Three kids on the same block were diagnosed with leukemia last year. That couldn’t happen just by chance, could it? There MUST be something in the environment that caused it (power lines, the chemical plant down the street, asbestos in their school, iPods, Twinkies?). Quick, let’s measure everything we can think of and compare exposures to other blocks and find an explanation.
That may be the common reaction, and it may seem plausible to the general public, but it’s not good science.
I have just read a book that does a great job of elucidating the pitfalls of epidemiologic studies, the problematic interface between science and emotion-laden public concerns, and the way environmental hazards have been hyped far beyond the evidence. Hyping Health Risks: Environmental Hazards in Daily Life and the Science of Epidemiology by Geoffrey C. Kabat.
He covers the uses, strengths and limitations of epidemiology, discusses the pros and cons of different study designs, and explains how to judge whether an association is causal.
The Ecological Fallacy and Determination of Causality
Ecological studies based on groups rather than individuals can be misleading. Early researchers found a strong correlation between per capita sales of cigarettes and lung cancer; but there was also a strong correlation between sales of silk stockings and lung cancer. They had to look for other types of evidence before they could conclude there was a causal relationship. A study showed a strong correlation between dietary fat consumption and age-adjusted breast-cancer mortality in 41 countries, but that could be due to confounding factors like economic and lifestyle differences. The “ecological fallacy” is
the unwarranted assumption that, just because a given factor is correlated with a disease based on aggregate data, this suggests (1) that the same correlation will hold at the individual level or (2) that the association is indicative of a causal relationship.
Statistical significance is often misunderstood.
Not every finding that is statistically significant is biologically significant, but too often the fact that a given result is statistically significant is used to imply precisely this. More relevant to gauging the importance of a finding is the consistency with existing evidence from other sources and the impact on a population basis.
The following criteria can be used to judge causality:
- The magnitude of the association
- Consistency (same effect shown in different studies carried out in different populations).
- Temporal relationship (cause must precede effect)
- Coherence of explanation (also known as biological plausibility).
- Dose-response relationship.
The relationship between smoking and lung cancer meets all those criteria. Kabat describes four cases in detail where those criteria were not met and where public concern and activism led to bad science and overblown fears.
Does the environment cause breast cancer?
There was a public perception that there was an epidemic of breast cancer in certain areas of the country such as Long Island, and there was a conviction that some form of environmental pollution must play a role. In reality, there was no epidemic, and there is little evidence to support a role for the environment as a cause of breast cancer. Activists got a law passed mandating scientific projects, which were poorly designed and focused on specific pollutants like DDT, and got negative results, disappointing everyone.
Clusters of illness are inevitable. If you spill rice on a grid, there will be squares with lots of grains of rice and squares with none. The challenge is to determine whether a cluster of illness is due to anything more than chance. The public doesn’t understand this, and they demand an explanation for every apparent cluster. In one of Kabat’s examples, a community that had the highest rate of breast cancer one year had the lowest rate the following year.
To put the concern about breast cancer into perspective, any environmental effect is bound to be small, while the rate of lung cancer is 1.5 times that of breast cancer and it is 90% preventable by smoking cessation.
Do Electromagnetic Fields Cause Cancer?
EMF concerns are demolished by an understanding of the physics involved. Any effects of EMF are well below the level of thermal noise. One physicist concluded that
there are good reasons to believe that weak ELF [extremely low frequency] fields can have no significant biological effect at the cell level – and no strong reason to believe otherwise…. He likened concern over weak EMF from power lines to the fear that leaves falling from trees could fracture a person’s skull.
The National Institute for Environmental Health Sciences reviewed the evidence for an association of EMF with cancer, acknowledged that it was all either negative or weak and inconsistent, yet they made a politically cautious decision to classify EMF as a possible carcinogen – a decision that cannot be scientifically justified.
Does residential radon exposure cause lung cancer?
Is there radon in your house? How much? In which rooms? How long have you lived there? How much time did you spend in which rooms? How much radon was in your previous residence? Residential radon studies are hampered by the near impossibility of quantifying an individual’s exposure over time. And most of the studies are flawed by the failure to correct for smoking. The risk of radon exposure to smokers is high; the risk to non-smokers is questionable. The way to reduce lung cancer deaths from radon is not to test everyone’s home and try to lower radon exposure – it is to get people to stop smoking. A 20% increase in risk of lung cancer from radon exposure (including smokers and non-smokers) must be put into perspective with the 2000% increase in risk from smoking.
Does Second-Hand Smoke Cause Lung Cancer and Heart Disease?
This is a little different because we are reasonably sure there is a risk and we know the mechanism. The problem is that the risk is small and difficult to measure accurately, and Kabat argues that it is practically impossible to quantify an individual’s exposure to second-hand smoke over long periods of time. Studies have compared the spouses of smokers to the spouses of non-smokers, but some smokers bombard their spouses with clouds of smoke while others considerately smoke out on the porch, and we have no way of knowing whether the non-smoking spouses were exposed to smoke at work or elsewhere. Other studies have measured cotinine (a nicotine derivative) in the urine of those exposed to second-hand smoke, but the test only represents one point in time.
The association of heart disease with passive smoking is much weaker than the association with lung cancer, and anti-smoking activists lost all credibility when they extrapolated from poor data and tried to claim that passive smoking caused 50,000 cardiac deaths a year in the US.
99.99% of the Pesticides in Our Diet are Natural and Unavoidable
There is a strong ideological environmental movement that is predisposed to find dangers from artificial chemicals in our environment. They seldom put those fears into perspective with natural and unavoidable chemicals. Bruce Ames, originator of the Ames test, a biological assay to identify carcinogens, co-wrote a paper showing that 99.99% of pesticides in the American diet are chemicals that plants produce to defend themselves. About half of the natural pesticides that have been tested are carcinogenic in rodents. They concluded that “natural and synthetic chemicals are equally likely to be positive in animal cancer tests. We also conclude that at the low doses of most human exposures the comparative hazards of synthetic pesticide residues are insignificant.”
Reading this book will give you a better understanding of what epidemiology can and can’t do, and insight into how the rational scientific process can be perverted by the press, politicians, and grass-roots activists. One unstated take-home lesson from this book is that we should worry less about potential small environmental dangers and do something about the very large and preventable environmental danger of smoking.