Greenland S, Poole C
Division of Epidemiology, UCLA School of Public Health 90024.
Scand J Work Environ Health. 1988 Apr;14(2):125-9. doi: 10.5271/sjweh.1945.
In two of his publications [Causal and preventive interdependence: Elementary principles. Scand J Work Environ Health 8 (1982) 159-168 and Theoretical Epidemiology, John Wiley & Sons, New York, NY 1985], Miettinen put forth basic definitions of causal and preventive interdependence of effects involving binary exposure indicators and outcomes. This paper shows that the identification of interdependence using Miettinen's definitions varies with the choice of the reference categories for the exposures. In particular, Miettinen's concepts of synergism and antagonism are not invariant under exposure recoding. It is also shown that, when both exposures affect risk in some individuals, the effects will appear interdependent under some choice of referent. In the deterministic case, invariant properties of joint effects may be identified through the formation of equivalence classes of response types. In the stochastic case, invariant properties may be identified through the averaging of individual hazards, rather than risks. In both cases, additivity of risk or rate differences emerges as an elementary criterion for the independence of effects.
在他的两篇出版物中[因果与预防的相互依存关系：基本原理。《斯堪的纳维亚工作环境与健康杂志》8（1982）159 - 168以及《理论流行病学》，约翰·威利父子出版公司，纽约，纽约州，1985年]，米耶蒂宁提出了涉及二元暴露指标和结局的效应的因果与预防相互依存关系的基本定义。本文表明，使用米耶蒂宁的定义来识别相互依存关系会因暴露参考类别的选择而有所不同。特别是，米耶蒂宁的协同作用和拮抗作用概念在暴露重新编码下并非不变。还表明，当两种暴露都影响某些个体的风险时，在某些参考选择下，效应会显得相互依存。在确定性情况下，联合效应的不变性质可通过形成反应类型的等价类来识别。在随机情况下，不变性质可通过个体风险率的平均来识别，而不是风险。在这两种情况下，风险或率差的可加性作为效应独立性的基本标准出现。
