Abstract
OBJECTIVES: The FACS, GILDA, and COLOFOL trials have cast doubt on the value of intensive extracolonic surveillance for resected nonmetastatic colorectal cancer and by extension metastasectomy. We reexamined this pessimistic interpretation. We evaluate an alternative explanation: insufficient power to detect a realistically sized survival benefit that may be clinically meaningful. METHODS: A microsimulation model of postdiagnosis colorectal cancer was constructed assuming an empirically plausible efficacy for metastasectomy and thus surveillance. The model was used to predict the large-sample mortality reduction expected for each trial and the implied statistical power. A potential recurrence imbalance in the FACS trial was investigated. Goodness of fit between model predictions and trial results were evaluated. Downstream life expectancy was estimated and power calculations performed for future trials evaluating surveillance and metastasectomy. RESULTS: For all 3 trials, the model predicted a mortality reduction of ≤5% and power of <10%. The FACS recurrence imbalance likely led to a large relative bias (>2.5) in the hazard ratio for overall survival favoring control. After adjustment, both COLOFOL and FACS results were consistent with model predictions (P>.5). A 2.6 (95% credible interval 0.5-5.1) and 3.6 (95% credible interval 0.8-7.0) month increase in life expectancy is predicted comparing intensive extracolonic surveillance-routine computed tomography scans and carcinoembryonic antigen assays-with 1 computed tomography scan at 12 months or no surveillance, respectively. An adequately sized surveillance trial is not feasible. A metastasectomy trial should randomize at least 200 to 300 patients. CONCLUSIONS: Recent trial results do not warrant de novo skepticism of metastasectomy nor targeted extracolonic surveillance. Given the potential for clinically meaningful life-expectancy gain and significant uncertainty, a trial of metastasectomy is needed.