Background
Carcinoembryonic antigen (CEA) monitoring facilitates the detection of recurrence in patients with colorectal cancer (CRC) after resection. False-positive CEA has been reported in CRC patients with certain comorbidities or smokers. However, limited information is currently available on the frequency of and changes in falsely elevated CEA levels in patients without these conditions.
Materials and methods
We retrospectively examined CRC patients who underwent surgical resection at our hospital between 2001 and 2017, had no recurrence for at least five years, and were free of known factors that may increase CEA. Postoperative CEA levels were retrieved until 2 years before the last contact. For comparison, we similarly selected patients who developed recurrence after resection of CRC during the same period, and CEA levels at initial presentation, at nadir, and at the time of recurrence were reviewed. The patterns of elevated CEA (>5 ng/ml) were classified as transient, repeated, or persistent based on longitudinal changes. The relationships between CEA and carbohydrate antigen 19-9, transaminases, creatinine, and C-reactive protein were examined.
Results
CEA elevation occurred in 90 (20%) out of 446 eligible patients without recurrence at least once during the mean postoperative period of 50.5 months, whereas CEA was >5 ng/ml in 117 (53%) of 221 patients when they developed recurrence. Twenty-seven patients without recurrence showed a transient elevation in CEA, 45 repeated elevations, and 18 a persistent elevation; the frequency of a high preoperative CEA level increased in this order. The majority (98%) of false elevations ranged between 5 and 15 ng/ml. CEA was not associated with other laboratory data.
Conclusions
Unexplained CEA elevations were observed in 20% of recurrence-free CRC patients after surgery, and were classified into three patterns based on longitudinal changes. A more detailed understanding of patient-specific fluctuations in CEA will prevent unnecessary imaging studies and reduce medical costs.