Why do multicenter randomized controlled trials not confirm the positive findings of single center randomized controlled trials in acute care?

Abstract

It is a common observation that many multicenter randomized controlled trials (mRCT) performed in critically ill patients do not achieve the positive findings often seen in single center studies (sRCT). This has, of course, relevant consequences for clinical practice, as mRCTs have higher scientific validity compared to sRCTs. The aim of this manuscript was to review and discuss the several potential causes of this phenomenon and to relate them to the future of mRCTs in critical care medicine. Overall, this seems to recall the old mythologic story of Achilles and the tortoise: although mRCTs (i.e. Achilles) are much more powerful, indeed, they always arrive later in time compared to the sRCTs (i.e. the tortoise) from which they were powered. However, sRCTs are more prone to several bias compared to mRCTs, such as local effect bias, selection and performance bias, detection and reporting bias, analysis and attrition bias, concomitant therapy bias, low fragility index and publication bias. In this sense, it is high time the critical care community should see the positive findings of sRTCs with a very high level of scientific caution, unless they are confirmed by mRCTs. MRCTs represent the final step of the process of evidence-based medicine and in the end (however slowly and painfully) such evidence catches up with sRCT and truly helps changes practice worldwide.