Hill's Criterion 'Experiment': The Counterfactual Approach in Non-Radiation and Radiation Sciences1

摘要

This review formalizes, refines, and extends the theoretical and practical aspects of the use of counterfactual concepts in the non-radiation and radiation sciences. The essence of the Hill's causality criterion "Experiment" (Hill, A.B., 1965), which is based on the "on the contrary" approach and which for epidemiology is the "natural experiment," is examined. It consists in observing an effect, the desired cause of which either decreases the intensity or is completely eliminated, regardless of the researcher (as opposed to controlled experiments in biology and medicine). This approach is called "counterfactual" in philosophy ("counterfactual" is "contrary-to-the-fact"). Hill called this methodology "the strongest support for the causality hypothesis." The philosophical meaning and history of the counterfactual concept in the humanitarian disciplines (D. Hume, J. Newman, D. Lewis, et al.) are described. Data on the use of the counterfactual approach in epidemiology are presented (development of a special theory and methodology in the 1980s and 1990s; S. Greenland, G. Maldonado, and K.J. Rothman). The frequent replacement of the term "counterfactual" with surrogate and verbose explanations (Western and Russian authors, some international organizations), such as "Reversibility," "Stop/recovery studies," "Intervention", "Prevention," "Manipulation," etc., was revealed. It is concluded that it is advisable to replace these "artisanal" constructions with the sole unified term "counterfactual." The concepts of "counterfactual ideal" and "counterfactual contrast" are considered. In essence, these are synonyms; however, there are differences in the practice of their use. The "counterfactual ideal" justifies the ideal control group when an exposed individual or a group of people is compared with the same individual or the same group without exposure at the same time. Since this approach is actually impossible, substitution with real comparison groups is used, that is, controls ("contrasts") imitating the "ideal." In this regard, the dualism of the term "counterfactual" can be observed: this is both a synonym for control or a comparison group (in experimental sciences) and a methodological approach based on eliminating the effect and then observing the effect (mainly in observational disciplines). In addition to examples of the counterfactual approach in general epidemiology, a number of relevant facts and scenarios from radiation epidemiology are considered. They are mainly associated with a decrease in carcinogenic effects with a decrease in the level of radiation exposure for one contingent or another (decrease in diagnostic and therapeutic doses, tightening of radiation safety standards, etc.). These are groups of pregnant women irradiated in utero during fluoroscopy in the 1940-1960s, children who exposed to radiotherapy for noncancerous pathologies in the 1920s-1950s, nuclear workers, radiologists, and some other contingents.