I think the choice of ACM can be both criticized and defended in the context of an observational study. Since it is black and white, there can be no ambiguity and no misclassification bias. And it eliminates the effects of residual and unmeasured confounding. However, as you note, competing risks is the issue. The authors are damned either way. Their first poor choice was to do an observational study at all. Having made that choice, it is what it is.
The main problem is the immediate separation of the curves. It is simply not biologically plausible. You’ve often mentioned the study which demonstrated how (to paraphrase) an observational dataset could be analyzed 100 different ways, to yield 100 different results. This is one where the authors ought to have seen their results, recognized it to be nonsense, and said “maybe we do this a different way”. Instead, they published something that is ridiculous on its face.
But your conclusion hits nail on head. The authors can study whatever they want, and submit whatever they want. That this study made it through the editors and peer reviewers at JAMA, to appear as a finished item in its current state, belies the rot in the medico-industrial complex. It does in fact seem nefarious to me.
Previous conclusions/reports of studies on LAAC(Watchman) vs DOACs excluding procedural bleeding events from the analysis and also reporting benefit in reducing strokes including patients that didn’t receive the device added to this observational report are misleading/biased. I hope that the Factor-XI “antagonists” are finally approved for clinical use in A. Fib. for anything sensible to be scientifically discussed for this patient population.
The point that some results should show up right away, but others should appear only slowly, is generalizable to many other studies.
In my own work on the impact of health insurance on health, there are many observational studies, with difference-in-difference designs, and even an RCT (Goldin et al., Quarterly Journal of Economics 2021) showing large immediate reduction in mortality for the newly insured. No one asks: What is the causal channel for an immediate effect?
It would appear that the authors demonstrate significant motivated reasoning. Could it be that those performing the LAAO procedure are benefiting the most from widespread adoption?
I think the choice of ACM can be both criticized and defended in the context of an observational study. Since it is black and white, there can be no ambiguity and no misclassification bias. And it eliminates the effects of residual and unmeasured confounding. However, as you note, competing risks is the issue. The authors are damned either way. Their first poor choice was to do an observational study at all. Having made that choice, it is what it is.
The main problem is the immediate separation of the curves. It is simply not biologically plausible. You’ve often mentioned the study which demonstrated how (to paraphrase) an observational dataset could be analyzed 100 different ways, to yield 100 different results. This is one where the authors ought to have seen their results, recognized it to be nonsense, and said “maybe we do this a different way”. Instead, they published something that is ridiculous on its face.
But your conclusion hits nail on head. The authors can study whatever they want, and submit whatever they want. That this study made it through the editors and peer reviewers at JAMA, to appear as a finished item in its current state, belies the rot in the medico-industrial complex. It does in fact seem nefarious to me.
Previous conclusions/reports of studies on LAAC(Watchman) vs DOACs excluding procedural bleeding events from the analysis and also reporting benefit in reducing strokes including patients that didn’t receive the device added to this observational report are misleading/biased. I hope that the Factor-XI “antagonists” are finally approved for clinical use in A. Fib. for anything sensible to be scientifically discussed for this patient population.
The point that some results should show up right away, but others should appear only slowly, is generalizable to many other studies.
In my own work on the impact of health insurance on health, there are many observational studies, with difference-in-difference designs, and even an RCT (Goldin et al., Quarterly Journal of Economics 2021) showing large immediate reduction in mortality for the newly insured. No one asks: What is the causal channel for an immediate effect?
Bernie Black (Northwestern Univ.)
I started reading this study and stopped when I read the methods. It was either designed by a med student or mathematician being intentionally obtuse
Great discussion. This remains a murky area of AF management which needs more study.
What’s the chance?
It would appear that the authors demonstrate significant motivated reasoning. Could it be that those performing the LAAO procedure are benefiting the most from widespread adoption?