Front Med (Lausanne) . 2021 Aug 27;8:644003. doi: 10.3389/fmed.2021.644003. eCollection 2021.

• PMCID: PMC8430211
• DOI: 10.3389/fmed.2021.644003

Abstract

Background and Aims: Intensifying therapy for Paediatric Crohn’s Disease (CD) by early use of immunomodulators and biologics has been proposed for cases in which predictors of poor outcome (POPO) were present. We investigated therapy stratifying potential comparing POPO-positive and -negative CD patients from CEDATA-GPGE®, a German-Austrian Registry for Paediatric Inflammatory Bowel disease. 

Methods: CD patients (1-18 years) registered in CEDATA-GPGE® (2004-2018) within 3 months of diagnosis and at least two follow-up visits were included. Disease course and treatments over time were analysed regarding positivity of POPO criteria and test statistical properties. 

Results: 709/1084 patients included had at least one POPO criterion (65.4%): 177 patients (16.3%) had persistent disease (POPO2), 581 (53.6%) extensive disease (POPO3), 21 (1.9%) severe growth retardation POPO4, 47 (4.3%) stricturing/penetrating disease (POPO6) and 122 (11.3%) perianal disease (POPO7). Patients with persistent disease differed significantly in lack of sustained remission >1 year (Odd Ratio (OR) 1.49 [1.07-2.07], p = 0.02), patients with initial growth failure in growth failure at end of observation (OR 51.16 [19.89-131.62], p < 0.0001), patients with stricturing and penetrating disease as well as perianal disease in need for surgery (OR 17.76 [9.39-33.58], p < 0.001; OR 2.56 [1.58-4.15], p < 0.001, respectively). Positive Predictive Value for lack of sustained remission was >60% for patients with initial growth failure, persistent or stricturing/penetrating disease. 

Conclusion: Predictors of poor outcome with complicated courses of disease were common in CEDATA-GPGE®. An early intensified approach for paediatric CD patients with POPO-positivity (POPO2-4, 6-7) should be considered, because they have an increased risk to fare poorly.

Der Beitrag Running Behind „POPO“ – Impact of Predictors of Poor Outcome for Treatment Stratification in Pediatric Crohn’s Disease erschien zuerst auf CEDATA GPGE – Patientenregister.

DATA FROM THE CEDATA GPGE REGISTRY

To date, failure of the first-line medical management of Crohn’s disease (CD) in children and adolescents has been followed by escalation treatment, with administration of immunomodulators, e.g., azathioprine, or biologics, such as infliximab (1). The most recent revision of the international guideline proposes stratification of the therapy, with selection of either accelerated step-up or primary treatment with biologics on the basis of predictors of poor outcome (POPO) (2). The objective of this study was to evaluate this concept of early immunosuppression with the aid of data from the largest registry of children and adolescents with chronic inflammatory bowel disease (IBD) in Europe.

Der Beitrag Early Immunosuppression in Children and Adolescents with Crohn’s Disease erschien zuerst auf CEDATA GPGE – Patientenregister.

Front Med (Lausanne). 2021; 8: 666190. 

Published online 2021 May 24. doi: 10.3389/fmed.2021.666190

PMCID: PMC8180568

PMID: 34109197

Abstract

Introduction: The rising incidence of pediatric inflammatory bowel diseases (PIBD) facilitates the need for new methods of improving diagnosis latency, quality of care and documentation. Machine learning models have shown to be applicable to classifying PIBD when using histological data or extensive serology. This study aims to evaluate the performance of algorithms based on promptly available data more suited to clinical applications. 

Methods: Data of inflammatory locations of the bowels from initial and follow-up visitations is extracted from the CEDATA-GPGE registry and two follow-up sets are split off containing only input from 2017 and 2018. Pre-processing excludes patients in remission and encodes the categorical data numerically. For classification of PIBD diagnosis, a support vector machine (SVM), a random forest algorithm (RF), extreme gradient boosting (XGBoost), a dense neural network (DNN) and a convolutional neural network (CNN) are employed. As best performer, a convolutional neural network is further improved using grid optimization. 

Results: The achieved accuracy of the optimized neural network reaches up to 90.57% on data inserted into the registry in 2018. Less performant methods reach 88.78% for the DNN down to 83.94% for the XGBoost. The accuracy of prediction for the 2018 follow-up dataset is higher than those for older datasets. Neural networks yield a higher standard deviation with 3.45 for the CNN compared to 0.83-0.86 of the support vector machine and ensemble methods. 

Discussion: The displayed accuracy of the convolutional neural network proofs the viability of machine learning classification in PIBD diagnostics using only timely available data.

Keywords: CEDATA-GPGE registry; convolutional neural network; diagnostic assistance; machine learning; pediatric inflammatory bowel disease.

Der Beitrag Machine Learning Classification of Inflammatory Bowel Disease in Children Based on a Large Real-World Pediatric Cohort CEDATA-GPGE Registry erschien zuerst auf CEDATA GPGE – Patientenregister.