An event rate of 2% vs 1% would be going from “very unlikely” to “trivially less likely”. And event rate of 93% Vs 92% is going from “very likely” to “trivially less likely”. The similar 1% ARR describes the similar “trivially less likely” nature of both of the purportedly superior treatments. I don’t see any difficulty in parsing that in either instance. It would be far more disingenuous in my book to report the “50% RRR” in the first instance.
Any trial result requires the characterization of the “average pt”. Any application of trial data requires extrapolating the results observed from an average pt (also under ideal and precise trial environments) onto Mr. Smith or Mrs. Jones, who lives in the messy real world. Any data you apply from any study onto any individual pt will espouse this uncertainty, regardless of what metric you use to characterize effect size.
I have to have a friendly disagreement about that. I think there is a chance that NNT is misleading even under ideal situations such as primary prevention. That's because people with mild risk factors may have NNT that is 10 fold greater than those with serious risk factors. NNT uses averages and may not apply to anyone. Primary prevention is best served by focusing on those with more risk, in some cases.
Please don't include NNT or NNH. These are harmful to thinking: https://discourse.datamethods.org/t/problems-with-nnt
An event rate of 2% vs 1% would be going from “very unlikely” to “trivially less likely”. And event rate of 93% Vs 92% is going from “very likely” to “trivially less likely”. The similar 1% ARR describes the similar “trivially less likely” nature of both of the purportedly superior treatments. I don’t see any difficulty in parsing that in either instance. It would be far more disingenuous in my book to report the “50% RRR” in the first instance.
Any trial result requires the characterization of the “average pt”. Any application of trial data requires extrapolating the results observed from an average pt (also under ideal and precise trial environments) onto Mr. Smith or Mrs. Jones, who lives in the messy real world. Any data you apply from any study onto any individual pt will espouse this uncertainty, regardless of what metric you use to characterize effect size.
I have to have a friendly disagreement about that. I think there is a chance that NNT is misleading even under ideal situations such as primary prevention. That's because people with mild risk factors may have NNT that is 10 fold greater than those with serious risk factors. NNT uses averages and may not apply to anyone. Primary prevention is best served by focusing on those with more risk, in some cases.