Nice article John. Once again, trial designers refused to use a full information endpoint (ordinal longitudinal endpoint) that would have (1) greater power and (2) been able to tell us on which treatment patients fared better.
Nice article John. Once again, trial designers refused to use a full information endpoint (ordinal longitudinal endpoint) that would have (1) greater power and (2) been able to tell us on which treatment patients fared better.
To clarify- from the article: The primary outcome was a composite of myocardial infarction or death from any cause up to 30 days after randomization." With the ordinal longitudinal endpoint are MI and death treated as 2 potential independent endpoints instead of as a composite?
The composite output you mentioned takes, as usual, 'composite' to mean 'union', ignoring different impacts/consequences of the different events. Ordinal endpoints break the ties. Death would be counted as worse than MI. The different kinds of events are not considered to be independent, but when using a hierarchical structure if death occurs on the same day as MI the MI is ignored. The outcome is the worst outcome in a given day. More at https://hbiostat.org/rmsc/markov - see the links at the top of that chapter. Ordinal analysis approximates a true patient utility-based outcome analysis.
Nice article John. Once again, trial designers refused to use a full information endpoint (ordinal longitudinal endpoint) that would have (1) greater power and (2) been able to tell us on which treatment patients fared better.
To clarify- from the article: The primary outcome was a composite of myocardial infarction or death from any cause up to 30 days after randomization." With the ordinal longitudinal endpoint are MI and death treated as 2 potential independent endpoints instead of as a composite?
The composite output you mentioned takes, as usual, 'composite' to mean 'union', ignoring different impacts/consequences of the different events. Ordinal endpoints break the ties. Death would be counted as worse than MI. The different kinds of events are not considered to be independent, but when using a hierarchical structure if death occurs on the same day as MI the MI is ignored. The outcome is the worst outcome in a given day. More at https://hbiostat.org/rmsc/markov - see the links at the top of that chapter. Ordinal analysis approximates a true patient utility-based outcome analysis.