Interesting article in December 2019 Circulation.(See here.) that discusses the "robustness" of a number of cardiovascular disease trials.
A purported measure of robustness is the Fragility Index (FI) which is the number needed to move from the non-event group to the event group to turn a statistically significant finding into a non-significant one. A FI of 1 would mean that if one subject was moved from the event to the non event group there would no longer be a significant difference.
A related concept is the comparison of the FI to the number of subjects lost to follow up. The trial's results would be considered less robust if the number lost is greater than the FI.
Ridgeon et published a review of 56 RCTs in critical care treatment topics.The median FI was 2 and slightly greater that 40% of trials had a FI of one or less.So does that suggest that clinical guidelines based entirely or in part on those low FI studies are built on shifting sands?
But wait, listen to this criticism of the concept of FI by Dr. F.Perry Wilson before rushing to trash a RCT simply because the FI is too low. Wilson shifts the problem to one of having too much reliance on the p value in the first place.
1)Ridgean EE et al Crit. Care Med 2016,vol 44 ,1278
No comments:
Post a Comment