The clinician treating colorectal cancer (CRC) should be able to predict which patients will live and which will die from their tumor. Many prognostic variables, significant in retrospective analyses, are never formally used in staging. The resultant imperfection of clinical predictions of survival are probably directly related to the relative inaccuracy of the staging system, a system inevitably dependent entirely on pathologic criteria. Potentially valuable clinical variables, sometimes associated with better or worse outcome despite the pathologic stage of disease, could improve the accuracy of prediction. Evolution to better clinico-pathologic staging systems would also help to better stratify patients in prospective randomized trials of new adjuvant therapeutic modalities.