Epidemiology of gastrointestinal cancer

Some 99,000 new cases of cancer of the colon are expected next year, an incidence rate higher than that for both cancer of the lung and cancer of the breast. Evidence from geographic pathology suggests that some environmental factors play a strong role in its etiology. Data obtained in the 1959 survey of one million people by the American Cancer Society and followed since, has failed to show correlation with any of the large number of factors listed. It is suggested that the etiology is one of multiple factors. The synergistic effect of exposure to asbestos and cigarette smoking in the production of bronchogenic carcinoma is demonstrated by data on cohorts of insulation workers. There was also a modest increase in the number of deaths from gastrointestinal cancer in asbestos workers, but smoking did not seem to act in synergistic fashion at that site, except perhaps in the esophagus. Deaths from cancer occurred almost entirely after a period of 20 years or more from initial exposure. The death rate from cancer tended to increase with duration of exposure, but a distinct rise over the expected was seen in those who had been exposed less than one year to amosite dust.


Epidemiology of Gastrointestinal Cancer
by Irving J. Selikoff* Some 99,000 new cases of cancer of the colon are expected next year, an incidence rate higher than that for both cancer of the lung and cancer of the breast. Evidence from geographic pathology suggests that some environmental factors play a strong role in its etiology. Data obtained in the 1959 survey of one million people by the American Cancer Society and followed since, has failed to show correlation with any of the large number of factors listed. It is suggested that the etiology is one of multiple factors. The synergistic effect of exposure to asbestos and cigarette smoking in the production of bronchogenic carcinoma is demonstrated by data on cohorts of insulation workers. There was also a modest increase in the number of deaths from gastrointestinal cancer in asbestos workers, but smoking did not seem to act in synergistic fashion at that site, except perhaps in the esophagus. Deaths from cancer occurred almost entirely after a period of 20 years or more from initial exposure. The death rate from cancer tended to increase with duration of exposure, but a distinct rise over the expected was seen in those who had been exposed less than one year to amosite dust.
Epidemiological studies can be used for a variety of purposes. Probably the most important is to identify factors of significance in the etiology of specific disease. Studies with this aim fall into two categories, generally called hypothesis seeking and hypothesis testing. I think it fair to say that, up to this point, most epidemiological studies associated with cancer of the large bowel have been hypothesis seeking. We have not had many good hypotheses to test.
It may be of interest to review the relative importance of cancer of the colon and rectum compared to other tumors in the United States projected for 1974. Judged in terms of incidence, we expect next year (1) some 99,000 new cases of cancer of the colon. This compares with 88,000 projected cases of cancer of the lung and some 90,000 cases of cancer of the breast. In terms of mortality, cancer of the lung is more important, We expect some 75,000 deaths from lung cancer in 1974, some 33,000 deaths from breast cancer, and some 48,000 from cancer of the colon. For comparison I should say that we expect some *Environmental Sciences Laboratory Mount Sinai School of Medicine of the City University of New York, New York 10029. 20,000 cases of cancer of the pancreas, and about the same number of cases of leukemia. The number of deaths from these two causes will be slightly less.
Of equal interest is the fact that, when we compare age-standardized death rates for cancer of the colon in various countries, they are not the same. For example, the elegant studies of Segei (2) showed that Scotland, Denmark, Canada and New Zealand were at the top of the list, with around 20 deaths per 100,000 of agestandardized population. At the bottom of the list were Chile, Finland, Japan, and Israel, with something between 5 and 10 per 100,000 of population. These rates are significantly different.
In the United States, our white population fell somewhere in between and our nonwhite population at a somewhat lower level. The time trends, when looked at from 1950 to the present time, seem to indicate that the highs and the lows are coming together, with England, Scotland and Wales declining and with Chile and Japan going up.
Another interesting set of data has been made available in the studies of Haenzel and others (3), in which they have shown that the cancer rates for those who migrate from country A to December 1974 country B tend to change after migration to those that exist in country B. Moreover, their children tend to have the rates of the country to which their parents migrated. This has been seen, for example, in the stomach cancer rates of Japanese who, after migration to the United States tend to have lower rates; and their children born in the United States have still lower rates (3). This strongly suggests that there is some exposure factor which is more prevalent in Japan than in the United States. The reverse situation obtains for cancer of the colon, and this may be an important factor in its etiology. There may be some dietary factor or some particular component of the diet that is responsible. Haenzel (3) and Berg (4) have both suggested that meat may be a factor. Others have considered that fat and cholesterol may be responsible (5). Burkett (6), on the basis of his studies in Africa, has suggested that intestinal stasis, the bulk of the stool and the consistency of the stools probably play a role.
We have looked at this in another way. In 1959 the American Cancer Society, under the direction of E. Cuyler Hammond, enrolled one million people in a prospective study in the United States. Some 200-odd questions were asked-where you were born, how old you are, whether you drink milk, whether you ever had tuberculosis, whether you eat fried foods, whether you smoke and if you smoke how old you were when you began, and so forth. All of these data were stored on computer tapes, and since that time Hammond has been following these million people with the assistance of some 68,000 volunteers of the American Cancer Society. All deaths are recorded and the data analyzsed to see whether any of the factors recorded in 1959 have a relationship to the death rates since then (7).
For our purposes today it may be of interest to look at the correlations with death rates for cancer of the colon and rectum. I can summarize the results by saying that they have been meager. High and low values for potatoes, candies, pancakes, cooked vegetables, meat and poultry, cheese, fruits, butter, fish, green salads, eggs; such characteristics as relative weight, exercise habits, religious groups, population densities, urban versus rural dwelling, hours of sleep, cigarette smoking, nervous tension, education, etc.; none of them even suggest a hypothesis worthy of further investigation.
To summarize to this point, in our studies and looking at what has been evaluated elsewhere, we have found no single factor to be of significance despite rather strenous attempts. On the basis of these experiences you can come to one of two hypotheses. Either, that there are many as yet unidentified factors, each of which acting alone is capable of producing the disease; or, that there are critical combinations of several different factors, probably including diet, heredity, and environmental agents, which are responsible for the occurrence of the disease. Hammond and I lean toward the second of these hypotheses; that multiple factors are involved.
Multiple factor etiology of cancer has been demonstrated in many experimental studies and experimental models, such as the development of mammary cancer in mice. I would like to review another example of multiple factor etiology which has been demonstrated in man. Table 1 gives the mortality data of a group of insulation workers in the New York Metropolitan area (8). The cohort was composed of 632 asbestos insulation workers who were members of a particular union on January 1, 1943. Hammond and I followed this group until December 31, 1962. Age, year, and sex specific rates indicated that there should have been some 203 deaths among them. They actually had 255. The total mortality was increased, and the reason for the excess mortality lay largely in deaths from intrathoracic cancer, with 6.6 expected but 45 observed. The large majority of the intrathoracic cancers were bronchogenic carcinomas, with a number of pleural mesotheliomas. Interestingly, and this was somewhat unexpected by us, we found that, whereas there should have been 9 or 10 deaths from cancer of the gastrointestinal tract, there actually were 29 -a modest increase. However, this did not seem to be conclusive, owing to the rather small numbers involved and the fact that this was only one cohort.
Deaths Total, 1943Total, -1948Total, -1953Total, -1958Total, -1943Total, -1947Total, 1952Total, 1957Total, 1962Total, 1962  were due to gastrointestinal cancer (Table 4). These data were inadequate for our purposes in one respect: they included only 87 nonsmokers, and that is not a very large number on 3 mail questionnaire survey of this group. The answers to the questionnaires gave us a list of 2066 male asbestos workers who had no history of cigarette smoking and 9590 with a history of ese 87 died of asbestosis cigarette smoking (Table 5). In five years we eir lungs had a great deal have seen only two cases of lung cancer among e of them died of lung the 2066 men with no history of smoking. hand, we would have ex-Asbestos workers who do not smoke cigarettes bronchogenic carcinoma do not often die of lung cancer. On the other history of cigarette smokhand, among the 9590 men with a history of were 24. Asbestos alone smoking, there have been 134 deaths from Lncer, and cigarette smokbronchogenic carcinoma, about five times the uced only 3 deaths from expected number. We found no relation between ,tos plus smoking produccigarette smoking and death from cancer of the combination exhibits the colon and rectum. There was modest increase, iultiple causes. but this was not associated with smoking. There December 1974   was, however, an increased risk of cancer of the esophagus. There also seemed to be some increased risk of death from asbestosis among the smokers (Table 6). When studied carefully, the deaths were seen to be almost all more than 20 years from onset of exposure. There were only a limited number of deaths from bronchogenic carcinoma less than 20 years from onset of exposure. The same was true of cancer of the colon and rectum: the increase began to occur more than 20 years after onset of exposure. When we looked at the deaths from pleural mesothelioma, we saw none in less than 10 years, and only a trivial increase in the interval 10 -14 years from onset of exposure. The increase really began only after 20, 25, 30, 35 years from onset, for both bronchogenic carcinoma and mesothelioma (Table 7).
To determine what duration of exposure is c United States data not available, but these are rare causes of death in the general population.  necessary, we examined the experience of workers at an amosite asbestos products factory. From 1941 to 1945 (the factory closed in 1954) 933 men were employed there. These people had varying durations of exposure; some worked for 1 day, others for the full 13 years. By the end of 1971 we found the same increase in cancers as was seen elsewhere. Where 11 lung cancers were expected, 73 occurred; there was the same modest number of pleural and peritoneal mesotheliomas, the same modest increase of gastrointestinal cancer. The data were then examined for duration of exposure (as distinct from time since onset of exposure), and they fell into three approximately equal groups: worked for less than 3 months; worked 3-11 months; worked for 1 year or more. We found that there was a significant increase in the risk of lung cancer in the group with less than 1 year of exposure, and even in the group with less than 3 month's exposure, although the increase was much greater in the group with more than one year's work ( Table 8). The total mortality in these three groups is very much the same, which, considering that the groups were not identical in age, provides an interesting coherence of results.
Finally, as regards tumors other than bronchogenic carcinoma, mesothelioma and gastrointestinal cancer, in 17,800 asbestos workers with 1092 deaths, we found 15 from genitourinary cancer, of which 11 were of the kidney and 4 of the bladder and urethra. It has been stated that lymphomas and leukemias are increased in asbestos workers. We feel that our experience is still too limited to make any statement on this matter. Of all deaths in this cohort, 16 deaths, somewhat over 1%, were due to leukemia and lymphoma. There were 11 deaths from pancreatic carcinoma. More were so listed on the death certificates, but many of them turned out on examination to be peritoneal mesothelioma, gastrointestinal cancer, or metastatic cancer from the lung. About 15 cases, or about 1.5% of all deaths, were due to cancer of the oropharynx of larynx. It is also interesting to note that 11 deaths, or about 1%, were due to cancer or the brain, of which 6 were astrocytomas and the others glioblastomas.