Tailoring Surveillance Colonoscopy in Patients With Advanced Adenomas

Abstract

Introduction: Patients with advanced adenomas (AA) consume considerable colonoscopy resources. Understanding their risk of subsequent colorectal cancer (CRC) and their comorbidity might improve use of surveillance colonoscopy (SC) by directing it to those with high subsequent CRC risk &/or low comorbidity and away from those with low CRC risk &/or high comorbidity. In this study, we derived a prediction rule that stratifies risk for non-CRC mortality (NCM) following removal of AA. Methods: We identified Veterans whose VA-based index colonoscopy (IC) (2002-2009) revealed ≥ 1 AA (adenoma ≥1 cm, villous histology or high-grade dysplasia). Excluded were those with CRC < 6 months of the IC and those with CRC prior to IC. Through 2015, we obtained subsequent CRC incidence, all-cause mortality (ACM), and CRC-specific mortality (CSM) from VA electronic sources, linked VA-Medicare data, and non-VA facilities. Comorbidity was measured with the Charlson comorbidity index (CCI). The National Death Index provided cause of death, while CRC incidence was obtained from VA oncology data. Logistic regression identified factors independently associated with NCM. We applied points to each factor based on its odds ratio (OR), calculated NCM risk at each point total, collapsed point totals into risk strata for which we calculated risk of NCM, CSM and CRC incidence. Results: 2,943 Veterans had AA (mean age [SD] 63.1 [8.6] years, 98% men, 74% White), of whom 2,080 had ≥ 1 SC. Mean [SD} follow-up was 7.5 [2.9] years, mean CCI was 1.4 [1.8]; 42% used tobacco. CRC incidence was 1.4%, CSM was 0.4%, ACM was 31.3% and NCM was 30.9%. Independently associated with NCM was age, CCI, APACHE score, several individual comorbidities and recent hospitalization (c-statistic=0.75, [CI, 0.73-0.77]; GOF test P-value=0.92). Risk scores (0-11) identified three risk groups for NCM: Low (score 0-2), intermediate (3-5), and high (≥ 6) with respective risks of 14.0%, 33.7%, and 64.3% and respective sample proportions of 40.3%, 43.1% and 16.6%. Foregoing SC in the high-risk group would reduce colonoscopy resources by nearly 17%, and would have missed 4 (10%) of 40 incident CRCs and 1 (8.5%) of 12 CRC deaths. Conclusion: Among Veterans with AA, patient-specific information may be used to tailor SC based on risk for NCM. While this prediction rule requires validation, this type of work combined with factors predicting subsequent risk for CSM may improve the clinical utility and efficiency of SC in this high-risk group.