Evaluating the usefulness of longitudinal CEA measurements to detect colorectal cancer recurrence

Talk Code: 

The problem

Colorectal cancer (CRC) is the fourth most common cancer in the UK, with approximately 40,000 new cases diagnosed each year. Unfortunately, around 30-40% of these patients go on to experience recurrence. In 2008, Tan et al. carried out a meta-analysis of 20 studies exploring the accuracy of CEA to detect CRC recurrence. Most of the primary studies in this review evaluated the accuracy of CEA based on the measurement taken closest to the time that recurrence was clinically detected, ignoring any prior CEA measurements collected. This focus on a single measurement is misleading as in practice all collected CEA values are interpreted.

The approach

Longitudinal CEA measurements taken during post-surgery follow-up of CRC patients were analysed from 2 arms of the FACS randomised controlled trial. In addition to evaluating the accuracy of interpreting raw CEA measurements, alternative methods of interpretation were also compared using ROC analysis: the difference in CEA from their post-operative baseline measurement, the ratio of CEA to baseline, and rate of change.


The area under the curve for interpreting raw CEA measurements was 0.74 (95% CI: 0.68-0.80). Interpreting the difference or ratio of CEA to baseline offered no improvement: 0.68 (95% CI: 0.61-0.75) and 0.66 (95% CI: 0.59-0.73), respectively. Interpreting the rate of change in CEA (absolute) over time did however significantly improve the accuracy of CEA: AUC = 0.91 (95% CI: 0.87-0.96).


Monitoring raw longitudinal CEA measurements, as recommended by many clinical guidelines, leads to a notable proportion of missed recurrences. The sensitivity of CEA was heavily compromised by a significant subset of the patients who developed recurrence and failed to produce elevated or increasing CEA levels. We would currently recommend an additional CT scan in the first year (when nearly half of recurrences occur, 47%) to ensure that these missed cases are picked up. There does appear to be greater discriminatory information available in the rate of change in CEA measurements however, but caution should be taken as this analysis included all CEA measurements available whereas in clinical practice CEA measurements would be collected cumulatively. The value of this method of CEA interpretation needs further prospective evaluation.


  • Bethany Shinkins, Department of Surgery, University of Southampton, Southampton, UK
  • Brian Nicholson, Department of Surgery, University of Southampton, Southampton, UK
  • Timothy James
  • John Primrose, Department of Clinical Biochemistry, Oxford University Hospitals Trust, Oxford, UK
  • David Mant, Department of Surgery, University of Southampton, Southampton, UK