Thread by @ZKForTre

I’d like to use this to highlight why defining terms matters, and refusing to do so severely reduces the utility of such conversations.

LDL risk stratification are best predicted by AUC. Point/cross-sectional sampling is only weakly predictive....

https://twitter.com/daveketo/status/1317329340791599104

Therefore, proposing this as a hypothetical without specification of age, first and foremost, limits the ability to extrapolate conclusions extensively. Is the person 70 with no history of events? Likely very mild difference? 30, with family history of ischemic events?

It is challenging to engage these types of arguments with data-driven follow up and analysis because terms and specifications are rarely provided. In exercises that are driven by cumulative statistics, which is exactly what this is, these sorts of specifications are **essential**

The response will, in all likelihood, be that individuals can be compared to age-matched controls across ages. However, this requires laborious assumptions. This requires assuming an exactly equivalent lipid profile throughout one’s entire life to extrapolate, which is

A dubious assumption based on any number of factors. Furthermore, it undermines the very nature of cumulative exposure metrics, which become extremely relevant in the case of a 30 year old with such a lipid panel.

Personally, a nuanced analysis of this likely requires several levels of stratification, especially looking at a genetic predisposition to atherosclerosis and ischemic events in the particular individuals. Even aside from that, though, asking completely out of context

With no critical analysis of data being used to make predictions is a false proposition, and I can lay out in immense detail why this sort of characterization is fraught with unaddressed limitations, despite attempts.

Anyway, good luck to all engaged parties. I would really like for the questions regarding extrapolation of the proposed but largely unstudied paradigms to be answered in a direct sense in a way that they have not been to date.