Comparing predictive decision rules in postoperative CEA monitoring.

To evaluate the usefulness of serial postoperative carcinoembryonic antigen (CEA) assays, seven previously published decision rules for predicting tumor recurrence were compared retrospectively using CEA values from 214 patients followed 36 to 120 months after surgery for colorectal carcinoma. Decision rules employing cutoff values to predict tumor recurrence were found inadequate for the asymptomatic patient. This attenuation of prognostic usefulness appeared attributable to inadequacies of CEA assays for predicting late recurrences. From these analyses, elevated CEA results without other objective evidence might be insufficient to justify second-look surgery. In addition, late recurring tumors tended not to cause elevated CEA levels.