Development and Validation of a Nomogram for Predicting Surgery in Newly-diagnosed Crohn's Disease: a Retrospective Cohort Study

Overview

A majority of patients with Crohn's disease undergo surgery during the disease course. We aimed to develop an easily available nomogram to predict the risk of surgery at diagnosis.

Crohn's disease (CD), a chronic inflammatory disorder involving all of the gastrointestinal tract, has a progressive and destructive course and is increasing in incidence worldwide.1 Although the most common disease behavior of patients with newly-diagnosed CD is inflammatory (B1 in the Montreal Classification2), a rapid and prominent progression in disease behaviour will be observed in approximate half of the patients within 10 years after diagnosis.3-5 Data from a population-based cohort show that nearly half of patients developed intestinal complications such as strictures and fistulae, in the 20 years following the diagnosis.3 In spite of the application of immunosuppressive maintenance therapies, more than half of the patients suffer from severe complications and required intestinal resection.6,7 In recent decades, with the advent of targeted biologic therapies such as tumor necrosis factor antagonists, gut-selective monoclonal anti-integrin antibody and inhibitors of IL-12 and IL-23 signaling, the medical management of CD has been revolutionized.8 Earlier and more aggressive application of biologics or novel small molecules and combination therapies have been demonstrated to induce a profound alteration of natural disease course and diminish the requirement for hospitalization and surgery among patients with newly-diagnosed CD.9,10 Nevertheless, one of the most difficult challenges in the so-called top-down treatment strategy is the identification of patients who are at high risk for disease progression and thus necessitate more intensive treatment pattern despite the therapy-related adverse events and heavy costs. From another perspective, failure to identify high-risk patients also delays the prescription of more effective therapies and accounts for an increase in the risk of disease progression. Much effort has been made in the field of baseline risk stratification for newly-diagnosed CD. Many clinical characteristics have been found to independently correlate with prognosis, including age at diagnosis, disease location, disease behavior, smoking status, and history of medication.9,11,12 Meanwhile, several prognostic biomarkers have been discovered in pilot studies, encompassing immune-related molecules and specific gene expression levels.13,14 Nonetheless, inconvenience and high expense has impeded their full validation and clinical application. Accordingly, the therapy selection is still tailored to the individual patient newly diagnosed with CD based on the clinical risk factors and patient comorbidities8, which is far from precision treatment. In this era of artificial intelligence, a lot of machine learning models have been developed for innovation in all fields of inflammatory bowel disease, such as diagnosis, monitoring, disease course prediction and management.15 Unfortunately, the majority of popular machine learning prediction models are essentially black boxes, rendering verdicts with a few accompanying justifications, which limits clinical reliability and hence obstructs clinical implementation.16 To balance effectiveness with convenience and interpretability, we aimed to construct a well-interpreted Cox statistical regression model together with a nomogram based on clinical characteristics and available serological indicators to predict the long-term prognosis of newly diagnosed CD.

Conditions

Interventions

Eligibility

Outcomes