Paper on person to person spread in Indonesia Karo cluster

A couple of weeks ago CDC's peer reviewed journal, Emerging Infectious Diseases, published online an ahead-of-print paper by Yang et al., "Detecting Human-to-Human Transmission of Avian Influenza A (H5N1)." The paper has now been published in the journal and predictably, it made news. It's an interesting paper, but we think some people are going beyond what it says. First, the gist according to Reuters:

A mathematical analysis has confirmed that H5N1 avian influenza spread from person to person in Indonesia in April, U.S. researchers reported on Tuesday.

They said they had developed a tool to run quick tests on disease outbreaks to see if dangerous epidemics or pandemics may be developing.

[snip]

Most have been infected directly by birds. But a few clusters of cases have been seen and officials worry most about the possibility that the virus has acquired the ability to pass easily and directly from one person to another. That would spark a pandemic.

Ira Longini and colleagues at the Fred Hutchinson Cancer Research Center in Seattle looked at two clusters -- one in which eight family members died in Sumatra in 2006, and another in Turkey in which eight people were infected and four died.

Experts were almost certain the Sumatra case was human-to-human transmission, but were eager to see more proof. (Reuters)

This account is not widely at variance with the facts but it is incomplete in our view. First, what's the problem here? The problem is that when you see a space-time cluster of cases, i.e., a collection of cases that appear together in a short time in a small area, there are (at least) two explanations. One is that they all got bird flu from one or more common non-human sources, like infected poultry. The other is that at least some of them got flu from another person. It is conventional wisdom at the moment that this virus is transmitted overwhelmingly from bird to human, not human to human. If easy and sustained human to human transmission appears, this will very likely lead to a spreading outbreak and possibly a pandemic. So being able to tell these two situations apart is important. The software tool is designed to help make this distinction.

This is a difficult technical problem because these clusters usually involve small numbers of cases (if they get large in size I don't think we will be worrying about software tools; we will be too busy getting ready for a big outbreak). With small numbers of cases it can become hard to chance from departures from chance. For example, if you flip a coin three times the chance it comes up heads three times in a row is 12.5%, so you might be uncertain if this is a fair coin (equal chance of coming up heads or tails) or an unfair one (say one that had heads on both sides). The software tool described in this paper is designed to tell you when it is unlikely the coin is not fair. It cannot tell you when the coin is fair (only that it might be).

So in this instance the tool says the Indonesian cluster is not likely due to chance, i.e., that the cluster isn't likely the result of a common source (non person to person). In the Turkish cluster the tool wasn't able to make that distinction. Maybe it was a common non-human source, maybe it wasn't. This is the three-heads coin flip example. If we had tossed the coin two or three more times we would have had much more evidence it wasn't a fair coin if it had come up heads consistently. So if the Turkey cluster had been a bit bigger we might have a better idea about whether there was person to person spread there. But you go to press with the evidence you have, not the evidence you'd like, as Donald Rumsfeld might say.

The authors are pretty clear about this in the paper (although the news reports are a bit hazy). But there are other uncertainties. The method requires quite a few assumptions, many of which are probably not very accurate. The authors report they tested how sensitive the tool was to two of these assumptions (the distribution of the incubation and infectious periods) and they reported it was not highly dependent on them. It is still not clear how sensitive the results are to other parameters like cluster size and other factors. If the Turkish cluster were one larger would that have made a difference? I note the p-value is .11, so it wouldn't take much to nudge this into a borderline area for "statistical significance." In general, we are talking about "hidden biases," unmeasured or uncontrolled covariates that affect both contact and infection. As a technical matter, there is the question of interference -- whether contact of one unit affects the outcome of other units, which it clearly does but which seems to violate the assumptions of the randomization test (Rubin's "stable unit treatment value assumption").

As an aside and as a comment on some ways this has been portrayed in the press, this paper doesn't "prove" anything. Proof is not attainable in science, only in mathematics. At best it shows the method is consistent with what we already knew, that there was almost certainly person to person spread in the Indonesian Karo cluster, which we know from other evidence. The importance of the paper is that it shows the statistical method concurs. This is more important for what it says about the method than what the method says about the cluster, although the two lines of evidence reinforce each other.

Does this mean the virus has mutated? Yes and no. The virus is constantly mutating. The real question is what any of the range of existing mutations means for sustained and facile human to human transmission. This paper doesn't help us much in that regard. Given the lower limit of estimated R0 of 1.14, the authors note that the expected chance of further spread for one introduced case with this basic reproductive value is about 12%, and there is much uncertainty in the 1.14 value itself, with a wide confidence interval.

This is a good paper and a valuable tool. Its authors are to be commended. But they should be careful what they say so as not to lead to interpretations like this:

It is no longer a debate whether if a possible bird flu plague is a hoax or not. A mathematical analysis has confirmed that H5N1 avian influenza spread from person to person in Indonesia in April, U.S. researchers reported on Tuesday.

[snip]

This means the virus has mutated to the degree that it now has the potential to race around the world in a matter of weeks, killing millions.

[snip]

When the virus spreads beyond the second generation, the virus will race around the world and millions could die in just a few weeks. (Donald Hughes, ASSIST News service)

Unfortunately, one of the authors is quoted in a way that invites this:

"It went two generations and then just stopped, but it could have gotten out of control," Longini said in a statement.

"The world really may have dodged a bullet with that one, and the next time, we might not be so lucky," he added.

Longini may well believe this is what happened, and he is entitled to say it. It may even be true.

But it's not a consequence of work reported in the paper.

More like this

Revere Thanks for the explanation.

I think it might be worthwhile to put ourselves in the position of the virus for a moment or two.

1) H5N1 has already infected homo sapiens, not once but probably thousands of times since 1997.

2) It has infected several other species, mammalian and non mammallian repeatedly as well...in fact, it has infected more species then any previously known infectious disease.

3)It has a large template of co-infected immunosuppressed homo sapiens to infect.

4) It has a large template of co-infected or other wise stressed ilmmunosuppressed farm animals and poultry to infect.

5) It has a large template due to the epizootic in a wide range of wild birds...the template in fact is larger and the virus is reacting differently in wild bird populations then any previous exotic influenza subtype.

6) It has a large template of exotic and continually changing seasonal influenza's to interact with which co-circulate in many mammalian species including homo sapiens.

7) All attempts at eradication with culls, vaccination and antivirals have failed miserably...in all species...and responsible authorities have not even been testing any of the other animals groups...bats, mice, rats and small shoreline mammals to see if there is a significant tertiary vector that could be controlled.

H5N1 must be pretty confident of its future opportunities and survival at this particular moment in time...as it has exploited every one of our civilizations weaknesses successfully.

Dancing on the head of a pin...of which there seems to be a lot of examples recently, isn't going to change the facts.

Funny thing is that we are not seeing anything from the other side...those who repeatedly stated last year that H5N1 cannot and will never achieve pandemic potential...I would like to see their evidence which has never been presented to my knowledge...by the way, an ethical requirement on their part.

Tom

That's a helluva toxic little nutshell you've presented us. But thanks for it.

Tom-I too have been worried a bit about the transgenic issue. With Tan06's multilingual skills, she has been perusing the Dutch/French/German archives for info on bird kills, pig kills, basically anything other than humans in the newspapers on her side if the pond.

So far she has only turned up a few things but there werent the eco-police running around like we have now. There is strong evidence that the flu in a weaker form was bugging around the trenches in WWI but it was cranked off most of the time to pneumonia, unknown causes. I am also researching to see how many "pneumonia" cases died from the Army archives in France.

Its by no stretch a study. I am looking to see if the CFR's match up with the graph that I first sent Revere and then the ones that he pulled and posted after that. So far they are pretty close and by the same age groups.

Do we have to run a computer model to prove H2H is underway in limited fashion? I dont know. It might provide some damning evidence but the true test will be if it breaks. I personally am expecting the "bad news" to arrive any day because it just doesnt go away and it makes me lean unscientifically towards your argument. Call it a mojo feeling. Used to get it in combat and I learned to abide by it and take cover.

I wont be dissapointed if it doesnt come. The food bank will reap the benefits of my preparedness.

By M. Randolph Kruger (not verified) on 31 Aug 2007 #permalink

Tom-I too have been worried a bit about the transgenic issue. With Tan06's multilingual skills, she has been perusing the Dutch/French/German archives for info on bird kills, pig kills, basically anything other than humans in the newspapers on her side if the pond.

So far she has only turned up a few things but there werent the eco-police running around like we have now. There is strong evidence that the flu in a weaker form was bugging around the trenches in WWI but it was cranked off most of the time to pneumonia, unknown causes. I am also researching to see how many "pneumonia" cases died from the Army archives in France.

Its by no stretch a study. I am looking to see if the CFR's match up with the graph that I first sent Revere and then the ones that he pulled and posted after that. So far they are pretty close and by the same age groups.

Do we have to run a computer model to prove H2H is underway in limited fashion? I dont know. It might provide some damning evidence but the true test will be if it breaks. I personally am expecting the "bad news" to arrive any day because it just doesnt go away and it makes me lean unscientifically towards your argument. Call it a mojo feeling. Used to get it in combat and I learned to abide by it and take cover.

I wont be dissapointed if it doesnt come. The food bank will reap the benefits of my preparedness.

By M. Randolph Kruger (not verified) on 31 Aug 2007 #permalink

Longini is a competent statistician, but talking well beyond his field of expertise. Unfortunately, he works in an environment where so many people (i.e., MDs) have no idea how to analyze data and let surprising amounts of authority go to people whose role is basically technical.

I'm usually a stickler for detail and fact verification, but frankly, I don't care how many liberties the media take with this report. It is finally mainstream "validation" of H2H, it's getting picked up by MSM, and that is something that is long overdue. Hopefully this will get some more people prepping.

anon: I read the study and I don't think it's a reach. Using likelihoods and a permutation test is quite a reasonable thing to do, but there are sharp corners to watch out for, which I mentioned. The key issue, I think, is that they came up with the same conclusion reached by another means, that there was H2H in the Karo cluster. This helps validate the method, which needs validation by means such as this for the reasons given. I don't know who your source is, but this is a specialized area that a virologist or flu maven might not understand that well.

Thanks for the summary, Revere.
The model doesn't seem to account for the genetic suscetibility theory. The exposed who contracted the disease were all blood relatives. The non related individuals did not get sick as well as other related ones. I have seen in other clusters a few who get sick and die and others in the same family unaffected. Perhaps it is exposure related but I feel the genetic link should be explored to a greater extent. If this genetic link is true, the reproductive rate in the genetically susceptible would be much higher than the 1.14 reported here. My feeling is that the number of susceptible people in the population remains very low, (hence the low number of human cases to the millions of potential contacts with infected birds) This in my view is why the disease did not progress rather than any controls implemented.

By Andrew Jeremijenko (not verified) on 03 Sep 2007 #permalink

andrew: Thanks for the comment. I addressed the family susceptibility issue a bit here. Suffice it to say the evidence is scanty at the moment. So much we don't know.