Novel trial method to address an absence of evidence in stress ulcer prophylaxis

The cluster randomised, crossover, registry-embedded clinical trial of proton pump inhibitors versus histamine-2 receptor blockers for ulcer prophylaxis therapy in the intensive care unit (PEPTIC study) which was posted in JAMA this month, evaluated the question of using histamine receptor blockers (ranitidine), versus using a proton pump inhibitor (PPI) to prevent gastrointestinal bleeding from stress ulceration of stomach and duodenum. Either of these agents are used in almost all critically ill patients - approximately 5 million globally per year - to prevent occurrence of this condition.

This study had the largest sample size and most statistical power ever conducted in critical care, at around 27,000 patients. Despite this, the trial cost was comparatively small at approximately $0.5M, thanks to the study design, which used randomisation of  ICUs (rather than individual patients) and registry data. An ICU trial would normally be considered ‘big’ with 4,000-8,000 participants, and would likely cost between $5M-$8M.

Ranitidine is an older drug and used originally for this purpose. Over the last two decades there was increase use of PPIs, as an alternative to ranitidine. However, this switch occurred without any evidence derived from relevant clinical trials.  PPIs have become much cheaper, but for much of the last twenty years, PPIs were substantially more expensive than ranitidine, resulting in substantial cost to the Australian Healthcare system.

The trial result exclude any mortality benefit from PPIs. Although not definitive, it is likely that use of ranitidine results in lower mortality—with the upmost benefit being a 10% relative reduction in mortality.

The trial illustrates the cost to patients - and the healthcare system - of the absence of evidence and the value of comparative effectiveness trials. This paper also highlights a novel and innovative trial method which emphasises that high quality trial evidence can be generated at low cost, particularly where there is existing registry data collection.