Abstract / Summary
We aimed to evaluate the association between intake of beverage-type ultra-processed foods (UPFs) and the risk of incident colorectal cancer (CRC), to derive a beverage-type UPF-related metabolomic signature, and to investigate the independent and joint associations of dietary exposure and its metabolomic signature with CRC risk. A total of 94,991 participants with at least two valid repeated 24-hour dietary recalls, complete covariate data, and available metabolomic data were included. Beverage-type and non-beverage-type UPF exposures were defined based on the NOVA framework. An elastic net regression model was used to construct a beverage-type UPF-related metabolomic signature. Cox proportional hazards models were applied to evaluate the associations of beverage-type UPF intake and its metabolomic signature with incident CRC, and joint models were further established. During a mean follow-up of 11.45 years, 958 incident CRC cases were identified. After multivariable adjustment, each 1-standard deviation (SD) increase in beverage-type UPF intake was associated with a 10.4% lower risk of CRC (hazard ratio [HR] = 0.896, 95% confidence interval [CI]: 0.831–0.966, P = 0.004); participants in the highest quartile (Q4) also had a lower risk than those in the lowest quartile (Q1) (HR = 0.835, 95% CI: 0.709–0.984, P = 0.031). No clear association was observed between non-beverage-type UPF intake and CRC risk. Based on 251 quality-controlled metabolites, the beverage-type UPF-related metabolomic signature included 29 metabolites. Each 1-SD increase in this signature was associated with a 10.4% higher risk of CRC (HR = 1.104, 95% CI: 1.022–1.193, P = 0.012). In the joint model, both beverage-type UPF intake and the metabolomic signature remained statistically significant, with HRs of 0.889 (95% CI: 0.824–0.959, P = 0.002) and 1.115 (95% CI: 1.032–1.205, P = 0.006), respectively. The joint model yielded the lowest Akaike information criterion (AIC) and the highest C-index (0.677). Sensitivity analyses showed consistent results. The beverage-type UPF intake was inversely associated with incident CRC risk, whereas the beverage-type UPF-related metabolomic signature was associated with higher CRC risk. Given the weak correlation between the dietary exposure and the metabolomic signature, these findings should be interpreted as distinct observational associations rather than as evidence of a single causal pathway linking beverage-type UPF consumption to CRC through metabolic alterations. The metabolomic signature may capture broader circulating metabolic characteristics beyond those specifically attributable to beverage-type UPF consumption. The inverse association should not be interpreted as evidence of a protective effect of beverage-type UPFs. Instead, the findings highlight the heterogeneity between dietary exposure and circulating metabolic characteristics and underscore the need for further studies to clarify the biological and causal basis of these associations.