Predicting Outcomes in Children with Ulcerative Colitis

摘要溃疡性结肠炎(UC)是一种慢性肠道疾病，是fl肠炎的一种。 (IBD)。UC会导致腹痛、腹泻、出血和体重减轻。每年大约有38,000个新的 UC患者中约25-30%为儿童。UC在儿童中通常比成人更严重。没有 管理所有UC患者的标准方案。UC的治疗包括5-氨基水杨酸(5-ASA)， 皮质类固醇、免疫调节剂、生物制剂和切除结肠的手术。最理想的是一种个性化的药物 应根据患者的病情严重程度和预测，采取因人而异的方法进行治疗 回应。然而，目前缺乏对临床医生的指导，即哪些儿童将表现良好 使用这些药物以及谁需要更多的药物暴露或手术。为了解决这一重要问题， 启动规范医疗治疗临床试验，纳入428名新诊断为UC的儿童 北美的29个儿科医疗中心(预测标准化儿科结肠炎治疗的反应： 保护研究)。保护研究收集了临床、遗传、环境和免疫因素 包括血液、粪便和结肠组织在内的生物标本。据预计，临床上， 在诊断时进行的遗传和免疫学测试可以为个性化诊断构建有价值的预测模型 可用于改善临床结果的药物，例如仅5-ASA的早期和晚期缓解 如果不同时使用类固醇药物。沿着这条线，最先进的机器学习算法可以 最终利用PROTECT研究数据，更准确地预测临床结果。发展的能力 使用基本的临床、内窥镜、组织学和实验室数据的准确预测模型将是极其重要的 有助于临床医生以及患者/家属更好地了解决策过程。 基于ML的模型将使用外部电子健康记录数据进行验证。此外，还缺乏容易的- TO-使用临床决策支持工具帮助临床医生在fi领域做出决策。这项研究将fi所有的差距。

Summary/Abstract Ulcerative colitis (UC) is a chronic intestinal disorder, a type of inflammatory bowel disease (IBD). UC causes abdominal pain, diarrhea, bleeding and weight loss. Each year there are about 38,000 new UC cases with about 25-30% being children. UC is often more severe in children than adults. There is no standard regimen for managing all people with UC. Treatment of UC includes 5-aminosalicylic acid (5-ASA), corticosteroids, immunomodulators, biologics and surgery to remove the colon. Ideally a personalized medicine approach should be taken to tailor therapy to the individual patient based on their disease severity and predicted response. However, there is currently a lack of guidance to clinicians as to which children are going to do well with these medications and who needs more medication exposure or surgery. To address this important issue, a clinical trial of standardized medical therapy was launched enrolling 428 children newly diagnosed with UC at 29 pediatric medical centers in North America (Predicting Response to Standardized Pediatric Colitis Therapy: The PROTECT Study). The PROTECT study collected clinical, genetic, environment and immune factors along with biospecimens including blood, stool, and colonic tissue. It was anticipated that a combination of clinical, genetic, and immunologic tests performed at diagnosis can construct a valuable predictive model for personalized medicine which can be implemented to improve clinical outcomes such as early and late remission on 5-ASA only without the concurrent use of steroid medications. Along this line, state-of-art machine learning algorithms can ultimately utilize the PROTECT study data and more accurately predict clinical outcomes. The ability to develop accurate predictive models using basic clinical, endoscopic, histologic, and laboratory data would be extremely helpful to clinicians, as well as to patients/families in helping them better understand the decision making process. ML-based models will be validated with external electronic health records data. Further, there is a lack of easy- to-use clinical decision support tool to help clinicians make decisions in the field. This study will fill the gap.