Clinical decisions based on CEA levels during colorectal cancer follow-up should be based on trend, not single measurements: results from the FACS randomised trial

Background: Most guidelines recommend that patients who have undergone curative resection for primary colorectal cancer are followed up for 5 years with regular blood carcinoembryonic antigen (CEA) tests to trigger further investigation for recurrence. However, CEA may miss recurrences, or patients may have false alarms and undergo unnecessary investigation. Methods: The diagnostic accuracy of trends in CEA measurements for recurrent colorectal cancer, taken as part of the FACS (Follow‐up After Colorectal Surgery) trial (2003–2014), were analysed. Investigation to detect recurrence was triggered by clinical symptoms, scheduled CT or colonoscopy, or a CEA level of at least 7 μg/l above baseline. Time‐dependent receiver operating characteristic (ROC) curve analysis was used to compare the diagnostic accuracy of CEA trends with single measurements. CEA trends were estimated using linear regression. Results: The area under the ROC curve (AUC) for CEA trend was at least 0·820 across all 5 years of follow‐up. In comparison, the AUCs for single measurements ranged from 0·623 to 0·749. Improvement was most marked at the end of the first year of follow‐up, with the AUC increasing from 0·623 (95 per cent c.i. 0·509 to 0·736) to 0·880 (0·814 to 0·947). However, no individual trend threshold achieved a sensitivity above 70 per cent (30 per cent missed recurrences). Conclusion: Interpreting trends in CEA measurements instead of single CEA test results improves diagnostic accuracy for recurrence, but not sufficiently to warrant it being used as a single surveillance strategy to trigger further investigation. In the absence of a more accurate biomarker, monitoring trends in CEA should be combined with clinical, endoscopic and imaging surveillance for improved accuracy.