Benefit-risk analysis of different risk-related surveillance schedules following colorectal polypectomy.

BACKGROUND/AIMS: For colorectal screening patients a gain of life time was previously calculated to be about 30-50 days. Different recommendations for recognizing at-risk groups and defining surveillance intervals after an initial finding of colorectal adenomas have been published. However, no benefit-risk analysis regarding specific long-term effects of follow-up patients has been reported to date.

METHODOLOGY

A Markov model based on time-dependent transition possibilities was developed to compare two surveillance schedules: recommendations based on the Erlangen Registry of Colorectal Polyps (ERCRP) and the National Polyp Study (NPS). The outcome was calculated for a 50-year-old patient with 30 years of follow-up after initial polypectomy. The data used in this model were taken from different sources, namely the ERCRP, the German Study Group of Colorectal Cancer, the German Statistical Yearbook, and from meta-analyses of studies reporting data on complications and sensitivity of colonoscopy.

RESULTS

Patients under surveillance have a mean lifetime gain of 98 (ERCRP) and 91 (NPS) days compared with those who do not come for surveillance. Approximately 84% and 79% of deaths from colorectal carcinoma (CRC) could be prevented if patients were followed up according to the recommendations of the ERCRP and the NPS, respectively. The risk of death due to colonoscopy for patients during followup is about 0.05% lifetime risk. Sensitivity analysis showed the stability of the results under a wide range of reasonable variations of relevant parameters. In a pessimistic one-way sensitivity analysis regarding compliance, effectiveness was reduced to one third.

CONCLUSIONS

Surveillance using colonoscopy is an effective tool for preventing CRC after colorectal polypectomy and similar to the screening procedure. The effectiveness is slightly higher when following the recommendations of the ERCRP, especially if a more realistic compliance is assumed.