|
|
|
|
|
|
by KempyKolibri
94 days ago
|
|
|
Yeah, this is a good example of what I’m talking about - you take a bunch of people at risk of diseases that lower LDL-c as a result of having that disease. They actually acknowledged that reverse causation is a risk here and so ran a sensitivity analysis by excluding patients with less than 5 years of follow up (a nice way of saying “patients who died within 5 years of the LDL-c reading”), the idea being that if these results were likely being driven by reverse causation, you’d expect to see an attenuation of the results. I’d point out that even if we didn’t, five years is a bit of a weird cutoff - plenty of LDL-c lowering diseases take much longer to kill the average person. Moot point though, because excluding those participants attenuated the result to the degree that the association with LDL-C and mortality became statistically insignificant. Quite why the authors said “we’re aware that reverse causation is a risk factor, let’s run a test to check if it’s likely influencing the results” and then completely ignored the fact the results suggested it was influencing the findings, is anyone’s guess. So yeah, basically huge confounder seems to be in play that likely explains the “paradox”. |
|
|