Development and Internal Validation of a Predictive Model for Prognostic Response to Proton Pump Inhibitors in Adults with Functional Dyspepsia

Abstract

Introduction. The response to proton pump inhibitor treatment in functional dyspepsia varies among patients. A decision tool that identifies the probability of response based on clinical data could optimize management in these patients. 

Objective. To develop and internally validate an easy-to-use predictive model that enables clinicians to establish the probability of response to proton pump inhibitors in patients with functional dyspepsia. 

Material and methods. We conducted a cross-sectional study on adults with functional dyspepsia. A pilot study identified variables related to response to proton pump inhibitors. Internal validation was performed by split-sample method, with 70% for model development and 30% for validation. We developed the model using multivariable logistic regression, and selected variables based on the best subset method. The final model was subsequently transformed into a scoring rule. Predictive ability was assessed using the area under the curve (AUC) and Hosmer-Lemeshow statistic. 

Results. A total of 192 patients were included. The best subset identified 4 variables with the best performance. The logistic model showed good calibration and discrimination with an AUC of 0.89 and a Hosmer-Lemeshow chi-square test with p = 1. The scoring rule assigned the following values: anxiety symptoms (3 points), fibromyalgia (2 points), depression (1 point), and female sex (− 1 point). A total score of 4 points or less was associated with a good response to proton pump inhibitor treatment, with a sensitivity of 96.5% and specificity of 56.6%. 

Conclusions. We present a novel tool that facilitates the early identification of patients with a better therapeutic response to the management of functional dyspepsia.