Objective: To derive and evaluate diagnostic determinants of gastro-oesophageal malignancy as a means of helping family physicians select patients for 'open-access' gastroscopy. In this multicenter study, 861 consecutive patients were investigated with first-time gastroscopy (study population). Another 1153 patients were studied during the next 6 years (validation population). General practitioners registered symptoms relevant to malignancy and patient histories; the results of gastroscopies were received from the endoscopists.
Methods: Univariate and multivariate analyses resulted in four relevant symptoms that were then compared with 'alarm symptoms' previously published in the literature. Receiver-operating characteristic analysis was used to evaluate the probability of finding malignancy using these two sets of symptoms.
Results: Positive answers regarding the symptoms, weight loss and dysphagia, together with negative answers on pain during the night and heartburn, predicted malignancy in the study population with an AUC (area under the curve) of 0.90. 'Alarm symptoms' performed less well in the study population (AUC 0.85), although reproducibility was better in the validation population (0.71 versus 0.63). If exclusion of malignancy had been chosen as the only valid reason for requesting gastroscopy, then pre-selection with the help of these symptoms would have reduced the number of requests by 10%.
Conclusion: Thorough evaluation of 'classical' alarm symptoms in dyspeptic patients might help minimize unnecessary gastroscopy requests by GPs seeking a safe treatment decision.