Pediatric patients who present with symptoms of gastroesophageal reflux and severe eosinophilic esophagitis may be unresponsive to aggressive anti-reflux medical therapy. In order to determine whether the degree of eosinophilia predicts anti-reflux treatment response and possibly distinguishes different etiologies, we reviewed the initial biopsies of patients with esophageal eosinophilia and compared the number of eosinophils with the response to anti-reflux treatment. Over a 1-year period, 102 patients with a biopsy demonstrating at least 1 intraepithelial eosinophil were identified among patients undergoing initial endoscopy for symptoms of reflux. All patients were treated with H2 blockers and prokinetic agents. Treatment response was classified into three categories: improvement, relapse, and failure. There were significant differences between the group who improved (mean eosinophil count [MEC] 1.1 +/- 0.3 SEM) and those who failed (24.5 +/- 6.1 SEM, P < 0.0025) or relapsed 6.4 +/- 2.4 SEM, P < 0.05). A threshold MEC value of > or = 7 provided a sensitivity of 61.3%, a specificity of 95.7%, and a predictive value for treatment failure of 86.1. A MEC value of < 7 provided an 85% predictive value of successful therapy. From these data we made the following conclusions: (1) The number of eosinophils has a predictive value of treatment response with > or = 7 per high power field offering a valuable clinical threshold for predicting outcome of conventional therapy. (2) The variable response to conventional reflux treatment may reflect different etiologies. (3) Alternate medical treatment modalities may be appropriate in the presence of severe eosinophilia, before considering surgical intervention.