FromtheDepartments ofSurgery, Diagnostic Radiology, Pathology, andMedicine, Royal Postgraduate Medical School, Hammersmith Hospital, DucaneRoad,London

SUMMARY Theinvestigation andtreatmentof131patients with'obscure' gastrointestinal bleeding hasbeenreviewed. Onehundred andsixpatients were assessed electively forrecurrent haemorrhage, 25presented as emergencies. Themajorpresenting feature was melaena(55 patients), anaemia (35), rectal bleeding (34), haematemesis (six) andileostomy bleeding (one). Thelesions responsible forhaemorrhage were colonic angiodysplasia (52patients), small bowel vascular anomalies (16), Meckel's diverticula (nine), smallbowelsmoothmuscle tumours (seven), gastric vascular anomalies (four), chronic pancreatitis (three), colonic diverticular disease (three) and16other miscellaneous lesions. Nolesion was foundin21cases.Lesions were first shownbyvisceral angiography (69patients), atlaparotomy (23), on endoscopy (11), on gastrointestinal contrast radiological studies (four), andatERCP (three). Lesions whichwere undetectable atoperation increased markedly with age(p<0O0001). Expert visceral angiography isstrongly recommended before surgeryinpatients over45yearsofageandafter laparotomy whenno causehasbeenfound. Exploratory laparotomy isrecommended atan early stagefor younger patients, andforolderpatients after non-diagnostic angiography. Thecauseofgastrointestinal haemorrhage isidentifiedbyroutine investigations inabout95% of cases.2 Theremaining patients with'obscure' gastrointestinal bleeding arefrequently difficult to manageandoften undergo extensive investigation andevenlaparotomy without diagnosis. Theaimof this paperistoreview theoverall experience ofthe investigation ofobscure gastrointestinal bleeding at this hospital, inparticular todetermine themethod ofdiagnosis orlocalisation ofhaemorrhage andthe relative incidence ofdifferent underlying conditions. Methods PATIENTS