For example, LDL cholesterol has RR of 0.743 for 90th percentile, and 0.664 for 99th percentile. That’s 26% and 34% lower chance per 1 SD (around 39 mg/dl) increase in LDL cholesterol. While a OR increase would be a higher chance. CI’s that don’t have the bar cross 1 are less likely to be due to chance.
“Predicted fasting glucose was not significantly linked to longevity. These results were sustained in the sensitivity analyses in both longevity (≥90th) and super-longevity (≥99th)”
All the rest is quite as expected: lower your LDL and blood pressure to live longer.
Ok, I’m not so good at this. Is educational attainment the most important thing of all? Are you better off with a phD even if it means you have to smoke?
I’m pretty sure smoking wasn’t compulsory at postgraduate level.
the graph says its “causal” but I doubt they were able to exclude all the confounders. This is a good example of how even Mendelian randomization (MR) studies can’t necessarily exclude all non-causative associations.
For example a MR marker for a certain ethnicity might correlate with both phd incidence and recent immigration status and therefore wealth. Very curious to see what gene markers they used for the phd group but I couldn’t seem to download the supplementary materials