Discriminant factors of gastric ulcer healing by colloidal bismuth

One hundred and twelve patients with 49 corpus and 35 prepyloric gastric ulcers and 28 duodenal ulcers associated with gastric ulcers were entered into a stratified, controlled double-blind randomized study comparing the healing efficacy of colloidal bismuth suspension with placebo. Healing of corpus and prepyloric ulcers, but not duodenal-ulcer associated gastric ulcers, was significantly better with the bismuth compound than with placebo. The clinical, personal, ulcer and acid secretory data of these patients were prospectively collected and evaluated by stepwise discriminant analysis in two phases. In the first phase, the collected data of 91 consecutive patients were entered into computer analysis. The derived discriminant function classified patients into healed and unhealed categories with a sensitivity of 82%, specificity of 77%, positive diagnostic value of 73%, negative diagnostic value of 85% and diagnostic efficiency of 79%. In phase two, the discriminant function from phase 1 was applied prospectively to 21 patients to predict the outcome of treatment. Five out of five healed and 11/16 unhealed gastric ulcers were correctly predicted, giving a diagnostic efficiency of 76%. The present study shows that healing or non-healing of gastric ulcer can be predicted with reasonable accuracy by discriminant analysis. In addition, discriminant analysis identified, aside from the efficacious drug, ulcer size, young age, co-existing duodenal ulcer, and concomitant medical condition as adverse factors for gastric ulcer healing.