Abstract
INTRODUCTION
Physical activity is a key modifiable protective factor for cognitive health, but randomized controlled trials initiating physical activity in late life have often failed to demonstrate significant cognitive benefits. The optimal window for prevention may be midlife, before irreversible pathology is established. We emulated a target trial to estimate counterfactual risks of cognitive impairment under hypothetical strategies of initiating or maintaining physical activity, alone or in combination with lifestyle and vascular risk modifications from midlife.
METHODS
We utilized data from the Whitehall II study (n = 3011) with baseline ages 45 to 64, following participants from midlife for up to 18 years. We applied the parametric g‐formula to handle time‐varying confounding and emulate hypothetical interventions. Strategies included physical activity, alone or combined with modifications to diet, smoking, alcohol consumption, and vascular risk modification (systolic blood pressure, total cholesterol, fasting glucose, and body mass index). The primary outcome was cognitive impairment based on cognitive tests. One thousand sets of bootstrapping were conducted to obtain 95% confidence intervals (CIs).
RESULTS
Initiating physical activity alone did not significantly reduce the 18‐year cumulative incidence of cognitive impairment. However, the estimated risk of cognitive impairment was lower under the strategy combining physical activity with healthy diet and no smoking, and with normal body mass index, compared with the natural course. Also, comprehensive strategies targeting physical activity, lifestyle factors, and all vascular risk factors achieved significant risk reduction.
DISCUSSION
While physical activity initiated in midlife may be insufficient as a single intervention in this cohort, its integration into a comprehensive multidomain strategy addressing lifestyle and vascular health significantly reduces the risk of cognitive impairment. These findings support holistic prevention strategies from midlife to improve long‐term cognitive outcomes.
Keywords: cognitive impairment, midlife risk, multidomain intervention, physical activity, target trial emulation factors, Whitehall II study
Highlights
Midlife physical activity alone did not reduce risk of cognitive impairment.
Addressing physical activity and other risk factors from midlife reduces risk.
Target trial emulation provides evidence on long‐term prevention from midlife.
1. INTRODUCTION
Physical activity is an important modifiable protective factor for cognitive outcomes. 1 However, causal evidence at the population level has been limited. It is notable that prior interventions of physical activity among older adults may have missed the optimal window to prevent cognitive outcomes, as cognitive decline starts from an earlier stage of life. For example, in the 24‐month Lifestyle Interventions and Independence for Elders (LIFE) trial of 1635 sedentary adults aged 70 to 89 years, physical activity did not significantly reduce incident dementia. 2 Irreversible cerebral and cognitive changes may already be present by late life, suggesting that earlier primary prevention is crucial.
Cognitive impairment and dementia are influenced by multiple modifiable lifestyle, metabolic, and vascular factors across life. Midlife hypertension, obesity, hypercholesterolemia, and diabetes or glucose dysregulation have been associated with poorer cognitive outcomes, whereas late‐life associations are often attenuated or reversed, possibly because of survival bias, reverse causation, or preclinical disease. 3 , 4 , 5 , 6 Smoking increases the risk of dementia, and earlier cessation may reduce excess risk. 7 Dietary patterns such as the Mediterranean‐Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay (MIND) diet have been associated in observational studies with slower cognitive decline and lower risk of dementia, 8 , 9 , 10 and alcohol use is recognized as a potentially modifiable risk factor for dementia. 11 Although a recent randomized controlled trial (RCT) of the MIND diet in older adults showed no significant between‐group difference in cognitive change over 3 years, 12 this does not exclude benefit from dietary or multidomain risk factor modification initiated earlier. Together, these findings support a lifecourse framework in which midlife is a critical window for modifying behavioral and vascular pathways before substantial neuropathology accumulates. Because cognitive impairment is multifactorial, multidomain strategies combining physical activity with dietary improvement, smoking cessation, avoidance of heavy alcohol use, and vascular/metabolic control may provide greater benefit than physical activity alone by targeting overlapping pathways.
The ideal evidence would come from multidomain randomized trials spanning midlife to late life, but multi‐decade RCTs are often infeasible. Target trial emulation applies causal inference methods to observational data to mimic a hypothetical trial when a trial is infeasible. 13 Furthermore, while dementia is a common endpoint, cognitive impairment defined by validated cutoffs remains under‐examined. As cognitive impairment typically precedes dementia and indicates underlying pathological processes amenable to intervention, it serves as an intermediate endpoint. 14 Therefore, we aimed to emulate a target trial to estimate the effects of hypothetical intervention strategies of physical activity and its combination with other modifiable factors initiated in midlife on reducing the risk of cognitive impairment.
2. METHODS
The Whitehall II study is a prospective cohort established in 1985 to 1987 and includes 10,308 British civil servants aged 35 to 55 years. Participants attended research clinics every 5 years, with validated questionnaires mailed between phases. 15 Phase 5 (1997 to 1999) served as baseline because this was when the cognitive test battery began for all participants. We included Phase 5 participants aged 45 to 64 years and excluded those who were cognitively impaired or without cognitive assessments or with missing essential covariates for the hypothetical interventions at Phase 5. Participants with missing Phase 5 cognitive data (3.9%) but available data from the Phase 3 experimental module were included using Phase 3 results as baseline (Figure S1).
2.1. Hypothetical interventions
2.1.1. Physical activity
RESEARCH IN CONTEXT
Systematic review: The authors reviewed the literature using traditional sources (e.g., PubMed). While physical activity is a known protective factor for cognition, RCTs initiated in late life often yield non‐significant results. No study has investigated the long‐term effectiveness of multidomain interventions initiated from midlife (aged 45 to 64 years) using causal inference methods to emulate long‐term trials.
Interpretation: Our findings suggest that midlife physical activity alone may not significantly reduce the risk of cognitive impairment in this cohort, but in combination with lifestyle (diet, no smoking) and vascular risk factor modifications (blood pressure, lipids, glucose, body mass index), it significantly reduces risk. This supports the midlife window hypothesis for risk factor modification.
Future directions: Future research should replicate these findings in more diverse populations with extended outcomes, including cognitive decline and incident dementia.
Physical activity was assessed at Phase 5 and at subsequent clinical phases included in the analysis using the Minnesota Leisure‐Time Physical Activity Questionnaire. 16 Participants reported the frequency and duration of 20 activities, including walking, cycling, sports, gardening, and do‐it‐yourself tasks, over the previous 4 weeks. Metabolic equivalent (MET) values were assigned from the 2011 Compendium of Physical Activities, 17 and activities were classified as light (<3 METs), moderate (3 to 5.9 METs), or vigorous (≥6 METs). 18 Moderate to vigorous physical activity (MVPA) was calculated by summing moderate and vigorous minutes. The questionnaire showed moderate validity against accelerometry (Spearman's r = 0.33; 95% confidence interval [CI]: 0.30 to 0.36). 19 Repeated measures were incorporated as time‐varying variables in the parametric g‐formula to emulate sustained strategies for maintaining MVPA thresholds during follow‐up.
2.1.2. Lifestyle factors
Diet was assessed using a 127‐item semi‐quantitative food‐frequency questionnaire 20 adapted from the Nurses' Health Study 21 and modified for the UK population. 22 Validation against 7‐day dietary diaries showed moderate correlations for energy and nutrient intake. 23 Participants reported consumption frequency from “never or less than once per month” to “six or more times per day” using standardized portion sizes. 23 Dietary quality was evaluated with the MIND diet score, 10 including 10 brain‐healthy components (green leafy vegetables, other vegetables, nuts, berries, beans, whole grains, fish, poultry, olive oil, and wine) and five unhealthy components (red meats, butter and stick margarine, cheese, pastries and sweets, and fried/fast food). Because the olive‐oil item was introduced while the berries item was omitted since Phase 11, the adjusted MIND score ranged from 0 to 14. Current smoking and alcohol consumption were self‐reported; heavy drinking was defined as >14 units/week.
2.1.3. Vascular risk factors
Blood pressure was measured at each clinical phase in the sitting position after 5 min of rest, using a Hawksley random‐zero sphygmomanometer in 1997 and an OMRON HEM 907 digital sphygmomanometer in 2003, 2007, 2012, and 2015. 24 Blood samples were collected after either an 8‐h overnight fast for morning appointments or at least a 4‐h fast after a light fat‐free breakfast for afternoon appointments. Total cholesterol and fasting glucose were measured at Phases 5, 7, 9, and 11; glucose used the glucose oxidase method. Weight was measured in underwear to the nearest 0.1 kg on Soehnle electronic scales with digital readout (Leifheit AS, Nassau, Germany). Height was measured in bare feet to the nearest 1 mm using a stadiometer with the participant standing erect and head in the Frankfurt plane. Body mass index (BMI) was calculated as kg/m2. 25
2.2. Cognitive tests
Cognitive function was assessed at Phases 5, 7, 9, 11, and 12 using a battery with good test‐retest reliability (range: 0.6 to 0.9). 26 Memory was assessed by a 20‐word free recall test in which one‐ or two‐syllable words were presented at 2‐s intervals and participants wrote as many as possible in any order within 2 min. 27 Reasoning used the Alice Heim 4‐I test, with 65 verbal and mathematical reasoning items of increasing difficulty. 28 Fluency included phonemic (“s”) and semantic (animal names) 1‐min written tests. 29 Vocabulary was assessed with the Mill Hill test. 30 Mean cognitive scores and standard deviations for each test were estimated using linear regression adjusted for age, sex, and years of education. Cognitive impairment was defined as performance ≥1.5 standard deviations below the cohort‐specific mean on any cognitive test after adjustment, 31 identifying incident low cognitive performance in an epidemiologic cohort rather than clinical mild cognitive impairment. The Mini‐Mental State Examination (MMSE) 32 was administered from Phase 5 but was not included in the primary outcome because of limited score variability. Sensitivity analyses used an alternative criterion of MMSE < 24 or performance ≥1.5 standard deviations below the mean on any other cognitive assessment.
2.3. Covariates
Covariates included age, sex, years of education, marital/cohabitation status, household income (£999 to £19,999, £20,000 to £39,999, £40,000 to £59,999, £60,000 to £200,000), and apolipoprotein E (APOE) ε4 allele count (0, 1, 2) at baseline, plus self‐reported depression and cardiovascular disease as time‐varying covariates. Cardiovascular disease included physician‐diagnosed angina, heart attack, stroke, transient ischemic attack, or other heart conditions.
Missing covariates are reported in Table S1. To address missingness, multiple imputation by chained equations (MICE) was performed using the mice and miceadds packages in R, assuming the data were missing at random. A total of 50 imputed datasets were generated, and the results were pooled using Rubin's Rules for the final results.
2.4. Statistical analysis
Baseline characteristics were described as means and standard deviations or frequencies and percentages. Physical activity strategies were maintaining MVPA ≥150 min/week and ≥75 min/week. Other hypothetical intervention components, combined with maintaining MVPA ≥150 min/week, were diet (MIND score ≥60th percentile), no current smoking, no heavy drinking, systolic blood pressure (SBP) <130 mmHg, total cholesterol <6.18 mmol/L, fasting glucose <5.6 mmol/L, and BMI 18.5 to 24.9 kg/m2. The evaluated multidomain strategies were physical activity combined with diet, smoking, alcohol use, diet+smoking, smoking+alcohol use, diet+smoking+alcohol use, blood pressure, cholesterol, glucose, BMI, blood pressure+glucose+cholesterol+BMI, diet+smoking+blood pressure+glucose+cholesterol+BMI, and diet+smoking+alcohol use+blood pressure+glucose+cholesterol+BMI. Thus, the lifestyle combinations covered diet, smoking, and alcohol components, whereas the vascular combinations covered blood pressure, cholesterol, glucose, and BMI individually and jointly.
The parametric g‐formula estimated the effects of hypothetical intervention strategies under assumptions of no unmeasured or residual confounding, no measurement error, and no model misspecification. It estimates outcome risks under hypothetical interventions, simulates counterfactual outcomes, and averages stratum‐specific estimates over the covariate distribution; for example, it compares all participants engaging in ≥150 min/week of MVPA with the natural course. Hypothetical interventions were applied among participants who did not meet the relevant target at baseline, and the method handles time‐varying confounding that standard regression cannot.
The g‐formula proceeded in three steps. First, at Phases 5, 7, 9, 11, and 12 (time index k = 0 to 4), we modeled each risk factor, censoring due to death or loss to follow‐up without cognitive impairment, and cognitive impairment risk as functions of prior history of risk factors. Second, Monte Carlo simulation followed a random sample of 10,000 participants from Phase 5 to Phase 12 under each strategy. At each phase, covariates were predicted using coefficients from the observational models, intervention values were assigned when required by the strategy, and probabilities of censoring and cognitive impairment at the next phase were estimated from the corresponding models. Each strategy was compared with the natural course, which served as the common reference condition, and no pairwise comparison was conducted among active intervention strategies. For each imputed dataset, effect estimates are presented as risk ratios (RRs). Further, 95% CIs were conducted using 1000 sets of bootstrapping. All reported estimates of statistical analysis were obtained from multiple‐imputed data. The gfoRmula R package was used for the analysis.
3. RESULTS
3.1. Baseline characteristics
A total of 3011 participants were included in the analysis. The mean age was 54.5±5.1 years at baseline, and 72.9% of the participants were men. Participants had, on average, 15.1±3.9 years of education, and 78.8% were married or cohabiting. In addition, 16.7% of participants had a household income less than £20,000, while 18.4% had an income between £60,000 and £200,000. With respect to genetic risk, 25.9% carried one APOE ε4 allele, and 2.1% carried two alleles. Cardiovascular disease was reported by 14% of participants, and depression was reported by 20.2%.
The baseline distribution of vascular and lifestyle risk factors showed that mean SBP was 121.5±16.0 mmHg, with 29% of participants having SBP ≥130 mmHg. Mean total cholesterol was 5.9±1.0 mmol/L, with 38.6% having total cholesterol ≥6.18 mmol/L. Mean fasting blood glucose was 5.2±1.0 mmol/L; 14% of participants had pre‐diabetes (5.6 to 6.9 mmol/L), and 1.9% had diabetes (≥7.0 mmol/L). The mean BMI was 26.0±4.0 kg/m2, with 0.9% of participants being underweight, 42.9% overweight, and 13.0% obese. Current smoking was reported by 8.8% of participants, and 33.9% reported heavy alcohol consumption (≥14 units/week). The mean MIND diet score was 7.8±1.6. Regarding physical activity, mean total physical activity was 654.6±311.4 min/week, and mean MVPA was 210.8±189.5 min/week; 27.9% of participants engaged in MVPA for <75 min/week, 17.3% for 75 to 149 min/week, and 54.7% achieved ≥150 min/week (Table 1).
TABLE 1.
Baseline characteristics of participants included in analysis.
| Characteristics | Statistics for non‐missing data | Missing, n (%) |
|---|---|---|
| Age (years) | 54.5 ± 5.1 | |
| Male, n (%) | 2194 (72.9) | |
| Years of education | 15.1 ± 3.9 | |
| Marital/cohabitation, n (%) | 2298 (78.8) | 95 (3.2) |
| Household income, n (%) | 396 (13.2) | |
| £999 to £19,999 | 437 (16.7) | |
| £20,000 to £39,999 | 987 (37.7) | |
| £40,000 to £59,999 | 709 (27.1) | |
| £60,000 to £200,000 | 482 (18.4) | |
| APOE ε4 allele, n (%) | 477 (15.8) | |
| No allele | 1824 (72.0) | |
| One allele | 657 (25.9) | |
| Two alleles | 53 (2.1) | |
| CVD, n (%) | 417 (14.0) | 22 (0.7) |
| Depression, n (%) | 598 (20.2) | 52 (1.7) |
| Systolic blood pressure (SBP) (mmHg) | 121.5 ± 16.0 | |
| Hypertension (SBP ≥ 130 mmHg), n (%) | 873 (29.0) | |
| Total blood cholesterol (mmol/L) | 5.9 ± 1.0 | |
| Hypercholesterolemia (≥ 6.18 mmol/L), n (%) | 1163 (38.6) | |
| Fasting blood glucose (mmol/L) | 5.2 ± 1.0 | |
| <5.6 mmol/L, n (%) | 2531 (84.1) | |
| 5.6‐6.9 mmol/L, n (%) | 422 (14.0) | |
| ≥7 mmol/L, n (%) | 58 (1.9) | |
| BMI (kg/m2) | 26.0 ± 4.0 | |
| Underweight (<18.5 kg/m2), n (%) | 27 (0.9) | |
| Normal weight (18.5 to 24.9 kg/m2), n (%) | 1302 (43.2) | |
| Overweight (25 to 29.9 kg/m2), n (%) | 1291 (42.9) | |
| Obese (≥30 kg/m2), n (%) | 391 (13.0) | |
| Current smoker, n (%) | 264 (8.8) | |
| Heavy drinking (≥14 units per week) | 1022 (33.9) | |
| MIND dietary score (range: 0 to 14) | 7.8 ± 1.6 | |
| MIND dietary score ≥60th percentile (8.5), n (%) | 1221 (40.6) | |
| MVPA (min/week) | 210.8 ± 189.5 | |
| MVPA <75 min/week, n (%) | 841 (27.9) | |
| MVPA 75 to 149 min/week, n (%) | 522 (17.3) | |
| MVPA ≥150 min/week, n (%) | 1648 (54.7) | |
| Total physical activity (min/week) | 654.6 ± 311.4 | |
| Total physical activity <75 min/week, n (%) | 30 (1.0) | |
| Total physical activity 75 to 149 min/week, n (%) | 44 (1.5) | |
| Total physical activity ≥150 min/week, n (%) | 2937 (97.5) |
3.2. Physical activity alone
For the emulation of physical activity strategies, two interventions were evaluated: maintaining MVPA ≥150 min/week and maintaining MVPA ≥75 min/week. These interventions were hypothetically applied to the 45.2% and 29.8% of participants who did not meet these respective targets. After 18 years of follow‐up, the cumulative incidence of cognitive impairment was similar for both strategies (24.9% vs 25.1%), which was close to the natural course incidence of 25.5%. No significant difference in the risk of cognitive impairment was found between either strategy compared to the natural course (Table 2).
TABLE 2.
Effectiveness of hypothetical physical activity interventions on incident cognitive impairment, 1997 to 2016.
| Hypothetical interventions | Average percentage with intervention | Cumulative incidence (%) | Risk ratio (95% CI) |
|---|---|---|---|
| Natural course | 0 | 25.5 | Reference |
| Maintain MVPA ≥150 min/week | 45.2 | 24.9 | 0.98 (0.94, 1.01) |
| Maintain MVPA ≥75 min/week | 29.8 | 25.1 | 0.98 (0.96, 1.01) |
Abbreviations: CI, confidence interval; MVPA, moderate to vigorous physical activity.
3.3. Physical activity plus lifestyle factors
For the emulation of hypothetical multidomain interventions, we evaluated combinations of physical activity and other lifestyle factors. Strategies including physical activity+diet, physical activity+smoking, physical activity+drinking, physical activity+diet+smoking, physical activity+smoking+drinking, and physical activity+diet+smoking+drinking were newly initiated in 67.3%, 46.1%, 53.9%, 67.3%. 54.4%, and 73.2% of participants, respectively. Over 18 years of follow‐up, compared to the natural course (25.5% cumulative incidence), the cumulative incidence of cognitive impairment was significantly reduced for physical activity+diet (23.6%; RR = 0.93, 95% CI: 0.87, 0.98) and physical activity+diet+smoking (23.2%; RR = 0.91, 95% CI: 0.86, 0.97) (Table 3).
TABLE 3.
Effectiveness of hypothetical multidomain interventions addressing physical activity and other lifestyle factors on incident cognitive impairment, 1997 to 2016.
| Hypothetical interventions | Average percentage with intervention | Cumulative incidence (%) | Risk ratio (95% CI) |
|---|---|---|---|
| Natural course | 0 | 25.5 | Reference |
| PA + diet * | 67.3 | 23.6 | 0.93 (0.87, 0.98) |
| PA + smoking † | 46.1 | 24.4 | 0.96 (0.91, 1.00) |
| PA + alcohol use ‡ | 53.9 | 26.3 | 1.03 (0.98, 1.09) |
| PA + diet + smoking | 67.3 | 23.2 | 0.91 (0.86, 0.97) |
| PA + smoking + alcohol use | 54.4 | 25.8 | 1.01 (0.95, 1.07) |
| PA + diet + smoking + alcohol use | 73.2 | 24.6 | 0.97 (0.89, 1.04) |
Abbreviations: CI, confidence interval; PA, physical activity.
PA: keeping MVPA ≥150 min/week; diet: keeping Mediterranean‐DASH Intervention for Neurodegenerative Delay (MIND) diet score >8.5 (60th percentile).
Smoking: no smoking.
Alcohol use: no heavy drinking (>14 units/week).
3.4. Physical activity plus vascular factors
We also examined combinations of physical activity and the control of vascular risk factors, specifically maintaining SBP <130 mmHg, total cholesterol <6.18 mmol/L, fasting blood glucose <5.6 mmol/L, and BMI 18.5 to 24.9 kg/m2. The strategies of physical activity+blood pressure, physical activity+cholesterol, physical activity+glucose, physical activity+BMI, and physical activity+blood pressure+cholesterol+glucose+BMI were newly initiated among 63.0%, 57.5%, 61.9%, 68.3%, and 86.3% of participants, respectively. Over 18 years of follow‐up, the cumulative incidence of cognitive impairment was significantly reduced for physical activity+BMI (23.7%; RR = 0.93, 95% CI: 0.88, 0.98) and physical activity+blood pressure+glucose+cholesterol+BMI (23.4%; RR = 0.92, 95% CI: 0.86, 0.98) (Table 4).
TABLE 4.
Effectiveness of hypothetical multidomain interventions addressing physical activity and vascular risk factors on incident cognitive impairment, 1997 to 2016.
| Hypothetical interventions | Average percentage with intervention | Cumulative incidence % |
Risk ratio(95% CI) |
|---|---|---|---|
| Natural course | 0 | 25.5 | Reference |
| PA + BP * | 63 | 24.8 | 0.97 (0.93, 1.02) |
| PA + cholesterol † | 57.5 | 25 | 0.98 (0.94, 1.03) |
| PA + glucose ‡ | 61.9 | 24.4 | 0.96 (0.92, 1.00) |
| PA + BMI § | 68.3 | 23.7 | 0.93 (0.88, 0.98) |
| PA + BP + glucose + cholesterol + BMI | 86.3 | 23.4 | 0.92 (0.86, 0.98) |
Abbreviations: CI, confidence interval; BMI, body mass index; BP, blood pressure; PA, physical activity.
BP: keeping systolic blood pressure <130 mmHg.
Cholesterol: keeping total cholesterol <6.18 mmol/L.
Glucose: keeping fasting glucose <5.6 mmol/L.
BMI: keeping body mass index 18.5‐24.9 kg/m2.
3.5. Comprehensive multidomain strategies
Finally, because alcohol consumption reduction was the least protective intervention in this cohort, a comprehensive strategy was assessed that combined physical activity, lifestyle factors (diet, smoking) except for alcohol, and vascular risk factors (blood pressure, cholesterol, glucose, BMI). This combined strategy was newly initiated among 91.3% of participants and showed a significant reduction in the risk of cognitive impairment (RR = 0.86, 95% CI: 0.79, 0.93). Also, the strategy that combined all components (physical activity, diet, smoking, alcohol, blood pressure, cholesterol, glucose, BMI) showed a significant reduction in the risk of cognitive impairment (RR = 0.91, 95% CI: 0.83, 0.99) (Table 5).
TABLE 5.
Effectiveness of hypothetical multidomain interventions addressing physical activity, other lifestyle, and vascular risk factors on incident cognitive impairment, 1997 to 2016.
| Hypothetical interventions | Average percentage with intervention | Cumulative incidence % |
Risk ratio (95% CI) |
|---|---|---|---|
| Natural course | 0 | 25.5 | Reference |
| PA+diet+smoking +BP+glucose+cholesterol+BMI * | 91.3 | 21.8 | 0.86 (0.79, 0.93) |
| PA+diet+smoking+alcohol+BP+glucose+cholesterol+BMI † | 92.3 | 23.1 | 0.91 (0.83, 0.99) |
Abbreviation: BMI, body mass index; BP, blood pressure; CI, confidence interval; PA, physical activity.
PA: keeping MVPA ≥150 min/week; diet: keeping Mediterranean‐DASH Intervention for Neurodegenerative Delay (MIND) diet score >8.5 (60th percentile); smoking: no smoking; BP: keeping systolic blood pressure <130 mmHg; cholesterol: keeping total cholesterol <6.18 mmol/L; glucose: keeping fasting glucose <5.6 mmol/L; BMI: keeping body mass index 18.5‐24.9 kg/m2.
Alcohol use: no heavy drinking (>14 units/week).
Additional sensitivity analyses that examined MMSE <24 or performance ≥1.5 standard deviations below the mean on any of the other cognitive assessments as an alternative indicator of cognitive impairment yielded similar results as the main analysis (Table S2‐S5, Figure S2).
4. DISCUSSION
In the Whitehall II cohort followed from midlife to late life for up to 18 years, physical activity initiated in midlife did not independently reduce the risk of cognitive impairment. However, using the parametric g‐formula, we found that several hypothetical multidomain strategies combining physical activity with other modifiable risk factors reduced the risk of cognitive impairment compared with the natural course (Figure 1). No RCT estimated the effects of initiating physical activity, alone or with other modifiable factors, from midlife on late‐life cognitive outcomes, and few studies used observational data and causal inference methods to estimate effects of such hypothetical interventions.
FIGURE 1.
Effectiveness of hypothetical multidomain interventions addressing physical activity, other lifestyle, and vascular risk factors on risk of cognitive impairment, 1997 to 2016. The right column shows the risk ratios and 95% confidence intervals (CIs) for each hypothetical intervention compared with the natural course. Dots represent the estimated risk ratios and horizontal bars extend from the lower to the upper bound of the 95% CI. The vertical dashed line indicates the null value of 1.00. Bolded interventions marked with an asterisk indicate statistical significance.
Our findings differ from results from the Health and Retirement Study using target trial emulation, where initiating physical activity from age 45 to 65 years showed a significant risk reduction of cognitive impairment, whereas initiating it after age 65 did not. 33 The discrepancy may be explained by the high prevalence of physical activity in Whitehall II, a civil‐servant cohort with generally better baseline health. This cohort consists of civil servants who generally maintain better baseline health. About 72% achieved ≥75 min/week of MVPA, which limited the potential impact of new physical activity interventions. Randomized trials and meta‐analyses suggest cognitive benefits of exercise in older adults, including for general cognition, memory, and executive function. 34 , 35 Our findings do not contradict this literature but address a different estimand: long‐term risk of incident cognitive impairment under sustained hypothetical strategies initiated in midlife. The null effect for physical activity alone may reflect high baseline activity, limitations of self‐reported activity, and the multifactorial etiology of cognitive impairment.
A previous Whitehall II analysis by Sabia et al. found no evidence that midlife physical activity was associated with lower risk of dementia and cognitive decline over 28 years. 18 The present study asked a different question: Using target trial emulation and the parametric g‐formula, we estimated counterfactual risks under sustained hypothetical strategies, accounted dynamically for time‐varying confounding, and evaluated clinically relevant combinations of physical activity with diet, smoking, alcohol, and modification of vascular risk factors.
Although physical activity alone was not significant, multidomain interventions involving physical activity, lifestyle, and vascular factors were beneficial, with estimated effects larger than diet alone (RR = 0.95, 95% CI: 0.91 to 0.99), no smoking alone (RR = 0.98, 95% CI: 0.96 to 1.00), or BMI alone (RR = 0.95, 95% CI: 0.91 to 1.00) (data not shown in tables). Multidomain interventions have shown promise in prior studies. The FINGER trial reported greater improvement in neuropsychological test scores after a multidomain intervention including diet, exercise, cognitive training, and vascular risk monitoring. 36 The US POINTER trial similarly found that a structured multidomain intervention increased global cognitive composite z‐scores compared with a self‐guided lifestyle intervention. 37 An Australian online multidomain intervention also improved outcomes compared with information‐only controls, 38 although other interventions did not show significant cognitive benefits. 39 , 40 , 41 While multidomain programs require more resources than single‐component advice, cost‐effectiveness evidence is emerging for FINGER‐based interventions. 42 Our study addresses an important knowledge gap by estimating long‐term counterfactual risks under hypothetical midlife intervention strategies, a question difficult to evaluate in conventional observational analyses or RCTs with short durations.
The estimated effects of strategies such as physical activity+diet, physical activity+diet+smoking, and the comprehensive vascular combination were moderate but meaningful, given that effects may be attenuated in a population with relatively healthy lifestyle and vascular profiles. Conversely, adding “no heavy drinking” to multidomain interventions reduced estimated effects. As a single hypothetical intervention, “no heavy drinking” had a statistically significant risk ratio greater than 1 (RR = 1.06, 95% CI: 1.02 to 1.10; data not shown in tables). In the Whitehall II cohort, where social drinking is prevalent, alcohol consumption may serve as a proxy for social interaction, a protective factor for cognitive impairment. 43 Reducing consumption without replacing this social mechanism could therefore remove a protective factor.
Several mechanisms may explain the benefits of multidomain intervention. Together, physical activity, lifestyle improvement, and control of vascular risk factors may reduce small vessel disease, including white matter hyperintensities, lacunes, microbleeds, silent infarcts, and atherosclerosis. 44 , 45 They may also improve cerebral hypoperfusion through vascular control 46 and reduce metabolic and inflammatory stress. 47 , 48 Because poor midlife lifestyle and vascular health are associated with dementia, earlier intervention may prevent adverse cognitive outcomes before irreversible cerebral damage occurs.
Our study has important clinical and public health implications. First, as cognitive impairment is a multifactorial condition, addressing multiple risk factors simultaneously may more effectively reduce risk. While physical activity is a promising protective factor for cognitive aging, the efficacy of preventive strategies is significantly enhanced when integrated into multidomain interventions. This suggests that, in addition to prescribing medications for vascular risk factors, clinicians should provide comprehensive lifestyle consultations, and community‐based events should be organized to promote holistic lifestyle optimization. Moreover, the data support the importance of the temporal window for intervention. While RCTs in older adults (age ≥65) often yield non‐significant results potentially due to established neuropathology, our results provide evidence that modifying risk factors during midlife (ages 45 to 64) may alter long‐term cognitive trajectories. This supports the hypothesis that mitigation must occur prior to the onset of substantial cerebral damage, and an earlier prevention can also be cost‐effective. 49
5. LIMITATIONS
Our study has limitations. First, Whitehall II participants were British civil servants with relatively good lifestyle and vascular health, which may have reduced estimated intervention effects and limits generalizability to less healthy or more socioeconomically diverse populations. The analytic sample was also predominantly men (73%), with the sample size of women below 900, and our findings may not be fully generalizable to women. Future studies should include broader and diverse populations to assess the effectiveness in different subgroups. Second, several key variables, including physical activity and diet, were based on self‐report. Although the physical activity questionnaire used in Whitehall II has demonstrated moderate validity against accelerometer‐assessed physical activity, 50 self‐reported measures remain subject to recall bias, social desirability bias, and imprecision in the classification of activity type, frequency, duration, and intensity. Similarly, dietary intake, smoking, and alcohol consumption may be affected by reporting error. Such measurement error may have led to exposure misclassification and could have attenuated the estimated effect size of the hypothetical interventions, particularly for physical activity alone. Third, there is a lack of repeating measurement of other modifiable factors, such as hearing loss, which could be subject to unmeasured confounding. Finally, cognitive impairment was defined as ≥1.5 standard deviations below the cohort‐specific mean on any test. This captures meaningful low performance across domains but may be more sensitive and less specific than clinical mild cognitive impairment criteria and may include false positives, although sensitivity analyses incorporating MMSE <24 were similar. Despite these limitations, the use of comprehensive longitudinal measurements and the parametric g‐formula allowed us to calculate the estimated effects of hypothetical multidomain interventions from midlife to late life. Analysis of cognitive impairment based on different sets of cognitive tests also made the results more robust.
6. CONCLUSIONS
In this target trial emulation, hypothetical multidomain strategies combining physical activity with modification of lifestyle and vascular risk factors initiated in midlife were estimated to reduce cognitive impairment risk compared with the natural course. These findings support further evaluation of multidomain prevention strategies beginning in midlife. Future studies should assess effects in more diverse populations and examine extended outcomes, including cognitive decline and incident dementia.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest. Author disclosures are available in the Supporting Information.
CONSENT STATEMENT
Written informed consent was obtained from all Whitehall II participants at baseline and renewed consent at each follow‐up.
Supporting information
Supporting Information
Supporting Information
ACKNOWLEDGMENTS
The authors have nothing to report. This work was supported by the Alzheimer's Association Research Grant (24AARG‐1196811).
REFERENCES
- 1. Iso‐Markku P, Kujala UM, Knittle K, Polet J, Vuoksimaa E, Waller K. Physical activity as a protective factor for dementia and Alzheimer's disease: systematic review, meta‐analysis and quality assessment of cohort and case–control studies. Br J Sports Med. 2022;56:701. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Sink KM, Espeland MA, Castro CM, et al. Effect of a 24‐Month Physical Activity Intervention vs Health Education on Cognitive Outcomes in Sedentary Older Adults: The LIFE Randomized Trial. Jama. 2015;314:781‐790. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Qiu C, Winblad B, Fratiglioni L. The age‐dependent relation of blood pressure to cognitive function and dementia. Lancet Neurol. 2005;4:487‐499. [DOI] [PubMed] [Google Scholar]
- 4. Fitzpatrick AL, Kuller LH, Lopez OL, et al. Midlife and late‐life obesity and the risk of dementia: cardiovascular health study. Arch Neurol. 2009;66:336‐342. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Solomon A, Kåreholt I, Ngandu T, et al. Serum cholesterol changes after midlife and late‐life cognition: twenty‐one‐year follow‐up study. Neurology. 2007;68:751‐756. [DOI] [PubMed] [Google Scholar]
- 6. Rawlings AM, Sharrett AR, Schneider ALC, et al. Diabetes in midlife and cognitive change over 20 years: a cohort study. Ann Intern Med. 2014;161:785‐793. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Zhong G, Wang Y, Zhang Y, Guo JJ, Zhao Y. Smoking is associated with an increased risk of dementia: a meta‐analysis of prospective cohort studies with investigation of potential effect modifiers. PLoS One. 2015;10:e0118333. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Morris MC, Tangney CC, Wang Y, et al. MIND diet slows cognitive decline with aging. Alzheimers Dement. 2015;11:1015‐1022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Hosking DE, Eramudugolla R, Cherbuin N, Anstey KJ. MIND not Mediterranean diet related to 12‐year incidence of cognitive impairment in an Australian longitudinal cohort study. Alzheimers Dement. 2019;15:581‐589. [DOI] [PubMed] [Google Scholar]
- 10. Morris MC, Tangney CC, Wang Y, Sacks FM, Bennett DA, Aggarwal NT. MIND diet associated with reduced incidence of Alzheimer's disease. Alzheimers Dement. 2015;11:1007‐1014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet. 2024;404:572‐628. [DOI] [PubMed] [Google Scholar]
- 12. Barnes LL, Dhana K, Liu X, et al. Trial of the MIND Diet for Prevention of Cognitive Decline in Older Persons. N Engl J Med. 2023;389:602‐611. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Hernán MA, Wang W, Leaf DE. Target Trial Emulation: A Framework for Causal Inference From Observational Data. JAMA. 2022;328:2446‐2447. [DOI] [PubMed] [Google Scholar]
- 14. Hugo J, Ganguli M. Dementia and cognitive impairment: epidemiology, diagnosis, and treatment. Clin Geriatr Med. 2014;30:421‐442. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Marmot M, Brunner E. Cohort Profile: the Whitehall II study. Int J Epidemiol. 2005;34:251‐256. [DOI] [PubMed] [Google Scholar]
- 16. Richardson MT, Leon AS, Jacobs DR, Jr., Ainsworth BE, Serfass R. Comprehensive evaluation of the Minnesota Leisure Time Physical Activity Questionnaire. J Clin Epidemiol. 1994;47:271‐281. [DOI] [PubMed] [Google Scholar]
- 17. AINSWORTH BE, HASKELL WL, HERRMANN SD, et al. 2011 Compendium of Physical Activities: a second update of codes and MET values. Med Sci Sports Exerc. 2011;43:1575‐1581. [DOI] [PubMed] [Google Scholar]
- 18. Sabia S, Dugravot A, Dartigues J‐F, et al. Physical activity, cognitive decline, and risk of dementia: 28 year follow‐up of Whitehall II cohort study. BMJ. 2017;357:j2709. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Sabia S, van Hees VT, Shipley MJ, et al. Association between questionnaire‐ and accelerometer‐assessed physical activity: the role of sociodemographic factors. Am J Epidemiol. 2014;179:781‐790. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Ozawa M, Shipley M, Kivimaki M, Singh‐Manoux A, Brunner EJ. Dietary pattern, inflammation and cognitive decline: The Whitehall II prospective cohort study. Clin nutr. 2017;36:506‐512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. WILLETT WC, SAMPSON L, STAMPFER MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol. 1985;122:51‐65. [DOI] [PubMed] [Google Scholar]
- 22. Bingham S Validation of dietary assessment methods in the UK arm of EPIC using weighed records, and 24‐hour urinary nitrogen and potassium and serum vitamin C and carotenoids as biomarkers. Int J Epidemiol. 1997;26(Suppl 1):S137‐S151. [DOI] [PubMed] [Google Scholar]
- 23. Brunner E, Stallone D, Juneja M, Bingham S, Marmot M. Dietary assessment in Whitehall II: comparison of 7 d diet diary and food‐frequency questionnaire and validity against biomarkers. Br J Nutr. 2001;86:405‐414. [DOI] [PubMed] [Google Scholar]
- 24. Abell JG, Kivimäki M, Dugravot A, et al. Association between systolic blood pressure and dementia in the Whitehall II cohort study: role of age, duration, and threshold used to define hypertension. Eur Heart J. 2018;39:3119‐3125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Singh‐Manoux A, Dugravot A, Shipley M, Brunner EJ, Elbaz A, Sabia S, Kivimaki M Obesity trajectories and risk of dementia: 28 years of follow‐up in the Whitehall II Study. Alzheimers Dement. 2018;14:178‐186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Singh‐Manoux A, Kivimaki M, Glymour MM, et al. Timing of onset of cognitive decline: results from Whitehall II prospective cohort study. BMJ. 2012;344:d7622. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Sabia S, Kivimaki M, Shipley MJ, Marmot MG, Singh‐Manoux A. Body mass index over the adult life course and cognition in late midlife: the Whitehall II Cohort Study. Am J Clin Nutr. 2009;89:601‐607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Heim AW. AH 4 Group Test of General Intelligence. Manual. By A (lice) W (inifred) Heim. Rev. Ed: NFER Publishing Company; 1970. [Google Scholar]
- 29. Borkowski JG, Benton AL, Spreen O. Word fluency and brain damage. Neuropsychologia. 1967;5:135‐140. [Google Scholar]
- 30. Raven JC. Guide to using the Mill Hill Vocabulary Scale with the Progressive Matrices Scales. H. K. Lewis & Co.; 1958. [Google Scholar]
- 31. Koscik RL, Jonaitis EM, Clark LR, et al. Longitudinal Standards for Mid‐life Cognitive Performance: Identifying Abnormal Within‐Person Changes in the Wisconsin Registry for Alzheimer's Prevention. J Int Neuropsychol Soc. 2019;25:1‐14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Cockrell JR, Folstein MF. Mini‐mental state examination, In: Abou‐Saleh MT, Katona CLE, Anand KA, eds. Principles and practice of geriatric psychiatry.. 2nd ed. John Wiley & Sons Ltd; 2002:140‐141. [Google Scholar]
- 33. Wei J, Lohman MC, Brown MJ, et al. Physical activity initiated from midlife on risk of dementia and cognitive impairment: The Health and Retirement Study. J Am Geriatr Soc. 2024;72(12):3668‐3680. [DOI] [PubMed] [Google Scholar]
- 34. Balbim GM, Falck RS, Barha CK, et al. Effects of exercise training on the cognitive function of older adults with different types of dementia: a systematic review and meta‐analysis. Br J Sports Med. 2022;bjsports‐2021‐104955. [DOI] [PubMed] [Google Scholar]
- 35. Singh B, Bennett H, Miatke A, et al.view and meta‐meta‐analysis. Br J Sports Med. 2025;59:866‐876. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at‐risk elderly people (FINGER): a randomised controlled trial. Lancet. 2015;385:2255‐2263. [DOI] [PubMed] [Google Scholar]
- 37. Baker LD, Espeland MA, Whitmer RA, et al. Structured vs Self‐Guided Multidomain Lifestyle Interventions for Global Cognitive Function: The US POINTER Randomized Clinical Trial. JAMA. 2025;334:681‐691. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Brodaty H, Chau T, Heffernan M, Ginige JA, et al. An online multidomain lifestyle intervention to prevent cognitive decline in at‐risk older adults: a randomized controlled trial. Nat Med. 2025;31:565‐573. [DOI] [PubMed] [Google Scholar]
- 39. Zülke AE, Pabst A, Luppa M, Roehr S, Seidling H, Oey A, et al. A multidomain intervention against cognitive decline in an at‐risk‐population in Germany: Results from the cluster‐randomized AgeWell.de trial. Alzheimers Dement. 2024;20(1):615‐628. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Andrieu S, Guyonnet S, Coley N, et al. Effect of long‐term omega 3 polyunsaturated fatty acid supplementation with or without multidomain intervention on cognitive function in elderly adults with memory complaints (MAPT): a randomised, placebo‐controlled trial. The Lancet Neurol. 2017;16:377‐389. [DOI] [PubMed] [Google Scholar]
- 41. Richard E, Moll van Charante EP, Hoevenaar‐Blom, et al. Healthy ageing through internet counselling in the elderly (HATICE): a multinational, randomised controlled trial. Lancet Digit Health. 2019;1:e424‐e34. [DOI] [PubMed] [Google Scholar]
- 42. Wimo A, Handels R, Antikainen R, et al. Dementia prevention: The potential long‐term cost‐effectiveness of the FINGER prevention program. Alzheimer's & Dementia. 2023;19:999‐1008. [DOI] [PubMed] [Google Scholar]
- 43. Wei CC, Hsieh MJ, Chuang YF. The Effects of Social Interaction Intervention on Cognitive Functions Among Older Adults Without Dementia: A Systematic Review and Meta‐Analysis. Innov Aging. 2024;8:igae084. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Nasrallah IM, Pajewski NM, Auchus AP, Chelune G, Cheung AK, Cleveland ML, et al. Association of Intensive vs Standard Blood Pressure Control With Cerebral White Matter Lesions. JAMA. 2019;322(6):524‐534. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Wardlaw JM, Smith C, Dichgans M. Mechanisms of sporadic cerebral small vessel disease: insights from neuroimaging. Lancet Neurol. 2013;12:483‐497. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. de la Torre JC. Cardiovascular risk factors promote brain hypoperfusion leading to cognitive decline and dementia. Cardiovasc Psychiatry Neurol. 2012;2012:367516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Gleeson M, Bishop NC, Stensel DJ, Lindley MR, Mastana SS, Nimmo MA. The anti‐inflammatory effects of exercise: mechanisms and implications for the prevention and treatment of disease. Nat Rev Immunol. 2011;11:607‐615. [DOI] [PubMed] [Google Scholar]
- 48. Craft S. Insulin resistance and Alzheimer's disease pathogenesis: potential mechanisms and implications for treatment. Curr Alzheimer Res. 2007;4:147‐152. [DOI] [PubMed] [Google Scholar]
- 49. Braun A, Höfler M, Auer S. Cost‐Effectiveness of Prevention for People at Risk for Dementia: A Scoping Review and Qualitative Synthesis. J Prev Alzheimers Dis. 2024;11:402‐413. [DOI] [PubMed] [Google Scholar]
- 50. Sabia S, van Hees VT, Shipley MJ, et al. Association between questionnaire‐ and accelerometer‐assessed physical activity: the role of sociodemographic factors. Am J Epidemiol. 2014;179:781‐790. [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
Supporting Information
Supporting Information