Decision analysis in elective operation for duodenal ulcer

A computerized data base with details of clinical history and examination as well as the results of a pentagastrin gastric acid stimulation test was established for 319 patients with radiological duodenal ulcers in the early 1970s. The patients were followed up during the next 10 years and divided into those who were, and were not, operated on. Originally 195 were operated on, as were a further 28 in the next 10 years up to the end of 1983. Multivariate analysis was done on the 1980 division into 218 who had, and 101 who did not have, an operation. Using all variables the most accurate prediction about outcome could be made in 84%. Six variables (length of history, night waking, weeks of illness away from work, length of attack, length of remission and maximum acid output after pentagastrin) gave a 70% correct prediction as to whether operation would be advised. With two variables (years of history and number of nights awoken each week by indigestion) 60% were allotted to their correct operative or non-operative subdivision. It is suggested that future operative trials for duodenal ulcer give details of at least these two factors for each patient.