Model to Select On-Therapy vs Off-Therapy Tests for Patients With Refractory Esophageal or Extraesophageal Symptoms

Background & Aims It is not clear whether we should test for reflux in patients with refractory heartburn or extraesophageal reflux (EER) symptoms, such as cough, hoarseness, or asthma. Guidelines recommend testing patients by pH monitoring when they are on or off acid-suppressive therapies based on pretest probability of reflux, determined by expert consensus. However, it is not clear what constitutes a low or high pretest probability of reflux in these patients. We aimed to develop a model that clinicians can use at bedside to estimate pretest probability of abnormal reflux. Methods We performed a prospective study of 471 adult patients with refractory heartburn (n = 214) or suspected EER symptoms (n = 257) who underwent endoscopy with wireless pH monitoring while they were off acid-suppressive treatment and assigned them to groups based on symptoms at presentation (discovery cohort). Using data from the discovery cohort, we performed proportional odds ordinal logistic regression to select factors (easy to obtain demographic criteria and clinical symptoms such as heartburn, regurgitation, asthma, cough, and hoarseness) associated with esophageal exposure to acid. We validated our findings in a cohort of 118 patients with the same features from 2 separate tertiary care centers (62% women; median age 59 years; 62% with cough as presenting symptom). Results Abnormal pH (>5.5% of time spent at pH 25 kg/m2. In patients with EER, the risk of reflux was independently associated with the presence of concomitant heartburn (odds ratio [OR] 2.0; 95% confidence interval [CI] 1.3–3.1), body mass index >25 kg/m2 (OR 2.1; 95% CI 1.5–3.1), asthma (OR 2.0; 95% CI 1.2–3.5), and presence of hiatal hernia (OR 1.9; 95% CI 1.2–3.1). When we used these factors to create a scoring system, we found that a score of ≤2 excluded patients with moderate to severe reflux, with a negative predictive value of 80% in the discovery cohort and a negative predictive value of 85% in the validation cohort. Conclusion We developed a clinical model to estimate pretest probability of abnormal pH in patients who were failed by proton pump inhibitor therapy. This system can help guide clinicians at bedside in determining the most appropriate diagnostic test in this challenging group of patients.