On 2015 Jul 23, Neil Davies commented:

Østergaard and colleagues report that fewer people with a high genetic risk of high blood pressure develop Alzheimer disease compared to those with a lower genetic risk of high blood pressure (1). This is consistent with evidence that use of anti-hypertensive medication is associated with incidence and progression of Alzheimer disease (2–4). We investigated these associations in an observational longitudinal cohort study using data from the Clinical Practice Research Datalink, a database of anonymous UK National Health Service primary and secondary health care data. We found that patients prescribed angiotensin-receptor blockers were less likely to be diagnosed with Alzheimer disease (OR 0.47, 95% CI: 0.37 to 0.58) than those prescribed other anti-hypertensives. Our results were based on observational data and may suffer from residual confounding, but they provide suggestive evidence that blood pressure or blood pressure medication is implicated in the aetiology of Alzheimer disease. A meta-analysis of randomised controlled trials has also suggested that treatment with blood pressure lowering medication may help slow cognitive decline (5).

However, there are alternative explanations for these associations. Although Mendelian randomization analyses may remove biases due to confounding and reverse causality, survival bias is still likely to be an issue. Individuals with a greater burden of high blood pressure SNPs may have higher mortality, which could cause them to die before developing Alzheimer’s disease and reduce the frequency of blood pressure increasing SNPs in cases compared to controls. Survival bias may also explain the apparent protective effect on Alzheimer’s disease risk of variants that are associated with heaviness of smoking. The smoking heaviness increasing allele located in the CHRNA5-A3-B4 gene cluster has been shown to be associated with increased mortality risk amongst ever smokers (6), which is likely to explain the reduced frequency of this allele amongst older populations of ever compared to never smokers.(7) This highlights the difficulties in using Mendelian randomisation for diseases of old age.

The authors discuss survival bias, and conclude it is unlikely to be a problem. However, in our opinion more research is needed to quantify the extent and impact of survival bias in Mendelian randomisation studies. It may be possible to assess the extent of survival bias by comparing risk allele frequencies to known allele frequencies in younger populations, or to track their change in a population as it ages. Survival bias is likely to only occur if there are fewer risk alleles in older populations. Simulations may allow us to estimate the size of survival bias and aid the interpretation of Mendelian randomisation studies.

Neil M. Davies, Amy E. Taylor, Marcus R. Munafò

[1] S. D. Østergaard et al., Associations between Potentially Modifiable Risk Factors and Alzheimer Disease: A Mendelian Randomization Study. PLOS Med. 12, e1001841 (2015).

[2] N. Li et al., Use of angiotensin receptor blockers and risk of dementia in a predominantly male population: prospective cohort analysis. Br. Med. J. 340, b5465 (2010).

[3] N. M. Davies, P. G. Kehoe, Y. Ben-Shlomo, R. M. Martin, Associations of anti-hypertensive treatments with Alzheimer’s disease, vascular dementia, and other dementias. J. Alzheimers Dis. JAD. 26, 699–708 (2011).

[4] P. G. Kehoe, N. M. Davies, R. M. Martin, Y. Ben-Shlomo, Associations of Angiotensin Targeting Antihypertensive Drugs with Mortality and Hospitalization in Primary Care Patients with Dementia. J. Alzheimers Dis. JAD. 33, 999–1008 (2013).

[5] N. Levi Marpillat, I. Macquin-Mavier, A.-I. Tropeano, A.-C. Bachoud-Levi, P. Maison, Antihypertensive classes, cognitive decline and incidence of dementia: a network meta-analysis. J. Hypertens. 31, 1073–1082 (2013).

[6] L. Rode, S. E. Bojesen, M. Weischer, B. G. Nordestgaard, High tobacco consumption is causally associated with increased all-cause mortality in a general population sample of 55,568 individuals, but not with short telomeres: a Mendelian randomization study. Int. J. Epidemiol. 43, 1473–1483 (2014).

[7] A. E. Taylor, M. R. Munafò, CARTA consortium, Commentary: Does mortality from smoking have implications for future Mendelian randomization studies? Int. J. Epidemiol. 43, 1483–1486 (2014).

On 2015 Jul 23, Neil Davies commented:

Østergaard and colleagues report that fewer people with a high genetic risk of high blood pressure develop Alzheimer disease compared to those with a lower genetic risk of high blood pressure (1). This is consistent with evidence that use of anti-hypertensive medication is associated with incidence and progression of Alzheimer disease (2–4). We investigated these associations in an observational longitudinal cohort study using data from the Clinical Practice Research Datalink, a database of anonymous UK National Health Service primary and secondary health care data. We found that patients prescribed angiotensin-receptor blockers were less likely to be diagnosed with Alzheimer disease (OR 0.47, 95% CI: 0.37 to 0.58) than those prescribed other anti-hypertensives. Our results were based on observational data and may suffer from residual confounding, but they provide suggestive evidence that blood pressure or blood pressure medication is implicated in the aetiology of Alzheimer disease. A meta-analysis of randomised controlled trials has also suggested that treatment with blood pressure lowering medication may help slow cognitive decline (5).

However, there are alternative explanations for these associations. Although Mendelian randomization analyses may remove biases due to confounding and reverse causality, survival bias is still likely to be an issue. Individuals with a greater burden of high blood pressure SNPs may have higher mortality, which could cause them to die before developing Alzheimer’s disease and reduce the frequency of blood pressure increasing SNPs in cases compared to controls. Survival bias may also explain the apparent protective effect on Alzheimer’s disease risk of variants that are associated with heaviness of smoking. The smoking heaviness increasing allele located in the CHRNA5-A3-B4 gene cluster has been shown to be associated with increased mortality risk amongst ever smokers (6), which is likely to explain the reduced frequency of this allele amongst older populations of ever compared to never smokers.(7) This highlights the difficulties in using Mendelian randomisation for diseases of old age.

The authors discuss survival bias, and conclude it is unlikely to be a problem. However, in our opinion more research is needed to quantify the extent and impact of survival bias in Mendelian randomisation studies. It may be possible to assess the extent of survival bias by comparing risk allele frequencies to known allele frequencies in younger populations, or to track their change in a population as it ages. Survival bias is likely to only occur if there are fewer risk alleles in older populations. Simulations may allow us to estimate the size of survival bias and aid the interpretation of Mendelian randomisation studies.

Neil M. Davies, Amy E. Taylor, Marcus R. Munafò

[1] S. D. Østergaard et al., Associations between Potentially Modifiable Risk Factors and Alzheimer Disease: A Mendelian Randomization Study. PLOS Med. 12, e1001841 (2015).

[2] N. Li et al., Use of angiotensin receptor blockers and risk of dementia in a predominantly male population: prospective cohort analysis. Br. Med. J. 340, b5465 (2010).

[3] N. M. Davies, P. G. Kehoe, Y. Ben-Shlomo, R. M. Martin, Associations of anti-hypertensive treatments with Alzheimer’s disease, vascular dementia, and other dementias. J. Alzheimers Dis. JAD. 26, 699–708 (2011).

[4] P. G. Kehoe, N. M. Davies, R. M. Martin, Y. Ben-Shlomo, Associations of Angiotensin Targeting Antihypertensive Drugs with Mortality and Hospitalization in Primary Care Patients with Dementia. J. Alzheimers Dis. JAD. 33, 999–1008 (2013).

[5] N. Levi Marpillat, I. Macquin-Mavier, A.-I. Tropeano, A.-C. Bachoud-Levi, P. Maison, Antihypertensive classes, cognitive decline and incidence of dementia: a network meta-analysis. J. Hypertens. 31, 1073–1082 (2013).

[6] L. Rode, S. E. Bojesen, M. Weischer, B. G. Nordestgaard, High tobacco consumption is causally associated with increased all-cause mortality in a general population sample of 55,568 individuals, but not with short telomeres: a Mendelian randomization study. Int. J. Epidemiol. 43, 1473–1483 (2014).

[7] A. E. Taylor, M. R. Munafò, CARTA consortium, Commentary: Does mortality from smoking have implications for future Mendelian randomization studies? Int. J. Epidemiol. 43, 1483–1486 (2014).