How much ulcer is ulcer-like? Diagnostic determinants of peptic ulcer in open access gastroscopy

A cross-sectional diagnostic prevalence study was carried out within a multi-centre experiment with open access gastroscopy in Utrecht, The Netherlands. The objective of the study was to contribute to improvement of patient selection for open access gastroscopy and to evaluate diagnostic determinants for peptic ulcer. Data were analysed in all 861 patients who were consecutively newly referred during the experiment to undergo gastrocopy. Patient characteristics and outcomes of gastroscopies were recorded. Univariate and multivariate (logistic) analyses were carried out and the results were evaluated with ROC (receiver operating characteristic) analysis. The most important clinical characteristics to be used for prediction of peptic ulcer are pain on an empty stomach, absence of pain after a meal and absence of obstructive complaints. The scoring list derived from the full model, comprising these characteristics together with age, sex, information on former dyspeptic diseases, medication and smoking ('basic characteristics'), predicted peptic ulcer with an AUC of 0.78. The 'ulcer-like' model, with characteristics known from the literature, had an AUC of 0.76. The amount of gastroscopy requests on patients suspected of a peptic ulcer could have been reduced from 60 to 44%. Forcing 'basic characteristics' into scoring lists on peptic ulcer improved the pre-diagnostic test capacities. The presented scoring list may improve gastroscopy requesting by GPs aiming at finding peptic ulcers. Practical manageability of the list should be prospectively evaluated in future experiments.