Somatostatin v placebo in bleeding oesophageal varices: Randomised Trial and meta-analysis

Peter C. Gotzsche, Ida Gjorup, Helen Bonnen, Niels Erik Bille Brahe, Ulrik Becker, Flemming Burcharth

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

Abstrakt

Objective: To study whether somatostatin or its derivative octreotide is more effective than placebo for treating bleeding oesophageal varices. Methods: Randomised, double blind trial and meta-analysis with blinded analysis of data and writing of manuscripts. Setting: Departments of medical and surgical gastroenterology in Copenhagen. Subjects: Patients suspected of bleeding from oesophageal varices and of having cirrhosis of the liver. Main outcome measures: Survival, number of blood transfusions, and use of Sengstaken-Blakemore tube. Results: 86 patients were randomised; in each group 16 died within six weeks (95% confidence interval for difference in mortality —19% to 22%). There were no differences between those treated with somatostatin or placebo in median number of blood transfusions (8 v 5, P=0.07, 0 to 4 transfusions) or in numbers of patients who needed balloon tamponade (16 v 13, P=0.54, -11% to 28%). In a meta-analysis of three trials involving 290 patients somatostatin had no effect on survival compared with placebo (P=0.59, odds ratio 1.16; 0.67 to 2.01). For blood transfusions and use of balloon tamponade there was heterogeneity between the trials with no convincing evidence in favour of somatostatin. No placebo controlled trials have been performed with octreotide. Conclusion: Within the limited power of this study and meta-analysis we were unable to show a clinical benefit of somatostatin in the emergency treatment of bleeding oesophageal varices. Key messages This study failed to show a benefit of somatostatin Meta-analysis of the placebo controlled studies also failed to show a benefit Treatment with somatostatin or octreotide cannot be recommended at present Larger studies are needed to give a definitive answer.

OriginalsprogEngelsk
TidsskriftBMJ
Vol/bind310
Udgave nummer6993
Antal sider1
ISSN0959-8138
DOI
StatusUdgivet - 10. jun. 1995

    Fingerprint

Citationsformater