VALIDATION OF RISK ASSESSMENT TOOLS FOR THE DETECTION OF BARRETT’S ESOPHAGUS IN PATIENTS WITHOUT CHRONIC REFLUX SYMPTOMS

Background:
Esophageal adenocarcinoma (EAC) has had a rising incidence in the US, prompting guidelines for endoscopic screening in those with risk factors for its precursor Barret’s esophagus (BE). Various risk assessment tools have been developed and validated in the general population to identify patients at high risk for BE. However, these tools have not been validated in patients without chronic reflux symptoms, a notable gap in care as up to 25% of patients with BE and 40% of patients with EAC are asymptomatic. Thus, we sought to evaluate two validated tools for risk of progression to EAC in patients without GERD symptoms.

Methods:
We conducted a retrospective case-control study to assess the accuracy of two previously validated risk tools: Kunzmann and Nord-Trøndelag Health Study (HUNT). Cases were patients with known BE-associated neoplasia who were obtained from the Treatment with Resection and Endoscopic Ablation Techniques for Barrett’s Esophagus (TREAT-BE) Consortium. These were matched with controls, defined as all patients between 2015-2023 at our academic tertiary health center who had undergone an endoscopy with neither GERD symptoms nor BE/EAC. Risk scores were calculated for the tools for each patient and compared to endoscopy/pathology reports as a reference standard. Receiver Operating Characteristics (ROC) curves were fit and compared using Chi Square tests. Areas under the ROC curve (AuROC) were calculated for each tool, and Youden’s Index was calculated to determine the score for each tool that had the highest sensitivity and specificity.

Results:
As expected, patients with BE in our study were on average older, more often male, and had higher Kunzmann and HUNT scores than those without. Both the Kunzmann and HUNT tools were significantly predictive of BE status in patients without reflux symptoms, with AuROCs of 0.727 (95% CI: 0.698, 0.756) and 0.679 (0.649, 0.709) respectively. The Kunzmann tool was more predictive than the HUNT tool in this population (difference in AuROC=0.048, p<0.001). With a score cut-off of 6.5, the Kunzmann tool is 73.7% sensitive and 59.9% specific for detecting BE. With a score cut-off of 21.20, the HUNT tool is 58.4% sensitive and 72% specific. The positive predictive values were low for both the Kunzmann and Hunt tools (10.8% and 12.1%, respectively) given the low rate of BE in our study. However, both tools had excellent negative predictive values (97.2% and 96.3%, respectively).

Conclusion:
The Kunzmann and HUNT tools had reasonable sensitivities for predicting BE in patients without reflux symptoms and as such might be used in clinical practice to identify which high-risk patients without chronic GERD symptoms would most benefit from screening.