Table 3.

Results from the two-step least squares Mendelian randomisation approach. Causal effects (b) were estimated using a proportional hazard Cox model from a regression of the instrumentally explained BHS against all-cause, cancer and CVD mortality, and cancer and CVD incidence. Hazard Ratios (HRs) are expressed for a 0.1 increase in the score, and we report the p-value assessing if the causal effect is different from 0. We present estimates for the model adjusted for age, sex and the first 10 principal components capturing the latent structure of the UK Biobank population (Base model), and for the model additionally adjusted for education. Results from the multivariable Mendelian randomisation were adjusted for age, sex and the 10 first principal components capturing the latent structure of the UK Biobank population.

	Base model			Base model + Education			Base model + $\widehat{\text{Education}}$
	β	HR	p-value	β	HR	p-value	β	HR	p-value
All-cause mortality	0.03	1.03	6.09 × 10−01	0.00	1.00	9.39 × 10−01	0.02	1.02	7.33 × 10−01
Cancer mortality	−0.01	0.99	8.91 × 10−01	−0.04	0.96	5.61 × 10−01	−0.02	0.98	8.14 × 10−01
CVD mortality	0.12	1.12	4.43 × 10−01	0.10	1.11	5.16 × 10−01	0.11	1.11	4.82 × 10−01
Cancer incidence	0.01	1.01	6.29 × 10−01	0.01	1.01	7.33 × 10−01	0.01	1.01	6.68 × 10−01
CVD incidence	0.27	1.31	3.32 × 10−11	0.26	1.30	3.18 × 10−10	0.27	1.30	1.23 × 10−10