In multivariate analyses, the relative risk estimates for digestive tract cancers in persons who specified their aspirin usage became somewhat stronger (i.e., further below 1.0) when persons with missing or incomplete data were excluded from the models.