An editorial in this issue explains that the degree of biological interaction between risk factors is measured as the deviation from additivity by the corresponding disease rates and not for example as deviation from multiplicativity. It is the purpose of this article to describe how a logistic regression model, or a Cox regression model, can be defined in order to produce the output that is needed for assessment of biological interaction. We will also demonstrate how common software can be programmed to deliver this output. Finally, we show how this output can be used as input in an Excel sheet that is set up to calculate the measures of biological interaction to be used for the assessment.
Ahlbom A, Alfredsson L. Interaction: Word with two meanings creates confusion. Eur J Epidemiol, in pressGoogle Scholar
Rothman, KJ 2002Epidemiology. An IntroductionOxford University PressNew YorkGoogle Scholar
Lundberg, M, Fredlund, P, Hallqvist, J, Diderichsen, F 1996A SAS program calculating three measures of interaction with confidence intervalsEpidemiology7655656PubMedGoogle Scholar
Hosmer, DW, Lemeshow, S 1992Confidence interval estimation of interactionEpidemiology3452456PubMedGoogle Scholar
Hallqvist, J, Ahlbom, A, Diderichsen, F, Reuterwall, C 1996How to evaluate interaction between causes: A review of practices in cardiovascular epidemiologyJ Intern Med239377382CrossRefPubMedGoogle Scholar
Assmann, SF, Hosmer, DW, Lemeshow, S, Mundt, KA 1996Confidence intervals for measures of interactionEpidemiology7286290PubMedGoogle Scholar
Padyukov, L, Silva, C, Stolt, P, Alfredsson, L, Klareskog, L 2004A gene–environment interaction between smoking and shared epitope genes in HLA-DR provides a high risk of seropositive rheumatoid arthritisArthritis Rheum5030853092CrossRefPubMedGoogle Scholar