Bleeding peptic ulcers.

Sir, ?Professor Stephen Bown is correct to remark on the recent advances in endoscopie treatment of bleeding peptic ulcers.1 With the emerging evidence of the efficacy of endoscopie techniques, and with an aging population, who tolerate laparotomy badly, these techniques should be more widely available. Professor Bown discusses the evidence of their efficacy in a traditional narrative manner, and I am confident that his con? clusions are correct. He has not, however, explored the results of three relevant meta-analyses.24 Two of these studies were reviewed at a workshop on meta-analysis in gastrointestinal haemorrhage held in Newcastle, New South Wales, in September last year.' This was a satellite meeting of the World Congress of Gastroenterology. The table summarises the results presented at the workshop. Several conclusions can be drawn. Firstly, evidence of benefit seems to be greatest in the case of laser treatments, with both meta analyses showing significant (p<005) reductions in rates of rebleeding and mortality. In the case of thermal contact devices (heater probes with monopolar and multipolar electrodes) the evidence is convincing for rebleeding, but for mortality the confidence intervals for the pooled odds ratio include one. We came to similar conclusions in a meta-analysis published in 1987.J Injection treat? ment seems to be associated with low rebleeding rates, but the meta-analyses disagreed somewhat on the extent and significance of the reduction in mortality. It should be noted that the numbers of subjects included in these studies are small. The data presented in the table are probably out of date now, which underscores the need for continual revision of meta-analyses of important topics such as this. The pooled odds ratio is only an approximate guide to the true magnitude of the benefits from endoscopie treatment. Sacks et al reported pooled rate differences for all endoscopie interventions combined, which indicate that in the relatively high risk patients who entered the trials the reduction in rebleeding was around 0 27 and in mortality around 0T0.' This means that it may be necessary to treat only four patients to prevent a case of further haemorrhage or 10 patients to prevent a death. Sacks et al also concluded that the data are insufficient to make direct comparisons between treatment modalities, and presently this is the crucial point. The pressing need now is for randomised trials to compare the efficacy of different endoscopie treatments for haemostasis. Because of their convenience and feasibility thermal contact devices and injection treatment should probably be the modalities that are com? pared in future trials. These trials will have to be large. To date, the controlled trials of endoscopie haemostasis have been too small, with total numbers of subjects ranging from 16 to 174.' Numbers in the thousands will be needed,6 and gastroenterologists need to look to the example of their colleagues in cardiovascular medicine, who have set new stan? dards in international multicentre trials when