Background-Although endoscopic injection therapy is effective in contr
olling initial haemorrhage from peptic ulcer, between 10% to 30% of pa
tients suffer rebleeding. Aim-To assess the factors that may predict t
he failure of endoscopic injection in patients bleeding from high risk
gastric ulcer. Subjects-One hundred and seventy eight patients admitt
ed for a gastric ulcer with a bleeding or a non-bleeding visible vesse
l were included. Methods-Patients received endoscopic therapy by injec
tion for adrenaline and polidocanol. Twelve clinical and endoscopic va
riables were entered into a multivariate logistic regression model to
ascertain their significance as predictive factor of therapeutic failu
re. Results-Eighty seven per cent (155 of 178) of patients had no furt
her bleeding after endoscopic therapy. Endoscopic injection failed in
23 (13%) patients: 20 (12%) continued to bleed or rebleed, and three (
1%) patients could not be treated because of inaccessibility of the le
sion. Logistic regression analysis showed that therapeutic failure was
significantly related to: (1) the presence of hypovolaemic shock (p=0
.09, OR 2.38, 95% CI: 0.86, 6.56), (2) the presence of active bleeding
at endoscopy (p=0.02, OR 2.98, 95% CI: 1.12, 7.91), (3) ulcer locatio
n high on the lesser curvature (p=0.04, OR 2.79, 95% CI: 1.01, 7.69),
and (4) ulcer size larger than 2 cm (p=0.01, OR 3.64, 95% CI: 1.34, 9.
89). Conclusion-These variables may enable identification of those pat
ients bleeding from gastric ulcer who would not benefit from injection
therapy.