OBJECTIVE: Alarm symptoms are present in 10% of patients consulting a general practitioner (GP) because of dyspepsia. The options for the GP are: perform "own" investigation, refer to a specialist or secondary care, or maintain watchful waiting. DESIGN: Postal questionnaire to GPs obtaining information from the records of patients who had consulted earlier owing to dyspepsia and reported one or more alarm symptoms at the consultation. SETTING: 92 general practices throughout Denmark. SUBJECTS: In 7240 consecutive consultations in 1991-1992 a total of 749 patients reported 881 alarm symptoms. Follow-up data 3 years later were obtained from 81%. MAIN OUTCOME MEASURES: The management of alarm symptoms by GPs: "own" investigation, referral to specialist, or expectance; procedures used in "own" investigations, patient characteristics related to such decisions, and gastrointestinal diseases diagnosed by this management. RESULTS: 67% of the patients with alarm symptoms were investigated promptly (8% by referral to specialists) and 33% were left uninvestigated, with great variations between alarm symptoms. The single most preferred GP investigation was endoscopy. A gastrointestinal diagnosis was obtained in 35% of the patients; 3% of these were malignancies while 9% were peptic ulcers. CONCLUSION: The decision on whether and how to investigate alarm symptoms is complex and relates to specific symptom, type of dyspepsia, age, gender and dwelling of the patient.