Lucibee thread - Apr 6, 2018
The more I poke, the more bits fall off.
#PACEtrial
#factfulness
This time I poked the fitness test, one of the few "objective" measures they used. They reported that they had used a version of the test by Petrella et al., but then quoted a different equation in their supplementary data. #PACEfail
None of the fitness equations had been validated in their patient population.
Looking closely at the paper they cited, the population studied was healthy older adults. When this study included unhealthy people, it stopped being a good estimate of VO2max. #PACEfail
And the equation they said they used gives very different results from those they actually quoted in their fitness graph in the only paper they reported them in (mediation analysis). #PACEfail
So I have absolutely no idea exactly what they did do, because they didn't report it properly. Given that the authors themselves had concerns about the fitness testing (it reportedly caused harm to pts), this is problematic. #PACEfail
Thread continues (Apr 6, 2018):
When a colleague tried to obtain the data for the fitness test, his request was deemed "vexatious".
I can now see why. They were probably too embarrassed to release it! #PACEfail
[colleague was Graham McPhee]
So not only did they use unreliable subjective measures in an unblinded trial, they also used unreliable objective measures too. #PACEfail
And @TheLancet
FastTracked this paper too.
#PACEfail
Mediation paper was published in @TheLancetPsych
Fitness data criticised here by Bob... http://thelancet.com/journals/lanpsy/article/PIIS2215-0366(15)00054-1/fulltext
Mediation analysis cannot be used to confirm a theory. You have to be damn sure that your theory is correct before you use it, because it won't tell you otherwise. #confirmationbias
But it can disprove it if the objective measures drop out. Which they did.
#PACEfail
Ref: "If the presumed causal model is not correct, the results from the mediational analysis are likely of little value." http://davidakenny.net/cm/mediate.htm
The whole idea that you can use a trial or mediation analysis to prove your theory correct doesn't make any sense. If that were so, then we would still think that scurvy is caused by a bacterium that is killed by the acid in lime juice.
Bookmarking... SMC [Science Media Centre] on mediation analysis paper:
More archiving...
This time, #PACEtrial mediation analysis (because it's on my mind).
Xitter thread from 6 April 2018.
"The more I poke, the more bits fall off.
#PACEtrial
#factfulness"