Background: Children with Crohn's disease (CD) are at a high risk for complications from both disease and treatment. The ability to predict risk and adverse outcomes at or close to diagnosis would allow patients to be stratified by risk, in order to avoid under treatment or overtreatment while reducing adverse outcomes from drug and disease. The goal of the current prospective study was to identify factors that would predict early complicated disease behavior or surgery.
Methods: The GROWTH CD study (Growth, Relapses and Outcomes With THerapy) is geared to identify factors that could predict early outcomes such as complications (stricturing, penetrating or perianal abscess)and surgery by 24 months. Newly diagnosed children underwent colonoscopy, gastroscopy and imaging. They were phenotyped by the Paris classification and followed at baseline, 8, 12, 26, 52, 78 and 104 weeks. Twenty dichotomous and continuous variables were assessed, including serum biomarkers (ASCA, CBIR1, OMPC), measures of inflammation (ESR, CRP, Calprotectin), disease activity (PCDAI and PGA) and serum albumin. Predictors at diagnosis and week 12 (post induction treatment) served as prediction time points. Complications and surgery were recorded at weeks 78 and 104. Logistic regression and risk modeling was performed for best fit models.
Results: 285 children, median age 13 yrs, 60% male, were followed prospectively for 2 years, of whom 78 (27.3%) developed complications and 28 (9.8%) required surgery. Use of immunomodulators by 12 weeks was not associated with decreased risk, (complications and surgery both p=0.9). Presence of a complication at diagnosis and PCDAI>30 at week 12 (OR 4.2, CI 1.16–16.77) p=0.03, were predictive of complications. Five parameters predicted increased risk of surgery (high ESR>50 wk 0, stricturing disease at diagnosis, ASCA, hypoalbuminemia or elevated PCDAI >10 at week 12). The first 4 remained in the best fit model. The combination of Paris B2 (OR 4.2, CI 1.3–12.7) p=0.01, ESR>50 (OR 2.9, CI1.004–8.642) p=0.049 at baseline, and Alb<3.6 (OR 4.9 CI 1.2–19.6), p=0.024, and ASCA IgA (OR 1.02, CI 1.007–1.03) p=0.001 after therapy week 12, had a sensitivity 10%, specificity 98.7%, PPV 50% and NPV 89.4%, the model is significant (p<0.0001) and correctly classifies 88.6% of the study population requiring early surgery. CRP and fecal calprotectin were not predictors of complications or surgery at either time point.
Conclusions: We identified a prediction model for early high risk disease based on risk for early surgery in paediatric CD. We could not accurately predict risk of early complications with the biomarkers used. Fecal calprotectin and CRP were not predictive at any time point.
Supported by grants from ECCO and the Thrasher Research Fund