Abstract
Socio-economic status (SES) and biological aging are risk factors for dementia, including Alzheimer’s disease, however, it is less clear if the associations with SES vary sufficiently across different biological age strata. We used data from 331,066 UK Biobank participants aged 38–73 with mean follow-up of 12 years to examine if associations between SES (assessed by educational attainment, employment status and household income) and dementia and Alzheimer’s disease are modified by biological age (assessed by leucocyte telomere length: LTL). Diagnosis of events was ascertained through hospital admissions data. Cox regressions were used to estimate hazard ratios [HRs]. A consistent dose–response relationship was found, with participants in low SES and shorter LTL strata (double-exposed group) reporting 3.28 (95% confidence interval [CI] 2.57–4.20) and 3.44 (95% CI 2.35–5.04) times higher risks of incident dementia and Alzheimer’s disease respectively, compared to those of high SES and longer LTL (least-exposed group). Of interest is a synergistic interaction between SES and LTL to increase risk of dementia (RERI 0.57, 95% CI 0.07–1.06) and Alzheimer’s disease (RERI 0.79, 95% CI 0.02–1.56). Our findings that SES and biological age (LTL) are synergistic risk factors of dementia and Alzheimer’s disease may suggest the need to target interventions among vulnerable sub-groups.
Similar content being viewed by others
Introduction
Dementia is a syndrome with multiple pathologies characterized by progressive neurodegeneration leading to cognitive and functional decline. It is estimated to affect 57 million people globally1. For the UK, estimated social and healthcare costs are currently around £35 billion, rising to £80.1 billion by 20402,3. It is also estimated that 35% of population dementia risk is attributable to modifiable social and environmental risk factors, including socioeconomic status (SES)4.
The role of SES in elevating dementia risk is well documented5. Epidemiologic studies have shown that higher education attainment (as a cognitive-stimulating activity)6,7 and household income (permitting more opportunities for healthy lifestyle)8,9 are protective for dementia. Conversely, low occupational employment was found to be associated with increased dementia risk10. Less clear, however, are the biological mechanisms underlying these associations.
Biological aging is the progressive loss of physiological function relative to chronological age, and captures the potential progressive impairment of metabolic and physiologic capabilities of human body at cellular and molecular levels11. Telomere length (TL) has been suggested as an indicator of biological aging. Telomeres are thousands of repeated TTAGGG hexa-nucleotide sequences that cap the ends of chromosomes and are responsible for chromosome stability. Being located at the tip of chromosomes, telomeres are not fully replicated during DNA replication leading to an attrition in length accompanying each successive cell division. Variation in TL between individuals can be used as a marker of biological aging12. Attrition in length also stems from the effects of life stressors13, inflammation, DNA damage, ageing, and lifestyle factors such as smoking14. Specifically, socioeconomic disadvantage constitutes an important marker of psychosocial stress and several studies have found evidence of associations between socioeconomic disadvantage and telomere shortening15,16. Evidence from the National Health and Nutrition Examination Survey (NHANES) study comprising n = 5,360 participants aged 20–84 years showed that participants completing high school education had relatively shorter telomeres than college graduates17. Consistent results were obtained from a sample of n = 416 participants aged 53–76 years from the Whitehall II epidemiological cohort, with lower educational attainment being associated with shorter leucocyte telomere length (LTL)18. In a study of n = 1,552 female twins aged 18–75 years, individuals belonging to lower SES category was found to have shorter telomeres relative to peers in higher SES, with the white blood cell telomeres in the lower SES category being shorter on average by 140.3 base pairs after accounting for other risk factors such as body size, smoking and physical activity19. A study of n = 1,026 found that higher parental SES was associated with 1.8% longer TL among newborn boys20. Another cross sectional study of n = 341 mixed race participants found that younger adults with higher burden of discrimination (race and sex) had shorter TL, while similar associations were reported among female from higher SES strata21. A recent longitudinal study of n = 1,031 participants from the Multi-Ethnic Study of Atherosclerosis (MESA) reported that improvements in neighbourhood-level SES were protective on the rate of telomere attrition22.
Telomere length (TL) has also been shown to be directly associated with risk of dementia and Alzheimer’s disease (see Supplementary Table S1 for study details). Briefly, one prospective study comprising n = 1,973 subjects aged ≥ 65 years from New York city found each kilobase pair attrition in TL was associated with 21% higher risk of dementia23. Consistently, longer TL was found to be protective on dementia risk among stoke survivors, aged ≥ 75 years in the UK24. A systematic review of 13 studies found consistent evidence of shorter TL among 860 Alzheimer’s disease patients relative to 2,022 controls with standardized mean difference of − 0.984 (95% CI −1.433 to −0.535)25. A study from Adelaide, Australia reported shorter TL in buccal and white blood cells in clinically diagnosed Alzheimer’s patients relative to healthy age-matched controls26. Recently, a two-sample Mendelian randomization study leveraging on summary statistics data extracted from largescale GWAS inferred that longer leucocyte TL have a protective effect on risk of Alzheimer’s disease27. Nonetheless, one large prospective study comprising n = 1,961 Dutch participants reported an U-shaped association, with higher risks of Alzheimer’s disease in the lowest and highest tertiles of TL relative to the middle tertile28. While, another cross sectional study comprising n = 2,210 participants demonstrating exceptional longevity from USA and Denmark reported null associations between leucocyte TL and cognitive performance in both cognitively unimpaired and demented sub-groups29.
Although independent positive associations of lower SES and shorter TL with risk of dementia have been established in several previous studies, there has thus far been no study examining synergistic interactions to inform targeted interventions for vulnerable subgroups. Here we aim to investigate how SES and TL interact to affect dementia risk. Using large-scale data from UK Biobank, we examine the effect-modification of biological age as measured by LTL on the associations of a composite measure of SES comprising highest educational attainment, employment status, and household income with incident dementia and Alzheimer’s disease.
Methods
Study population
This prospective study included participants from the UK Biobank, a nationwide population-based cohort of over 500,000 adults aged 37–73 years at baseline30. Participants were recruited between April 2006 and December 2010 by postal invitation from the National Health Service patient registry (response rate of 5.5%) within approximately 25 miles of 22 assessment centres across England, Scotland and Wales. Participants completed touchscreen questionnaires, underwent nurse-administered face-to-face interviews, had anthropometric assessments, and provided biological samples (blood, saliva and urine). Detailed protocol of the UK Biobank are available online31.
Study variables
Ascertainment of clinically diagnosed dementia and Alzheimer’s disease
Date and cause of hospital admissions were derived from Hospital Episode Statistics Admitted Patient Care (England), Patient Episode Database (Wales), and General/ Acute Inpatient and Day Case—Scottish Morbidity Record (Scotland). Incident clinically diagnosed all-cause dementia diagnosis was defined as primary or secondary diagnosis of dementia from the International Classification of Diseases—10th Revision (ICD 10), coded as F00 (Dementia in Alzheimer disease), F01 (vascular dementia), F02 (dementia in other diseases), F03 (unspecified dementia), G30 (Alzheimer disease), G31 (other dementias) or A81 (Sporadic Creutzfeld Jacob disease). Incident Alzheimer’s disease diagnosis was defined as ICD10 code F00 (Dementia in Alzheimer disease) or G30 (Alzheimer Disease) (see Appendix Table S3). Follow-up was censored at the date of hospital admission, date of all-cause death (from death certificates), or study end date (31 March 2021), whichever occurred earlier.
Measurement of composite socioeconomic status
Individual-level SES, our primary exposure was derived from a latent class model based on three socioeconomic factors; highest educational attainment, employment status and household income in accordance with a prior study32. Latent class analysis (LCA) is a well-validated data-driven technique that employs observed categorical domain variables to identify unobserved latent variable with mutually exclusive latent classes33. The three socioeconomic factors were collected from cohort participants based on structured self-reported questionnaires. Highest educational attainment was included as a seven-category variable based on the International Standard Classification of Education in the latent class model and categorized as college or university degree, A levels/AS levels, O levels/GCSEs, CSEs, NVQ or HND or HNC, other professional qualifications (for example nursing, teaching), and none of the above equivalent to less than high school. Employment status was a three-category variable coded as employed or self-employed, retired, and others (which included looking after home, being unable to work due to sickness or disability, unemployed, doing unpaid or voluntary work or student). Average total household income before tax was divided into five groups (< £18,000, £18,000 to 30,999, £31,000 to 51,999, £52,000 to 100,000, and > £100,000). Three SES classes (low, medium and high) were subsequently identified based on item-response probabilities using gsem command in Stata34. Maximum likelihood method, which permits the use of all data available including those with missingness, was employed in the LCA models. Iterations were undertaken before the final model was selected. Model selection was determined by using Bayesian information criteria (BIC) and Akaike information criterion (AIC). The convergence and performance of the latent class models were checked for two, three and four or more classes32. In our iterations, models with two and three classes performed the best. The AIC and BIC values for the latent class model with two classes was reported as 3.78 × 106 (both statistic), while that for the model with three classes are 3.76 × 106. Based on these fit statistic, we selected the latent class model with three classes to define individual-level SES for this study. The mean posterior probability values of the latent class model with three classes are 0.73, 0.81 and 0.84 for classes 1, 2 and 3 respectively (Supplementary Methods S1 and Table S2).
Measurement of biological age
Leucocyte telomere length (LTL), expressed as the ratio of telomere repeat copy number to the single copy gene (T/S), was extracted from peripheral blood leucocytes of the UK Biobank participants and measured via validated qPCR methods35. LTL was adjusted for technical parameters (including enzyme, PCR machine, primer, operator, temperature, humidity, hours from 6 a.m., pipetting robot and extraction method), loge transformed and Z-standardized. We developed a composite measure of biological age which accounted for the potential correlations between chronological age and LTL as well as the combined effects of inflammation and chronic diseases. As in previous study36, biological age was defined as residuals of LTL additionally adjusting for chronological age, serum C-reactive protein (from immunoturbidimetric high-sensitivity analysis) and any of the 46 curated chronic comorbidities (Supplementary Table S4) comprising cardiovascular diseases, cancer, diabetes, hypertension, psychiatric disorders and respiratory diseases. Serum CRP value expressed as mg/L was measured by immunoturbidimetric high-sensitivity analysis with the use of a Beckman Coulter AU580037.
Covariates
Following a socio-ecological perspective, model covariates were identified a priori, informed from published literature4. Demographic covariates included age at baseline, sex and ethnicity. A composite healthy lifestyle score was developed based on smoking status, alcohol intake frequency and dietary factors assessed at baseline as in previous literature38,39,40. A score of 1 was assigned to participants who achieved the healthy standard in each lifestyle component (non-current smoker; alcohol not consumed daily/almost daily; consumed ≥ 4 of 7 types of food following dietary recommendations as previously). The healthy score ranged between 0 and 3 was subsequently categorized as favorable (a score of 3), intermediate (a score of 2) and unfavorable (a score of 0 or 1) following the data distribution. The number of leisure and social activities per week was coded as none, one or more. Hearing difficulty was coded as self-reported deaf or having difficulty with hearing (coded as 1), and without hearing difficulty (coded as 0). Abdominal obesity was defined in terms of waist-hip ratio (WHR) ≥ 0.85 cm for women and ≥ 0.90 cm for men. Frailty was assessed at baseline from five phenotypes; weight loss, exhaustion, low grip strength, low physical activity and slow gait speed and coded as non-frail, frail/pre-frail41,42. Among the environmental variables, urbanicity was a composite measure of densities of housing, retail and public transport, and urban centrality around the participants’ residence (coded as quintiles 1, 2–4 and 5) as previously43,44. Neighbourhood deprivation (coded as quintiles 1, 2–4 and 5) was measured by the Townsend deprivation index (Supplementary Methods S2 for additional details).
Statistical analyses
Descriptive characteristics across the three SES categories are presented as mean and standard deviation for continuous variables, and frequency and percentage for categorical variables. We employed Nelson-Aalen hazard method to calculate the cumulative hazard for incident dementia based on SES categories (3 groups), and subgroups by SES and LTL strata (9 groups) over time.
Cox proportional hazard regression models estimating hazard ratios (HR) and 95% confidence interval (CI) for associations of SES and TL with risks of incident dementia and Alzheimer’s disease were developed. Person-time in months were calculated from the date of entry into the cohort until the date of hospitalization or date of death, or the study end date, whichever occurred earlier. The proportional hazards assumption was tested by using the Schoenfeld residual test. We excluded events prior to baseline assessment. To minimize chance of reverse causality, we performed landmark analyses by excluding events within 3 years of entry in to the study.
We examined independent associations of SES and LTL with incident dementia and Alzheimer’s disease diagnosis. HRs in the medium and low SES categories were reported in reference to the high SES category. Linear test was conducted to estimate the associational trend across the three exposure categories. To examine if biological age moderated the associations, we conducted stratified analyses by LTL. LTL was coded as a three-level categorical variable based on cut-offs at the 25th and 75th percentiles, with participants at the top 25th percentile with longer LTL taken as the reference group. Lastly, we examined joint associations between SES and LTL on incident events by stratifying participants into 9 groups by SES (low, medium and high) and LTL (long medium, short), with participants in the high SES and long LTL strata acting as the reference group. Biologic interaction on an additive scale was expressed as relative excess risk due to interaction (RERI) with the corresponding 95% CIs obtained by estimating the separate and joint effects of SES and LTL on incident dementia and Alzheimer’s disease. In the RERI calculations, we defined participants with low SES and short LTL as the double-exposed group, while those with high SES and long LTL strata as the reference category45,46. The models included age and sex, ethnicity, healthy lifestyle score, social activities, hearing difficulty, clinical phenotypes (abdominal obesity and frailty index), urbanicity, neighbourhood deprivation and SES/LTL. To fulfill the proportionate hazards assumption, sex was inputted using the strata option throughout the analysis.
We performed several sensitivity tests. First, we repeated our models with single socioeconomic factors (highest education attainment, employment status and household income) to examine their independent associations with incident dementia and Alzheimer’s disease. Second, we repeated our primary analyses using combined cases of both incident dementia diagnosis and mortality. Data on dementia mortality was extracted from death register, with the date of death obtained from death certificates held by the National Health Service (NHS) Information Centre (England and Wales) and the NHS Central Register in Scotland. Third, we reran the LCA model comprising the three individual socioeconomic factors as well as the Townsend deprivation index to additionally take into account the effect of neighbourhood deprivation within the composite SES measure. Fourth, we further reran analyses with adjustment for number of household members (coded as living alone, two, three, and more than three) to take into account variability in household size. Fifth, we reran models after additionally adjusting for comorbidities in the main models as covariates to examine robustness of our findings (with comorbidities being removed from the residual analysis for LTL). Sixth, for the employed subpopulation, we additionally reran models examining associations between occupation categories and incident dementia and Alzheimer’s disease. Occupation categories were coded as a five-factor variable comprising other occupations, managers and senior officials, professional occupations, associate professional and technical occupations, and administrative and secretariat occupations. Other occupations included manual occupations such as skilled trades occupations/personal service occupations/sales and customer service occupations/process, plant and machine operatives/elementary occupations, and was taken as a reference category. Seventh, we examined potential modifying effect of age by conducting stratified analyses by age groups (< 65 versus ≥ 65 years). Lastly, we performed multiple imputation by chained equations with 20 imputation sets47 to account for missing data across covariates on a sample of 422,907 participants with complete data on outcome and exposure variables. All covariates, exposure and outcome variables in the fully-adjusted model and the follow-up time were used to impute missingness.
Data were analyzed from November 2021 to August 2022. All analyses were performed in Stata software version 16.1.
Ethics approval and consent to participants
Ethical approval of UK Biobank was obtained from the National Health Service National Research Ethics Service (Ref: 11/NW/0382) and subsequently renewed (Ref: 16/NW/0274 and 21/NW/0157). We also obtained institutional ethical approval from The University of Hong Kong’s Human Research Ethics Committee (Refs: EA220302 and EA1904019). Informed electronic consent was acquired from all cohort participants when they visited the assessment centres.
Results
From the UK Biobank population sample of 502,650 individuals, participants were omitted from the analysis due to withdrawal (n = 191), diagnosed with dementia during the first three years of follow-up (the landmark period) or at baseline (n = 377), and missingness on one or more SES exposure (n = 79,175) or covariates (n = 91,841). The remaining 331,066 participants were used in a complete sample analysis (Supplementary Figure S1). There was no systematic difference between analytic sample and excluded sample (Supplementary Table S5). Over 3,989,206 person-years of follow-up (mean follow-up period = 12 years; IQR: 11.5–13.0 years), 3,313 incident cases of dementia were recorded, of which 1,423 were of Alzheimer’s disease.
Mean age of the sample at baseline was 56 (SD = 8), with 52% women (Table 1). The LCA model for composite SES categorized participants into three SES groups for high (n = 111,525, 34%), medium (n = 138,373, 42%) and low (n = 81,168, 25%) socioeconomic status. In general, participants with dementia belonged to the older age sub-group and with educational qualification less than high school, lower household income, obese and with lower grip strength, gait speed, physical activity levels, shorter leucocyte telomere length (LTL) and more comorbidities. The mean LTL of analytic sample was 0.01 (SD = 1.00, range = − 15.28–12.14). As expected older participants belonging to low SES strata had shorter LTL. The mean LTL among participants aged < 65 years and of low, medium and high SES strata were − 0.10 (SD = 1.00), 0.04 (SD = 0.99) and 0.15 (SD = 0.99) respectively. While, the mean LTL in sub-group aged ≥ 65 years and belonging to low, medium and high SES strata were − 0.30 (SD = 0.98), − 0.27 (SD = 0.98) and − 0.25 (SD = 0.97) respectively.
Cumulative hazards of incident dementia and Alzheimer’s disease across SES and LTL strata
Nelson-Aalen hazard estimates showed consistently higher cumulative risks of incident dementia for lower SES strata (Fig. 1a). The cumulative risks for incident dementia at 12 years were 2.1% (95% CI 1.99–2.20) for low SES, 0.88% (95% CI 0.83–0.94) for medium and 0.30% (95% CI 0.27–0.33) for the high SES stratum. Further stratification by SES and LTL revealed a cumulative risk for incident dementia of 2.33% (95% CI 2.12–2.56) in participants of low SES and short LTL, while it was only 1.67% (95% CI 1.49–1.87) for the low SES and long LTL category (Fig. 1b). A similar pattern was observed for Alzheimer’s disease with the cumulative risks at 12 years being 0.94% (95% CI 0.87–1.01) for low SES, 0.37% (95% CI 0.34–0.40) for medium and 0.12% (95% CI 0.10–0.14) for the high SES stratum.
Associations of individual-level socioeconomic status (SES) with incident dementia and Alzheimer’s disease
Multivariate analysis found strong associations of dementia and Alzheimer’s disease with SES (Table 2). Relative to high SES, after adjustment for age and including sex as strata, the hazard ratio (HR) for incident dementia increased with lower SES, going from HR 1.66 (95% CI 1.47–1.88) for medium SES to HR 3.02 (95% CI 2.68–3.40) for low SES strata (p trend < 0.001) (model 1). Further adjustment for ethnicity, lifestyle, social activities, hearing difficulty, obesity, frailty, urbanicity and neighbourhood deprivation moderated these associations slightly with HR 1.60 (95% CI 1.42–1.80) for medium SES and HR 2.59 (95% CI 2.29–2.92) for low SES strata without making any material difference to the significance of the trend (p trend < 0.001) (model 2). Further adjustment for biological age (as modelled from residual of LTL) resulted in little change in the effect estimates with HR 1.59 (95% CI 1.41–1.80) for medium SES and HR 2.58 (95% CI 2.28–2.91) for low SES, indicating independence of SES and LTL exposures. Relative to long LTL, after adjustment for age, ethnicity, lifestyle, social activities, hearing difficulty, obesity, frailty, urbanicity, and neighbourhood deprivation and including sex as strata, the hazard ratio for incident dementia increased with shorter LTL, going from HR 1.17 (95% CI 1.07–1.28) for medium LTL and HR 1.20 (95% CI 1.09–1.32) for short LTL strata (p trend < 0.001) (Supplementary Table S6). A comparable pattern of association was found for Alzheimer’s disease.
Associations of individual-level socioeconomic status (SES) with incident dementia and Alzheimer’s disease stratified by biological age
To investigate the interaction effect of LTL and SES on dementia risk, dementia was regressed on SES stratified by LTL (Table 3). There was a closely similar pattern of elevated risks of dementia and Alzheimer’s disease towards low SES across the LTL stratum. Relative to participants of high SES, those of low SES had HRs for incident dementia of 2.49 (95% CI 1.91–3.24) for long LTL stratum and 2.96 (95% CI 2.32–3.77) for short LTL stratum. Similarly, participants of low SES were also associated with higher risks of incident Alzheimer’s disease with HRs of 2.27 (95% CI 1.51–3.43) for long LTL and 3.17 (95% CI 2.18–4.59) for short LTL category.
Joint associations of individual-level SES and LTL showed a consistent dose–response relationship for LTL and SES independently (Fig. 2). Compared with participants of high SES and longer LTL (least-exposed group), those of low SES and shorter LTL (double-exposed group) had higher HRs of 3.28 (95% CI 2.57–4.20) for incident dementia and 3.44 (95% CI 2.35–5.04) for incident Alzheimer’s disease. We found evidence of positive biologic interaction between LTL and SES, with the combined effect of the two factors in the double-exposed group exceeding the sum of their separate effects relative to the least-exposed group (Supplementary Table S7). The relative excess risk due to interaction (RERI) for both dementia (RERI = 0.57, 95% CI 0.07–1.06), and Alzheimer’s disease (RERI = 0.79, 95% CI 0.02–1.56) were positive, indicating a departure from additivity towards excess risk in double-exposed groups (i.e. synergism).
Sensitivity analyses
Repeating analyses with single socioeconomic factors of educational attainment, household income and employment status, showed elevated risks for both incident dementia and Alzheimer’s disease among participants of lower education attainment, lower household income and those who were retired and belonging to the others category (including unemployed) (Supplementary Tables S8–10). Consistent with our primary analysis, models combining cases of incident dementia diagnosis and mortality as the outcome showed that participants of low SES and shorter LTL (double-exposed group) had higher HRs of 3.21 (95% CI 2.53–4.08) in reference to those of high SES and longer LTL (least-exposed group) with synergistic interaction (RERI = 0.62, 95% CI 0.16–1.09) (Supplementary Table S11). Similarly, the analysis with composite SES derived from LCA models comprising both multiple individual socioeconomic factors and neighborhood deprivation (measured by Townsend index of deprivation) showed consistent results as in our primary analysis, with participants of low SES and shorter LTL (double-exposed group) reporting higher HRs of 2.75 (95% CI 2.27–3.34) and 2.99 (95% CI 2.19–4.07) for incident dementia and Alzheimer’s disease respectively, in reference to their counterparts belonging to high SES and longer LTL (least-exposed group) (Supplementary Table S12). Our results also remained robust after additional adjustment for number of household members (Supplementary Table S13) and comorbidities taken as covariates (Supplementary Tables S14) with increased risk of incident dementia and Alzheimer’s disease being associated with lower SES and shorter LTL. Rerunning analyses for occupational categories showed lower risk of incident dementia and Alzheimer’s disease for participants engaged in professional occupations, and administrative and secretariat occupations relative to those who belonged to other occupational (manual) category, nonetheless the results remained insignificant (Supplementary Table S15).
Stratified analyses by age in the study subgroups showed that relative to participants aged < 65 years belonging to high SES-longest LTL tertile, those in low SES-shortest LTL tertile had higher HRs of 4.57 (95% CI 2.46–8.50) for dementia and 3.45 (95% CI 1.26–9.42) for Alzheimer’s disease. The effect sizes for age-group ≥ 65 years were more pronounced in the double-exposed category relative to least-exposed, with HRs of 6.37 (95% CI 4.84–8.38) for dementia and 7.42 (95% CI 4.82–11.44) for Alzheimer’s disease (Supplementary Tables S16, 17). Sensitivity analyses with imputed data accounting for missingness across covariates produced consistently robust results further supporting the main analyses (Supplementary Tables S18).
Discussion
In a large prospective study, we found SES (a composite metric developed from household income, educational attainment and employment status) and leucocyte telomere length (LTL), were each independently monotonically associated with higher risks for incident dementia and Alzheimer’s disease. We also found that SES and LTL demonstrated modest interaction effects, suggesting that biological aging as measured by LTL and SES are likely to represent separate but synergistic risk pathways. Several sensitivity analyses such as rerunning models with single socioeconomic factors, combined cases of incident dementia diagnosis and mortality, LCA models based on both individual socioeconomic factors and neighborhood deprivation, and multiple imputations showed largely consistent results.
These findings are consistent with previous work. A 12-year follow-up study of 6,220 participants aged ≥ 65 from the English Longitudinal Study of Aging reported a dementia hazard ratio of 1.50 (95% CI 1.05–2.13) for lower SES9. Another 12-year follow-up study of 2,457 US elderly found increased hazard for incident dementia among participants with less than high school education (HR 1.47, 95% CI: 1.17–1.86)6. A Swedish cohort study of 931 participants aged ≥ 75 reported that low education and occupation-based SES were associated with elevated risks of dementia (RR = 2.4, 95% CI 1.5–4.0) and Alzheimer’s disease (RR = 3.1, 95% CI 1.6–5.7)48. For LTL, the Washington Heights study of 1,983 participants followed for 8 years reported that each kilobase pair decrease in LTL was associated with 21% increased risk for dementia49. However, a study of 1,961 participants followed for 8 years in Rotterdam found a U-shaped relationship between LTL and Alzheimer’s risk with HR 1.59 (95% CI 1.13–2.23) for the first tertile in reference to the middle. Of particular interest are cross-sectional findings from the HANDLS study. Among 325 individuals, interactions between LTL and SES for cognitive performance were found (b = − 2.06; p = 0.008) suggesting that for low SES shorter LTL is associated with poorer cognitive performance50. Our study extends this evidence by providing longitudinal data, at-scale, for clinically relevant outcomes.
Potential mechanisms for these associations are complex and remain an active area of research. Lower SES is associated with poorer cognitive reserve, higher levels of psychosocial stress, poorer lifestyle and behaviour that increases cognitive risk51. It has also been suggested that low SES may act as a proxy for reduced exposure to intellectually stimulating environments with inadequate cognitive resources to protect against neurodegeneration9,47,52. Biological age, as measured by telomere length is considered a composite marker of current health status encapsulating cumulative burdens of chronic comorbidities and psychosocial stress from allostatic load13,53. Shorter TL is associated with age-related morbidity, lower immune function, elevated oxidative stress-dependent senescence, and pro-inflammatory mediators54. Given the potential overlap between these two constellations of pathways, our study suggests SES and LTL were synergistic for dementia. These findings may suggest that both SES and biological aging potentially contribute to ‘downstream’ mechanisms underlying increased dementia risk. For example, inflammatory processes are increasingly implicated as systemic influences in neurodegeneration, and both biological ageing and SES may act separately to upregulate neuro-inflammatory pathways.
Previous evidence mostly showed significant associations between individual-level SES and dementia in elderly cohorts comprising late-onset cases. However, research on early-onset dementia has been relatively scarce. In our UK Biobank analytic sample, 233 of 3,313 (7%) dementia cases presented themselves as early-onset cases. Our stratified analysis by age in the study sub-groups found that the effect sizes for socioeconomic status on the risk of dementia and Alzheimer’s disease in the younger age subgroup (< 65 years) comprising early-onset cases remained significant, as were the older sub-groups (≥ 65) with late-onset cases. Joint associational analysis found that the HR for dementia in the double-exposed group (relative to the least-exposed) were 4.57 for the subgroup with early-onset cases (< 65 years) and 6.37 for subgroup with late-onset cases (≥ 65 years). In early-onset dementia (including frontotemporal dementia), symptom onset occurs at a relatively younger age when the patients are still in a crucial life stage, being inactive employment with family and social responsibilities, thereby making them vulnerable. Patients generally experience a steeper decline in cognitive and functional capacities as compared to late-onset dementia55,56. It is noteworthy to mention that the observed significant associations between early-onset dementia and SES are of value, and may plausibly point to the protective role of socioeconomic position (and associated lifestyle) on dementia, potentially via lowering stress and chronic inflammation. Further research needs to be conducted in this direction.
In our models, biological age was defined as a composite measure derived from residual LTL adjusting for chronological age, inflammation (CRP) and 46 curated chronic comorbidities including risk factors of dementia. Sensitivity tests in which models included residual LTL adjusted for age and serum CRP only, and additionally adjusted for the comorbidities as covariates also produced consistent results. The generalizability of our findings to other cohorts in diverse settings also needs to be explored.
With increasing proportion of elderly population and stronger impetus towards ensuring healthy longevity, evidence as generated from population stratification by SES and biological age may help identify vulnerable sub-groups for allocating crucial public health and social care resources. Risk factors such as SES may act as important considerations for community screening of neurodegenerative diseases as well as designing of tailored intervention programs to prevent against cognitive impairment. Further research is needed to improve technologies for measuring biological age (telomere length) at-scale, accuracy and cost effectiveness57,58.
Strengths
Strengths of the study include scale and follow-up period. This analysis is an order of magnitude larger than previous studies linking either SES or LTL to dementia, enabling smaller effect sizes and interactions to be detected with confidence. The 12-year follow-up period, also longer than many previous studies, has enabled the analysis to focus on incident cases and implement a landmark period, to minimise the impact of reverse causation. Latent class modelling is a parsimonious approach to developing a composite SES construct. Complimentary sensitivity analyses using individual base SES variables further confirmed the primary analysis. Biological age was comprehensively measured from LTL which underwent rigorous quality checks.
Limitations
We acknowledge several limitations. Firstly, ascertainment of dementia is a challenge for health services across the world. We defined dementia and Alzheimer’s disease through electronic health record linkage as in previous studies9,59,60, which would have potentially led to an underestimation through omitting undiagnosed conditions. Previous cohort studies have estimated the low sensitivities of dementia diagnosis from hospital discharge records; for example 51% in Finland61, 78% in England62 and 70% in USA63. A recent systematic review concluded that the positive predictive value (PPV) for all-cause dementia from routinely collected health datasets were > 75% for 16 of the 27 studies, while the sensitivities were reasonable ranging between 21 and 86%64. Such an under-ascertainment would be more likely among ethnic minority, older sub-groups and in participants with milder dementia and poorer health-seeking behaviour. Nonetheless, routinely collected and coded hospital records linked to large population cohorts constitute a cost-effective means of case ascertainment for dementia with minimum attrition rate in prospective analysis relative to standardized clinical assessment needing face to face examination. Within the UK Biobank, PPV for all-cause dementia from hospital admissions records remained high at 87.3%65. Secondly, the UK Biobank population sample is not representative. Compared to the general population, cohort participants were less likely to be sampled from socioeconomically deprived areas and had lower prevalence of obesity, smoking, drinking and self-reported health conditions due to healthy volunteer selection bias, and as such, summary statistics are not generalizable66. However, being a large and heterogeneous cohort, it is designed specifically for etiologic analyses and sampling bias is unlikely to have a material effect on the estimates reported67. Thirdly, as in any observational study, we cannot completely rule out residual confounding and reverse causation and causality cannot be assumed. However, our results remained robust subsequent to exclusion of cases over the 3-year landmark period and adjustments for chronic morbidities. Fourthly, our primary SES exposure was based on self-reported measures of household income, educational attainment and employment status. Lastly, we excluded participants due to missingness across covariates. Nonetheless, there was no systematic difference between analytic sample and excluded sample (Supplementary Table S5) and our primary results were consistent with those of sensitivity analyses using imputed data.
In conclusion, in a large prospective study we find that SES and biological aging are synergistic risk factors for dementia and Alzheimer’s disease suggesting the importance of targeted tailor-made preventive interventions in vulnerable at-risk population groups.
Data availability
Data are available from the UK Biobank (https://www.ukbiobank.ac.uk/) for researchers who meet the criteria for access to de-identified UK Biobank data.
References
Nichols, E. 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. Lancet Public Health 7, e105–e125. https://doi.org/10.1016/S2468-2667(21)00249-8 (2022).
Wittenberg, R., Hu, B., Barraza-Araiza, L. & Rehill, A. Projections of Older People with Dementia and Costs of Dementia Care in the United Kingdom, 2019–2040 (London School of Economics, 2019).
Wittenberg, R. et al. Projections of care for older people with dementia in England: 2015 to 2040. Age Ageing 49, 264–269. https://doi.org/10.1093/ageing/afz154 (2019).
Livingston, G. et al. Dementia prevention, intervention, and care. The Lancet 390, 2673–2734. https://doi.org/10.1016/S0140-6736(17)31363-6 (2017).
Fratiglioni, L., Marseglia, A. & Dekhtyar, S. Ageing without dementia: Can stimulating psychosocial and lifestyle experiences make a difference?. Lancet Neurol. 19, 533–543. https://doi.org/10.1016/S1474-4422(20)30039-9 (2020).
Yaffe, K. et al. Effect of socioeconomic disparities on incidence of dementia among biracial older adults: Prospective study. BMJ 347, f7051. https://doi.org/10.1136/bmj.f7051 (2013).
Kivimäki, M. et al. Association between socioeconomic status and the development of mental and physical health conditions in adulthood: A multi-cohort study. Lancet Public Health 5, e140–e149. https://doi.org/10.1016/S2468-2667(19)30248-8 (2020).
Petersen, J. D. et al. Association of socioeconomic status with dementia diagnosis among older adults in Denmark. JAMA Netw. Open 4, e2110432–e2110432. https://doi.org/10.1001/jamanetworkopen.2021.10432 (2021).
Cadar, D. et al. Individual and area-based socioeconomic factors associated with dementia incidence in England: Evidence from a 12-year follow-up in the english longitudinal study of ageing. JAMA Psychiat. 75, 723–732. https://doi.org/10.1001/jamapsychiatry.2018.1012 (2018).
Rusmaully, J. et al. Contribution of cognitive performance and cognitive decline to associations between socioeconomic factors and dementia: A cohort study. PLOS Med. 14, e1002334. https://doi.org/10.1371/journal.pmed.1002334 (2017).
Adams, J. M. & White, M. Biological ageing: A fundamental, biological link between socio-economic status and health?. Eur. J. Public Health 14, 331–334. https://doi.org/10.1093/eurpub/14.3.331 (2004).
Belsky Daniel, W. et al. Quantification of biological aging in young adults. Proc. Natl. Acad. Sci. 112, E4104–E4110. https://doi.org/10.1073/pnas.1506264112 (2015).
Epel Elissa, S. et al. Accelerated telomere shortening in response to life stress. Proc. Natl. Acad. Sci. 101, 17312–17315. https://doi.org/10.1073/pnas.0407162101 (2004).
Fyhrquist, F., Saijonmaa, O. & Strandberg, T. The roles of senescence and telomere shortening in cardiovascular disease. Nat. Rev. Cardiol. 10, 274–283. https://doi.org/10.1038/nrcardio.2013.30 (2013).
Puterman, E. et al. Lifespan adversity and later adulthood telomere length in the nationally representative US health and retirement study. Proc. Natl. Acad. Sci. 113, E6335–E6342. https://doi.org/10.1073/pnas.1525602113 (2016).
Mitchell, C. et al. Social disadvantage, genetic sensitivity, and children’s telomere length. Proc. Natl. Acad. Sci. 111, 5944–5949. https://doi.org/10.1073/pnas.1404293111 (2014).
Needham, B. L. et al. Socioeconomic status, health behavior, and leukocyte telomere length in the national health and nutrition examination survey, 1999–2002. Soc. Sci. Med. 85, 1–8. https://doi.org/10.1016/j.socscimed.2013.02.023 (2013).
Steptoe, A. et al. Educational attainment but not measures of current socioeconomic circumstances are associated with leukocyte telomere length in healthy older men and women. Brain Behav. Immun. 25, 1292–1298. https://doi.org/10.1016/j.bbi.2011.04.010 (2011).
Cherkas, L. F. et al. The effects of social status on biological aging as measured by white-blood-cell telomere length. Aging Cell 5, 361–365. https://doi.org/10.1111/j.1474-9726.2006.00222.x (2006).
Martens, D. S. et al. Association of parental socioeconomic status and newborn telomere length. JAMA Netw. Open 3, e204057–e204057. https://doi.org/10.1001/jamanetworkopen.2020.4057 (2020).
Beatty Moody, D. L. et al. Interpersonal-level discrimination indices, sociodemographic factors, and telomere length in African-Americans and Whites. Biol. Psychol. 141, 1–9. https://doi.org/10.1016/j.biopsycho.2018.12.004 (2019).
Brown, R. et al. Neighborhood social environment and changes in leukocyte telomere length: The multi-ethnic study of atherosclerosis (MESA). Health Place 67, 102488. https://doi.org/10.1016/j.healthplace.2020.102488 (2021).
Honig, L. S., Schupf, N., Lee, J. H., Tang, M. X. & Mayeux, R. Shorter telomeres are associated with mortality in those with APOE ϵ4 and dementia. Ann. Neurol. 60, 181–187. https://doi.org/10.1002/ana.20894 (2006).
Martin-Ruiz, C. et al. Telomere length predicts poststroke mortality, dementia, and cognitive decline. Ann. Neurol. 60, 174–180. https://doi.org/10.1002/ana.20869 (2006).
Forero, D. A. et al. Meta-analysis of telomere length in Alzheimer’s disease. J. Gerontol. Ser. A 71, 1069–1073. https://doi.org/10.1093/gerona/glw053 (2016).
Thomas, P., O’Callaghan, N. J. & Fenech, M. Telomere length in white blood cells, buccal cells and brain tissue and its variation with ageing and Alzheimer’s disease. Mech. Ageing Dev. 129, 183–190. https://doi.org/10.1016/j.mad.2007.12.004 (2008).
Rodríguez-Fernández, B. et al. Genetically predicted telomere length and its relationship with neurodegenerative diseases and life expectancy. Comput. Struct. Biotechnol. J. 20, 4251–4256. https://doi.org/10.1016/j.csbj.2022.08.006 (2022).
Fani, L. et al. Telomere length and the risk of Alzheimer’s disease: The rotterdam study. J. Alzheimers Dis. 73, 707–714. https://doi.org/10.3233/JAD-190759 (2020).
Ashrafi, A. et al. Leukocyte telomere length is unrelated to cognitive performance among non-demented and demented persons: An examination of long life family study participants. J. Int. Neuropsychol. Soc. 26, 906–917. https://doi.org/10.1017/S1355617720000363 (2020).
Sudlow, C. et al. UK biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLOS Med. 12, e1001779. https://doi.org/10.1371/journal.pmed.1001779 (2015).
UK Biobank. UK Biobank: Protocol for a large-scale prospective epidemiological resource. Protocol No: UKBB-PROT-09–06. Available at: http://www.ukbiobank.ac.uk/wp-content/uploads/2011/11/UK-Biobank-Protocol.pdf. (2007).
Zhang, Y.-B. et al. Associations of healthy lifestyle and socioeconomic status with mortality and incident cardiovascular disease: Two prospective cohort studies. BMJ 373, n604. https://doi.org/10.1136/bmj.n604 (2021).
Lanza, S. T. & Rhoades, B. L. Latent class analysis: An alternative perspective on subgroup analysis in prevention and treatment. Prev. Sci. 14, 157–168. https://doi.org/10.1007/s11121-011-0201-1 (2013).
StataCorp, L. (College Station: Stata Press, 2017).
Codd, V. et al. Measurement and initial characterization of leukocyte telomere length in 474,074 participants in UK Biobank. Nat. Aging 2, 170–179. https://doi.org/10.1038/s43587-021-00166-9 (2022).
Wang, Q. et al. Shorter leukocyte telomere length is associated with adverse COVID-19 outcomes: A cohort study in UK Biobank. EBioMedicine 70, 103485. https://doi.org/10.1016/j.ebiom.2021.103485 (2021).
Fry, D., Almond, R., Moffat, S., Gordon, M. & Singh, P. UK biobank biomarker project companion document to accompany serum biomarker data. Version 1. (2020).
Cao, L. et al. Dietary patterns and risk of dementia: A systematic review and meta-analysis of cohort studies. Mol. Neurobiol. 53, 6144–6154. https://doi.org/10.1007/s12035-015-9516-4 (2016).
Sabia, S. et al. Alcohol consumption and risk of dementia: 23 year follow-up of Whitehall II cohort study. BMJ 362, k2927. https://doi.org/10.1136/bmj.k2927 (2018).
Lourida, I. et al. Association of lifestyle and genetic risk with incidence of dementia. JAMA 322, 430–437. https://doi.org/10.1001/jama.2019.9879 (2019).
Fried, L. P. et al. Frailty in older adults: Evidence for a phenotype. J. Gerontol. Ser. A 56, M146–M157. https://doi.org/10.1093/gerona/56.3.M146 (2001).
Hanlon, P. et al. Frailty and pre-frailty in middle-aged and older adults and its association with multimorbidity and mortality: A prospective analysis of 493 737 UK Biobank participants. Lancet Public Health 3, e323–e332. https://doi.org/10.1016/S2468-2667(18)30091-4 (2018).
Sarkar, C., Webster, C. & Gallacher, J. UK biobank urban morphometric platform (UKBUMP)—A nationwide resource for evidence-based healthy city planning and public health interventions. Ann. GIS 21, 135–148. https://doi.org/10.1080/19475683.2015.1027791 (2015).
Sarkar, C. et al. Environmental correlates of chronic obstructive pulmonary disease in 96 779 participants from the UK Biobank: A cross-sectional, observational study. Lancet Planet. Health 3, e478–e490. https://doi.org/10.1016/S2542-5196(19)30214-1 (2019).
Knol, M. J., Egger, M., Scott, P., Geerlings, M. I. & Vandenbroucke, J. P. When one depends on the other: Reporting of interaction in case-control and cohort studies. Epidemiology 20, 161–166 (2009).
Li, R. & Chambless, L. Test for additive interaction in proportional hazards models. Ann. Epidemiol. 17, 227–236. https://doi.org/10.1016/j.annepidem.2006.10.009 (2007).
Sterne, J. A. C. et al. Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls. BMJ 338, b2393. https://doi.org/10.1136/bmj.b2393 (2009).
Karp, A. et al. Relation of education and occupation-based socioeconomic status to incident Alzheimer’s disease. Am. J. Epidemiol. 159, 175–183. https://doi.org/10.1093/aje/kwh018 (2004).
Honig, L. S., Kang, M. S., Schupf, N., Lee, J. H. & Mayeux, R. Association of shorter leukocyte telomere repeat length With Dementia and mortality. Arch. Neurol. 69, 1332–1339. https://doi.org/10.1001/archneurol.2012.1541 (2012).
Leibel, D. K. et al. Telomere length and cognitive function: Differential patterns across sociodemographic groups. Neuropsychology 34, 186–198. https://doi.org/10.1037/neu0000601 (2020).
Weng, P.-H. et al. The effect of lifestyle on late-life cognitive change under different socioeconomic status. PLoS ONE 13, e0197676. https://doi.org/10.1371/journal.pone.0197676 (2018).
Stern, Y. Cognitive reserve in ageing and Alzheimer’s disease. Lancet Neurol. 11, 1006–1012. https://doi.org/10.1016/S1474-4422(12)70191-6 (2012).
Zalli, A. et al. Shorter telomeres with high telomerase activity are associated with raised allostatic load and impoverished psychosocial resources. Proc. Natl. Acad. Sci. 111, 4519–4524. https://doi.org/10.1073/pnas.1322145111 (2014).
Eitan, E., Hutchison, E. R. & Mattson, M. P. Telomere shortening in neurological disorders: An abundance of unanswered questions. Trends Neurosci. 37, 256–263. https://doi.org/10.1016/j.tins.2014.02.010 (2014).
Stanley, K. & Walker, Z. Do patients with young onset Alzheimer’s disease deteriorate faster than those with late onset Alzheimer’s disease? A review of the literature. Int. Psychogeriatr. 26, 1945–1953. https://doi.org/10.1017/S1041610214001173 (2014).
Wattmo, C. & Wallin, Å. K. Early- versus late-onset Alzheimer’s disease in clinical practice: Cognitive and global outcomes over 3 years. Alzheimer’s Res. Ther. 9, 70. https://doi.org/10.1186/s13195-017-0294-2 (2017).
Luo, Y., Viswanathan, R., Hande, M. P., Loh, A. H. P. & Cheow, L. F. Massively parallel single-molecule telomere length measurement with digital real-time PCR. Sci. Adv. 6, eabb7944. https://doi.org/10.1126/sciadv.abb7944 (2020).
Tham, C.-Y. et al. High-throughput telomere length measurement at nucleotide resolution using the PacBio high fidelity sequencing platform. Nat. Commun. 14, 281. https://doi.org/10.1038/s41467-023-35823-7 (2023).
Sommerlad, A., Sabia, S., Singh-Manoux, A., Lewis, G. & Livingston, G. Association of social contact with dementia and cognition: 28-year follow-up of the Whitehall II cohort study. PLOS Med. 16, e1002862. https://doi.org/10.1371/journal.pmed.1002862 (2019).
Petermann-Rocha, F. et al. Associations between physical frailty and dementia incidence: A prospective study from UK Biobank. Lancet Healthy Longev. 1, e58–e68. https://doi.org/10.1016/S2666-7568(20)30007-6 (2020).
Solomon, A. et al. Validity of dementia and Alzheimer’s disease diagnoses in finnish national registers. Alzheimers Dement. 10, 303–309. https://doi.org/10.1016/j.jalz.2013.03.004 (2014).
Sommerlad, A. et al. Accuracy of general hospital dementia diagnoses in England: Sensitivity, specificity, and predictors of diagnostic accuracy 2008–2016. Alzheimers Dement. 14, 933–943. https://doi.org/10.1016/j.jalz.2018.02.012 (2018).
Knopman, D. S., Petersen, R. C., Rocca, W. A., Larson, E. B. & Ganguli, M. Passive case-finding for Alzheimer’s disease and dementia in two U.S. communities. Alzheimer’s Dement. 7, 53–60. https://doi.org/10.1016/j.jalz.2010.11.001 (2011).
Wilkinson, T. et al. Identifying dementia cases with routinely collected health data: A systematic review. Alzheimers Dement. 14, 1038–1051. https://doi.org/10.1016/j.jalz.2018.02.016 (2018).
Wilkinson, T. et al. Identifying dementia outcomes in UK Biobank: A validation study of primary care, hospital admissions and mortality data. Eur. J. Epidemiol. 34, 557–565. https://doi.org/10.1007/s10654-019-00499-1 (2019).
Fry, A. et al. Comparison of sociodemographic and health-related characteristics of UK biobank participants with those of the general population. Am. J. Epidemiol. 186, 1026–1034. https://doi.org/10.1093/aje/kwx246 (2017).
Rothman, K. J., Gallacher, J. E. & Hatch, E. E. Why representativeness should be avoided. Int. J. Epidemiol. 42, 1012–1014. https://doi.org/10.1093/ije/dys223 (2013).
Funding
The research used the UK Biobank resource (approved application numbers: 11730 and 44469). C.S. is a recipient of a fellowship in Global Health Leadership from the National Academy of Medicine (NAM), Washington DC. J.E.J.G. acknowledges the support of Medical Research Council: MR/T0333771 award for the Dementias Platform UK. The built environment metrics used in this study was supported by funding from a seed grant from the UK Biobank, and UK Economic & Social Research Council’s Transformative Research grant (ES/L003201/1). The funders had no role in the study design; collection, analysis, and interpretation of data; writing of the paper; and/or decision to submit for publication.
Author information
Authors and Affiliations
Contributions
K.Y.L., C.S., J.E.J.G, and C.W. developed the research questions. All the authors contributed to study design. K.Y.L. and S.K. were responsible for accessing, cleaning and verifying the data. K.Y.L. and C.S. conducted the data analysis and wrote-up the manuscript with inputs from all authors. J.E.J.G., C.W., S.K., C.S. and K.Y.L. further critically revised the manuscript. C.S. is the guarantor.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Lai, K.Y., Webster, C., Kumari, S. et al. The associations of socioeconomic status with incident dementia and Alzheimer’s disease are modified by leucocyte telomere length: a population-based cohort study. Sci Rep 13, 6163 (2023). https://doi.org/10.1038/s41598-023-32974-x
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-023-32974-x
- Springer Nature Limited