Background: Incidence of colorectal cancer (CRC), for which diet is an important modifiable risk factor, is among the highest in Europe. However, epidemiological evidence on the role of some bioactive dietary components, such as dicarbonyl compounds, is limited. These reactive molecules, with both potentially deleterious and protective metabolic effects, exist in a wide variety of foods. 

Objective: We evaluated CRC risk associated with the intake of three main food-derived dicarbonyls: methylglyoxal (MGO, mainly from coffee and cereals), glyoxal (GO, from cereals, fruits, and vegetables), and 3-deoxyglucosone (3-DG, from confectionery and baked goods), in European adults.

Methods: Dietary intakes of MGO, GO, and 3-DG were estimated using a detailed food composition database from 367,462 participants of the prospective EPIC cohort (median follow-up=13.9 years; 4,687 accrued CRC cases). Hazard ratios (HRs) and 95% confidence intervals (CIs) for the associations between dietary dicarbonyls and CRC were computed using multivariable adjusted Cox proportional hazard models, with exposures analyzed across quintiles (comparing the highest vs lowest quintile, Q5 vs Q1) and per log-transformed standard deviation increase (ln-SD). Subgroup analyses by anatomical subsite (colon vs rectum) and biological sex were also conducted. Interactions of risk associations with hs-CRP, a marker of chronic inflammation, were also assessed in 1,054 cases with available biomarker data using a nested, matched case-control design.

Results: CRC risk was inversely associated with dietary GO (HRQ5-Q1=0.82, 95%CI=0.74-0.91, P-trend=0.0003) and positively associated with 3-DG (HRQ5-Q1=1.17, 1.06-1.28, P-trend=0.002) but not associated with MGO (HRQ5-Q1=0.92, 0.83-1.01, P-trend= 0.097). Significant heterogeneity by sex was observed for dietary MGO and GO (P-value-for-heterogeneity=0.036 and 0.007, respectively), with inverse associations among men only: HRMGO ln(SD)=0.94 (95%CI=0.9-0.98), HRGO ln(SD)=0.87 (0.83-0.91). No associations were observed in women. Dietary MGO was associated with higher CRC risk among participants with lower hs-CRP levels (OR MGO ln(SD) = 1.24, CI 1.01–1.53, P-heterogeneity = 0.027).

Conclusions: Our findings suggest differential associations with CRC risk depending on specific dietary dicarbonyls, level of systemic inflammation and biological sex. Further research is needed on the potentially divergent roles of these common food components in CRC development, especially regarding sex-specific potential pro-or anti-carcinogenic mechanisms.