Mild bird flu: no news is bad news

For a long time I (and many others) were of the opinion that the reported deaths from H5N1 and the extraordinaraily high Case Fatality Ratio (CFR; proportion of all infections that end fatally) was an over estimate due to underascertainment of infections that were mild, inapparent or just undiagnosed because they weren't severe enough to come to the attention of the medical care system. The reason for thinking this was that this is the pattern for most other infectious diseasesk, even serious ones like TB and cholera. Most of the infections are asymptomatic or at least undiagnosed. It is estimated that half of all seasonal flu infections are also asymptomatic. So this was expected to be true of H5N1 as well. But for some time there has been no or little evidence of the hypothesized "missing" cases. Every time we look hard for them we don't find them. The latest search was reported in the form of an abstract and oral presentation at the big flu meeting in Toronto. I'd like to say they have now been found. But I can't:

A study of 901 people who lived near confirmed victims of the H5N1 avian flu strain showed no serological evidence that they had been infected and fought off the disease, Rapeepan Dejpichai, M.D., of the Thai Ministry of Health told attendees at the Options for Influenza Control meeting.

Even those who were in close contact with both infected birds and infected people showed no sign of ever having been infected, Dr. Dejpichai and colleagues found.

[snip]

The participants were mostly male, with a median age of 40; 85.5% were selected by convenience sampling, the rest by random sampling.

Study participants were asked about respiratory illnesses and risk factors, such as exposure to infected birds or the human patients.

The researchers also took blood samples, which were tested for antibodies to the H5N1 virus, using the so-called microneutralization assay and confirmed by ELISA. An antibody titer of greater than 1:40 was considered positive.

The researchers found:

• There were no significant differences between the random and convenience samples.
• 68.1% reported contact with backyard poultry.
• 33.3% reported contact with sick or dead chickens without use of any protective equipment.
• 7.1% reported contact with the infected humans.
• All participants tested negative for H5N1 neutralizing antibodies.

During the time that the human victims were developing the disease, 118 study participants recalled having their own acute respiratory symptoms, Dr. Dejpichai said. Of those, 33% had had contact with sick birds but all had antibody titers less than 1:40 and were considered seronegative for H5N1 infection. (MedPageToday)

What do we learn from this, acknowledging the results are preliminary and have yet to undergo the usual peer review process that is part of normal scientific publication. Let's divide it into The Good News, The Bad News, and Questions and Uncertainties.

The Good News. The virus circulating in Thailand in 2004 - 2005 (so-called clade 1) was transmitted only rarely and with difficulty from infected poultry (assuming that was their source, which is likely but not 100%). Failure to find evidence of infection elsewhere means that surveillance can focus by using seriously ill cases in hospitals as sentinel events, as the session moderator, CDC's Joseph Bresee, observed.

The Bad News. The current Case Fatality Ratio of over 60% may not be the huge overestimate we expected it to be. This disease is Ebola-like in its lethality. The hoped for phantom cases are looking more and more like they don't exist.

Uncertainties and questions. The surveyed population is relatively large (over 900) but it was a "convenience sample," meaning it was people that were available to the investigators and willing to participate, not a random sample of the population at risk. Thus the results are not a true estimate of the base population. On the other hand, it is hard to see why this would produce a negative selection bias, although this is possible. If the results are from biased sampling, it must be either a very strong bias or a fairly small seroprevalence. An extremely strong selection seems unlikely. A very small seroprevalence has much the same implication as the results.

Another question relates to the test (microneutralization) and the threshold for positivity. Microneutralization is essentially a demonstration that the subject's serum is able to prevent influenza infection in a cell culture. We think this is a good measure of both whether someone has been infected and whether they are protected against the test virus. It is possible that it is not sufficiently sensitive and that people can be infected but not have a positive microneutralization test. The other question is exactly how "positive" the test has to be. The threshold in this paper was a conventional one of a titer of 1:40. A recent paper looking at seroprevalence against avian virus subtypes in veterinarians used a lower threshold of 1:20. When we posted on it recently there was a discussion in the comments about the appropriateness and validity of this lower titer, but the fact that there was a difference between veterinarians and a control population that was very unlikely a chance finding suggests there might be other criteria for positivity. We have not seen the data from the Thai paper so we don't know if lower titers were detected but considered "negative."

On balance, is the news good or bad? My take is pessimistic. This disease may never become easily transmissible from person to person or animals to people. We are still pretty much in the dark as to what it would take for this to happen. But if it does, this is a really, really nasty bug and even a ten fold reduction in virulence would be result in a major public health disaster.

If a ten fold reduction in virulence is still bad news, then on balance I'd say this is bad news, too.