Hill’s “Biological Plausibility” Criterion: Integration of Data from Various Disciplines for Epidemiology and Radiation Epidemiology

• Published in: Biology bulletin of the Russian Academy of Sciences 2021-12, Vol.48 (11), p.1991-2014
• Main Authors: Koterov, A. N., Ushenkova, L. N., Biryukov, A. P.
• Format: Article
• Language: English
• Subjects: Bayesian analysis; Biochemistry; Biomedical and Life Sciences; Carcinogenesis; Causality; Cell Biology; Ecology; Epidemiology; Integration; Life Sciences; Meta-analysis; Public health; Radiation; Scientific Search Methodology; Zoology
• ISSN: 1062-3590; 1608-3059
• DOI: 10.1134/S1062359021110054

In this review, various aspects of the causality criterion “Biological Plausibility,” which is sometimes replaced by the criterion “Coherence” (consistency with well-known medical and biological knowledge), are considered. The importance of the criterion for epidemiological evidence of causation, especially for disciplines such as ecology, toxicology, and carcinogenesis, in which there are difficulties not only to perform experiments, but even to observe the effect, is noted. Only statistical approaches in epidemiology are incapable of proving the true causality for association (possibly the effect of chances, confounders, biases, and reverse causation). Without knowledge of the biological mechanism and a plausible model, such a relationship (especially for weak associations) cannot be regarded as confirmation of the true causation of the effect of the factor. The essence of the criterion is the integration of data from various biomedical disciplines, including molecular and experiments on animals and in vitro. There are three (D.L. Weed and S.D. Hursting, from 1998) and four (M. Susser from 1977 and 1986) levels of attaining biological plausibility and coherence with the existing knowledge. The methodologies for integrating data from various disciplines through Bayesian meta-analysis, based on Weight of Evidence (WoE) and teleoanalysis, are considered. The latter is a combination of data from different types of studies to quantify the causal relationship between two such associations, each of which can be proved. However, it is difficult for several reasons, including an ethical plan, to determine the relationship between the causality of the first and the final effect of the second. The approach by teleoanalysis seems doubtful. Despite the need for the “Biological Plausibility” criterion, it, similarly to almost all of Hill’s criteria (except for “Temporality”), is neither necessary nor sufficient for evidence. Examples are presented (including the effects of radiation) that show that, first, “Biological Plausibility” depends upon the biological knowledge of the day and, second, there are real though seemingly implausible associations, and vice versa. This is basis for criticism by some authors (A.R. Feinstein; K.J. Rothman and S. Greenland; B.G. Charlton; K. Goodman and C.V. Phillips) both specifically for the criterion “Biological Plausibility” and for the entire inductive approach based on causal criteria. However, the “Biological Plausibility” criter