HD-OPM Archive
Topics › How to read research

U-shaped curves in observational data usually reflect reverse causality

Louis Nervegna · Aug 16, 2025 · ♥ 2 reactions · 1 comments · original Facebook post
How to read researchLDL, ApoB & lipid panel
↻ This is a repost. The main version (highest ranked) is U-shaped mortality curves for cholesterol, BMI, A1C and BP reflect reverse causality · Jun 12, 2024.
AI summary & key facts

Louis explains that U-shaped associations for cholesterol, BMI, A1C and blood pressure are common in epidemiology and are explained by reverse causality, not by true optimum levels at high values.

(no text — see the shared post below)

U-shaped curves happen with lots of health-related parameters, not just cholesterol.

This is a very common phenomenon in epidemiology.

You see it with cholesterol, BMI, A1C, blood pressure, just to name a few.

Don’t be fooled by this. The consensus explanation is reverse causality.

Do you really think the optimal BMI is 28, per observational studies? That the optimal A1C is 6 to 8.5, per observational studies? That the optimal systolic blood pressure is 140 to 150, per observational studies?

Of course not.

These look optimal in observational studies because of reverse causality.

For the same reason, the optimal total cholesterol is not 220 to 240.

The charts in the photos below show hazard ratio (risk of dying) for various values of total cholesterol, BMI, A1C, and blood pressure. All of these charts come from observational studies.

The reference for these particular U-shaped charts is this video. They are explained in more detail in the video:

↪ Video: cholesterol and risk of death, new evidence

Note: Reverse causality means that increased risk of dying (e.g. illness) causes the parameter in question to go down. As opposed to the other way around. For example, cancer is well known to cause lower LDL and BMI. Liver or kidney diseases cause lower A1C. Uncontrolled diabetes causes lower blood pressure.

#ushapedcurveHD #ldlHD #a1cHD #bmiHD #bloodpressureHD #reversecausalityHD #cholesterolHD #lowldlHD

No photo description available.No photo description available.No photo description available.No photo description available.
← newer: Adding Lp(a) does not improve PREVENT 10-year risk predictioolder: Low LDL in old age is linked to mortality through reverse ca →