We started a small series on the criteria of AB Hill for causal inference some time ago we talked about the use of these criteria, the original paper by AB Hill that describe these so called criteria, and a discussion on their orginis in comparison to Hume by A Morabia. In this post we will discuss some examples of review articles that not discuss the use of the criteria, but who apply the criteria in a search for an answer on the presumed causal mechanism discussed in that particular article. We discuss some examples in the hope that there is one that is related to your favorite topic of research. Some examples might be outdated, but it is interesting to see how these criteria are used in the course of history.
// breast cancer research
in this 1987 paper Godwin and Boyd appraise the evidence for a causal relationship between dietary fat intake and breast cancer. They conclude “The remaining criteria, with the exception of temporality and epidemiological coherence, were not satisfied. Insufficient evidence existed to conclude a causal association existed between dietary fat and breast cancer risk in humans.” PMID: 3476790
Another example, also by Godwin, but this time together with Steinberg, discusses the evidence for a causal relationship between alcohol and breast cancer. PMID: 1838016
// genital ulcer disease and HIV transmission
Dickerson and colleagues discuss the causal relationship between genetic ulcer disease as a causal factor in the transmission of HIV, of course independent of all other factors. It is interesting to see how the authors do a review of all evidence by first discussion the individual studies, their design and analyses etc, before discussing the Hill criteria. This combination is nice, because the authors conclude “When applying the literature to Hill’s criteria, all nine criteria for causal inference were met, providing additional evidence that genital ulcers are associated with an increased risk for the development of HIV infection.” Additional evidence…interesting phrasing when it comes to providing a definite answer on the causal question: apparently, these authors do not regard the Hill criteria to be sufficient? PMID: 8885077
// antipsychotic drugs as cause of Neuroleptic malignant syndrome
in this article by Gilman, a classic question comes to mind: are the drugs we give to a particular kind of patient also the cause of one of the related disorders, of tis the other disorder just a common disorder found in these patient irrespective of treatment. Gilman writes “The formal criteria for substantiating cause-effect relationships in medical science, established by Hill, are applied to NMS and, for comparison, also to malignant hyperthermia and serotonin toxicity.” Whereas it hurts a little to see him use the term “formal criteria”, Gilman applies a nice negative control concept. Perhaps we should read some more on the concept of negative controls later on this blog? Ah…. we already did! PMID: 20623765