Exploring the effect of pre-clinical Alzheimer's disease on blood pressure using Mendelian randomisation and parental dementia as an instrumental variable in UK Biobank

Abstract

Background: Evidence suggests there may be a bidirectional relationship between high blood pressure (BP) and Alzheimer's disease (AD). It is hypothesised that this is due to cerebral changes during pre-clinical AD that cause elevation of systemic BP. We aimed to test this by exploring the effect of risk of pre-clinical AD on blood pressure.

Methods: We used data from the UK Biobank, including adults without prevalent or incident (within first 5 years of follow-up) clinical AD (N = 501,420, mean age 56.6, SD 8 years). We used two instrumental variables, an age-weighted parental dementia instrument score and a participant genetic instrument score, that are vulnerable to differing biases, to instrument risk of pre-clinical AD (the exposure). We tested the association of both instrument scores with systolic BP (SBP), diastolic BP (DBP), and hypertension. Sensitivity analyses were undertaken to explore different biases.

Results: Both the higher parental dementia instrument and participant genetic instrument score were associated with higher mean SBP (difference in mean SBP mmHg per 1SD higher score: 0.12, 95% CI 0.06 to 0.17, p < 0.0001, and 0.07, 95% CI 0.00 to 0.13, p=0.037, respectively) but not DBP. Sensitivity analyses were largely consistent with these findings. CONCLUSIONS: Our findings provide preliminary evidence that pre-clinical AD increases SBP. Further research is required to determine whether this increase in SBP is due to increased cerebrovascular resistance as a result of pre-clinical AD. Obtaining a better understanding of the changing relationship with BP at different stages of AD may enable effective optimisation and targeting of therapies.

Keywords: Blood pressure; Human population; Instrumental variables; Pre-clinical Alzheimer’s disease; Prospective cohort.

Fig. 1

Directed acyclic graphs depicting the hypothesised causal pathways using A parental dementia instrument score as an instrumental variable and B participant genetic instrument score as an instrumental variable for risk of pre-clinical Alzheimer’s disease. Directed acyclic graphs (DAGs) place arrows between a variable at the base of the arrow and one at the head (point of the arrow) when there is a known or plausible causal effect of the variable at the base on the variable at the top. DAGs are useful for illustrating analysis assumptions and here we use them to show that our two instrumental variables have different key sources of bias)

Fig. 2

Difference in mean systolic blood pressure (SBP, mmHg) (black squares), and 95% confidence intervals (horizontal lines) per one standard deviation higher parental dementia instrument score (PDIS, A) as instrumental variable for risk of pre-clinical Alzheimer’s disease and participant genetic instrument score (PGIS, B) as instrumental variable for risk of pre-clinical Alzheimer’s disease. Each model (from crude to fully adjusted) is shown to indicate the differing effects of confounding with each instrument (due to each being vulnerable to differing biases (see main text for details))

Fig. 3

Difference in mean diastolic blood pressure (DBP, mmHg) (black squares), and 95% confidence intervals (horizontal lines) per one standard deviation higher parental dementia instrument score (PDIS, A) as instrumental variable for risk of pre-clinical Alzheimer’s disease and participant genetic instrument score (PGIS, B) as instrumental variable for risk of pre-clinical Alzheimer’s disease. Each model (from crude to fully adjusted) is shown to indicate the differing effects of confounding with each instrument (due to each being vulnerable to differing biases (see main text for details))

Fig. 4

Difference in mean ratio of odds of being hypertensive over odds of being normotensive (odds ratio, black squares), and 95% confidence intervals (horizontal lines) per one standard deviation higher parental dementia instrument score (PDIS, A) as instrumental variable for risk of pre-clinical Alzheimer’s disease and participant genetic instrument score (PGIS, B) as instrumental variable for risk of pre-clinical Alzheimer’s disease. Each model (from crude to fully adjusted) is shown to indicate the differing effects of confounding with each instrument (due to each being vulnerable to differing biases (see main text for details))