What question did this study set out to answer?

This research aims to evaluate the criteria for effective Crohn’s disease prevention trials, focusing on predictive performance and intervention efficacy.

January 23, 2026

P1184High predictive performance, cohort enrichment and intervention efficacy are critical for Crohn’s disease prevention trials

Key Points

This research aims to evaluate the criteria for effective Crohn’s disease prevention trials, focusing on predictive performance and intervention efficacy.
Simulated models based on 5-year CD risk estimates for general and high-risk populations.
Assessed predictive model performance using AUC metrics.
Calculated positive predictive values at fixed sensitivity levels.
Estimated required sample sizes for two-arm prevention trials based on targeted risk groups and intervention efficacy.
General population screening yields a low positive predictive value of 1.6%, while first-degree relatives have 12.6%.
Targeting the top 1% of first-degree relatives shows a 5-year risk of 27%.
To achieve adequate power, trials need sample sizes of 106 for high-efficacy interventions, requiring up to 10,000 screenings.
Lower intervention efficacy increases necessary sample sizes significantly, necessitating screening of up to 124,000 individuals.

Abstract

Abstract Background Prevention of Crohn’s disease (CD) is a key unmet need, with increasing demand from patients and their relatives. However, the low incidence of CD, even in high-risk groups like first-degree relatives (FDRs), poses a significant challenge for designing interventional trials1. We modelled the necessary criteria for trial population enrichment and preventive intervention efficacy, exploring how the predictive performance of screening tools and intervention efficacy influence the sample sizes required for a well-powered CD prevention trial. Methods We performed simulations based on published peak Western European 5-year CD risk estimates for the general population (no known IBD FDRs, 0.11%) and FDRs (1%)2. We modelled the performance of a ‘high but realistic’ predictive model (AUC=0.90)3. We calculated the precision (Positive Predictive Value, PPV) at a fixed 50% sensitivity for both 5-year risk estimates4, and estimated prediction in the high-risk tails (top 1, 5, 10%). Finally, we modelled the required sample sizes for two-arm prevention trials (80% power, target p-value α = 0.05) by varying three factors: screened cohort (population vs FDRs), top risk percentile targeted, and intervention efficacy (25–100%). Results At 50% sensitivity, screening the general population and FDRs yields low precision of 1.6% and 12.6% respectively, illustrating how low incidence limits PPV even for high-AUC models. Focusing on screening within the top risk tails yields groups with much higher risk of developing CD – the top 1% of FDRs has a 5-year risk of 27% (Fig. 1). Even with a high efficacy intervention (Fig. 2), for example preventing 75% of cases, planning an interventional trial requires a sample size of 106 participants in the top 1% risk group, implying that ∼10,000 FDRs must be screened to identify this top 1% risk group. With lower intervention efficacies (50% and 25%), the required sample sizes increase to 276 and 1,242 respectively, necessitating screening of ∼27,000 and ∼124,000 individuals to achieve the same target sample within the top risk stratum. These results highlight the demanding combination of intervention efficacy and scale required for CD prevention trials. Conclusion Feasible prevention trials in CD require high-performance models (AUC ≥ 0.9) and high-efficacy interventions. Furthermore, to avoid underpowered trials, these cohorts must be stratified to prioritise the highest-risk individuals (e.g., top 1–5%) to substantially reduce the required sample size for interventional trials. Designing CD prevention trials needs the trifecta of careful at-risk population selection, further enrichment for the highest-risk individuals, and an efficacious treatment to demonstrate a significant risk reduction. References: 1. Jostins L, Barrett JC. Genetic risk prediction in complex disease. Hum Mol Genet. 2011;20(R2):R182-R188. 2. Moller FT, Andersen V, Wohlfahrt J, Jess T. Familial risk of inflammatory bowel disease: a population-based cohort study 1977-2011. Am J Gastroenterol. 2015;110(4):564-571. 3. Grännö O, Bergemalm D, Salomon B, et al. Preclinical protein signatures of Crohn’s disease and ulcerative colitis: A nested case-control study within large population-based cohorts. Gastroenterology. 2025;168(4):741-753. 4. Bravo AC, Christensen SG, Julsgaard M, et al. Patient and first-degree relatives perceptions about prediction and prevention of inflammatory bowel disease-A multinational survey. United European Gastroenterol J. 2025;13(7):1226-1238. Conflict of interest: Ms. Sollok, Michelle: No conflict of interest Tinoco da Silva Torres, Joana: Grant: Abbvie, Janssen Personal Fees: Pfizer, Janssen, Abbvie, Sandoz, Lilly, Sanofi, Takeda Non-financial Support: Janssen, Abbvie Weersma, Rinse K: No conflict of interest Colombel, Jean-Frédéric: Grant: AbbVie, Janssen Pharmaceuticals, Takeda, Prometheus and Bristol Myers Squibb Lectures from: AbbVie, Roche and Takeda Other: AbbVie, Amgen, AnaptysBio, Allergan, Apini, Arena Pharmaceuticals, Astellas, Boehringer Ingelheim, Bristol Myers Squibb, candidrx Celgene, Celltrion, Clearview Curogen, Eli Lilly, Envision Pharma Ferring Pharmaceuticals, Galmed Research, Glaxo Smith Kline, Roche, Janssen Pharmaceuticals, Kaleido Biosciences, Immunic, Iterative Scopes, Landos, Microba Life Science, Merck, Mirador, Novartis, Otsuka Pharmaceutical, Owkin, Pfizer, Protagonist Therapeutics, Sanofi, Sun Pharma, Takeda, Teva, TiGenix, and is holding stock options in Intestinal Biotech Development Jess, Tine: Personal Fees: Consultancy for Ferring, Pfizer, Johnson&Johnson Sazonovs, Aleksejs: No conflict of interest

Mark Helpful

Bookmark

Relay