Abstract
BACKGROUND & AIMS: Although many early life exposures have been associated with inflammatory bowel disease (IBD), prediction of risk based on these variables is not yet available. METHODS: We conducted a nationwide population-based cohort study including offspring in the Danish National Birth Cohort with available interview data at 4 time points on early life exposures. Follow-up continued until IBD diagnoses, emigration, death, or end of the study period (September 1, 2022). We used elastic net with underlying Cox regression model to predict IBD diagnosis at <13 and <17 years of age and for the duration of follow-up. We used the entire dataset for model development and internal validation with optimism correction in the fit metrics by bootstrapping. RESULTS: A total of 50,592 offspring were included in the study. During a median follow-up of 21.8 years (interquartile range, 20.4-22.9 years), 310 offspring (0.6%) developed IBD, including 171 with Crohn's disease (CD) and 139 with ulcerative colitis (UC). The area under the receiver operating characteristic curve (AUC) for prediction of IBD overall, at age <17 years, and at <13 years was 0.56, 0.60, and 0.69, respectively. The best prediction models were for UC with onset at age <17 years and <13 years (AUC, 0.64 and 0.76, respectively). Cumulative incidence rates for each outcome for those in the top 5%, compared with the lower 95% of predicted risk, were between 1.6- and 26-fold higher for each model. The main variables contributing to the models included parental IBD status, hygiene-related variables, and those relating to early life diet and infections. CONCLUSIONS: In a nationwide cohort with long-term follow-up data, early life exposures were predictive of UC, especially younger-onset disease, whereas prediction of CD was less accurate.