Association of Metabolic Syndrome With Incident Dementia: Role of Number and Age at Measurement of Components in a 28-Year Follow-up of the Whitehall II Cohort Study

Metabolic syndrome (MetS) is defined by a cluster of interrelated risk factors, consisting of abdominal obesity, hypertriglyceridemia, low HDL cholesterol (HDL-C), elevated blood pressure, and elevated fasting glucose (1). It has been shown to be associated with risk of coronary artery disease (1,2) and stroke (3), possibly due to a synergistic effect of its components. MetS is amenable to change by lifestyle modification (4), making it an important target for preventive strategies.

MetS components are included in prevention guidelines for dementia (5,6), but whether MetS itself increases the risk of dementia remains uncertain. Most previous studies have used the standard definition of MetS (i.e., the presence of three or more risk factors). However, risk of dementia may increase for each additional MetS component (7), rather than only in those with three or more components. A meta-analysis of longitudinal studies (4) did not find an association between MetS and all-cause dementia, but there was some evidence of an increased risk of vascular dementia, although most studies had a follow-up duration of <10 years as MetS was assessed in late life. More recent studies with a follow-up of up to 10 years show an increased risk of all-cause dementia in individuals with MetS (7–9). It is possible that like obesity and high blood pressure (10,11), late-life MetS may not necessarily increase dementia risk. Two previous studies had a follow-up >20 years, but included only men (12,13) or were based on a small number of dementia cases (13).

Besides age at measurement of MetS components and limited follow-up in previous studies, the role of cardiovascular disease (CVD) in the association between MetS and dementia also remains unclear. The aim of our study, using data spanning ∼30 years, was to examine the role of age at measurement of MetS and its components for incident dementia. Using multistate models, a further aim was to examine whether MetS was associated with incident dementia in those free of CVD over the follow-up.

Data were drawn from the ongoing Whitehall II cohort study, established in 1985–1988 among 10,308 London-based civil servants aged 35–55 years (14). Participants responded to a comprehensive questionnaire and underwent a structured clinical examination at recruitment and thereafter every 4 to 5 years. Data on MetS components were first collected during the 1991–1993 study wave and repeated in 1997–1999, 2002–2004, 2007–2009, 2012–2013, and 2015–2016. In addition, linkage to electronic health records of the U.K. National Health Service (NHS) was used to record data on health outcomes over the follow-up for all but 10 participants until 31 March 2019. Research ethics approval and written informed consent from participants were renewed at each contact; the latest approval was from the Joint University College London/University College London Hospital Committee on the Ethics of Human Research (reference number 85/0938).

MetS was defined using the latest harmonized definition (1). Participants were classified as having MetS when they met three or more of the following metabolic criteria: 1) elevated waist circumference (waist circumference ≥102 cm in men and ≥88 cm in women); 2) elevated triglycerides (serum triglycerides level ≥150 mg/dL [1.7 mmol/L] or use of lipid-modifying drugs); 3) low HDL-C (<40 mg/dL [1.0 mmol/L] in men and <50 mg/dL [1.3 mmol/L] in women or use of lipid-modifying drugs); 4) elevated blood pressure (systolic blood pressure ≥130 mmHg and/or diastolic blood pressure ≥85 mmHg or use of antihypertensive drugs); and 5) elevated fasting glucose (serum fasting glucose level ≥100 mg/dL [5.6 mmol/L] or use of glucose-lowering drugs).

Waist circumference was measured as the smallest circumference at or below the costal margin. Fasting venous blood samples were collected and centrifuged, and serum was stored in aliquots at −80 °C. Blood pressure was the mean of two measurements using a sphygmomanometer with the participant in a sitting position after 5 min of rest. Use of medication was reported by participants.

MetS components were assessed six times between 1991 and 1993 and 2015 and 2016, and, for each participant, we extracted data on all components at <60 (range 40–59.9 years), 60 to <70 (range 60–69.9 years), and ≥70 years of age (range 70–84 years). When data were available at several time points within an age range, the measure closest to 55, 65, and 75 years of age was chosen for the <60, 60 to <70, and ≥70 age groups, respectively.

Dementia cases were ascertained until 31 March 2019, by linkage to three national registers (the national Hospital Episode Statistics, the Mental Health Services Data Set, and the National Statistics Mortality Register) using the unique NHS identification number. ICD-10 codes F00-F03, F05.1, G30, and G31 were used to identify dementia cases. The NHS provides most of the health care in the country, including in- and outpatient care. Assessment of dementia based on Hospital Episode Statistics data has a sensitivity and specificity of 78.0% and 92.0%, respectively (15). The sensitivity in our study is likely to be higher, as we also used data from the Mental Health Services Data Set and the mortality register. Date of dementia was set at the first record of dementia diagnosis using all three data sets.

CVD included stroke (Multinational Monitoring of Determinants and Trends in Cardiovascular Disease [MONICA]-Augsburg stroke questionnaire; ICD-10 codes I60–64), coronary heart disease (CHD) (12-lead resting electrocardiogram recording; ICD-10 codes I20–25 or procedures K40–49, K50, K75, and U19), and heart failure (ICD-10 code I50).

Sociodemographic variables included age, sex, ethnicity (White and non-White), and education (university or higher degree, secondary school, or lower secondary school or less). Lifestyle factors included smoking (never, former, and current smoker), alcohol consumption (no consumption, 1–14 units/week, and >14 units/week), consumption of fruits and vegetables (less than daily, once a day, and twice or more a day), and time spent in moderate and vigorous physical activity (hours per week). Data on covariates were extracted in a similar manner to measure of MetS components at <60, 60 to <70, and ≥70 years of age.

All participants were included in the analyses except those without linkage to electronic health records or with incomplete data on MetS components or covariates and prevalent dementia at the start of follow-up in the three age-based analyses (Supplementary Fig. 1). Having verified the proportional hazards assumption, we used Cox proportional regression with age as the timescale and incidence of all-cause dementia as the outcome. Age at entry in the follow-up was age at assessment of MetS components and covariates. Participants were censored at date of record of dementia, death, or end of follow-up (31 March 2019), whichever came first. Cause-specific hazard models were used to account for competing risk of death. All analyses were first adjusted for sex, education, ethnicity, and birth-cohort effects using 5 year groups (model 1) and subsequently for smoking, alcohol consumption, fruits and vegetables consumption, and physical activity (model 2).

The analyses consisted of first examining the association between each individual component of MetS measured at <60, 60 to <70, and ≥70 years of age (separate models) with incidence of dementia. Then, the association between the number of MetS components on a scale from zero to five (categorical variable) at <60, 60 to <70, and ≥70 years (separate models) and risk of incident dementia was examined with participants without MetS components as the reference. We repeated these analyses using the number of MetS components as a linear variable to estimate the increase in dementia risk associated with one additional component. We subsequently examined whether alternate cutoff points to define “high metabolic risk” influenced the association with incident dementia using one or more components and then two or more and three or more components (current clinical definition of MetS) at <60, 60 to <70, and ≥70 years of age. We prefer the term “high metabolic risk” rather than MetS, as one or even two components cannot be considered to represent a “syndrome.” In these analyses, the reference was composed of participants without metabolic risk in each definition of high metabolic risk.

In the final analyses, we used multistate models (Weibull distribution) to examine the role of CVD (stroke, CHD, or heart failure) over the follow-up in the association between high metabolic risk and dementia. These analyses were based on participants free of CVD and dementia at baseline, and to allow a sufficient number of CVD and dementia cases in these analyses, they were only undertaken using MetS components assessed at <60 years of age. As previously, high metabolic risk was defined as presence of one or more, two or more, or three or more MetS components. The multistate models allow simultaneous estimation, expressed as a hazard ratio (HR), of the transition from: 1) high metabolic risk to CVD; 2) CVD to dementia; and 3) high metabolic risk to dementia in those free of CVD over the follow-up. Age was used as the timescale, and analyses were adjusted for all covariates.

To examine the robustness of findings, we performed several sensitivity analyses. First, the shape of the association between the number of metabolic syndrome components and incidence of dementia was examined using restricted cubic splines with three knots, Stata command xblc, with zero components as the reference. Second, we repeated the primary analyses using inverse probability weighting to account for missing data (16). This involved first calculating the probability of being included in the analytical sample using logistic regression that included demographic, socioeconomic, and behavioral factors, as well as MetS components at the 1991–1993 wave, chronic diseases during the follow-up including dementia, and stepwise-selected interactions between covariates. The inverses of these probabilities were used as weights in the Cox regression.

Multistate models were performed using R software. All other analyses were undertaken using STATA version 16.1 (StataCorp). A two-sided P value <0.05 was considered statistically significant.

Whitehall II data cannot be shared publicly because of constraints dictated by the study’s ethics approval and institutional review board restrictions. The Whitehall II data are available for sharing within the scientific community. Researchers can apply for data access at https://www.ucl.ac.uk/epidemiology-health-care/research/epidemiology-and-public-health/research/whitehall-ii/data-sharing.

Of the 10,308 participants recruited to the study in 1985–1988, 159 (1.5%) died before 40 years of age and 1,708 (16.6%) had dropped out of the study when they were 40–59.9 years of age. We also excluded 1,175 (11.4%) participants with missing data on MetS components and 1 with missing data on covariates, leading to 7,265 (70.5%) participants free of dementia in the analyses at <60 years (mean age at clinical examination was 55.1 [SD 2.9] years) (flowchart in Supplementary Fig. 1). Over a mean follow-up of 19.6 (SD 5.9) years, 393 (5.4%) incident dementia cases were recorded. Mean age at dementia diagnosis was 74.7 (SD 6.3) years. Supplementary Fig. 1 also describes the sample selected in the analysis for MetS at 60 to <70 and ≥70 years of age. Participants’ characteristics at <60, 60 to <70, and ≥70 years of age, overall and according to incident dementia, are shown in Supplementary Table 1. In general, participants with dementia had lower education and were more likely to be women and of non-White ethnicity. There was no interaction between MetS components and sex or ethnicity (all P values for interaction >0.05), leading to analyses without stratification.

Table 1 presents the results of the analyses on individual MetS components measured at <60, 60 to <70, and ≥70 years with incident dementia over a mean (SD) follow-up of 19.6 (5.9), 10.9 (5.8), and 5.7 (3.2) years, respectively. When measured at <60 years, elevated waist circumference (HR 1.39 [95% CI 1.07, 1.81]), low HDL-C (HR 1.30 [95% CI 1.02, 1.66]), and elevated blood pressure (HR 1.34 [95% CI 1.09, 1.63]) were associated with a higher dementia risk in analyses adjusted for all covariates. For MetS components measured at 60 to <70 years, low HDL-C (HR 1.26 [95% CI 1.02, 1.57]) and elevated fasting glucose (HR 1.40 [95% CI 1.12, 1.74]) were associated with dementia in the fully adjusted analyses. For MetS components measured at ≥70 years, only elevated fasting glucose was associated with incident dementia in the fully adjusted analyses (HR 1.38 [95% CI 1.07, 1.79]).

Table 2 shows the association between the number of MetS components with incident dementia. Treating the MetS scale as a linear variable showed that a 1 point increment in MetS score at <60 (HR 1.13 [95% CI 1.05, 1.23]) and 60 to <70 years (HR 1.08 [95% CI 1.00, 1.16]) but not at ≥70 years of age (HR 1.04 [95% CI 0.96, 1.13]) was associated with higher dementia risk in analyses adjusted for all covariates. Linearity of associations using cubic splines was found for MetS components measured at <60 (Supplementary Fig. 2A) and 60 to <70 (Supplementary Fig. 2B), but not ≥70 years of age (Supplementary Fig. 2C).

Analyses on alternative dichotomous definitions of high metabolic risk are shown in Table 3. High metabolic risk, defined as the presence of one or more components at <60 (HR 1.38 [95% CI 1.09, 1.74]) and 60 to <70 years of age (HR 1.35 [95% CI 1.05, 1.73]), was associated with incident dementia compared with those without any MetS components. When high metabolic risk was defined as presence of two or more components, an increased risk of dementia was observed at <60 years of age, with an HR (95% CI) of 1.32 (1.08, 1.62) but not for high metabolic risk in the fully adjusted analyses in those 60 to <70 and ≥70 years of age. Analyses of high metabolic risk defined as the presence of three or more components (the current definition of MetS) showed no association with incident dementia in the fully adjusted analyses, irrespective of the age of measurement of MetS components. Changing the reference group to those with no MetS components showed MetS (three or more components) at <60 years (HR 1.50 [95% CI 1.11, 2.02]) and 60 to <70 years (HR 1.40 [95% CI 1.04, 1.89]) to be associated with higher risk of dementia (Supplementary Table 2).

Results from the multistate analyses (Fig. 1) showed high metabolic risk at <60 years of age defined as the prevalence of one or more (HR 1.99 [95% CI 1.08, 3.66]) and two or more (HR 1.69 [95% CI 1.12, 2.56]) components, but not MetS itself (three or more components; HR 1.51 [95% CI 0.99, 2.29]) to be associated with higher risk of dementia in those free of CVD during the follow-up. As expected, high metabolic risk was associated with higher risk of incident CVD, irrespective of the number of components used to define risk (pathway A, Fig. 1).

Inverse probability weighting to account for missing data yielded results similar to those in the main analyses (Supplementary Fig. 3 and Supplementary Tables 3–5).

This longitudinal study using repeated measures of MetS components and incidence of dementia presents three key findings. First, the conventional case definition of MetS that requires three or more of five risk components was not associated with incident dementia, irrespective of age at assessment of MetS. The five MetS components measured before 70 years of age, considered on a continuous scale, showed the risk of dementia to increase with each additional component. Second, individual MetS components were more strongly associated with incident dementia when measured before 60 years rather than after 70 years. Third, the association between high metabolic risk, defined as one or two components at high risk, and dementia was partly mediated by CVD, but the association was also present in those free of CVD over the follow-up. Taken together, these results highlight the importance of controlling MetS risk factors in midlife, particularly when the interest is in dementia at older ages.

The current evidence on the association between MetS and dementia is characterized by a single assessment of MetS (9,12,13,17–21), primarily in older adults (17–20). The follow-up in these studies was <10 years (4,7–9,17–22). Given the long preclinical phase of dementia (23), the results from these studies may be prone to reverse causation biases, as changes in multiple metabolic processes may occur during this phase. A long follow-up, such as in the current study spanning 28 years, allows the assessments of MetS and onset of dementia to be separated by over two decades, minimizing the risk of reverse causation bias. A 25-year follow-up study on 3,555 men aged 45–68 years at baseline showed higher dementia risk in participants with a cluster of cardiometabolic risk factors (12). However, that study did not use a standardized definition of MetS, and competing risk of death was not considered. Another study on 1,225 men, mean age 52.8 years at baseline, showed an imprecisely estimated, statistically nonsignificant association between MetS (using the International Diabetes Federation definition) and incident dementia over a 20 year follow-up, possibly due to small number of dementia cases (N < 100) in the analyses (13). Studies with a short follow-up that have shown an association with dementia used a retrospective measure of history of MetS (22) or were based on Asian populations (7–9). The typical approach consists of including participants with a wide age range at baseline. The disadvantage of this approach is that older participants at baseline who are more likely to develop dementia during the short follow-up are also more likely to be in the preclinical phase of dementia. Two previous studies performed age-stratified analyses (8,20): one showed no association between MetS and all-cause dementia before or after 75 years of age (20), and the other study on 4,106,590 Korean participants found higher dementia, Alzheimer disease, and vascular dementia risk in participants <65 and >65 years of age (8).

The current study adds to the understanding of the association between MetS and dementia due to three novel features. First, alternative thresholds to define high metabolic risk were used, and findings show increased risk of dementia to start with the presence of one MetS component. Second, assessment of MetS components in midlife and later life allowed the examination of the role of age at prevalence of metabolic risk for incident dementia at older ages. Third, the use of multistate models permitted the role of CVD in the association between high metabolic risk and incident dementia to be examined.

Our results suggest that the current definition of the MetS (i.e., presence of three or more components) may not be optimal to define dementia risk, as it is already present in those with one or two risk factors. The comparison group, in studies using the conventional MetS definition, involves comparing incidence of dementia in those with three or more risk factors to those with zero to two risk factors. The heterogeneity in metabolic risk in the reference category offers a possible explanation for inconsistent results in previous studies on MetS and risk of dementia, irrespective of the age at assessment of MetS (4,13,17–21). A previous study on 1,492,776 Korean adults, mean age 53.1 years at baseline, also found higher risk of dementia in those with one or more MetS components (7). Our results showed an increased risk of dementia with an increasing number of MetS components starting with one component, notably at <60 years of age, highlighting the importance of all MetS components in midlife for risk of dementia at older ages.

Individual components of MetS have previously been associated with dementia, notably when assessed in midlife, highlighting the role of age in the association between cardiometabolic risk factors and the risk of late-life dementia. In the current study, high waist circumference, decreased HDL-C, and elevated blood pressure, when prevalent before 60 years but not after 70 years of age, were associated with risk of dementia when considered individually. Obesity/high waist circumference (11,24) and hypertension (10) in mid- but not late life are known to be associated with an increased dementia risk. Similar to our results, a study of >8,000 men and women found higher dementia risk among those with elevated systolic blood pressure at 50 but not at 70 years of age (25). Obesity and central obesity, defined by high BMI and elevated waist circumference, respectively, have also been shown to be associated with a higher risk of late-onset dementia when measured at 50 but not after 60 years of age (11). Our results show low HDL-C, but not elevated triglycerides, to be associated with incident dementia. Although dyslipidemia, particularly in midlife, has been associated with an increased risk of dementia in previous studies (7–9,12,18), the precise blood lipid that is pertinent for dementia remains the subject of debate (26). An increment of 15 mg/dL in blood glucose measured during middle (51–60 years) but not in late adulthood (≥61 years) has previously been associated with higher risk of Alzheimer disease (27). Nevertheless, elevated fasting glucose in our analyses was associated with dementia only after 60 years of age; it is possible that the explanation lies in the threshold for fasting glucose used in MetS definition that combines individuals with prediabetes and diabetes. There is now robust evidence that diabetes, but not prediabetes, is associated with risk of dementia (5,28), possibly driven by poor glycemic control (29).

Given the increased risk of dementia among those free of CVD during follow-up, our results suggest that the association between MetS and dementia is not fully explained by CVD. The mechanisms linking MetS and dementia are likely to be due to multifactorial pathogenesis linked to its components and based on both vascular injury and neurodegeneration (30). All MetS components may jointly contribute to dementia through the development of atherosclerotic lesions and/or microvascular dysfunction (30,31). Both elevated glucose and obesity have been associated with insulin resistance, as well as with low-grade systemic inflammation, which could increase the expression of proinflammatory cytokines, contributing to neurodegeneration and neurotoxicity leading to dementia (30,32).

The strengths of this study include repeated, objective measures of MetS components from midlife to late-life, long follow-up duration for incident dementia, and linkage to multiple electronic health records for dementia ascertainment. A further strength was the use of inverse probability weighting to account for missing data.

This study also has several limitations. First, data come from the Whitehall II study, a longitudinal cohort study in which participants are known to be healthier than the general population. While this setting precludes estimation of the incidence and prevalence of risk factors and disease outcomes, it is unlikely to affect associations between risk factors and disease of interest (33), as has been previously using these data (34). Second, data on dementia subtypes were incomplete and did not allow analyses of MetS and its components. Third, analyses on MetS components measured after 70 years and subsequent dementia were based on a relatively small number of events, which may have led to an imprecise estimate. Fourth, we cannot exclude the possibility of residual confounding, although the analyses were adjusted for a broad range of sociodemographic and behavioral covariates.

In conclusion, the current study suggests that MetS components when present before 60 years of age are associated with an increased risk of late-onset dementia. Our results also suggest that the conventional definition of MetS, requiring the prevalence of three or more components, is not optimal, as risk for dementia is linear, with risk accumulating over the entire scale of MetS components. Dementia is a public health problem with tremendous personal and societal implications; our results show the importance of targeting all metabolic risk factors rather than the presence of metabolic risk clusters.

Acknowledgments. The authors thank all of the participating civil service departments and welfare, personnel, and establishment officers, the British Occupational Health and Safety Agency, the British Council of Civil Service Unions, all participating civil servants in the Whitehall II study, and all members of the Whitehall II study team. The Whitehall II study team comprises research scientists, statisticians, study coordinators, nurses, data managers, administrative assistants, and data entry staff, who make the study possible.

Funding. The Whitehall II study has been supported by grants from the National Institute on Aging, National Institutes of Health (R01AG056477 and R01AG062553), UK Medical Research Council (R024227 and S011676), and the Wellcome Trust (221854/Z/20/Z). M.K. is supported by the Academy of Finland (311492), Helsinki Institute of Life Science, and NordForsk. S.S. is supported by the French National Research Agency (ANR-19-CE36-0004-01).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.D.M.-F., S.S., and A.S.-M. were responsible for conceptualization. M.D.M.-F., A.F., S.S., and A.S.-M. were responsible for methodology. M.D.M.-F., A.F., M.S.Y., T.T.v.S., A.G.T., M.K., S.S., and A.S.-M. performed investigation. M.D.M.-F., A.F., S.S., and A.S.-M. were responsible for validation. M.D.M.-F. and A.F. were responsible for formal analysis. M.D.M.-F. and A.F. were responsible for access and verified the data. A.F. performed data curation. M.D.M.-F. and A.S.-M. performed writing of the original draft preparation. M.D.M.-F., A.F., M.S.Y., T.T.v.S., A.G.T., M.K., S.S., and A.S.-M. were responsible for review and editing of the manuscript. M.D.M.-F., A.F., S.S. and A.S.-M. were responsible for visualization. S.S., and A.S.-M. performed supervision. M.K. and A.S.-M. were responsible for funding acquisition. A.S.-M. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.