After adjustment for other CV risk factors or diseases, diabetes, stroke, HF and AF remained significantly associated with an increased dementia risk in all aged groups studied. The strength of these associations decreased with advanced age for diabetes, stroke and HF. From 66 to over 90 years, the increased dementia risk associated with respectively diabetes, stroke and HF passed from 1.84, 3.54, 2.08 folds in the younger age group to 1.20, 1.13, 1.16 among patients ≥ 90 years old. The association of AF and dementia was more or less stable with advanced ages. Our results suggest that preventing diabetes, stroke, HF and AF before early advanced age is crucial to prevent dementia. So, cognitive function assessment should be performed in the routine clinical management of patients with CV risk factors, especially in early ages, and a decline in cognitive function should militate for a more aggressive control of CV risk factors31. It also suggests that pathophysiology of dementia in younger and older populations seems to be different with less impact of CV risk factors or disease in advanced age.
The age-dependent effect of dementia in older populations has been described previously19,32. Recently, Legdeur et al. found a similar pattern of decrease in the risk of CV risk factors or diseases studies for incident dementia with aging in patients ≥ 65 years, but only up to 90 years19. And, our results are also well aligned with the study of McGrath et al., where vascular risk factors and dementia risk prediction across later-life from Framingham Heart Study for hypertension, diabetes, AF and stroke. For instance, the age group 75–80, the HRs of hypertension, diabetes, AF and stroke were at 0.74, 1.40, 1.43 and 1.63 compared to our results 0.99, 1.39,1.23, 1.73, respectively32. Previous studies had also reported age-dependant effects on incident dementia risk for hypertension20,33,34,35,36, dyslipidemia20,37, AF, MI38, and a risk score that included diabetes mellitus, hypertension, dyslipidemia and MI39. They found that these risk factors are associated with dementia in age 40–65 years, but not in age > 65 years, excepted for the risk score that was associated with dementia up to the age of 79 years. For diabetes mellitus, it is known that diabetes increases dementia risk, that the risk is stronger when diabetes occurs at mid-life than in late life40; and that this association remains in patients ≥ 85 years20,41,42,43. For stroke as well, we noted that the dementia risk decreased with aging, and the association of stroke with incident dementia persisted in patients ≥ 85 years33. It should be noted that study design used, adjustments for CV risk factors or diseases, medication exposure, and the type of outcome (all cause dementia, versus vascular dementia or Alzheimer’s disease) differed between studies, prohibiting an adequate direct comparison of effect estimates over studies.
Many specific risk-factors have been put forward to explain the observed age-dependant effect. First, it is suggested that atherosclerosis resulting from high blood pressure, and cerebral hypoperfusion related to severe atherosclerosis and to low blood pressure could explain the pathways linking high blood pressure in younger patients and low blood pressure in older adults to cognitive decline and dementia33,35. Additionally, this could also be explained by the complex perspective of age-associated changes in cardiovascular structure and its cross talk with blood pressure variability. Thickening and stiffening of the large arteries develop with aging due to collagen and calcium deposition and loss of elastic fibers in the medial layer44. In the left ventricle, modest concentric wall thickening occurs with ageing due to cellular hypertrophy44. These arterial changes cause systolic blood pressure to rise with age, while diastolic blood pressure generally declines, resulting in an increased pulse pressure variability. Simultaneous multi-organ dysfunction (defined as simultaneous occurrence of left ventricular hypertrophy, common carotid artery disease and increased thickness, and chronic kidney disease) has been associated with low diastolic blood pressure rather than systolic hypertension45. So, the decrease in diastolic blood pressure with age might explain more damageable than systolic hypertension, explaining the decrease in the association between hypertension and dementia age45. However, in our study, hypertension and dyslipidemia are not associated with dementia risk, when adjusted for the other CV risk factors or diseases. Second, higher cholesterol values was associated with a decreased dementia risk in late life, and may indirectly indicate a better overall health status in older patients46. Third, there might be a reverse causality for the age-dependant effect of hypertension, dyslipidemia and diabetes with blood pressure, cholesterol and glucose levels may drop before the onset of the neurodegenerative process35,47.
Hypothesis generalizable to all risk factors may explain the observed age-dependent effect as well. First, there may be a selective survival of patients who are less likely to develop dementia as consequence of the risk factors. Second, the higher prevalence of CV risk factors and other non-CV risk factors of dementia in elderly patients attenuates the distinction of patients with and without a CV risk factors48. Nonetheless, a similar decrease in the dementia risk was observed when additional adjustment for any of the other CV risk factors or diseases included in our study was performed, but other CV and non-CV risk factors may confound our observations. Third, patients with CV risk factors may die earlier than those without the risk factor, and are consequently less likely to develop dementia before their death. Notwithstanding, adjusting for mortality as a competing risk led to similar results in our study, with the risk of CV risk factors or diseases for incident dementia diminishes with increasing age (except for hypertension, dyslipidemia and MI). Fourth, neuropathological studies have shown that the vascular risk burden was predictive of vascular dementia (cortical infarcts, subcortical infarcts, atherosclerosis, and arteriosclerosis), but not Alzheimer’s disease (Braak stage, cerebral amyloid angiopathy and, neuritic plaque score)49. However, neuropathological studies have also revealed that, in late life, the presence of multiples pathologies (Alzheimer’s disease, microinfarcts, hippocampal sclerosis, Lewy body disease, macroinfarcts, cerebral amyloid angiopathy, white matter disease, and others) were associated with an increased risk of dementia50. Accordingly, the observed decreasing risk of CV risk factors or diseases for incident dementia with increasing age might be due to the fact that our studied outcome was all-cause dementia and that the presence of multiple pathologies, rather than vascular dementia on its own, seems to be involved in the development as age increase. Fifth, we should have been taking in account the CV risk level in the risk prediction of dementia with longitudinal data since, the improvement of CV modifiable risk was associated with a reduction of dementia risk51,52,53. Finally, we did not take in account the impact of level adherence of CV medication.
Our study is one of the largest to examine the effect of age on risk of CV risk factors or diseases on the incidence of dementia in a population ranging from 66 to over 90 years. The strengths and limitations of our study are greatly related to the nature of the Quebec RAMQ administrative databases. As there is single payer public health care system in Quebec (Canada), Quebec RAMQ administrative database includes all individuals over 65 years old, limiting selection bias, and captures all diagnosis, procedures and drugs. However, several limitations must be taken into consideration. First, administrative claims data depend on the exhaustive, accurate recording and coding of diagnoses, procedures, and drugs. Second, this observational study of administrative data might have been subject to confounding bias by unadjusted factors (e.g., auto-immune diseases, pro-inflammatory conditions, race, level of education, smoking, sedentary lifestyle, major depressive disorder, etc.). Third, administrative claims data depend on the exhaustive, accurate recording and coding of diagnoses and procedures. Fourth, the accurate starting for CV risk factors or diseases may not always accurately recorded in administrative claims data, so the duration of the CV disorders could not be accounted for, but we have used a 5-year period to assess risk CV risk factors or diseases. Fifth, no distinction between types of dementia was performed, as the type of dementia is not reliably specified in administrative claims data. Sixth, quantitative measurements of blood pressure, glucose and cholesterol were not available and we could not take in account appropriate control of hypertension, diabetes mellitus and cholesterol. Seventh, a patient may be included in more than one age group during the follow-up in order to consider the cohort risk according to each age-group, which may introduce a survival bias. We thus provided a sensitivity analysis using competing risks correcting for mortality, and this led to similar results. Again, the risk factors of dementia exhibit similar results by stratifying by sex across age-groups. Last, we could not adjust for factors such as, economic, health status, severity of disease, lifestyle and education that could be also linked with dementia risk54.