Abstract
A substantial proportion of dementia risk may be attributable to modifiable factors, yet these are often examined in isolation despite their interrelated nature and tendency to co-occur. It remains unclear whether the relationship between modifiable factors and dementia risk is influenced by individual characteristics such as sex and genetic susceptibility. We investigated longitudinal associations between the Lifestyle for Brain health (LIBRA) score and risk of dementia, cognitive performance, and brain structure, and whether relationships differed by sex and APOE ɛ4 carrier status.
Participants were aged > 50 years, dementia-free at baseline, 50% female and predominantly (97%) white/Caucasian. The LIBRA score included 11 modifiable factors (e.g., hypertension, obesity, physical inactivity). Magnetic resonance imaging estimated brain volume, domain-specific cognitive composite scores were calculated, and dementia diagnoses were determined based on self-reported and linked healthcare data.
Across a mean follow-up of 10.2 years, a higher LIBRA score was associated with greater odds of developing dementia (OR = 1.20, 95% CI 1.18–1.22). This association was stronger in APOE ɛ4 non-carriers compared to ɛ4 carriers. Cross-sectionally, higher LIBRA scores related to poorer cognition, smaller whole-brain gray and white matter volumes, and increased ventricular cerebrospinal fluid (CSF), however, only the association with increased ventricular CSF persisted longitudinally (mean follow-up 3.4 years).
Each one-point increase on the LIBRA score was associated with 20% increased odds of developing dementia. These results reinforce the need to target modifiable dementia risk factors and to tailor dementia prevention strategies to individual risk profiles to maximize the impact on brain health.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1007/s11357-025-02056-z.
Keywords: Dementia, Cognition, Brain structure, Lifestyle, Modifiable, Older adults
Introduction
Lifestyle interventions are at the forefront of dementia prevention, with 45% of dementia risk potentially attributable to modifiable factors [1]. Although modifiable risk factors (e.g., physical activity, diet) are often studied individually, they are frequently correlated and co-occur (e.g., regular exercise can delay or prevent type 2 diabetes; both of which are independent dementia risk factors) [2]. Understanding the additive and synergistic effects of various risk factors for dementia is critical for advancing knowledge of how to best alter modifiable factors to preserve brain health in aging.
The LIfestyle for BRAin health risk (LIBRA) score was developed based on a systematic review and Delphi consensus [3] weighing the importance of several modifiable factors for dementia prevention. The score comprises 12 lifestyle and health factors which can be directly targeted through intervention (e.g., hypertension, obesity, physical inactivity). Estimates of the association between higher LIBRA scores (indicative of poorer lifestyle habits and health) and increased odds of developing dementia vary between 8–22% [4–9], requiring further study in large samples for greater effect size precision. A higher LIBRA score has been associated with greater cognitive decline prior to dementia onset [4, 5], and has been cross-sectionally associated with smaller gray matter volume, greater white matter hyperintensity burden, and older brain age [10, 11]. However, few studies have examined whether the LIBRA score is predictive of longitudinal changes in brain morphology.
The priority target population for lifestyle interventions for dementia prevention remains unclear. To determine priority populations, it is important to understand how individual characteristics may influence the effectiveness of preventative interventions. Female sex and apolipoprotein E (APOE) ɛ4 allele carriage are two of the leading identified risk factors for dementia, but studies investigating how sex and genetics influence associations between LIBRA score and dementia risk have yielded inconsistent results [4, 7, 11]. Therefore, additional research is required to understand whether the comprehensive set of lifestyle factors included in the LIBRA score differentially associate with brain health dependent upon individual characteristics (sex and APOE ɛ4 allele carriage), which will inform the personalization of future multi-domain dementia prevention trials.
Using data from the UK Biobank repository, we examined whether a score of lifestyle and health (i.e., the LIBRA score) was associated with cognition and brain volume, both cross-sectionally and longitudinally, as well as with dementia risk. We hypothesized that a higher LIBRA score would be associated with increased odds of developing dementia, poorer cognitive performance, and smaller brain volumes. Additionally, because of associations between high LIBRA score and increased risk for dementia, we aimed to determine whether higher LIBRA scores were associated with a shorter time to dementia diagnosis. Finally, we investigated whether any of these associations varied by demographics and genetic risk, namely sex and APOE ɛ4 allele carriage, hypothesizing that associations would be stronger for males [11] and in APOE ɛ4 non-carriers [7].
Methods
Study design and participants
This study used participants from the UK Biobank repository, which consists of half a million middle-aged adults (40–69 years) recruited from England, Scotland, and Wales who completed baseline measures from 2006–2010. Participants returned from 2014 onward for additional cognitive assessments and the first neuroimaging scans. Additional follow-up visits began in 2019. We excluded anyone < 50 years of age and those reporting a diagnosis of dementia at baseline (via self-report or health-linked data; n = 343). To maximize sample size, participant numbers varied for each analysis (see Fig. 1). For analyses treating the imaging timepoint as baseline (Analyses 2 and 3, see below), an additional 28 participants diagnosed with dementia before their first imaging visit were excluded (Fig. 1). Participants with missing data for any LIBRA score component or any covariates were excluded. For longitudinal analyses, participants were required to have non-missing outcome data for at least two visits.
Fig. 1.
Sample size for each analysis from the full UK Biobank cohort
All study procedures involving human participants were approved by the North West Multi-Center Research Ethics Committee and all participants gave informed consent to participate in the UK Biobank and be followed up with repeated assessments. The UK Biobank protocol is available online (https://www.ukbiobank.ac.uk/media/3sbeknnz/ukbiobank_protocol.pdf).
LIfestyle for BRAin Health (LIBRA) Score Calculation
A LIBRA score was calculated as a weighted sum of 11 lifestyle and health factors (Table 1). We had available data for all factors detailed in the original LIBRA score calculation [5], except cognitive stimulation which was not available from the UK Biobank Repository. Some items composing the score required study-specific definitions, for example a modified Mediterranean diet score was created based on previous research [12] and available variables from the UK Biobank (see Appendix 1). Table 1 details each lifestyle factor and their measurement; Appendix 1 includes International Classification of Diseases (ICD) codes, medication codes, and field IDs for each variable where relevant. For variables using ICD codes, which were drawn from health-linked data, conditions were considered present if they were reported any time before baseline. Each lifestyle factor was weighted based on the original LIBRA scoring system; these weightings were calculated based on the relative risk of each factor [5]. A summed score with a possible range of −3.7 to 12.7 was created for each participant, where higher scores indicate poorer lifestyle habits and health. All variables contributing to the LIBRA score were drawn from the initial UK Biobank visit (2006–2010). For participants with available data, an updated score was calculated using values from the first imaging visit for analyses using imaging and cognitive outcomes (Analyses 2 and 3 detailed below).
Table 1.
Lifestyle variables included in LIBRA score, their measures in UK Biobank and weightings in final score
| Lifestyle variable | Measure | Weighting |
|---|---|---|
| Hypertension | Coded in ICD 9 or 10; blood pressure > 140/90; or self-reported taking blood pressure medication or self-reported high blood pressure | + 1.6 |
| Obesity | Body mass index ≥ 30 kg/m2 | + 1.6 |
| High Cholesterol | Taking cholesterol-lowering medication | + 1.4 |
| Type 2 Diabetes | Coded in ICD 9 or 10; or self-reported diabetes diagnosed by a doctor | + 1.3 |
| Coronary Heart Disease | Coded in ICD 9 or 10 | + 1 |
| Chronic Kidney Disease | Coded in ICD 9 or 10 | + 1.1 |
| Physical Inactivity | Self-reported via the International Physical Activity Questionnaire; <1492 MET-min/week (lowest tertile) | + 1.1 |
| Smoking | Self-reported current smoking (occasional or on most days) | + 1.5 |
| Moderate Alcohol Intake | Consumption of >0 and ≤14 units of alcohol per week (coded as 2); no alcohol consumption (coded as 1); >14 units per week (coded as 0) | −1.0 |
| Healthy Diet | Modified Mediterranean diet score ≥21 | −1.7 |
| Depressive symptomology | Coded in ICD 9 or 10; or score of ≥3 on the patient health questionnaire | +2.1 |
All variables calculated from ICD codes were considered present if they were reported any time before baseline. Abbreviations: ICD, International Classification of Diseases; LIBRA, LIfestyle for BRAin Health
Outcomes
Cognitive measures were administered through a computerized touchscreen interface [13]. The tests were designed for the UK Biobank to allow unsupervised administration and have shown moderate-to-high test–retest reliability and moderate correlations with other validated measures [14]. Based on previous research [15], composite scores were created for the cognitive domains of: processing speed (reaction time and digit symbol), reasoning (fluid intelligence and matrix pattern completion), executive function (Trail Making A, Trail Making B, tower rearranging), and memory (pairs matching, numeric memory, paired associative learning), where higher composite scores indicate better performance. See Appendices 2 and 6 for additional information.
Volumetric MRI data were collected using a Siemens Skyra 3T scanner (Siemens Healthcare, Erlangen, Germany) equipped with a 32-channel RF head coil, following protocols established by the UK Biobank [16]. In brief, the 3D MPRAGE T1-weighted images were pre-processed and analyzed using version 5 of the FSL software (FMRIB Software Library, Oxford, UK). Imaging-derived phenotypes, such as total white and gray matter volumes and gray matter regions of interest, were produced through the image-processing pipeline developed and implemented on behalf of the UK Biobank [17]. For the current study, we examined total gray and white matter volumes (normalized for head size; field IDs 25005, 25007) and ventricular cerebrospinal fluid (CSF; field ID 25003).
We used the algorithmically defined variable in UK Biobank to determine all-cause dementia diagnoses (Field ID 42018). This variable was created based on self-reported and linked routinely collected healthcare data (ICD-9 and ICD-10 codes available at https://biobank.ndph.ox.ac.uk/showcase/refer.cgi?id=460). This algorithmically defined variable has shown acceptable accuracy for identifying dementia cases [18].
Statistical analyses
All statistical analyses were performed in R version 4.3.2. Normality was assessed by visual inspection and by examining skewness and kurtosis statistics. For executive function, one clear outlier with a score > 13 standard deviations from the mean was identified and excluded. Means, standard deviations and frequencies for demographic and lifestyle variables are presented in Table 2. A p-value of < 0.05 was deemed significant and the false discovery rate correction for multiple comparisons was applied within each outcome variable (e.g., processing speed, 6 models; dementia risk, 18 models). Covariates for all analyses included baseline age, sex (self-reported as male or female), Townsend deprivation index (a measure of socioeconomic status based on postcode and calculated from levels of employment, home and car ownership, and household overcrowding in a given area) [19], self-reported race/ethnicity, APOE ɛ4 allele carrier status (carriers or non-carriers), and qualification (held a college/university degree or not). There were some missing data for age at assessment visit (Field ID 21003) for the repeat imaging assessment. For these instances, age was calculated as assessment center visit date minus month and year of birth (day was imputed as 15).
Table 2.
Demographic characteristics of UK Biobank participants at baseline, aged 50 and over, and with data required for current analyses
| Characteristic | Analysis 1a n = 45,694 |
Analysis 2b n = 27,174 |
Analysis 3c n = 2,793 |
|---|---|---|---|
| Sex, female | 22,118 (48.4) | 13,059 (48.0) | 1,316 (47.1) |
| Age, years; mean (SD) | 59.1 (5.4) | 68.0 (5.5) | 66.3 (4.9) |
| BMI; mean (SD)# | 26.7 (4.2) | 26.5 (4.2) | 26.3 (4.1) |
| APOE ε4 allele carriers | 13,693 (30.0) | 7,299 (26.9) | 721 (25.8) |
| Townsend deprivation index; mean (SD) | −1.98 (2.7) | −2.05 (2.6) | −2.15 (2.5) |
| Qualification, College/University | 20,144 (44.1) | 13,398 (49.3) | 1419 (50.1) |
| Ethnicity | |||
| White/Caucasian | 44,431 (97.3) | 26,494 (97.5) | 2,745 (98.3) |
| Mixed | 168 (0.4) | 89 (0.3) | 8 (0.3) |
| Asian | 515 (1.1) | 274 (1.0) | 17 (0.6) |
| Black | 233 (0.5) | 117 (0.4) | 11 (0.4) |
| Unknown | 347 (0.8) | 200 (0.7) | 12 (0.4) |
| Hypertension, present# | 14,249 (31.2) | 11,753 (43.2) | 1,129 (40.4) |
| Hyperlipidemia, present# | 8,102 (17.7) | 8,556 (31.5) | 753 (27.0) |
| Type 2 Diabetes, present# | 1,894 (4.1) | 1,692 (6.2) | 156 (5.6) |
| Coronary Heart Disease, present# | 2,702 (5.9) | 2,591 (9.5) | 215 (7.7) |
| Chronic Kidney Disease, present# | 110 (0.2) | 308 (1.1) | 24 (0.8) |
| Physical activity# | |||
| Active ≥1492 MET-min/week | 30,465 (67.0) | 18,149 (66.9) | 1893 (67.8) |
| Inactive<1492 MET-min/week | 15,229 (33.0) | 9,025 (33.1) | 900 (32.2) |
| Smoking status# | |||
| Never | 43,095 (94.3) | 26,513 (97.6) | 2,724 (97.5) |
| Occasionally or on most or all days | 2,599 (5.7) | 661 (2.4) | 69 (2.5) |
| Alcohol intake# | |||
| No Alcohol | 10,490 (23.0) | 7,153 (26.3) | 676 (24.2) |
| Moderate (0–14 standard drinks/week) | 18,906 (41.4) | 12,044 (44.3) | 1,263 (45.2) |
| Excessive (>14 standard drinks/week) | 16,298 (35.7) | 7,977 (29.3) | 854 (30.6) |
| Modified Mediterranean Diet Score; mean (SD)# | 22.0 (3.5) | 22.5 (3.3) | 22.5 (3.3) |
| Depression# | 1,856 (4.1) | 1,123 (4.1) | 105 (3.7) |
| Total LIBRA Score; mean (SD) | −0.55 (2.2) | −0.37 (2.3) | −0.53 (2.3) |
| Time to follow-up, years; mean (SD) | 10.2 (3.3) | - | 3.7 (1.4) |
a LIBRA score at baseline (2006–2010) and risk for dementia across follow-up
b Cross-sectional analysis of LIBRA score with cognition and brain volume
c Longitudinal analysis of LIBRA score with cognition and brain volume (participants with at least two measurements)
# Included in LIBRA score
Unless otherwise described, data are presented as n (%). Descriptive data for the sample in each analysis includes data from participants that had any imaging or cognitive data available. Because not every participant had missing data for imaging or cognitive outcomes, there were slightly different samples in the analysis for each outcome. Thus, there is a larger overall sample reported here compared to those included in each individual analysis.
Abbreviations: APOE, apolipoprotein E; BMI, body mass index; SD, standard deviation; LIBRA, LIfeystle for BRAin health; MET, metabolic equivalent of task
Risk for all-cause dementia
A logistic regression analysis with baseline LIBRA score as the predictor and dementia (yes/no) as the outcome was conducted, adjusting for covariates as listed above, represented as odds ratios and their 95% confidence intervals (Analysis 1). To determine whether a higher LIBRA score was associated with shorter time to dementia diagnosis, we estimated hazard ratios (HR) using Cox proportional hazards (CpH) models. For both the logistic regression and CpH models, participants without follow-up data or those who died from another cause were removed from the analyses. For Cox models, time was calculated as number of days from the initial visit to dementia diagnosis or last follow-up for those who did not develop dementia. The proportional hazards assumption was evaluated using Schoenfeld residuals and was violated by age (χ2 = 32.2, p < 0.001), therefore we stratified the analysis by age tertiles (< 57, 57–62 and > 62 years), and added age as a time-dependent covariate. Interaction terms for sex and APOE ɛ4 allele carrier status were added to each model to determine if any associations differed based on these two factors.
Cognitive and MRI outcomes
The first imaging timepoint (2014 +) was the first UK Biobank assessment visit to include our outcome variables of interest (i.e., volumetric brain data and in-depth cognitive data), thus we treated this timepoint as ‘baseline’ for Analyses 2 and 3. Analysis 2 was a series of cross-sectional linear regression models, with LIBRA score as the predictor, and cognitive (processing speed, reasoning, executive functioning, memory) or MRI (total gray matter, total white matter, ventricular CSF) variables as outcomes. Separate models were run for each outcome for both Analyses 2 and 3. For Analysis 3, we conducted linear mixed models (LMMs) to evaluate associations between baseline LIBRA score and change in each outcome (i.e., both cognitive and MRI outcomes), with baseline LIBRA score × time (in years) as the primary predictor of interest. Few participants in the UK Biobank have > 2 timepoints of imaging data, and our longitudinal analyses illustrate a change between two isolated timepoints rather than a trajectory of change over time (which requires three timepoints for reliable estimation). Data are reported as standardized betas with standard errors. LIBRA score and outcomes were z-scored before analysis, so each β represents the expected SD difference in the cognitive or MRI outcome per 1 SD higher LIBRA score. LMMs included a random intercept for participant ID to account for between-participant variability in the outcome variable at baseline. Models included a baseline LIBRA × time interaction term, so the interaction β reflects the association of baseline LIBRA with longitudinal change in the outcome. We tested whether any associations varied by sex and APOE ɛ4 allele carriage status by adding each variable as an interaction term (e.g., sex*LIBRA*time) to the cross-sectional and longitudinal models.
Results
Descriptive statistics for participant characteristics are presented in Table 2. The samples for all analyses were approximately 50% female and predominantly white/Caucasian. Participants were younger and had lower prevalence of chronic conditions (hypertension, lipidemia, etc.) in Analysis 1, which was expected given this was the sample at baseline (2006–2010) and Analyses 2 and 3 were based on the imaging visit data (2014 +). Differences between males and females in total LIBRA score and LIBRA score components are detailed in Appendix 3.
Within the included dataset (Fig. 1), there were 4715 dementia cases across the course of follow-up (mean 10.2±3.3 years, range: 1–16.4 years). After covariate adjustment, a higher LIBRA score (indicative of poorer lifestyle habits and health) was associated with greater odds of developing dementia (OR=1.20, 95% CI 1.18–1.22). Each one-point increase on the LIBRA score was associated with 20% increased odds of developing dementia. Associations did not vary by sex (b=0.00, SE=0.01, p=0.972) but did vary by APOE ɛ4 allele carrier status (b=−0.10, SE=0.01, p<0.001). Higher LIBRA scores were associated with greater odds of developing dementia in APOE ɛ4 allele non-carriers (OR=1.25, 95% CI 1.22–1.27), compared to APOE ɛ4 allele carriers (OR = 1.15, 95% CI 1.12–1.17). All associations remained significant after adjusting for multiple comparisons.
Cox proportional hazard analyses indicated that in all age groups (i.e.,<57, 57–62 and >62 years), a higher LIBRA score was associated with a shorter time to a dementia diagnosis (Fig. 2). The largest effect size was observed in those aged <57 years at baseline (Fig. 2). In those aged 57–62 years and >62 years, but not <57 years, the effect of LIBRA score on dementia risk was moderated by APOE ɛ4 allele carriage (b = −0.12, p<0.001; b=−0.11, p<0.001, respectively), such that APOE ɛ4 allele non-carriers demonstrated a stronger association between LIBRA score and shorter time to dementia than APOE ɛ4 allele carriers (Fig. 2). There were no significant moderating effects by sex for any age group. All associations remained significant after adjusting for multiple comparisons.
Fig. 2.
Hazard Ratios and 95% CIs for dementia risk by age, sex, and APOE ɛ4 carrier status. Significant differences exist between APOE ɛ4 allele carriers and non-carriers in both the mid and older groups but not the younger group. There were no sex differences in any age group.
Cross-sectionally, a higher LIBRA score was associated with poorer processing speed, reasoning, executive function, memory, smaller gray matter volume and greater ventricular CSF; associations remained significant after adjusting for multiple comparisons (Table 3). However, only associations between higher LIBRA scores and greater ventricular CSF persisted longitudinally (β = 0.004, SE=0.001, p=0.002; Table 3). In the whole sample, LIBRA score was not associated with white matter volume cross-sectionally or longitudinally (Table 3).
Table 3.
Linear regression (cross-sectional) and linear mixed models (longitudinal) of associations between LIBRA score and cognition and brain volume with moderation analyses by sex and APOE ɛ4 allele carriage status
| Cross-sectional | Longitudinal | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LIBRA | LIBRA*Sex | LIBRA*APOE | LIBRA*Time | LIBRA*Sex*Time | LIBRA*APOE*Time | |||||||
| β (SE) | p | β (SE) | p | β (SE) | p | β (SE) | p | β (SE) | p | β (SE) | p | |
| Processing Speed | -0.05 (0.01) | <.001* | -0.001 (0.014) | .965 | 0.005 (0.006) | .481 | 0.01 (0.00) | .241 | 0.005 (0.009) | .616 | -0.001 (0.010) | .916 |
| Reasoning | -0.02 (0.01) | <.001* | -0.008 (0.014) | .578 | 0.005 (0.007) | .437 | 0.002 (0.004) | .685 | 0.009 (0.009) | .350 | 0.006 (0.010) | .538 |
| Executive Function | -0.02 (0.01) | <.001* | 0.001 (0.014) | .929 | 0.009 (0.007) | .172 | 0.000 (0.007) | .994 | -0.005 (0.016) | .745 | 0.010 (0.018) | .564 |
| Memory | -0.03 (0.01) | <.001* | 0.031 (0.014) | .030 | 0.016 (0.007) | .022 | -0.015 (0.008) | .072 | 0.013 (0.017) | .442 | 0.003 (0.019) | .890 |
| Females | - | -0.050 (0.01) | <.001 | - | - | - | - | |||||
| Males | - | -0.018 (0.009) | .050 | - | - | - | - | |||||
| ɛ4 carriers | - | - | -0.005 (0.013) | .688 | - | - | - | |||||
| ɛ4 non-carriers | - | - | -0.041 (0.008) | <.001 | - | - | - | |||||
| Total Gray Matter | -0.13 (0.01) | <.001* | -0.05 (0.01) | <.001* | 0.01 (0.01) | .258 | 0.004 (0.002) | .124 | -0.006 (0.005) | .297 | -0.002 (0.006) | .746 |
| Females | - | -0.096 (0.009) | <.001 | - | - | - | - | |||||
| Males | - | -0.151 (0.008) | <.001 | - | - | - | - | |||||
| Total White Matter | -0.01 (0.01) | .054 | -0.04 (0.01) | .001* | -0.01 (0.01) | .621 | 0.006 (0.003) | .089 | -0.008 (0.007) | .264 | 0.005 (0.007) | .495 |
| Females | - | 0.012 (0.010) | .235 | - | - | - | - | |||||
| Males | - | -0.033 (0.009) | <.001 | - | - | - | - | |||||
| Ventricular CSF | 0.06 (0.01) | <.001* | 0.04 (0.01) | .003* | -0.01 (0.01) | .299 | 0.004 (0.001) | .002* | 0.001 (0.002) | .620 | -0.003 (0.003) | .226 |
| Females | - | 0.039 (0.010) | <.001 | - | - | - | - | |||||
| Males | - | 0.077 (0.009) | <.001 | - | - | - | - | |||||
Reported as standardized betas (SE). Simple slopes analyses were conducted for significant interaction terms only. Covariates include age, sex, Townsend deprivation, ethnicity, APOE ɛ4 allele carriage status, and qualification.
*significant after correction for multiple comparisons.
Abbreviations: APOE, apolipoprotein E; CSF, cerebrospinal fluid; LIBRA, LIfestyle for BRAin health risk; SE, standard error.
Cross-sectionally, males demonstrated stronger associations between higher LIBRA scores and smaller gray matter volume, white matter volume, and greater ventricular CSF compared to females (Table 3 and Appendix 4). Cross-sectionally, there was a trend (FDR-corrected p = 0.060) towards stronger associations between higher LIBRA scores and poorer memory for females compared to males and APOE ɛ4 non-carriers compared to carriers (Table 3; Appendices 4 and 5). There were no other cross-sectional or longitudinal moderations by sex or APOE ɛ4 allele carriage (Table 3).
Discussion
We investigated associations between potentially modifiable dementia risk factors, quantified by LIBRA score, and dementia risk, cognition, and brain volume, in a sample of adults >50 years from the UK Biobank repository. Consistent with our predictions, we found that a higher LIBRA score (indicative of poorer lifestyle habits and health) was associated with both an increased risk for dementia and shorter time to dementia diagnosis, associations which were more pronounced in APOE ɛ4 allele non-carriers compared to carriers. Importantly, we demonstrated that the strongest association between higher LIBRA scores and shorter time to dementia was in those aged <57 years at baseline, suggesting midlife may be a crucial period for lifestyle intervention.
Those aged <57 and 57–62 years at baseline showed a stronger association between poorer lifestyle and health with shorter time to dementia diagnosis compared to those aged >62 years. For those aged <57, this association existed regardless of APOE ɛ4 allele carrier status. This suggests that poor lifestyle and health at midlife relates to earlier dementia onset, and suggests that conversely, healthier habits within this age range may contribute to delayed dementia onset, highlighting midlife as a crucial period for lifestyle intervention [1]. Importantly, our results represent between-person differences rather than within-person differences, and further longitudinal studies are required to confirm the optimal intervention window for dementia risk reduction.
We extend previous literature by showing that the magnitude of the association between LIBRA score and both memory and dementia risk was dependent on APOE ɛ4 allele carrier status, and specifically, that associations were stronger in APOE ɛ4 allele non-carriers. These results may suggest that a larger proportion of dementia risk is attributable to modifiable factors in those with lower genetic risk. Additionally, these results support the notion that poorer health and lifestyle may exacerbate genetic risk for dementia [20, 21], since LIBRA score was still significantly associated with dementia risk in APOE ɛ4 allele carriers.
Males demonstrated stronger associations between higher LIBRA scores and smaller gray and white matter volume, and larger ventricular CSF compared to females. However, there was a trend (FDR-corrected p = 0.060) for females to demonstrate stronger associations between higher LIBRA score and poorer memory performance compared to males. Complex differences in male and female cognitive aging and their relation to modifiable factors likely underlie these findings. For example, older adult females tend to show better verbal memory performance compared to males but show faster declines in white and gray matter volumes [22, 23]. This decline in brain volume may be due to reduced neuroprotection in menopause [23, 24], caused by a variety of changes including sharp declines in estrogen and other neurotrophins such as brain-derived neurotrophic factor [25]. In addition, previous research [26–28] and our own data (Appendix 3) suggest the factors contributing to a higher LIBRA score differ between males and females. Vascular risk factors, which may more likely contribute directly to neuroinflammatory pathways and brain volume reductions [29], are more prevalent in males and non-vascular risk factors (e.g., depression) are more likely to be present in females.
Consistent with prior work [4, 5, 10, 11], our cross-sectional findings demonstrate that poorer health and lifestyle relate to worse brain structure and cognitive function. However, although these associations were statistically significant, effect sizes were relatively small (e.g., a 1 SD increase in LIBRA score, approximately 2.2 points, associated with a 0.02 SD decrease in reasoning). A clinically meaningful threshold for change in cognitive composite scores has not been established for the specific cognitive assessments used in this study; therefore, it is difficult to determine the clinical significance of these associations. Longitudinally, only the association between higher LIBRA score and greater ventricular CSF remained significant. This contrasts with previous studies demonstrating that higher LIBRA scores are associated with more precipitous decline in cognitive function [4, 5], however longitudinal studies investigating associations between LIBRA scores and volumetric brain outcomes are lacking. Future work should focus on investigating associations between LIBRA scores and region-specific volumetric changes in areas which may be especially impacted in preclinical disease stages (e.g., hippocampus).
The current findings should be interpreted in the context of several limitations. Some measures for the factors that composed the LIBRA score (e.g., chronic kidney disease) relied on health-linked data, and it is possible that undiagnosed cases of these conditions could have been missed. Additionally, for factors which included health-linked data, we did not apply a cut-off date for when the diagnosis occurred. Thus, some diagnoses may have preceded our baseline assessments by several years. Although this is a limitation, we believe that applying a cut-off date at this point would have been arbitrary and the effects of chronic conditions on brain health may still manifest many years later. Although the LIBRA score was developed based on a Delphi consensus [3], it should be considered that some individual components comprising the score are not direct measures of lifestyle, but rather are potentially modifiable through lifestyle (e.g., diabetes, hypertension). Therefore, a small proportion of the observed associations between LIBRA score and dementia risk may reflect effects of potentially modifiable health conditions, alongside directly measured lifestyle factors. The observational design means we cannot draw conclusions regarding causation, and although many components included in the LIBRA score are established dementia risk factors [1], some may also be early symptoms of cognitive decline (e.g., physical activity declines in the years preceding dementia diagnosis) [30]. Our longitudinal analyses included only two timepoints per participant, limiting our ability to detect changes in the trajectory of brain health outcomes over time. In addition, our results may not be generalizable to other populations since the UK Biobank sample is predominantly white/Caucasian (97%), and additional research with more diverse cohorts is critical.
This study highlights the importance of lifestyle and health factors and their contribution to cognition, brain volume, and dementia risk in older adulthood. We demonstrate that the association between lifestyle and health factors and dementia differs based on person-level factors, with shorter time to dementia observed in younger participants (<62 years) and APOE ɛ4 allele non-carriers. Ultimately, our results reinforce the urgent need to target modifiable behaviors early in the lifespan, and to tailor dementia prevention strategies to individual risk profiles for maximum impact on brain health.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
This work was conducted using the UK Biobank Resource under Application Number 108907, Applicant Murdoch University. We are grateful to all UK Biobank participants for generously contributing their time and data to this research.
Author contributions
KRS, JDD, SLG, KIE and BMB contributed to the conception and design of the study; KRS, JDD and BMB contributed to the acquisition and analysis of the data; KRS, KIE, BS and BMB contributed to drafting the text and preparing the figures. All authors are responsible for the decision to submit for publication.
Funding
This work was supported by a Western Australian Near-Miss Award granted to BMB. Western Australian Future Health and Innovation Fund, Government of Western Australia.
Data availability
All data used in this study are available through registration on the UK Biobank at https://www.ukbiobank.ac.uk/use-our-data/applyfor-access/
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Livingston G, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet. 2024;404(10452):572–628. 10.1016/S0140-6736(24)01296-0. [DOI] [PubMed] [Google Scholar]
- 2.Schellenberg ES, Dryden DM, Vandermeer B, Ha C, Korownyk C. Lifestyle interventions for patients with and at risk for type 2 diabetes: a systematic review and meta-analysis. Ann Intern Med. 2013;159(8):543–51. 10.7326/0003-4819-159-8-201310150-00007. [DOI] [PubMed] [Google Scholar]
- 3.Deckers K, et al. Target risk factors for dementia prevention: a systematic review and Delphi consensus study on the evidence from observational studies. Int J Geriatr Psychiatry. 2015;30(3):234–46. 10.1002/gps.4245. [DOI] [PubMed] [Google Scholar]
- 4.Neuffer J, et al. Association of lifestyle for brain health risk score (LIBRA) and genetic susceptibility with incident dementia and cognitive decline. Alzheimers Dement. 2024;20(6):4250–9. 10.1002/alz.13801. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Schiepers OJG, et al. Lifestyle for Brain Health (LIBRA): a new model for dementia prevention. Int J Geriatr Psychiatry. 2018;33(1):167–75. 10.1002/gps.4700. [DOI] [PubMed] [Google Scholar]
- 6.Röhr S, Stephens C, Alpass F. Lifestyle for brain health and cognitive functioning in midlife to early late-life New Zealanders: utility of the LIBRA index. Int J Geriatr Psychiatry. 2024;39(5):e6091. 10.1002/gps.6091. [DOI] [PubMed] [Google Scholar]
- 7.Deckers K, et al. Long-term dementia risk prediction by the LIBRA score: a 30-year follow-up of the CAIDE study. Int J Geriatr Psychiatry. 2020;35(2):195–203. 10.1002/gps.5235. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Rosenau C, Köhler S, van Boxtel M, Tange H, Deckers K, Barbera M. Validation of the updated “LIfestyle for BRAin health” (LIBRA) index in the English longitudinal study of ageing and Maastricht aging study. J Alzheimers Dis. 2024;101(4):1237–48. 10.3233/jad-240666. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Vos SJB, et al. Modifiable risk factors for prevention of dementia in midlife, late life and the oldest-old: validation of the LIBRA index. J Alzheimers Dis. 2017;58(2):537–47. 10.3233/jad-161208. [DOI] [PubMed] [Google Scholar]
- 10.DeJong NR, et al. Brain structure and connectivity mediate the association between lifestyle and cognition: the Maastricht Study. Brain Commun. 2024. 10.1093/braincomms/fcae171. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Heger IS, et al. Associations of the lifestyle for brain health index with structural brain changes and cognition. Neurology. 2021;97(13):e1300-12. 10.1212/WNL.0000000000012572. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Bountziouka V, Nelson CP, Wang Q, Musicha C, Codd V, Samani NJ. Dietary patterns and practices and leucocyte telomere length: findings from the UK Biobank. J Acad Nutr Diet. 2023;123(6):912-922.e26. 10.1016/j.jand.2023.01.008. [DOI] [PubMed] [Google Scholar]
- 13.Lyall DM, et al. Cognitive test scores in UK biobank: data reduction in 480,416 participants and longitudinal stability in 20,346 participants. PLoS ONE. 2016;11(5):e0156366. 10.1371/journal.pone.0156366. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Fawns-Ritchie C, Deary IJ. Reliability and validity of the UK Biobank cognitive tests. PLoS One. 2020;15(4):e0231627. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Campbell T, Cullen B. Estimating the effect of physical activity on cognitive function within the UK Biobank cohort. Int J Epidemiol. 2023;52(5):1592–611. 10.1093/ije/dyad009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Miller KL, et al. Multimodal population brain imaging in the UK Biobank prospective epidemiological study. Nat Neurosci. 2016;19(11):1523–36. 10.1038/nn.4393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Alfaro-Almagro F, et al. Image processing and quality control for the first 10,000 brain imaging datasets from UK biobank,. Neuroimage. 2018;166:400–24. 10.1016/j.neuroimage.2017.10.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wilkinson T, et al. Identifying dementia outcomes in UK Biobank: a validation study of primary care, hospital admissions and mortality data. Eur J Epidemiol. 2019;34:557–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.J. Mackenbach, Health and deprivation. Inequality and the North: by P. Townsend, P. Phillimore and A. Beattie (eds.) Croom Helm Ltd, London, 1987 221 pp., ISBN 0–7099–4352–0, £8.95. Health Policy (New York). 1988;10: 207. 10.1016/0168-8510(88)90006-1. [DOI]
- 20.de Frutos Lucas J, Sewell KR, García-Colomo A, Markovic S, Erickson KI, Brown BM. How does apolipoprotein E genotype influence the relationship between physical activity and Alzheimer’s disease risk? A novel integrative model. Alzheimers Res Ther. 2023;15(1):22. 10.1186/s13195-023-01170-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Angelopoulou E, Paudel YN, Papageorgiou SG, Piperi C. APOE genotype and Alzheimer’s disease: the influence of lifestyle and environmental factors. ACS Chem Neurosci. 2021;12(15):2749–64. 10.1021/acschemneuro.1c00295. [DOI] [PubMed] [Google Scholar]
- 22.Sang F, Chen Y, Chen K, Dang M, Gao S, Zhang Z. Sex differences in cortical morphometry and white matter microstructure during brain aging and their relationships to cognition. Cereb Cortex. 2021;31(11):5253–62. 10.1093/cercor/bhab155. [DOI] [PubMed] [Google Scholar]
- 23.Than S, et al. Interactions between age, sex, menopause, and brain structure at midlife: a UK Biobank study. J Clin Endocrinol Metab. 2020;106(2):410–20. 10.1210/clinem/dgaa847. [DOI] [PubMed] [Google Scholar]
- 24.Barha CK, Galea LAM. Influence of different estrogens on neuroplasticity and cognition in the hippocampus. Biochimica et Biophysica Acta (BBA) - General Subjects. 2010;1800(10):1056–67. 10.1016/j.bbagen.2010.01.006. [DOI] [PubMed] [Google Scholar]
- 25.Bagit A, Hayward GC, MacPherson REK. Exercise and estrogen: common pathways in Alzheimer’s disease pathology. Am J Physiol Endocrinol Metab. 2021;321(1):E164–8. 10.1152/ajpendo.00008.2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Mosca L, Barrett-Connor E, Wenger NK. Sex/gender differences in cardiovascular disease prevention. Circulation. 2011;124(19):2145–54. 10.1161/CIRCULATIONAHA.110.968792. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Kim ESH, et al. Sex differences in peripheral vascular disease: a scientific statement from the American Heart Association. Circulation. 2025;151(14):e877–904. 10.1161/CIR.0000000000001310. [DOI] [PubMed] [Google Scholar]
- 28.Kessler RC, et al. The epidemiology of major depressive disorderresults from the National Comorbidity Survey Replication (NCS-R). JAMA. 2003;289(23):3095–105. 10.1001/jama.289.23.3095. [DOI] [PubMed] [Google Scholar]
- 29.Lane CA, et al. Associations between vascular risk across adulthood and brain pathology in late life: evidence from a British birth cohort. JAMA Neurol. 2020;77(2):175–83. 10.1001/jamaneurol.2019.3774. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Sabia S, et al. Physical activity, cognitive decline, and risk of dementia: 28 year follow-up of Whitehall II cohort study. BMJ. 2017;357:j2709. 10.1136/bmj.j2709. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data used in this study are available through registration on the UK Biobank at https://www.ukbiobank.ac.uk/use-our-data/applyfor-access/