Practice of Epidemiology Asymptotically Unbiased Estimation of Exposure Odds Ratios in Complete Records Logistic Regression

Missing data are a commonly occurring threat to the validity and efficiency of epidemiologic studies. Perhaps the most common approach to handlingmissing data is to simply drop those records with 1 or moremissing values, in so-called “complete records ” or “complete case ” analysis. In this paper, we bring togetherearlier-derived yet perhaps now somewhat neglected results which show that a logistic regression complete records analysis can provide asymptot-ically unbiased estimates of the association of an exposure of interest with an outcome, adjusted for a number of con-founders, under a surprisingly wide range of missing-data assumptions. We give detailed guidance describing how the observed data can be used to judge the plausibility of these assumptions. The results mean that in large epide-miologic studies which are affected by missing data and analyzed by logistic regression, exposure associations may be estimated without bias in a number of settings where researchers might otherwise assume that bias would occur. complete case analysis; logistic regression; missing data; odds ratio Abbreviations: BMI, body mass index; CRA, complete records analysis; MAR, missing at random; MCAR, missing completely at random; MNAR, missing not at random. Missing data are a pervasive issue in epidemiologic stud-ies, reducing statistical power to detect associations of interest and potentially biasing estimates. In a recent systematic re