Development of a model to identify patients who do not need oesophageal biopsies when eosinophilic oesophagitis is suspected.

Background: Predictive models for eosinophilic oesophagitis (EoE) may not fully rule in the diagnosis.

Aim: To develop a reverse model that predicts against EoE to eliminate the need for oesophageal biopsies.

Methods: In this two-centre study, a predictive model was developed (Mayo Clinic) and validated (University of North Carolina [UNC]). Cross-sectional data from consecutive adult patients without prior EoE who underwent upper enoscopy with oesophageal biopsies were used. EoE cases had ≥15 eosinophils/high-power field while controls had no eosinophils. Data were collected on patient clinical and endoscopic features. Multiple variable logistic regression was used to identify predictors of non-EoE status while maintaining specificity ≥95%. A secondary model was developed to predict against the need for endoscopy in patients suspected of having EoE without alarm symptoms.

Results: The Mayo and UNC cohorts consisted of 345 (EoE = 94, non-EoE = 251) and 297 patients (EoE = 84, non-EoE = 213), respectively. A primary model based on clinical and endoscopic features predicted against EoE with c-statistic 0.92 (95% CI: 0.88-0.96), specificity 95%, and sensitivity 65%. This model was validated (UNC) with c-statistic 0.87 (95% CI: 0.82-0.92). A simplified scoring system was created and a threshold of ≥12 points excluded EoE with 95% specificity and 50% sensitivity. A secondary model based on clinical characteristics alone predicted against EoE with c-statistic 0.86 (95% CI: 0.82-0.90), specificity 95% and sensitivity 39% and validated (UNC) with c-statistic 0.78 (95% CI: 0.71-0.85).

Conclusion: A simplified scoring system accurately identified a group of patients with a low likelihood of EoE where unnecessary oesophageal biopsies can be avoided, potentially resulting in resource and cost savings.