This article is more than 5 years old. Information may no longer be current.
Prediction model may help identify candidates for Barrett’s esophagus screening
A novel risk prediction model showed promise as a tool to help clinicians determine which patients with gastroesophageal reflux symptoms should be considered for endoscopic screening for Barrett’s esophagus, according to study results.
Researchers in Australia compared data from 285 patients with incident nondysplastic Barrett’s esophagus with 313 patients who had esophageal inflammatory changes without Barrett’s esophagus. The latter group served as “inflammatory controls,” the researchers wrote.
The researchers used two phases of stepwise backward logistic regression to identify predictors of Barrett’s esophagus in men and women.
They first included all significant covariates from the univariate analyses. Then they used non-significant covariates from the univariate analyses, which allowed them to identify the effects that were detectable only after adjusting for other factors.
The researchers’ final model pooled the predictors.
Data from a Barrett’s esophagus study conducted in Washington State helped validate the model for discrimination and calibration.
The predictive factors observed included age, sex, smoking status, BMI, highest level of education and frequency of use of acid suppressant medications (area under the receiver operating characteristic [ROC] curve=0.70; 95%CI, 0.66-0.74).
A moderate discrimination was observed between the model and the external comparator data set (area under the ROC curve=0.61; 95% CI, 0.56-0.66).
Results of a Hosmer–Lemeshow test also indicated that the model was well calibrated (P=.75). Predicted probability and observed risk were highly correlated, the researchers wrote.
“The prediction model performed reasonably well and has the potential to be an effective and useful clinical tool in selecting patients with gastroesophageal reflux symptoms to refer for endoscopic screening for [Barrett’s] esophagus,” the researchers concluded.
Aaron Thrift, MD, of the Cancer Control Laboratory of the Population Health Department at the Queensland Institute of Medical Research, discussed the findings with HemOnc Today. “Predicting Barrett’s esophagus based solely on symptoms (e.g., frequency of reflux) and other characteristics and lifestyle exposures is potentially feasible, but our analyses show that we need to do more work in this area,” he said. “Getting the balance between sensitivity and specificity is the challenge, so that we are not missing too many cases. We also need to make sure that the tool has some clinical value in that it sufficiently reduces the number of unnecessary endoscopies performed for Barrett’s esophagus.”
Perspective
Back to Top
I read with interest the recent article published by Thrift and colleagues. As is now well known, esophageal adenocarcinoma has risen dramatically over the last 40 years. The precursor appears to be chronic acid reflux. Given the fact that most people who are diagnosed with esophageal adenocarcinoma do not survive, it is important to try to come up with strategies to help clinicians decide who is at risk for this potentially deadly disease.
The good news is that most patients with Barrett’s esophagus do not get cancer. There has been no consensus among experts or medical societies about who should be screened. However, the American Gastrointestinal Society, based on expert opinion, does recommend screening for Barrett’s esophagus in patients with accepted risk factors. These have included: Caucasian race, obesity, male sex, chronic gastroesphoageal reflux disease (GERD) symptoms, age of 50 years or older, and a hiatal hernia. When looking at the published literature on Barrett’s esophagus, the estimated frequency ranges from 0.9% to more than 20%.
An even bigger problem is that, in many studies, 40% to 50% of patients did not have any significant reflux symptoms. In a VA study, patients who were being screened for colon cancer had an upper endoscopy at the same time (none had GERD symptoms at the time) and 25% of the 110 volunteers had Barrett’s esophagus. The authors in this model study looked at many risk factors in men and women who had a diagnosis of Barrett’s esophagus vs. those who only had inflammatory changes in the esophagus. Of all the variables studied, the best predictors in both men and women were highest level of education and use of acid suppressant medications. Other predictors that met statistical significance in men included BMI and tobacco use. They suggest clinicians use these predictors to decide who should and should not get endoscopy.
Depending on the number of such predictors used, they could change the sensitivity and specificity of the prediction model. This is a very important topic. These models have worked for other disease processes. The problem with Barrett’s esophagus is that fact that many patients do not have GERD symptoms. There also is a considerable amount of sampling error when biopsies are obtained. Many pathologists disagree what is and what is not Barrett’s esophagus. There are even more discrepancies when we get into Barrett’s esophagus with dysplastic changes. It is possible that the control group of patients with “inflammation” could have had Barrett’s that was missed, changing the results.
Newly developed technologies that we have instituted in our Barrett’s center have been shown to markedly increase our diagnostic yield for this disease and/or dysplasia. Given the above facts, I think these issues must be clarified before moving to this prediction model alone for screening. I believe many patients may not be screened and, therefore, a percentage will end up with esophageal adenocarcinoma. This study is a great first step in trying to get a validated group of risk factors to help clinicians, but I suspect as genetic data becomes available and we get better at properly diagnosing Barrett’s and/or dysplasia, it will need to be modified. In the meantime, clinicians should add the highest level of education achieved and use of acid suppressant agents to their list of questions when evaluating patients. These questions, along with the other known risk factors mentioned above, will at least give clinicians a good starting point.