonal virus, containment is quite unlikely, but the point about surveillance is still critically important. We not only need to strengthen the surveillance system but include in it data mechanisms for easier and more timely estimates of how well an agent is transmitting. This is just as important as its virulence, which has gotten the lion's share of interest. It also suggests that the current poorly transmitting swine flu could well have been circulating for a long time in another closed population with intense contact, factory hog farms, until it acquired some feature that suddenly raised its R₀ for humans to a level where it could circulate among us. But there are other possibilities, so at this point we don't know the answer to this pesky beast came from.