Identification of prediagnostic proteomic biomarkers for colon and rectal cancer in 3 prospective cohorts

Abstract

BACKGROUND: Large-scale proteomics may reveal circulating biomarkers of colorectal cancer (CRC) years before diagnosis, yet cross-cohort reproducibility and anatomic site-specificity remain unclear. METHODS: We conducted nested case-control studies in three prospective cohorts with banked blood and cancer follow-up (median 7 years), totaling 851 incident CRC cases matched 1:1 to controls, with 1,801 pre-diagnostic plasma proteins from Olink panels. RESULTS: Twelve proteins showed significant paired differences between CRC cases and controls (FDR <0.1), with modest effect sizes. These included previously implicated CRC-biomarkers (e.g., higher GDF15 and TFF2) as well as less established associations (e.g., lower LEPR and BCAN). Paired protein differences are significantly distinct in colon vs. rectal cancer pairs (P = 0.005), driven by genes enriched in leukocyte activation, cell activation, and extracellular components. In stratified analysis, for colon cancer, 11 out of 12 proteins were lower in cases (FDR <0.1), including BCAN with 0.83 (0.74-0.94) odds ratio (OR, with 95% CI) per 1-unit increase in quantile-normalized NPX level, with cross-cohort consistency (I2 <0.01). For rectal cancer, 25 out of 28 proteins were higher in cases, including HDGF (OR 1.75 [1.15-2.66], I2 <0.01). HDGF could partially mediate BMI-rectal cancer risk (25.0%, FDR=0.044). Overall, we found that pre-diagnostic plasma proteomics identified proteins and pathways with modest effect sizes and clear differences between future risk of colon and rectal cancer. CONCLUSION: These findings provide insight for future mechanistic validation of pre-diagnostic plasma proteins as CRC biomarkers and motivate site-specific biomarker discovery that distinguishes colon from rectal cancer.