The comparisons of treatments in randomized clinical trials may use the ana
lysis of covariance to adjust for patient characteristics. We present theor
etical results that describe when such an adjustment would be expected to b
e beneficial. A distinction is made between covariates that are balanced in
the design and those that are assigned by the randomization process. The r
esults support the commonly held view that features balanced in the design
of the trial (e.g., by stratification) and those that are strongly predicti
ve of the outcome, and thus considered clinically prognostic, should normal
ly be included in the analysis. For other covariates that are not balanced
in the design, the potential benefits of including them in the analysis wil
l depend on the number of patients in the trial. However, there is frequent
ly a set of variables whose relevance is unknown and for which data-depende
nt methods of selection, based on the data for the current trial, have been
proposed. A review of the literature has shown that these methods can prod
uce misleading inferences. The decision as to which covariates to include i
n the analysis should be specified in the protocol on the basis of data fro
m previous trials on similar patient populations. The methods are illustrat
ed with data from a trial comparing two therapies for treating scalp psoria
sis where the clinical importance of patients' age and sex as prognostic fa
ctors for efficacy is unknown. We show for what size of future trials it wo
uld be beneficial to adjust for these covariates and for what size trials i
t would not. In all cases, prespecification of variables to be included in
the analysis is essential in order to avoid bias. (C) Elsevier Science Inc.
2000.