Table 3.
Multivariate MR of iron-related traits on healthspan, parental lifespan, and longevity shows a protective effect for transferrin and a deleterious effect for serum iron.
| Exposure | βMR | SE | P | Padj | βhealthspan | βlifespan | βlongevity |
|---|---|---|---|---|---|---|---|
| Serum iron | −0.79 | 0.242 | 1E−03 | 4E−03 | −1.10 (0.58) | −1.17 (0.63) | −5.07 (2.42) |
| Transferrin saturation | 0.80 | 0.252 | 1E−03 | 4E−03 | 1.11 (0.61) | 1.16 (0.66) | 5.15 (2.52) |
| Transferrin | 0.32 | 0.100 | 2E−03 | 4E−03 | 0.48 (0.24) | 0.46 (0.26) | 2.02 (1.00) |
| Ferritin | −0.01 | 0.024 | 0.5380 | 1.0000 | 0.13 (0.06) | −0.02 (0.06) | −0.26 (0.24) |
The effects of 15 SNPs genome-wide significant for one or more iron-related traits were tested against the effects of our GWAS meta-analysis and individual healthspan, parental lifespan, and longevity GWAS in an inverse variance-weighted regression.
Coefficients are derived from a model with a fixed regression intercept, as a sensitivity analysis showed a non-significant intercept centred around zero for all traits (Pintercept ≥ 0.76). Although the causal effect sizes appear large, in practice, homeostatic effects prevent large variation in one of the exposures independent of the others.
βMR: the causal effect of one standard deviation increase in the exposure on the healthspan/parental lifespan/longevity meta-analysis (in standard deviation units), conditional on the other exposures, P: nominal P value for the MR effect, Padj: multiple testing-corrected P value, βhealthspan, βlifespan, βlongevity: the conditional effect of one standard deviation increase in the exposure on healthspan (in -logHR units), parental lifespan (in -logHR units), or longevity (in logOR units), with the standard error reported in parentheses.