Sex differences in dementia risk and risk factors: an IPD meta-analysis

Dementia is expected to affect an anticipated 150 million people globally by 2050 (Nichols et al, 2019). The Lancet Commission on dementia prevention has previously calculated that up to 40% of cases of dementia may be prevented or delayed through 12 potentially modifiable risk factors (Livingston et al, 2020). In the absence of readily available disease-modifying treatments for dementia, understanding these risk factors is essential to tackle this public health challenge and to guide the best preventive practices.

Previous work has suggested that women may be at increased risk of dementia, but this evidence is conflicting and has largely been carried out in High-Income Countries (HICs) (Nebel et al, 2018). Today, 60% of people with dementia live in low-and-middle income countries (LMICs) and this is expected to increase to 80% by 2050 (Prince et al, 2007), but the association between sex and dementia risk in these settings is not well understood. In addition, many studies of potentially modifiable dementia risk factors adjust for sex as a covariate, meaning we don’t know whether their effect varies according to sex. This has potential implications for targeting dementia prevention efforts.

In this study (Gong et al, 2023), the authors used an individual-patient data meta-analysis of a diverse, global population across 21 cohorts to ask:

1. Is there is a sex difference in the overall risk of all-cause dementia?
2. Does sex influence the impact of risk factors for dementia?

Gong et al. conducted an individual participant data meta-analysis to explore whether there is a sex difference in the overall risk of all-cause dementia, and whether sex influences the impact of risk factors for dementia.

Methods

Globally, population-based ageing cohorts vary substantially in their methods and techniques. This makes it difficult to accurately compare across studies or identify true differences. To address this, the Cohort Studies of Memory in an International Consortium (COSMIC) collaboration synchronises international cohort studies of cognitive ageing with the aim to boost their comparability and statistical power (Sachdev et al, 2013).

This was an individual participant data meta-analysis. In this study, Gong et al (2022) pooled and re-analysed individual participant data from the 21 cohorts of the COSMIC collaboration to understand sex-differences in dementia risk and rate. They calculated:

• Hazard Ratios for sex as a risk factor for all-cause dementia in each included study (Cox proportional hazards regression, age- and education-adjusted) and a pooled estimate across all studies (Cox proportional hazards mixed effect model, Gaussian random effects to account for study variability).
• Pooled sex-specific Hazard Ratio’s for each Lancet Commission 2020 risk factor and the risk of all-cause dementia (mixed-effect Cox regression models accounting for the study as a random effect, age- and education-adjusted). An interaction term was then fitted for each risk factor and sex to calculate women-to-men ratios of hazard ratios (RHRs).
• Sex-specific age-adjusted dementia incidence rates per 1,000 person-years (Poisson regression).
• Subgroup analyses for whether sex differences varied by prespecified subgroups: age (≥ or < 80 years), education (≥ or < 9 years), birth cohorts (born before 1925, 1925-1934, 1935 and after), country-level income, region, and Apolipoprotein E (APOE) genotype. Certain APOE genotypes are known to increase the risk of dementia.
• Sex-specific dementia-free survival probability (Kaplan-Meier survival curve).

This study analysed data from the Cohort Studies of Memory in an International Consortium (COSMIC) collaboration, which synchronises international cohort studies of cognitive ageing with the aim to boost their comparability and statistical power.

Results

Of 29,850 eligible participants from 21 studies across 18 countries, 2,089 developed all-cause dementia (66% women) over a median of 4.6 years of follow up (range 0.01 to 19.6 years). The largest cohorts were from Spain (n= 3,237) and China (n= 2,097). The smallest cohorts were from the Netherlands (n= 519) and Sweden (n= 732). 64% of analysed participants were from HICs, 29.3% from upper-middle-income countries, and 6.6% from low-income countries.

Sex differences for all-cause dementia:

• The age-adjusted incidence rate of dementia was higher in women (16.4 per 1,000 person-years, 95% CI 15.2 to 17.6) than in men (12.3 per 1,000 person-years, 95% CI 11.1 to 13.5).
• The pooled age- and education-adjusted risk of all-cause dementia was higher in woman than men (HR 1.12, 1.01 to 1.23). The pattern varied across individual cohorts: 14/21 studies reported risk to be higher in women, 7/14 studies reported the opposite association, and for many, the 95% CI crossed the line of no effect.
• However, in sensitivity analysis, if only cohorts using a Diagnostic and Statistical Manual of mental disorders (DSM) based dementia definition were included (15/21 studies), the age- and education-adjusted sex difference in dementia risk was no longer significant.
• Survival probability for all-cause dementia was higher for men than women (log-rank p value <0.0001).

Subgroup analysis:

• Consistently, the age-adjusted incidence rates for dementia were higher in women than men across all categories of the subgroups considered.
• Interestingly, when stratified by country-level income, incidence rates were highest among women from low to lower-middle income countries, but similar across country-level incomes for men.
• The greater risk in women was most pronounced in low to lower-middle income countries (HR 1.73, 1.25 to 2.39).
• By region, the greater risk in women was found in other countries (South America and Africa) (HR 1.65 1.29 to 2.11), but not Western (HR 1.05, 0.93 to 1.19) or Asian (HR 1.10, 0.86 to 1.40) countries.

Sex differences in risk factor associations:

• In men, lower BMI was associated with lower risk of dementia (HR 0.88, 0.83 to 0.93) and higher LDL cholesterol was associated with higher risk of dementia (HR 1.07, 1.02 to 1.13), but these associations were not replicated in women.
• Among all risk factors considered, sex differences were only clear for former alcohol use, with the relationship being stronger in men than women (women to men RHR 0.55, 0.38 to 0.79). A less strong sex difference was found for years spent in education, with a stronger protective effect for women (women-to-men RHR 1.01, 1.00 to 1.03).
• There was a trend towards a stronger association between APOE e4 carriage and dementia in men than in women, but the women-to-men RHR included the no effect value (0.88, 0.72 to 1.07). APOE e4 carriage is a risk factor for Alzheimer’s disease.

Conclusions

The authors conclude:

A greater risk of dementia for women was observed, contributing to accruing evidence for sex differences in dementia risk.

They found almost no evidence of sex differences for most risk factors for dementia and that the excess dementia risk in women was more pronounced in poorer countries. The authors suggest that their study justifies ongoing efforts to support programmes to improve sex and gender equity in brain health throughout the life course.

Gong et al. observed a greater risk of dementia for women, contributing to accruing evidence for sex differences in dementia risk. They found almost no evidence of sex differences for most risk factors for dementia and that the excess dementia risk in women was more pronounced in poorer countries.

Strengths and limitations

A major strength of this study is its geographic and ethnic diversity, with data from 21 cohorts across six continents. This should increase the global representativeness of the study’s results and includes participants that have traditionally been excluded from dementia research. The study was also large, including a total of 29,850 participants, allowing improved precision of estimates. Authors were able to harmonise multiple risk factors across cohorts.

However, the breadth of available data was also a limitation of the study. Protocols, population characteristics, and timing of studies varied across cohorts. In particular, the authors point out the variability in dementia definition and measurement tools used. When authors carried out a sensitivity analysis examining only cohorts using the DSM definitions, there was no evidence of a difference in risk of dementia by sex.

Variability was also a limitation for harmonisation of risk factors. For example, a clinical diagnosis of depression is not comparable to self-reported depressive symptoms. Several cohorts did not contain data at all for specific risk factors, such as hearing loss. In addition, as a result of harmonisation many variables lost nuance.

Effect estimates for sex and dementia risk varied widely by cohort, even when these were in the same country, such as two Brazilian cohorts that found opposite results (HR 1.51 and 0.54). According to Cochrane, a pooled effect estimate might not be informative when there is high variability and “inconsistency in the direction of effect” (Deeks and Altman, 2023). The authors could have explicitly tested for differences between cohorts using meta-regression to explore this inconsistency.

Though one of the major strengths of the study was the use of a global population, only two included cohorts and 6.6% of the total included population came from low-income countries. Many of the cohorts had high loss to follow-up, though this didn’t appear to differ between men and women, apart from in the Gothenburg study. Loss to follow up was highest among people born before 1925 and people in low-income cohorts.

The authors were able to explore country-level economic data but these were unfortunately not available at the participant level. Ideally, analyses would have adjusted for socioeconomic factors at the individual level and compared these between men and women, as this is a potential important confounder of the association between sex and dementia.

Data harmonisation across datasets is a challenge for many individual participant data meta-analyses, with important variables losing valuable nuance.

Implications

This is an important study that contributes to the evidence-base for a higher incidence rate of dementia among women compared to men, and a pooled higher risk of dementia among women. Other research has attributed increased levels of dementia among post-menopausal women to a potential detrimental neurological effect of lack of oestrogens (Rahman et al, 2020), while androgens are purported to have neuroprotective effects (Cai et al, 2020). However, the inconsistency of effect estimates seen by Gong et al. (2022) across cohorts suggests that, if a difference in dementia risk between men and women truly exists, this may not be fully explained by hormonal or biological factors.

This study is particularly valuable as it includes data from a large globally representative consortium of studies (COSMIC), at a time when evidence on dementia from LMICs is sorely lacking. The authors found that the higher risk of dementia among women was strongest in LMICs. However, the low proportion of participants from low-income countries in this study reinforces the low amount of available data in these settings. Data from England has shown that social disadvantage and adversity are risk factors for dementia regardless of sex. Gong et al. (2022) argue that socioeconomic and gender-related risk factors might be more pronounced among women than men in LMICs, leading for example, to differential access to education and occupational attainment, which might in turn modulate dementia risk.

The authors report effect modification by sex for former alcohol use and longer education. However, the 95% confidence interval of the RHR for education contained the null (1.00 to 1.03), suggesting there is no evidence of an association. In addition, the prevalence of former alcohol use was likely low as the confidence intervals for these estimates were wide, reducing our certainty in this result. The Lancet Commission cites excessive alcohol use as a risk factor for dementia but Gong et al (2022) were not able to categorise alcohol use into mild, moderate, or excessive consumption. Unfortunately, like many other risk factors examined in this COSMIC population, the available data when harmonised did not allow for exploration of nuances such as amount of alcohol consumed.

Overall, the authors found a pooled higher risk of dementia among women, though results were inconsistent. It was not possible to explain the higher risk of dementia in women through different effects of risk factors by sex. However, the frequency of these risk factors varies by sex and a helpful next step might be to calculate the population attributable fraction (PAF) separately for women and men to help guide targeted public health initiatives. The authors conclude that efforts to increase health equity for women will help reduce dementia cases. Further work to disambiguate differences in cohorts and potential reasons for inconsistencies in associations is needed. In addition, this study reinforces the need to establish research cohorts and registries in LMICs to better understand dementia in these settings.

Gong et al. (2022) argue that socioeconomic and gender-related risk factors might be more pronounced among women than men in low-middle income countries, leading for example, to differential access to education and occupational attainment, which might in turn modulate dementia risk.

Links

Primary paper

Gong J, Harris K, Lipnicki DM, et al. (2023) Sex differences in dementia risk and risk factors: Individual-participant data analysis using 21 cohorts across six continents from the COSMIC consortium. Alzheimer’s Dement. 2023; 19: 3365–3378. https://doi.org/10.1002/alz.12962

Other references

1. Nichols E, Steinmetz JD, Vollset SE, et al. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. The Lancet Public Health 2022; 7(2): e105-e25.
2. Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 2020; 396(10248): 413-46.
3. Nebel RA, Aggarwal NT, Barnes LL, et al. Understanding the impact of sex and gender in Alzheimer’s disease: A call to action. Alzheimers Dement 2018; 14(9): 1171-83.
4. Prince M, Ferri CP, Acosta D, et al. The protocols for the 10/66 dementia research group population-based research programme. BMC Public Health 2007; 7: 165.
5. Sachdev PS, Lipnicki DM, Kochan NA, et al. COSMIC (Cohort Studies of Memory in an International Consortium): An international consortium to identify risk and protective factors and biomarkers of cognitive ageing and dementia in diverse ethnic and sociocultural groups. BMC Neurology 2013; 13(1): 165.
6. Deeks JJ HJ, Altman DG Chapter 10: Analysing data and undertaking meta-analyses. In: Higgins JPT TJ, Chandler J, Cumpston M, Li T, Page MJ, Welch VA ed. Cochrane Handbook for Systematic Reviews of Interventions version 64 (updated August 2023): Cochrane; 2023.
7. Rahman A, Schelbaum E, Hoffman K, et al. Sex-driven modifiers of Alzheimer risk. A multimodality brain imaging study 2020; 95(2): e166-e78.
8. Cai Z, Li H. An Updated Review: Androgens and Cognitive Impairment in Older Men. Front Endocrinol (Lausanne) 2020; 11: 586909.

Photo credits