January 16, 2020

Predicting presence of Barrett's

Predicting presence of Barrett's

Strengths of this risk prediction model for Barrett's includes the relatively large number of cases and an external validation dataset. A limitation is that only those with symtomatic GERD (reflux) were included, so this is not necessarily generalizable to the general adult population. The authors observed that age, sex, cigarette smoking, waist circumference, frequency of stomach pain, duration of heartburn and acidic taste, and taking antireflux medication were predictive with an AUC of 0.81 in validation set.

Lancet Digital Health. Volume 2, ISSUE 1, Pe37-e48, January 01, 2020. Published online December 5, 2019 https://doi.org/10.1016/S2589-7500(19)30216-X

Development and validation of a risk prediction model to diagnose Barrett’s oesophagus (MARK-BE): a case-control machine learning approach.

Rosenfeld A, Graham DG, Jevons S, et al. on behalf of the BEST2 study group

Abstract

Background: Screening  for  Barrett’s  oesophagus  relies  on  endoscopy,  which  is invasive  and  few  who  undergo  the  procedure are found to have the condition. We aimed to use machine learning techniques to develop and externally validate a simple risk prediction panel to screen individuals for Barrett’s oesophagus.

Methods:  In  this  prospective  study,  machine  learning  risk  prediction  in Barrett’s  oesophagus  (MARK-BE),  we  used  data from two case-control studies, BEST2 and BOOST, to compile training and validation datasets. From the BEST2 study,  we  analysed  questionnaires  from  1299  patients,  of  whom  880  (67·7%) had  Barrett’s  oesophagus,  including  40  with  invasive  oesophageal adenocarcinoma,  and  419  (32·3%)  were  controls.  We  randomly  split  (6:4)  the cohort  using a computer algorithm into a training dataset of 776 patients and a testing dataset of 523 patients. We compiled an  external  validation  cohort  from the  BOOST  study,  which  included  398  patients,  comprising  198  patients  with Barrett’s oesophagus (23 with oesophageal adenocarcinoma) and 200 controls. We identified independently important diagnostic features of Barrett’s oesophagus using the machine learning techniques information gain and correlation-based feature  selection.  We  assessed  multiple  classification  tools  to  create  a multivariable  risk  prediction  model.  Internal validation of the model using the BEST2 testing dataset was followed by external validation using the BOOST external validation dataset. From these data we created a prediction panel to identify at-risk individuals.

Findings:  The BEST2 study included 40 diagnostic features. Of these, 19 added information gain but after correlation-based feature selection only eight showed independent diagnostic value including age, sex, cigarette smoking, waist circumference, frequency of stomach pain, duration of heartburn and acidic taste, and taking antireflux medication, of which all were associated with increased risk of Barrett’s oesophagus, except frequency of stomach pain, with was inversely associated  in  a  case-control  population.Logistic  regression  offered  the  highest prediction  quality  with  an  area under the receiver-operator curve (AUC) of 0·87 (95% CI 0·84–0·90; sensitivity set at 90%; specificity of 68%). In the testing dataset, AUC was 0·86 (0·83–0·89; sensitivity set at 90%; specificity of 65%). In the external validation dataset, the AUC was 0·81 (0·74–0·84; sensitivity set at 90%; specificity of 58%).

Interpretation:  Our  diagnostic  model  offers  valid  predictions  of  diagnosis  of Barrett’s  oesophagus  in  patients  with  symptomatic  gastro-oesophageal  reflux disease,  assisting  in  identifying  who  should  go  forward  to  invasive confirmatory testing. Our predictive panel suggests thatoverweight men who have been taking antireflux medication for a long time might merit particular consideration for further testing. Our risk prediction panel is quick and simple to administer but will need further calibration and validation in a prospective study in primary care.