Monday, February 28, 2005

Another new case in Thai Binh and the question of "clusters"

The 14 year old sister of the 18th case, originally reported as "negative" for H5N1 infection, has now been confirmed positive, according to both AFP and AP reports. Hospital officials say she has a serious lung infection but is not on a respirator. The 21 year old brother remains in critical condition. The siblings are from Thai Binh province, where a 69 year old man died last week. A 36 year old man from Thai Binh was reported as the 19th case this weekend. The family of the siblings did not raise poultry but there have been outbreaks in the area and they were said to have eaten chicken during the New Year holiday meal.

Thus there have been four reported cases from Thai Binh province in the last week. This raises the issue of "clusters" which I would like to discuss briefly. There is no technical definition of cluster in epidemiology. It is a word used to describe cases that seem to be related by some underlying factor, usually geographic area, time of occurrence or (most commonly) both simultaneously (a time-space cluster). A cluster may also be related to a common source (for example a restaurant meal) or contact with each other within specific time windows. The Thai Binh cases represent a cluster if you use "province" as a marker for some geographic factor, as Niman (Recombinomics) does here, or in the case of two of the siblings, family, which could represent either a common source (a holiday meal) or person-to-person transmission.

Familial clusters need not be time, space or time-space clusters. The well-known Thai person-to-person case reported in The New England Journal was none of these. The interpretation of human transmission depended upon the fact that the mother did not become ill after residing in the same area as her daughter or at the same time as her daughter. Thus a common source was eliminated. The timing of her illness, a suitable interval after her daughter's disease onset, also contributed to the interpretation of person-to-person transmission. This is all to say that the use of the word "cluster" in describing the bird flu cases in Vietnam or elsewhere may be confusing. The underlying question is whether the cases are related in some way besides contact with diseased poultry within the incubation period of their disease onset.

There is one final point about clusters worth mentioning. In a series of email communications with Dr. John Field, a physicist trained in quantum optics and atomic physics, he pointed out an analogy between cooperative phenomena that occur in laser action, liquid crystals, and a large variety of other situations in physics that involve "phase transitions." He is not the first to notice this and there is a small group of researchers who are attempting to apply a variety of mathematical techniques from cooperative phenomena and also from network theory, percolation theory and cellular automata to the problem of infectious disease spread. Without going into the technical details (which are quite interesting), Field notes that an early warning sign of an impending pandemic might not be the number of cases that appear or the number of clusters but the increase in the average size of the clusters that appear. This makes biological sense, since the size of the cluster should be related to the ease of transmission of the virus involved in that cluster.

In that regard, Field points out that some of the outbreaks that might be H5N1, like the Baguio meningo-like cases Niman has been discussing at Recombinomics, are potentially of great importance as early warning signs because they are large clusters. Ruling out H5N1 in these instances, especially as we now know that the virus can have an atypical presentation, should be a priority.

There has been no response to my Open Letter to WHO requesting confirmation or disconfirmation of several important questions, including that one.