Thread by @realDaveFeldman

1/ Yes-- I've been meaning to literally do an entire talk on Simpson's Paradox because the concept is really quite simple.

Here's an example:
✅Hypothesis: higher body weight causes higher blood pressure.

💡If we only had scales and BP cuffs, we'd certainly see the relationship

https://twitter.com/LowCarbEyeDoc/status/1633873097034956828

2/ Let's further point out we could demonstrate it over three lines of evidence: observational (epi), interventional (Rx), and genetic (🧬)

Epi: When randomly grabbing a large population of Americans, we find those who generally weigh more generally have higher BP ✅

3/ Rx -- interventions that generally reduce weight, also generally reduce BP ✅

And lastly, 🧬 -- most genes that associate with lower weight will generally have lower BP ✅

Now obviously we have more advanced tools these days than weight scales and BP cuffs...

4/ You are probably predicting the problem...

Most Americans have higher weight due to poor metabolic health (more total & visceral overall)
🔴🔴🔴🔴🔴

But a much smaller % of the population have higher weight due to more muscle mass and low fat (athletes, body builders). 🟩

5/ But if we're lumping them all together (ungrouped)...
(🔴🔴🔴🔴🔴 + 🟩)
... Then it is deceptively suggesting *everyone* with higher weight will have higher BP and thus should take steps to lose weight (even the athlete / body builders)

6/ Simpson's Paradox is really just pointing out that not only does the effort in grouping these data often tell us a different story - it can actually result in the *opposite* story.

Like in this example -- the athletes may well have lower BP than even the avg weight population

7/ My concern in lumping remnants with LDL under their shared ApoB is the potential for this very phenomenon.

I think it's crucial to consider how much metabolic health can impact both the lipid profile and ASCVD risk independently. (Hence the value of our #LMHRstudy, ofc)