False-positive elevations of carcinoembryonic antigen in patients with a history of resected colorectal cancer

Abstract

Routine monitoring of carcinoembryonic antigen (CEA) levels is standard in patients with resected colorectal cancer (CRC). The incidence of false-positives and the upper limits of false-positive elevations have not been previously well characterized. A search of medical records at Memorial Sloan-Kettering Cancer Center identified 728 patients who underwent an R0 resection of locoregional CRC between January 2003 and December 2012 and who had an increase in CEA level above the normal range after a normal perioperative CEA level. Of these, 358 had a false-positive elevation of CEA level, 335 had a true-positive elevation indicative of recurrent CRC, and 35 had a true-positive elevation indicative of the development of a new, non-CRC malignancy. Of those with false elevations, 111 had a single isolated CEA level elevation (median highest CEA level of 5.5 ng/mL) with no further abnormal measurements, whereas 247 had elevations on 2 or more readings, with a median highest level of 6.7 ng/mL. Of these 247 patients with confirmed false-positive CEA level elevations, only 5 (2%) had measurements greater than 15 ng/mL, and no confirmed elevation greater than 35 ng/mL was a false-positive. False-positive CEA test results in the range of 5 to 15 ng/mL are common. Confirmation of CEA elevation in this range before initiating imaging studies may be appropriate. False-positive results greater than 15 ng/mL are rare, and all confirmed CEA levels greater than 35 ng/mL were associated with cancer recurrence.

Figure 1

Receiver operating curves (ROC) comparing manufacturer’s expected performance with performance at Memorial Sloan-Kettering Cancer Center (MSKCC). Abbreviations: AUC, area under the curve; CEA, carcinoembryonic antigen; M-L, MSKCC lower (limit of the confidence interval); M-U, MSKCC upper; T-L, Tosoh lower; T-U, Tosoh upper.

Figure 2

Comparison of results at each carcinoembryonic antigen (CEA) level for patients with and without actual cancer. Levels > 5 ng/mL represent actual data; values < 5 ng/mL are calculated.