A197 A PREDICTION MODEL OF RISK OF HARBOURING ADVANCED COLORECTAL NEOPLASMS IN LOW TO MODERATE RISK PERSONS OVER AGE 50

Colonoscopy decreases the incidence of colorectal cancer (CRC) and CRC-related death, primarily through timely detection and treatment of advanced colorectal neoplasms (ACNs), including CRC and high-risk adenomas (HRA). Unfortunately, risk stratification methods for colonoscopy are poor, and less than 20% of persons over age 50 who undergo colonoscopy are diagnosed with ACNs. Current guidelines do not adequately account for the simultaneous contribution of multiple major and minor risk factors and protective factors for developing ACNs. Combined with increasing demands for colonoscopy, Canadians now faces wait times that greatly exceed recommended targets, escalating colonoscopy-related costs and poor value for the money spent on these procedures. Widespread implementation of population-based FIT screening in average-risk patients in coming years will compound these problems. To derive prediction models that discriminate between individuals who are likely or unlikely to harbour ACNs. We studied 11,719 consecutive persons aged 50 years or older who underwent outpatient colonoscopy at The Ottawa Hospital between 2008 and 2012 for low-to-moderate risk indications, including non-life-threatening signs or symptoms, personal history of adenomas, family history of CRC and average-risk screening. We excluded individuals who had high risk or rare indications, as well as those who had incomplete colonoscopy, poor bowel preparation, or important missing information. We obtained model variables through chart review and linkage to Ontario health administrative databases. We tested 22 candidate predictors, encompassing colonoscopy indication, age, sex, residential setting, household income, co-morbidity burden, cancer history, and prior colonoscopy and polypectomy exposure. We used multivariable logistic regression with stepwise selection to derive our final models. We tested the performance of our primary models in multiple subgroups. Our final models retained eight variables that are easily ascertainable in an office setting. The models showed excellent discriminatory capacity (c-statistic > 0.95) and calibration (p-value > 0.5 for goodness-of-fit test) for CRC in the main cohort and all subgroups, and improved the specificity of colonoscopy for detecting ACNs without significantly impacting sensitivity. Applying the models to our derivation cohort would have allowed for a 25% reduction in colonoscopy volume with a CRC miss rate of < 1% and a HRA miss rate of < 10%. We have derived predictive models with high discriminatory capacity for ACNs that could help optimize the use of colonoscopy resources in clinical practice. If successfully validated, these models have the potential to improve the clinical utility and cost-effectiveness of colonoscopy. Figure 1. Receiver operating curve for model of colorectal cancer (area under curve = 0.96) Academic Health Sciences Centres Alternate Funding Plan Innovation Fund (administered by The Ottawa Hospital Academic Medical Association)

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.