Disease clusters: chasing after shadows

Clusters of anything – suicides, birth defects, diseases – are a magnet to journalists. They demand explanation, and it is seldom enough to observe that events can be perfectly random yet still occur in clusters.
In the US, the Senate is considering a new law, the Strengthening Protection for Children and Communities from Disease Clusters Act. According to compelling testimony given to the Senate Committee on Environment and Public Works by Dr Richard Belzer yesterday, far from protecting communities or children, the law will cause alarm and confusion.
To coincide with the hearing, the Natural Resources Defense Council and the National Disease Clusters Alliance produced a report detailing 42 disease clusters across the US. It was reported by The Daily Mail under the headline “’Tip of the iceberg’: 42 disease clusters found in 13 US states, but researchers say this is just the beginning”. That’s an accurate summary of what the report says.
The clusters described are not new. Many are old, and have been investigated thoroughly. Some, possibly most, are simply chance events which local activists find impossible to accept as chance. But the report has no such qualms: it is convinced that toxic chemicals are the cause.
It says: “Investigations of disease clusters are complex, expensive, and often inconclusive, partly due to limitations in scientific tools for investigating cause-and-effect in small populations.” No quibble with that, but it then goes on: “Preventing pollution is the best way to avoid creating additional disease clusters.”
So they’re hard to study, but we know what causes them.
A few examples, chosen at random from the 42 clusters detailed in the report, show just how inconclusive investigations can be. 
McFarland, a town in California, suffered 13 childhood cancers between 1984 and 1989, a rate 3-4 times higher than expected. The California Department of Health Sciences investigated, and in a 1991 report said it had found no evidence of environmental contamination that could be plausibly linked to the cancers. It made some recommendations about pesticide use, and remarked that at the time of writing there had been no more childhood cancers in the area.
The same California department investigated six cases of leukaemia and lymphoma in children up to 19 years old in Montecito between 1981 and 1989 – five times greater than expected in a population this size. The investigation “found nothing that would increase cancer risk in the area”, and remarked that the cases involved several different types of cancer “with, probably, different causes”. 
The community around the Rocketdyne Facility in southern California believed it had suffered elevated rates of cancer as a result of hazardous substances released from the plant. This time the report did suggest that chance and bias might affect its estimate of elevated incidence rates, and that smoking and diet might be confounders. Again, no smoking gun was detected. 
Finally, three mothers working for the same tomato farmer in Collier County, Florida, had babies with congenital abnormalities within eight weeks of one another in 2005. The abnormalities were different, and in two cases there were possible links to either family history or a genetic abnormality. No evidence could be adduced that the three mothers had suffered pesticide poisoning.
I have not attempted to read all the reports into the 42 clusters. Some appear to rely on press reports rather than authoritative data for their very existence. But it is clear that it is very seldom possible to be sure that a cluster is more than a random event, or, if it is not, what was responsible for it. Most of the reports, no doubt to allay local fears or to indicate the investigation teams had not been complacent, make recommendations about the better control of the very chemicals they have failed to link to the clusters.
The law before the Senate will not help resolve the issue, according to Dr Belzer, who is the President of Regulatory Checkbook, a nonprofit organisation based in Virginia that aims to improve the scientific and economic information used to make political decisions.
It will make things worse, he said, by defining what is a cluster so sloppily that the Environmental Protection Agency would have “essentially unlimited discretion” to define clusters. The potential causes of disease both “environmental pollutants and toxic substances” but also “any other form of environmental pollution or toxic substance that is a known or potential cause of an adverse health effect”. Dr Belzer remarked: “It is difficult to imagine what is not included in this expansive definition.”
The law mandates the creation of a national database of disease clusters. Unfortunately, the EPA would be obliged to include not only clusters that had been scientifically validated, but those deemed to be clusters by state and federal legislatures, and potential causes also picked by legislators.
“The predictable consequence of a database of this design is public misinformation and unwarranted alarm” he said. “The public would be encouraged to misinterpret legislative definitions and scientific and misconstrue association with causation, something that science consistently teaches against.”
The law, he says, would address the legitimate issue of disease clusters with a combination of science bias, statistical bias and the politicisation of science. It would lead to the massive misallocation of resources, because under the law understanding what causes a disease only matters if it is a chemical.
He showed the senators some diagrams of how 100 cases might be randomly distributed across a geographical matrix of 100 squares. One block had four times the expected number of cases, seven blocks three times the expected number, 19 blocks double the expected number – a total of 27 potential “clusters” out of just 100 squares.
Did the senators listen? I wasn’t there, but I certainly hope so.