A methodologic assessment of radiation epidemiology studies.

Epidemiologic studies of the late effects of ionizing radiation have used the entire spectrum of situations in which man has been exposed. These studies have provided insights into the dependence of human effects upon not only dose to target tissues but also other dimensions of exposure, host characteristics, and time following exposure. Over the past three decades studies have progressed from the mere identification of effects to their measurement. Because investigators of human effects have no control over the exposure situation, validity must be sought in the consistency of findings among independent studies and with accepted biologic principles. Because exposure may be confounded with factors that are hidden from view, bias may enter into any study of human exposure. Avoidance of bias and attainment of sufficient power to detect relationships that are real--these are the methodologic challenges. Many methodologic issues, e.g. those associated with the definition and measurement of specific end points, or with the selection of appropriate controls, permeate epidemiologic work in all fields. Others, especially those concerned with the measurement of exposure, the patterning of events in time after exposure, and the prediction of events beyond the scope of existing observations--these give radiation epidemiology its distinctive character.