Exclusive enteral nutrition in adult-onset Crohn’s disease, big-data and machine learning to predict mortality in C. Diff, and a structured approach to management of non-coeliac enteropathies

In this month’s blog we will focus on three articles, including two state of the art reviews and a highly innovative and important study involving artificial intelligence. To aid with the understanding of this article we highlight a recently published review in our sister journal, Frontline Gastroenterology, detailing the potential uses of machine learning in IBD.

Diagnosis and management of seronegative (most commonly TTG negative) enteropathies is challenging due to the lack of standardisation, poor characterisation and rarity. In their article, Schiepatti et al detail enteropathies with villous atrophy but negative coeliac serology in adults and discuss a structured approach to clinical care of these patients, including a pragmatic guide for investigation and clinical management in adults based on the available literature. Seronegative coeliac disease is the most frequently seen condition, although the patient group and range of conditions is highly heterogenous. Correct identification and targeted management of seronegative enteropathies is mandatory because of the variation in terms of clinical outcomes and prognosis.

Exclusive enteral nutrition is a standard part of care in paediatric Crohn’s disease. The use of this induction therapy in adults has long been more complex, but the evidence underpinning it’s use in adults has always been more controversial. Mitrev et al identified 79 articles in their review. Importantly, EEN in adult patients been shown to improve clinical, biomarker, endoscopic and radiologic measures of disease activity, whilst avoiding steroid side-effects and replenishing nutritional deficits. Additional benefits include preoperative EEN reducing postoperative complications and recurrence, alongside benefits in combing EEN with antitumour necrosis factor agents. The authors conclude that despite potential limitations in evidence EEN should be considered in treating adults with CD.

In their article on prediction of in-hospital mortality of C. difficile infection using a critical care database, the authors use a big data-driven, machine learning approach to identify factors associated with poor outcomes. The paper uses three machine learning models: logistic regression (LR), random forest (RF) and gradient boosting machine (GBM) in 61532 intensive care stays and 1315 infections. Mortality was 18.33%. AUROC was 0.69 (95% CI, 0.60 to 0.76) for LR, 0.71 (95% CI, 0.62 to 0.77) for RF and 0.72 (95% CI, 0.64 to 0.78) for GBM. Albumin, lactate and bicarbonate were significant mortality factors for all models, whilst calcium, potassium, white blood cell count, urea, platelet and mean blood pressure were present in at least 2/3 models. These data demonstrate the potential utility of machine learning to aid clinicians in identification of at-risk patients.

Brooks-Warburton and the BSG artificial intelligence IBD task force present a narrative review in Frontline Gastroenterology, looking at artificial intelligence and inflammatory bowel disease, including practicalities and future prospects. The authors discuss the science and mathematics underpinning machine learning and AI, alongside how this impacts management currently and how this may impact prediction and personalised therapy for IBD in the future. This paper is an ideal reading companion to better understand the methods used in the previous manuscript, or is a great read by itself.

(Visited 273 times, 1 visits today)