Objective‘Environmental’ factors associated with colorectal cancer (CRC) risk include modifiable and non-modifiable variables. Whether those with different non-modifiable baseline risks will benefit similarly from reducing their modifiable CRC risks remains unclear.DesignUsing 7945 cases and 8893 controls from 11 population-based studies, we combined 17 risk factors to characterise the overall environmental predisposition to CRC (environmental risk score (E-score)). We estimated the absolute risks (ARs) of CRC of 10 and 30 years across E-score using incidence-rate data from the Surveillance, Epidemiology, and End Results programme. We then combined the modifiable risk factors and estimated ARs across the modifiable risk score, stratified by non-modifiable risk profile based on genetic predisposition, family history and height.ResultsHigher E-score was associated with increased CRC risk (ORquartile, 1.33; 95% CI 1.30 to 1.37). Across E-scores, 30-year ARs of CRC increased from 2.5% in the lowest quartile (Q1) to 5.9% in the highest (Q4) quartile for men, and from 2.1% to 4.5% for women. The modifiable risk score had a stronger association in those with high non-modifiable risk (relative excess risk due to interaction=1.2, 95% CI 0.5 to 1.9). For those in Q4 of non-modifiable risk, a decrease in modifiable risk reduced 30-year ARs from 8.9% to 3.4% for men and from 6.0% to 3.2% for women, a level lower or comparable to the average population risk.ConclusionsChanges in modifiable risk factors may result in a substantial decline in CRC risk in both sexes. Those with high inherited risk may reap greater benefit from lifestyle modifications. Our results suggested comprehensive evaluation of environmental factors may facilitate CRC risk stratification.

Background There is substantial evidence that use of non-steroidal anti-inflammatory drugs (NSAIDs) reduces the risk of colorectal cancer (CRC), but no subgroup has been identified for which the chemoprevention effect outweighs the risk of side effects. Methods We tested the interaction between NSAID use and multiple risk factors on CRC risk in the VITAL cohort. A total of 73,458 individuals aged 50-76 completed a questionnaire between 2000 and 2002, and 674 incidental colorectal cancer cases were identified through 2010. Results In stratified analysis, high use of any type of NSAIDs (4+days/week for 4+ years) was statistically significantly associated with a lower risk of CRC across all subgroups stratified by sex, BMI, physical activity, smoking, alcohol intake, screening and dietary factors. There was a suggestion of stronger associations among men, obese individuals, and heavier drinkers; however, none of these tests for interaction reached statistical significance. The associations were almost identical for subjects with higher overall CRC risk scores (HR: 0.62; 95% CI: 0.49-0.79) and those with lower risk scores (HR: 0.61; 95% CI: 0.42-0.88). Differential effects by cancer subsites and stages were tested. NSAID use was associated with a greater risk reduction of proximal colon cancer vs. distal (p for difference = 0.06) and distant stage vs. local (p for difference = 0.04). Conclusion The association between high use of NSAIDs and CRC risk does not differ significantly among subgroups. Impact: Our results suggest that NSAIDs have a generally beneficial role in colorectal cancer prevention, largely unmodified by other exposures.

We present a general framework for developing a machine learning (ML) tool that supports clinician assessment of patient risk using electronic health record-derived real-world data and apply the framework to a quality improvement use case in an oncology setting to identify patients at risk for a near-term (60 day) emergency department (ED) visit who could potentially be eligible for a home-based acute care program. Framework steps include defining clinical quality improvement goals, model development and validation, bias assessment, retrospective and prospective validation, and deployment in clinical workflow. In the retrospective analysis for the use case, 8% of patient encounters were associated with a high risk (pre-defined as predicted probability ≥20%) for a near-term ED visit by the patient. Positive predictive value (PPV) and negative predictive value (NPV) for future ED events was 26% and 91%, respectively. Odds ratio (OR) of ED visit (high- vs. low-risk) was 3.5 (95% CI: 3.4–3.5). The model appeared to be calibrated across racial, gender, and ethnic groups. In the prospective analysis, 10% of patients were classified as high risk, 76% of whom were confirmed by clinicians as eligible for home-based acute care. PPV and NPV for future ED events was 22% and 95%, respectively. OR of ED visit (high- vs. low-risk) was 5.4 (95% CI: 2.6–11.0). The proposed framework for an ML-based tool that supports clinician assessment of patient risk is a stepwise development approach; we successfully applied the framework to an ED visit risk prediction use case.