“Cardio” Causes Heart Disease.
I haven’t carefully read this article but it seems interesting. In general, the author is an M.D. so it is pretty interesting reading:
=== Quoted article starts here ====
The image is not the thing itself
As a radiologist, this is my coffee-mug mantra for the epistemology of medical imaging. It reminds me that the picture or the test result is not the patient or even the disease.
Here is one of the heirarchies I use as a heuristic (a teaching device) when thinking about coronary disease. It’s a rough ordering from left to right of how far away you are from “the thing itself” when looking for evidence of a previous myocardial infarct.
From useless to definitive:
Framingham risk score – modern blood tests (HDL , LDL-P, oxidized LDL, etc.)
calcium score (Heart scan for coronary calcium)
Intravascular Coronary Ultrasound
Abnormal coronary angiogram
Abnormal coronary CTA
EKG evidence of infarct
Stress SPECT Nuclear Medicine perfusion
Cardiac perfusion MRI with dobutamine
Late Gadolinium enhancement cardiac MRI (LGE)
Death with autopsy evidence of MI
I have ordered these roughly according to how definitive they are in establishing myocardial injury – whether you have actually had a heart attack sometime in the past (not acutely, as that can often be shown with serum enzymes). I have not accounted for sensitivity, which for some of these tests is so poor as to make them useless.
…
I am obliged to point out that by the conventional completely arbitrary criteria used in biomedical publishing, the difference was “not statistically significant”. To meet the standard definition, there would have to be a 95% chance the difference is real. Instead, the significance level was 8% by McNemar’s test, so there is only a 92% probability that the difference is not due to chance.
That’s a relief.
Sidebar: Does anyone else think it’s strange that if your doctor said “you have a 92% chance you are disease free” you would find that reassuring, but we are supposed to dismiss a mere 92% probability that a study result is real? Stop doing what you are told and read the statistics without letting the authors or editors tell you what is “significant”. Yes, P = .00001 is better than .05. But decide for yourself, it is not written in stone.
[Maybe this weekend, I will geek out and try to refresh my failing memory regarding statistics and probability]