Background: In molecular epidemiologic studies biospecimen data are collected on only a proportion of subjects eligible for study. This leads to a missing data problem. Missing data methods, however, are not typically incorporated into analyses. Instead, complete-case (CC) analyses are performed, which result in biased and inefficient estimates.
Methods: Through simulations, we characterized the bias that results from CC methods when interaction effects are estimated, as this is a major aim of many molecular epidemiologic studies. We also investigated whether standard multiple imputation (MI) could improve estimation over CC methods when the data are not missing at random (NMAR) and auxiliary information may or may not exist.
Results: CC analyses were shown to result in considerable bias while MI reduced bias and increased efficiency over CC methods under specific conditions. It improved estimation even with minimal auxiliary information, except when extreme values of the covariate were more likely to be missing. In a real study, MI estimates of interaction effects were attenuated relative to those from a CC approach.
Conclusions: Our findings suggest the importance of incorporating missing data methods into the analysis. If the data are MAR, standard MI is a reasonable method. Under NMAR we recommend MI as a tool to improve performance over CC when strong auxiliary data are available. MI, with the missing data mechanism specified, is another alternative when the data are NMAR. In all cases, it is recommended to take advantage of MI’s ability to account for the uncertainty of these assumptions.
Desai, Manisha; Esserman, Denise; Gammon, Marilie; and Terry, Mary Beth, "The use of multiple imputation in molecular epidemiologic studies assessing interaction effects" (November 2010). COBRA Preprint Series. Working Paper 73.