STATS.org - The Challenge of Identifying Cancer Clusters

The Challenge of Identifying Cancer Clusters
August 01, 2002
VitalSTATS
Anyone can find a cancer cluster if they want to

Over the last 20 years, reports of cancer clusters have soared. Movies like "A Civil Action" and "Erin Brokovich" have laid the template of a murder mystery for news coverage of cancer clusters. Such stories involve sympathetic victims, grisly murder weapons (cancers), and the hunt for a villain (e.g., industrial solvents).

For instance, the Richmond Times-Dispatch (Jul. 22) featured a series of articles under the title of "Toxic Burial Ground." It profiled residents like Wade Sprouse, who "may never know for sure what made his wife sick" with osteoporosis, but "suspects it could be the water she drank for many of the 37 years that they lived ... in Rayon Park," and Sheila Wilmoth, who "believes the water she played in as a child is to blame for her" colon cancer and her mother's lung and liver cancers. The not-so-subtly hinted cause of such cancer clusters in Rayon Park? The Defense Supply Center, a Superfund site. According to the Times-Dispatch, "at least one person in each of Rayon Park's 75 homes has developed cancer."

But what can be learned from these articles? Are cases of cancer clusters as straightforward as the plots of dime-store mystery novels?

Cancer is an unfortunately common ailment. The National Cancer Institute estimates there are more than 8 million cancer cases in the US. So finding several cases within the same neighborhood or even the same family is not unusual. For one thing, we see more cancer today than in generations past because we live longer, increasing our likelihood of contracting the disease. Nonetheless, we are still not certain what causes most cancers. Researchers continue to grapple with a wide range of possible causes, including heredity, diet, lifestyle and environmental exposure.

So where do cancer clusters fit into this puzzle? They represent the occurrence of a greater than expected number of cases of a particular cancer within a group of people, a geographic area, or a period of time. But clusters tend to exist more in our minds than in reality.

This is sometimes referred to as the 'Texan sharp-shooter effect.' The Texan sharp-shooter shoots at the side of a barn and then walks up and artfully paints a bull's-eye around his shots large enough that all of them fall inside it. People have a natural tendency to encircle groups of events into clusters. While drawing a circle around a block where three kids have cancer may show a cluster, drawing the circle around the town as a whole shows none.

According to science writer Atul Gawande (The New Yorker, Feb. 8 1999), humans are programmed to see patterns which do not necessarily exist, making it hard for us to comprehend or accept random chance. Flipping a coin a hundred times does not guarantee an equal distribution of heads and tails, but most people become suspicious if the results deviate greatly from that pattern.

For example, three male co-workers in Albuquerque contracted breast cancer, a disease that is very rare in men. ABC News' "Good Morning America" (Jun. 4) observed that "you're more likely to win the power ball lottery two days in a row than find two men... in the same workplace with breast cancer, statistically." Unless winners are disqualified from future games, however, odds on winning the power ball lottery should be independent from game to game, person to person.

Diseases works the same way. We cannot expect cancer cases to be equally distributed across all demographics and locations, anymore than we can expect them to all fall in, say, Los Angeles. Daniel Kahneman and Amos Tversky, back in 1971, dubbed this problem the Belief in the Law of Small Numbers - the assumption "that the pattern of a large population will be replicated in all its subsets."

A lot of cancer clusters are just random noise. Many of them do not encompass enough cases to allow useful conclusions. Even when they do, statistical analysis often finds no significant increase in disease incidence.

Proven cause and effect in cancer clusters is uncommon. Reasonably conclusive proof can usually only be found in cases where chemical exposure was extremely high, the disease is extremely rare, the population does not move much and is of an age not usually affected by the cancer. In the hundreds of investigations of U.S. residential cancer clusters, no conclusive underlying environmental cause has been unearthed.

This brings us back to the Times-Dispatch. Data collection in Rayon Park has been conducted primarily by local residents. No one has given much thought to the locals' lifestyles, which could more easily explain some of the cancer cases (e.g., lung and throat cancer cases might be linked to heavy smoking). And what about the age range of the cases? Many of the paper's anecdotes focus on older sufferers, who are much more likely to contract cancer than younger people.

The Times-Dispatch provided a compelling story, with emotional anecdotes of innocent people suffering, but the plural of anecdote is not data. A scientific study of the area is due this month. Until more concrete information is available, the Times-Dispatch has shown little more than its skills as a Virginian sharp-shooter.


	home
	The Challenge of Identifying Cancer Clusters August 01, 2002 VitalSTATS Anyone can find a cancer cluster if they want to Over the last 20 years, reports of cancer clusters have soared. Movies like "A Civil Action" and "Erin Brokovich" have laid the template of a murder mystery for news coverage of cancer clusters. Such stories involve sympathetic victims, grisly murder weapons (cancers), and the hunt for a villain (e.g., industrial solvents). For instance, the Richmond Times-Dispatch (Jul. 22) featured a series of articles under the title of "Toxic Burial Ground." It profiled residents like Wade Sprouse, who "may never know for sure what made his wife sick" with osteoporosis, but "suspects it could be the water she drank for many of the 37 years that they lived ... in Rayon Park," and Sheila Wilmoth, who "believes the water she played in as a child is to blame for her" colon cancer and her mother's lung and liver cancers. The not-so-subtly hinted cause of such cancer clusters in Rayon Park? The Defense Supply Center, a Superfund site. According to the Times-Dispatch, "at least one person in each of Rayon Park's 75 homes has developed cancer." But what can be learned from these articles? Are cases of cancer clusters as straightforward as the plots of dime-store mystery novels? Cancer is an unfortunately common ailment. The National Cancer Institute estimates there are more than 8 million cancer cases in the US. So finding several cases within the same neighborhood or even the same family is not unusual. For one thing, we see more cancer today than in generations past because we live longer, increasing our likelihood of contracting the disease. Nonetheless, we are still not certain what causes most cancers. Researchers continue to grapple with a wide range of possible causes, including heredity, diet, lifestyle and environmental exposure. So where do cancer clusters fit into this puzzle? They represent the occurrence of a greater than expected number of cases of a particular cancer within a group of people, a geographic area, or a period of time. But clusters tend to exist more in our minds than in reality. This is sometimes referred to as the 'Texan sharp-shooter effect.' The Texan sharp-shooter shoots at the side of a barn and then walks up and artfully paints a bull's-eye around his shots large enough that all of them fall inside it. People have a natural tendency to encircle groups of events into clusters. While drawing a circle around a block where three kids have cancer may show a cluster, drawing the circle around the town as a whole shows none. According to science writer Atul Gawande (The New Yorker, Feb. 8 1999), humans are programmed to see patterns which do not necessarily exist, making it hard for us to comprehend or accept random chance. Flipping a coin a hundred times does not guarantee an equal distribution of heads and tails, but most people become suspicious if the results deviate greatly from that pattern. For example, three male co-workers in Albuquerque contracted breast cancer, a disease that is very rare in men. ABC News' "Good Morning America" (Jun. 4) observed that "you're more likely to win the power ball lottery two days in a row than find two men... in the same workplace with breast cancer, statistically." Unless winners are disqualified from future games, however, odds on winning the power ball lottery should be independent from game to game, person to person. Diseases works the same way. We cannot expect cancer cases to be equally distributed across all demographics and locations, anymore than we can expect them to all fall in, say, Los Angeles. Daniel Kahneman and Amos Tversky, back in 1971, dubbed this problem the Belief in the Law of Small Numbers - the assumption "that the pattern of a large population will be replicated in all its subsets." A lot of cancer clusters are just random noise. Many of them do not encompass enough cases to allow useful conclusions. Even when they do, statistical analysis often finds no significant increase in disease incidence. Proven cause and effect in cancer clusters is uncommon. Reasonably conclusive proof can usually only be found in cases where chemical exposure was extremely high, the disease is extremely rare, the population does not move much and is of an age not usually affected by the cancer. In the hundreds of investigations of U.S. residential cancer clusters, no conclusive underlying environmental cause has been unearthed. This brings us back to the Times-Dispatch. Data collection in Rayon Park has been conducted primarily by local residents. No one has given much thought to the locals' lifestyles, which could more easily explain some of the cancer cases (e.g., lung and throat cancer cases might be linked to heavy smoking). And what about the age range of the cases? Many of the paper's anecdotes focus on older sufferers, who are much more likely to contract cancer than younger people. The Times-Dispatch provided a compelling story, with emotional anecdotes of innocent people suffering, but the plural of anecdote is not data. A scientific study of the area is due this month. Until more concrete information is available, the Times-Dispatch has shown little more than its skills as a Virginian sharp-shooter.