The science of cancer clusters is no simple matter
Disease surveillance systems should include public consultation and education
Recent concerns over a cancer cluster at Hamilton’s Cathedral High School trigger feelings of empathy, concern and even fear, and raise questions about the safety of our environment.
Officials have concluded that cancer cases at Cathedral can’t be attributed to air quality in the school, but understandably, the fear of cancer remains.
While it’s inappropriate to ignore the concerns of people affected by or worried about cancer clusters in their community, it is often challenging for public officials to adopt a rigorous and transparent process for investigating these apparent clusters that also satisfies all community stakeholders.
As a thought experiment, imagine a community of a 1,000 people, in which a person has recently been diagnosed with brain cancer, and where people have had fears about an industrial facility for decades.
As news of the diagnosis spreads, other members of the community disclose their own encounters with cancer — another brain tumour, a few cases of breast and prostate cancer, and a rare form of leukemia.
All told, there are 12 cases of cancer in the community that year. Compared to the national cancer rate (around 5 in 1,000 per year) this seems very high.
Together, the cancer and the environmental concerns create local buzz: news stories, water-cooler conversations and pressure on officials to do something about it.
Typically we might want to determine whether this cluster is a statistical anomaly or a genuine concern by determining the “statistical significance” of the 12 cases of cancer. However, this proves to be more of a challenge than you might think.
Consider the casino game of roulette. Based on how roulette wheels are designed, we know with certainty that in the long run players will lose money, but every once in a while a person gets lucky and wins 10 or more favourable spins in a row.
These players aren’t winning because they are prognosticating geniuses. With tens of thousands of people playing roulette every year we should expect a few to have amazing runs of good luck, even if the longrun probability is losing.
In fact, if we know how many roulette games are played every year, we can actually determine approximately how many really lucky roulette players to expect.
Unfortunately, we have no equivalent reference for cancer clusters. Canada is a country of about 35 million, within which there is a seemingly infinite number of possible “communities” of people, depending on how you define them. Some communities are city neighbourhoods, some are groups of small villages, and some are apartment complexes. Some are people working within an industry or teachers in a school.
For any single community of 1,000 people, 12 cases would seem like an anomaly that can’t be explained by randomness. But with many, many thousands (or even millions) of possible communities in the country, it is also a statistical certainty that some of them would have at least 12 cases of cancer in a given year, even if the true risk of cancer were the same across the country.
In response, then, it might be tempting to dismiss all but the most striking cancer clusters as statistical anomalies, but this does lit- tle to comfort members of communities that directly or indirectly experience the effects of cancer. Moreover, it could contribute to an already worrisome tension between concerned citizens and seemingly unsympathetic technocrats.
Fortunately, there are things that can be done to improve the situation. Provincial governments have been working for years on developing disease surveillance systems, including some that are co-ordinated at the federal level, to routinely monitor infectious and non-infectious diseases.
An important dimension of this process should include public consultation to define communities worth monitoring ahead of time; that is, before the suspected clusters of disease emerge.
This would provide information equivalent to the number of “roulette players,” and serve as an important reference point for distinguishing between real clusters and expected statistical anomalies.
Once communities are defined, officials can routinely monitor — and publicly disclose — the status of target diseases in these communities.
Consultation and transparency could help build public trust, and the routine monitoring of a known set of communities of concern would improve the scientific rigour necessary to determine whether or not a cluster is a statistical accident or a serious health concern.