Gender differences in risk of neuropathology, cognitive decline and dementia in the Cognitive Function and Ageing Study (CFAS)

Abstract

There is renewed research interest in whether women experience a greater risk of dementia than men, and in how understanding such differences may inform prevention and treatment. Early epidemiological evidence suggested a higher incidence of dementia in women in older age groups; however, differences vary between countries and across generations. Key themes in the literature suggest that women experience faster cognitive decline, a higher burden of Alzheimer’s disease neuropathological change (ADNC) and a greater impact of the Apolipoprotein E (APOE) ε4 risk allele. A focus on sex-linked biological factors, particularly oestrogen decline following the menopause, may overlook how social norms and structural factors shape opportunities, exposures and behaviours by gender. These in turn may influence group differences in the accumulation of cardiovascular risk factors and protective ‘cognitive reserve’ across the life course. The Cognitive Function and Ageing Study (CFAS) is a multi-centre population-representative study of ageing and dementia. This longitudinal study collected data on physical and mental health, disability, cognition and used a validated algorithm for dementia assessment. It also recruited a subsample of brain donors, allowing detailed post-mortem neuropathological assessment. A narrative review synthesizing published CFAS findings showed patterns of gender differences across cognitive, physical and mental health in older age which mostly favoured men; patterns which have changed across generations. However, despite adjusting for age, many of these findings may still be shaped by differential survival between men and women (known as ‘survival bias’). Prominent findings from the literature such as more ADNC pathology and a greater impact of APOE ε4 in women have not previously been replicated in CFAS. The quantitative analyses in this thesis therefore focused on survival bias, a systematic analysis of gender differences in neuropathology, and consideration of cardiovascular risk factors and cognitive reserve. The effects of dropout and death on estimates of gender differences in cognitive decline were explored using subject-specific and population-average statistical models. Over 12 years of follow up in CFAS (n=13004) there was higher dropout of women and higher mortality of men. Population-average models of cognitive decline were biased by survival (even if weighted to account for dropout) and over time showed an increasing advantage in verbal fluency for men. However, in subject-specific models the gender difference seen at baseline was reduced to no difference by 12 years in the joint model (which adjusts for dropout and death). Harmonised data from the neuropathology cohorts of CFAS and the Cambridge City over-75s Cohort (CC75C) (n=804) were analysed to examine the prevalence of dementia and neuropathologies by age and gender, and their associations. There was strong evidence that many pathologies were more common in women (adjusting for age at death), including cortical and hippocampal neurofibrillary tangles and neuritic plaques, small and large vessel atherosclerosis and arteriolosclerosis, and atrophy. The correlation between pathologies, their patterns of co-occurrence and their effect on risk of dementia showed minimal differences between men and women. The higher prevalence of dementia at death in women was mediated by ADNC pathology, atrophy, and vascular pathology to a lesser degree. However, it was not mediated by the available cognitive reserve indicators (education, occupation duration and social class). Clinical data showed that men who died at younger ages had a higher burden of vascular risks and diseases, mirrored by vascular pathologies, suggesting a possible survival bias selecting for men with better cardiovascular health in older age groups. APOE ε4 carriers showed an increased risk of a range of pathologies: primarily ADNC pathologies and cerebral amyloid angiopathy (CAA), but also other neurodegenerative pathologies and some vascular pathologies. Interaction tests found no differences between genders. APOE ε2, by contrast, had a protective effect against ADNC pathologies, predominantly in women. The effect of APOE ε4 on dementia was mediated by both ADNC and other mainly neurodegenerative pathologies (with the inclusion of CAA), with a larger mediating effect of ADNC in women. By contrast, the protective effect of ε2 was only mediated by reduced ADNC pathology in women. Retrospectively reported age at menopause was categorised and its associations with health and cognition were descriptively analysed in the full CFAS cohort at baseline, and in the neuropathology cohort, though missing data were common. There was an overall pattern of worse physical and mental health, functional ability and cognition in premature and early menopause groups, as well as lower occupational social class and education. Comparing these baseline characteristics across age at menopause groups in the neuropathology cohort did not show the same gradients, suggesting some selection bias in becoming a donor. Interpretation of neuropathology outcomes were challenging in light of small numbers, missing data, and the absence of any APOE ε4 carriers in the early/premature menopause group. Finally, the neuropathology cohort was divided into four birth cohorts to explore trends in clinical and pathological outcomes. Several vascular pathologies increased after the first birth cohort while arteriolosclerosis decreased in the latest birth cohort. All amyloid measures (including plaques and CAA) increased in the latest birth cohort. Atrophy declined across several birth cohorts, while average brain weight consistently increased across birth cohorts, slightly more pronounced in women. The prevalence of dementia at death remained similar across birth cohorts, and across all outcomes there was very little evidence of trends differing by gender. The analyses in this thesis have leveraged the strengths of these rare population-based cohorts, and both replicated and challenged some established findings from the literature. The synthesis of new and existing evidence has highlighted some of the social, biological and methodological factors which have shaped observed gender differences in dementia incidence; it also makes the case for the importance of epidemiological theory and methods to ensure research produces robust evidence to further our understanding of dementia, and to inform policy action to address gender inequalities and promote healthy ageing.