Covariate adjustment of treatment effects in clinical trials.

It has been previously shown that the magnitude of adjustment of a treatment effect Z value for a baseline covariate depends on both the degree of disparity of the covariate between the treatment groups and the degree of correlation between the covariate and the outcome variable. In this article, the magnitude of adjustment is further explored and tabulated for a variety of situations. Theoretically, the magnitude of adjustment may be great, for example, from Z = 3.0 unadjusted to 13.1 adjusted, or Z = 2.0 unadjusted to -1.6 adjusted. However, in more practically realistic situations the magnitude of adjustment will generally fall in the range of 1 to 2 standard error units, when the Z value for disparity of the covariate between the treatment groups is as great as +/- 3. If there is perfect comparability of treatment groups with respect to baseline covariate, there may still be a sizable difference between adjusted and unadjusted Z value for treatment effect, with adjusted almost always greater in absolute value than unadjusted Z value.