Environmental Exposures and Cancer Risks

One of the major justifications for environmental regulation is concern about the carcinogenic potential of air and water pollution, pesticide residues on foodstuffs, and other so-called environmental exposures. In the 1970s, it was common to hear claims that the environment was responsible for 80 to 90 percent of all cancers.

The term environment, however, has two quite different meanings. As used by epidemiologists—scientists who study the occurrence and causes of human disease—it refers to many of the things that humans come into contact with: food, drink, and anything smoked; drugs and medicines; and air, water, and soil. In this context, the environment means things outside the body as distinct from a person's genetic makeup. Thus, when epidemiologists say that 80 to 90 percent of cancer is associated with the environment, they mean that factors other than a person's own genetics are almost always involved in cancer causation, not that some particular exposures contributed to the onset of cancer.

In contrast, the term environment is more commonly taken to mean air, water, and soil in the sense conveyed by the focuses of the Environmental Protection Agency (EPA). Many people hear "environment" used in that sense and interpret the claim concerning cancer causation to mean that 80 to 90 percent of cancer is caused by exposures to substances in air, water, and soil.

In 1981, epidemiologist Richard Doll and statistician Richard Peto, both at Oxford University, clarified the two uses of the term with a comprehensive study that analyzed U.S. national cancer mortality records from 1933 through 1978. They attributed only about 2 percent of total cancer mortality to environmental exposures (or pollution, as they referred to it) and another 3 percent to geophysical factors, including sunlight and other kinds of natural radiation.

Doll and Peto divided all exposures to carcinogens into twelve distinct categories. Two of these, pollution and geophysical factors, encompass the exposures that are subject to regulation by the EPA and that are discussed in Unfinished Business. Two more of Doll and Peto's exposure categories, occupation and industrial products—the latter is called consumer products in the United States—are also discussed in Unfinished Business.

Because of differences in the quality and quantity of data available, Doll and Peto relied upon somewhat different methods to estimate the cancer mortality associated with each of their exposure categories. Under the pollution category, for instance, they referred to others' work that compared the concentrations of known carcinogenic substances in urban air to concentrations of the same materials that were associated with cancer in workplace atmospheres. Assuming that a direct relationship exists between concentrations and cancer risk, Doll and Peto estimated that urban air pollution might cause 1 percent of cancer mortality, or 4,000 deaths annually.

In calculating risk from geophysical factors, Doll and Peto considered two kinds of radiation—ultraviolet (UV), the source of which is sunlight, and ionizing radiation, which is emitted by various natural sources such as cosmic rays and radon. Distinguishing between current and projected cancer mortality from UV radiation, they estimated that 1 to 2 percent of cancer mortality (4,000 and 8,000 cancer deaths, respectively) was associated with exposures to sunlight in 1981 and that fewer than 120 of those deaths were related to deterioration of the ozone layer. They also estimated that future depletion of the ozone layer, which filters out UV radiation, might increase the total of UV-related skin cancer deaths by 20 percent.

Doll and Peto estimated that 1.4 percent of cancer mortality was associated with ionizing radiation. They arrived at this estimate by calculating the background radiation dose for the entire U.S. population and assuming a direct proportionality between the carcinogenic potency of that dose and the higher doses of radiation found by researchers in people exposed in medical settings or at Nagasaki and Hiroshima during World War II.

In estimating risks to exposures under the occupation category, Doll and Peto studied the epidemiologic literature to determine what proportion of various cancers could be associated with workplace exposures. Based on their review, Doll and Peto attributed 17,000 annual cancer deaths to occupational exposures, 11,000 of which were lung cancer deaths.

In comparing Doll and Peto's estimates with those of the Environmental Protection Agency, it should be noted that most, but not all, of the problem areas considered in the EPA report fit neatly into one or another of the Doll-Peto categories. Some EPA categories such as drinking water, which includes both chemical and radiation risks, must be subdivided in order to fit within Doll and Peto's categories of pollution and geophysical factors.

An additional adjustment is necessary because Doll and Peto used mortality data as the basis for their calculations, whereas the EPA generally estimated cancer incidence rather than mortality. For the purpose of comparing estimates, cancer incidence data can be converted into cancer mortality data by assuming that between half and all the people who develop cancer will die of that disease. The assumption that half of the people who develop cancer will survive the disease is based on the National Cancer Institute's claim that the overall five-year survival rate for cancer patients is 48 percent and the fact that five-year survival is considered a "cure."

The estimated number of cancer deaths derived from the EPA's analysis and grouped into Doll and Peto's four categories can be divided by 485,000—the annual U.S. death toll from cancer—and then multiplied by 100 to calculate percentages of cancer mortality. Agreement between the two sets of estimates is good.

Doll and Peto associated pollution and geophysical factors with 2 and 3 percent, respectively, of total cancers. Similar risk estimates are derived when the EPA's estimates of individual risks within these categories are added together, resulting in an association of 1 to 3 percent of total cancers with pollution and 3 to 6 percent with geophysical factors. The EPA's estimate that workplace exposures are associated with 1 to 4 percent of total cancers agrees with Doll and Peto's estimate for those exposures. Both the Doll-Peto and EPA estimates for cancer risks from consumer products are less than 1 percent (see table 1). In this case, neither estimate is based on extensive analysis.

Implications for pollution control

Doll and Peto's estimates of preventable causes of cancer in the United States do not support the idea that control of environmental pollution would have a major impact on human cancer rates. Instead of pointing to pollution as a major cause of cancer, their analysis of national cancer mortality data and trends finds that smoking is a greater factor in cancer causation, associated with 30 percent of all cancers. Their analysis also draws renewed attention to diet, which they estimate is associated with 35 percent of all cancers.

Reliance on data from human studies has been criticized for a number of reasons. For one thing, many small increases in cancer are undetectable by epidemiologic methods. Moreover, because many years may elapse between exposures and cancer manifestation, analysis of epidemiologic data cannot reveal the effects of recent exposures but only those of exposures years ago. Risk assessment methods, which depend on the results of animal tests for the prediction of human risks, circumvent these purported deficiencies of epidemiology.

The agreement between the Doll-Peto and EPA estimates of cancer risks associated with pollution and geophysical factors can be regarded as surprising because the basis of Doll and Peto's analysis is actual cancer deaths, in contrast to the EPA's toxicology-based risk assessment methods, which generate upper confidence limits on risk. The use of these limits makes it unlikely that risks are underestimated, though they are likely to be overestimated. While the quantitative relationship between upper confidence limits and best estimates of risk would vary from chemical to chemical, we would expect that the accumulation of data from many chemicals would exaggerate the cancer risk. That expected bias might, of course, contribute to the EPA's estimates being higher than Doll and Peto's. Another factor also mitigates the effect of using the upper confidence level for toxicology-based risk estimates: the EPA uses best estimates for risks that are based on data from human studies; these include risks from indoor radon and sunlight.

While both analyses lead to the conclusion that little aggregate impact on overall cancer rates can be expected from attention to environmental exposures, this does not mean that pollution should go unchecked. Two to three percent of cancer is a small proportion of the total cancer risk, but it represents 9,000 to 13,000 deaths annually. Both the percentage of the total cancer risk and the estimated number of cases should be considered in making decisions about where to invest resources to combat cancer.

Pollution is involved in other significant health impairments beside cancer. Eye irritation from pollution, for example, is quite painful and can cause damage if prolonged. A far more serious problem is exposure to lead, which poses significant risk of injury to the nervous system. And quite apart from health effects, the ecological and aesthetic damages that accompany pollution must be considered.

More important, even with regard to cancer, cost-effective controls that use current technology may be available. For instance, the modification of homes on radon-contaminated soil could reduce the most significant current cancer risk in the EPA's catalog. The phasing in of chlorofluorocarbon substitutes, scheduled for the next dozen years, may help to check the expected increase in deaths from skin cancer resulting from continued depletion of stratospheric ozone.