Prediction of major pathologic conditions in dyspeptic patients referred for endoscopy. A prospective validation...

This study aimed to validate the use of a decision support system previously developed on answers to a structured interview of dyspeptic outpatients and designed to identify patients at low risk of organic dyspepsia. We evaluated the performance of the scoring system in two cohorts of dyspeptic outpatients: 878 consecutive Danish patients (study group) referred for upper endoscopy and 1279 British patients whose results had previously been reported (validation group). Performance of the scoring system was analysed by receiver-operating characteristic (ROC) curves and comparison of loss in detection rate of organic dyspepsia, defined as cancer, ulcer, and complicated oesophagitis. The performance of the scoring system in the study group was less favourable over the entire span of cut-off points evaluated. This was reflected by a significant decrease in area under the ROC curve (65.1% versus 75.0%). The only cut-off point with an acceptable loss in detection rate (3.1%) led to a reduction in endoscopic activity of only 7.5%. Despite optimal working conditions the scoring system could not be used as a safe method of extracting dyspeptic patients with a low probability of organic dyspepsia. Before adopting a predictive scoring system, clinicians must evaluate its applicability in their own setting.