Flipping a coin gives a 0.5 probability of heads. If one flips a coin twice, the probability that both times it will turn up heads is 0.5×0.5—0.25, and if done three and four times it goes to 0.125 and 0.0625, respectively, by chance alone. At about 80 years of age a man has a roughly 50% chance of having had a cancer, and if there are four men aged 80 in the average block, then in about 6.25% of such blocks (one in every 16), all four older men are likely to each have had a cancer, by chance alone.

Although we have an incomplete understanding of the causation of individual cancers, we have an even more imperfect ability to predict whether a given chemical agent is a carcinogen. We can’t expose people to potential danger just to see if a chemical can cause human cancer, so the information has to come from already available observations. Determining whether or not a particular agent actually causes cancer is not a simple process, however, much suspicion has been raised by its characteristics. Each such review must inevitably consider three forms of evidence: studies of exposed humans, studies conducted by exposing members of nonhuman species, and laboratory studies of individual cells or tissues. Each of these methods offers some advantages but may lead to frustrating pitfalls.

Investigators may often suspect that genetic variants enhance the risk of a cancer, but rarely they can use them to predict a truly high cancer risk. When scientists first recognized that a single kind of cancer tended to cluster in the members of a family, methods were devised to zero in on the specific determinants, in the hope that the investigation would result in specific forms of protection. Often the first step was to compare relatives of cases in general to relatives of similar healthy persons, giving an idea of the importance of the family connection.

Three goals justified this book: to provide information about cancer in a form suitable for interested laypersons, to address the concerns of the residents of Los Angeles County about local risk, and to use the abundant information to advance our understanding of cancer causation. Research into the causes of human cancer is a complicated business that ultimately requires the kind of analytic studies of carcinogens and gene variants described in Chapter 5, How a Cancer Can Be Caused by a Carcinogenic Toxin?. To be successful, studies must be capable of ruling out bias, confounding, and chance as alternative explanations for apparent links.

A brief discussion precedes each set of figures. It provides a background summary of known or probable causes, a description of the pattern presented by the figures, and an interpretation of this pattern in light of the background, pointing out those discrepancies that are not easily explained.

In real life, we use math all the time without realizing it, whenever a hard decision is necessary. You leave a piece of cake on the table while you go do an errand, and when you come back it has disappeared. Should you call the police to report an intruder? You put together all the facts you can dredge up: the probability that unfamiliar people are nearby, that the door was safely locked, that the people nearby are really hungry, that someone in your family would take it without telling you, that you might have forgotten that you put it in the refrigerator, that the dog is in the next room, that the dog is tall and clever enough to get onto the table, etc. Then you put all these probabilities together and decide how they compare to the probability that it was a cake thief. If you wind up deciding that the not-intruder possibility is at least 40 times more likely than the intruder possibility, you relax. You have just crudely rejected the intruder hypothesis with a precision about the same as the conventional scientific confidence level.

This chapter and the next one treat the detailed information provided in the figures describing the pattern of each malignancy. For each cancer, six pages of figures, plus a series of maps detail the pattern. Each set contains up to twenty-four different figures, followed by up to five maps. Some figures are empty and maps are absent because the number of cases is insufficient.

The short answer is that a cancer is the result of a single cell that has escaped normal controls and continues to proliferate. There are longer and longer answers. The middle answer is that a cancer is started when several events come together: an unexpected change (mutation or functional change) in a growth gene within a particular cell that is still unspecialized and can readily change course. This change can be delivered by a random error in cellular replication, by an environmental exposure, or uncommonly, by an inherited error. Whatever the initiation, it must be in a DNA unit that regulates growth. The damage is always likely to be repaired by normal functions, and even if not, the odds are that such a damaged cell will just die. If additional gene changes, environmental exposures, and simple bad luck conspire to not only keep the damaged cell alive but also enable it to thrive and grow faster than its sister cells, a cancer may result.

Los Angeles County is the largest county in the United States, with a population of over 10 million persons, much larger than the majority of states. It is also one of the most ethnically and socioeconomically diverse conurbations in the world (those of European ancestry are now in the minority) and is heterogeneous in both of these respects at the neighborhood level. It serves as the economic center for a population two or three times that size, with a wide range of industries, occupations, cultural practices, and levels of education, and a GDP as large as most European countries. Medical care for cancer patients is uniformly available and accessible, and standards of quality for medical records are high.

When groups are compared and show differences in incidence, after assessing the possibility of bias, we ask whether they are dissimilar with respect to one or more characteristics already known to be linked to risk of that cancer. “Confounding” refers to the artifact created by such a familiar or obscure true predictor of cancer occurrence: Confounding is caused by expected disparities, preventing us from recognizing unexpected disparities. Age, race/ethnicity, or social class commonly act as confounders because these factors are surrogate measures of many well understood exposures. Disparities in the analysis of any factor, but in particular of calendar time and neighborhood, commonly reflect such confounding, because these characteristics are often closely linked to seemingly innocuous but influential characteristics such as wealth and education. If an unmeasured “confounding” variable cannot be recognized from surrogate characteristics, further speculation may be needed. If the putative exposure is thought linked to a potentially confounding characteristic that has been recorded, adjustment can be made in the analysis. For example, when groups differ in incidence but the exposure is thought dependent on age, and a subgroup has an unusual number of older residents, further age “adjustment” or “standardization” of the comparison should correct the confounding. As described earlier, this is done routinely in advance when calculating cancer incidence data, but it only can be effective if the recorded age information is known to be accurate.