As someone who spent 13 years running a clinical trial repository of my own design, I’d about given up trying to catalogue the ways in which the ‘ethical pharmaceutical’ industry and Google have conspired to lead a disinterested press corps around by the nose. I greatly appreciate your work here.
Snap. Though it took me 32y & retirement before I fully understood how badly astray we’d gone. Was pharma ‘clean’ when I was “in the belly of the beast”?
Of course not. I merely failed to notice, perhaps because there are far fewer opportunities to cheat in research & greater professional sanctions if caught doing it. No one would work again with a cheat. I only encountered two episodes of outright fraud during my career.
But up in HQ, there clearly was much fraud because I was there when Pfizer negotiated a $2.3B settlement after being caught doing very bad things with one of its products.
I was expecting to see a discussion of the 2nd reason for RCT preference, besides serving as barrier to entry. It's also easier to twiddle RCT design parameters to obtain desired goal -- either to boost one's own drug or to suppress another one. Is this discussed anywhere in the book?
“A superior statistician with an inferior data set is nearly always preferable to an inferior statistician wielding the most pristine data possible (and no respectable statistician would argue that point).”
This is quite the claim and I don’t think it’s close to true.
Obviously getting any sort of a bead on this would require clarifying what we mean by an inferior statistician and data set but this seems deeply suspect.
“the average results of the observational studies were remarkably similar to those of the randomized, controlled trials.“
I’d expect observational studies to be deployed in cases where they were unlikely to introduce biases relative to RCTs so I’d be cautious about extrapolating from this. Researchers obviously select consider which situations are most amenable to observational design and potential objections of reviewers.
“Plenty of RCTs have failed in their goal, whereas much great science has been done without them.”
I mean of course, but so what? What is the relative rate of failure for a given methodology? Plenty of sober drivers have been in car accidents while many drunk drivers arrive home safe.
As someone who spent 13 years running a clinical trial repository of my own design, I’d about given up trying to catalogue the ways in which the ‘ethical pharmaceutical’ industry and Google have conspired to lead a disinterested press corps around by the nose. I greatly appreciate your work here.
Snap. Though it took me 32y & retirement before I fully understood how badly astray we’d gone. Was pharma ‘clean’ when I was “in the belly of the beast”?
Of course not. I merely failed to notice, perhaps because there are far fewer opportunities to cheat in research & greater professional sanctions if caught doing it. No one would work again with a cheat. I only encountered two episodes of outright fraud during my career.
But up in HQ, there clearly was much fraud because I was there when Pfizer negotiated a $2.3B settlement after being caught doing very bad things with one of its products.
I was expecting to see a discussion of the 2nd reason for RCT preference, besides serving as barrier to entry. It's also easier to twiddle RCT design parameters to obtain desired goal -- either to boost one's own drug or to suppress another one. Is this discussed anywhere in the book?
Perhaps in a different way. I'm currently working on several articles about rigging research.
“A superior statistician with an inferior data set is nearly always preferable to an inferior statistician wielding the most pristine data possible (and no respectable statistician would argue that point).”
This is quite the claim and I don’t think it’s close to true.
Obviously getting any sort of a bead on this would require clarifying what we mean by an inferior statistician and data set but this seems deeply suspect.
“the average results of the observational studies were remarkably similar to those of the randomized, controlled trials.“
I’d expect observational studies to be deployed in cases where they were unlikely to introduce biases relative to RCTs so I’d be cautious about extrapolating from this. Researchers obviously select consider which situations are most amenable to observational design and potential objections of reviewers.
“Plenty of RCTs have failed in their goal, whereas much great science has been done without them.”
I mean of course, but so what? What is the relative rate of failure for a given methodology? Plenty of sober drivers have been in car accidents while many drunk drivers arrive home safe.