Leveraging EHRs and a phenome-wide association study to identify pre-diagnostic clinical markers in early-onset colorectal cancer.

Abstract

3622 Background: The incidence of colorectal cancer (CRC) has undergone a significant demographic shift, with diagnosis in individuals under 55 years escalating from 11% in 1995 to 20% in 2019 (Siegal et al., 2019). Traditional risk factors, such as obesity, diabetes, and colon microbiome changes, fail to fully explain this emerging clinical phenomenon, necessitating new approaches to identify pre-diagnostic markers in younger patients. Our study leverages electronic health records (EHR) and employs a Phenome-Wide Association Study (PheWAS) to characterize age-specific manifestations prior to diagnosis. Methods: A PheWAS was conducted with a cohort of 2,799 CRC patients from Northwestern Medicine (NM, 2012 – 2022). ICD-9/10 codes were categorized into 527 phenotypes using the PheCode framework. Logistic regression models were employed to examine associations between age groups ( < 55 vs. ≥55 years) and each phenotype during the 6 months preceding diagnosis. A control cohort of 294,590 non-cancer NM patients adjusted for age-related and baseline disease patterns. Results: Our approach revealed significant associations with age in 9 of 527 analyzed phenotypes, identifying well-documented CRC features and less recognized clinical markers. Younger patient cohorts exhibited elevated probabilities of critical clinical indicators, including lower gastrointestinal hemorrhage (22% [95% CI: 19-25%] vs. 11% [95% CI: 10-13%]; p = 1.0 x 10 -11 ), hepatic dysfunction (6% [95% CI: 4-8%] vs. 4% [95% CI: 3-4%]; p = 2.7 x 10 -6 ), and diverticular disease (4% [95% CI: 3-6%] vs. 3% [95% CI: 2-3%]; p = 2.4 x 10 -11 ) compared to older patients. Furthermore, an analysis of 1,866 NM pathology reports revealed single-institution trends consistent with global CRC literature. Notably, younger patients showed higher prevalence of hereditary CRC (28% vs. 24%; χ 2 p = 6.3 x 10 -8 ), greater left-sided tumor localization (51% vs. 45%; χ 2 p = 0), and increased proximal tumors (58% vs. 53%; χ 2 p = 0), corroborating studies that show a rise in proximal tumor incidence in patients under 50 and a decline in those aged 50–79. Conclusions: Our PheWAS study uncovered clinicopathological distinctions, revealing statistically significant variations in pre-diagnostic clinical markers that suggest a more aggressive and complex disease progression mechanism in early-onset CRC. The increased probabilities of clinical indicators—such as gastrointestinal hemorrhage, hepatic dysfunction, and diverticular disease—underscore the need for tailored diagnostic and screening strategies for younger populations. These results also highlight the value of utilizing phenotype-based methodologies to identify nuanced clinical features that may enable earlier detection and improve outcomes for younger CRC patients. Towards this goal, ongoing research aims to further delineate CRC clinical profiles between age groups.