Epidemiology of Cancer: Spatial-Temporal Aggregation MICHAEL B. SHIMKIN Feis Research Institute, Temple University School of Medicine, Philadelphia, Pennsylvania SUM1\IARY Studies of spatial and temporal distribution patterns of cancer in man and in ani mals have yielded much information and have identified many topics for more in tensive, more rigidly designed epidemiologic investigations. In fact, the wealth of such leads makes the determination of priority and the logistics of the research inn portant considerations. In addition to specific investigations, long-term ecologic observations of cancer in human and animal populations are to be encouraged. There are also unexploited opportunities in experimental cancer epidemiology that deserve biometrically con trolled studies. AIMS AND PURPOSES Functional judgment then has to be made regarding the During my salad days, when I was more daring about priority of the projected investigations, relative not only being bloodied by the spur of the moment, I once wrote, to similar investigations but to a wider spectrum of bio “The wearing of skirts would undoubtedly correlate medical research. The scientists should involve them strongly with the occurrence of mammary cancer, yet has selves in these judgments, or they will be made for them no possible direct significance in the etiology― (64). by their financial sponsors. Here Weinberg's (102) criterion Twenty years later, my plea for clemency is based solely is appealing: “That field has the most scientific merit on the word “direct.―I prefer now to begin with the which contributes most heavily to and illuminates most thought that in this infinite universe, there are no com brightly its neighboring scientific disciplines.― pletely unrelated or independent entities and events. This cursory presentation of spatial and temporal This approach is appropriate to my assignment of dis aggregations of cancer in human and animal populations cussing spatial and temporal aggregation of cancer. “Ag has 3 purposes. The 1st is descriptive : to indicate some gregation― means the entire number, the sum, the total. examples of epidemiologic excursions into neoplastic In the total, all units of an entity must be considered. disease. The 2nd is evaluative : to suggest the type and Spatial amalgamations, or areas of greater density, are design of further studies that seem fruitful. The 3rd is apparent because they contrast with areas of lesser density. to provide material for deliberations regarding scientific Thus, the rarity of units in some areas may be as impor priority and the illuminative capacity of epidemiology. tant as their frequency elsewhere. Temporal changes of The topic is discussed according to the outline indicated increment or decrement attract our attention, but steady in Table 1, which starts with the broadest stage available states may be equally significant. to man until he plants his feet on other planets, focuses In the investigation of the complex patterns that must down to the smallest epidemiologic units, and then be involved in the disease processes we call cancer, the widens the field to ecology and experimental epidemiology. science of epidemiology has a key role. But it cannot The preceding sessions, devoted to viral, physicochemical, stand alone. It must relate to pathology and statistics, and genetic factors, allow this portion of the conference and, in this age, to genetics, chemistry, virology, and to serve as a correlative summary. molecular biology. These relationships must be continual and intimate, and they must flow in both directions. THE PROBLEM OF IDENTITY Such generalizations have meaning, of course, only as Two assumptions regarding cancer appear to be sound. frameworks for functional decisions regarding what kind The 1st is that cancer is a great group of diseases and not of and which specific investigations are to be undertaken. one disease. The 2nd is that cancer occurs in all human In these decisions, as in all science, it is not enough to and animal populations, although the types of cancers outline broad problems. Rather, having specffied the and the frequencies of occurrence differ widely. question, one must determine whether the requisite Since cancer is a group of diseases, a basic requirement methods and materials are at hand and whether the is the correct identification of separate entities that cancer experimental design is likely to yield convincing results. encompasses. Up to now, our foundation for such identi Among the irreplaceable requirements are the interest fication has been descriptive morphologic pathology. and the availability of competent investigators. This has served us well, but the day may not be too distant 1363 Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 1965 American Association for Cancer Research. 1364 Cancer Research Vol. 25, September 1965 TABLE 1 SPATIAL-TEMPORAL STUDIES OF CANCER ApproachMaterialsMethodsGeneral epidemiologyGlobalTotal populations National features Census, vital statistics Case recordsDescriptive and corn demographicRegionalPopulation parative groups Geographic regions Census, vital statistics Case records SurveysstatisticsSpecific epidemiologyClass and groupOccupational, social, other subgroups Disease states retrospec Physiologic featuresCase-control tive studies studiesMicro Prospective Family records ClustersEcologyDefined populations, human and animal laboratory studiesExperimental Defined environmentRelated . epidemiologyIdentified groups trials Purposeful modificationsField when we may be able to place its limitations in the archives velopment of vital statistics, the census, central tumor of history. It is no longer a flight of imagination, for registries, along with continual improvements in diagnosis example, to predict that some forms of leukemia and and in records have placed such data on a more reliable mongolism may be reclassified as reactions to a viral footing. Segi's (79, 80) collection and analysis of cancer invasion of the ovum, rather than as separate, unrelated mortality rates in 24 countries is a useful baseline for these diseases. An analogy is afforded by tabes dorsalis and considerations. some forms of aortic aneurism, unified as 2 reactions to Tables 2A and 2B summarize the data, using the United infection with Spirochaeta pallida. States white population as a reference figure, and the In 1919 James Ewing (25) prophetically wrote: “No highest and lowest rates for comparison. Some striking one would think of confusing lobar pneumonia with features immediately attract attention. The first is that pneumonic plague, although both are examples of acute the total mortality differs within a factor of 2, whereas exudative pneumonitis, but it is quite the rule to identify for a number of sites such as esophagus, stomach, larynx, for statistical studies several equally different forms of prostate, bladder, and breast the differentials between the . cancer.― highest and the lowest rates are 5- to 10-fold. Secondly, The corollary to the requirement of correct identifica the male rates exceed the female rates in all countries tion of separate entities is that entities seemingly distinct except Israel and for all sites common to both sexes, with by one set of criteria may be closely related in regard to the exception of thyroid and possibly liver. Also, the other characteristics. We could miss entirely some im differences by countries and by sites are more pronounced portant relationships of cancer either by the dilution among males than among females. This point is reiterated effect of grouping diverse entities, or by overisolating by considering the male-female ratios, as indicated in such entities from presumably different but really asso Table 3. ciated conditions. Investigations of cancer are impossible It is important to emphasize the gross and limited without appropriate attention to morphologic pathology, nature of these data. The 24 countries represented but also they may be seriously weakened by limiting our include Australia, Austria, Belgium, Canada, Chile, attention to situations that meet its present definitions. Denmark, England and Wales, Finland, France, Germany, Ireland, Israel, Italy, Japan, the Netherlands, New Zea WORLD DISTRIBUTION OF CANCER IN MAN land, Northern Ireland, Norway, Portugal, Scotland, A century ago, August Hirsch (39) accumulated some Sweden, Switzerland, the Union of South Africa, and the data on the frequency of cancer in a few countries of the United States, the last being divided by its white and world. In 1915, Frederick Hoffman's (40) compilation of nonwhite populations. The total population encompassed @ cancer mortality throughout the world revealed some of is approximately 567 million, or of the peoples of the the major patterns of occurrence. Since then, the do earth. Perhaps the recent statistics from the Soviet Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 1965 American Association for Cancer Research. SrnMi@iN—Spatial-temporal Aggregation 1365 TABLE 2A CANCER MORTALITYIN 24 COUNTRIES: MALES 100,000U. AGE-ADJUSTEDRATESPER (white)HighestLowestAll S. neoplasms (189.8) (97.7) Buccal cavity 4.7 France (7.7) Israel (1.2) Esophagus 3.4 France (12.0) Sweden (2.2) Stomach 12.7 Japan (70.6) (U. S.) Rectum 6.4 Denmark (12.6) Portugal (3.6) Liver 4.6 Japan (14.7) Norway (2.8) Pancreas 7.7 U. S. (nonwhite) (9.0) Japan (2.5) Larynx 2.0 France (9.0) Norway (0.4) Lung 29.4 Scotland (61.9) Portugal (7.9) Prostate 13.9 U. S. (nonwhite) (20.6) Japan