The epidemiologic features of cancers among children have stimulated abundant descriptive and analytic investigation. The descriptive work has demonstrated consistent differences in the incidence rates of these cancers by anatomic site, age, race, and gender. It is clear that the various forms of cancer during childhood have distinctive patterns of occurrence. To a large extent, the characteristic population distributions of these diseases may represent differences in the underlying etiologic processes. Analytic studies of cancer during childhood have addressed possible genetic and environmental risk factors for these diseases. Single gene mutations, as manifested through autosomal recessive or dominant inheritance, have been shown to predispose to cancers of children. Some of the children affected by these syndromes also exhibit specific chromosomal abnormalities. The demonstration of cancers induced by transplacental exposure to diethylstilbestrol has confirmed the speculation that the prenatal environment may influence subsequent carcinogenesis. Although possible leukemogenic effects of intrauterine diagnostic irradiation remain controversial, the issue may become unimportant clinically as prenatal irradiation is replaced by other diagnostic modalities (194). To date, studies of prenatal ultrasound have provided no evidence of an overall excess of subsequent malignancies. Postnatal exposure to high doses of irradiation is known to produce considerable excesses of leukemias and other cancers. At present, there are insufficient data available to reach a firm conclusion on the possible carcinogenic effects of exposure during childhood to low doses of irradiation, fringe magnetic fields, or chemicals. The evaluation of parental occupations as risk factors for cancer in children will require more detailed characterization of the timing, duration, and intensity of exposures. The many unresolved questions about the epidemiology of cancer in children will provide challenging areas for future investigation. In contemplating such research, care must be taken to avoid the shortcomings of earlier studies. As a general rule, because of sample size limitations, the unconfirmed risk factors for cancer in children are those with threefold excess risk or less. Therefore, it is essential that future investigations have sufficient statistical power to detect even modest associations. That is, studies with small sample sizes, which are likely to lead to inconclusive results on these putative risk factors, should be avoided. There are several possible resources available for conducting large epidemiological studies of cancer in children.(ABSTRACT TRUNCATED AT 400 WORDS)