This cohort study investigates the association of meat consumption with risk of cognitive decline and dementia among older adults with APOE genotypes ε3/ε4 and ε4/ε4.
Key Points
Question
Is higher meat consumption associated with better cognitive health among individuals with APOE genotypes ε3/ε4 and ε4/ε4, and does this association differ from that observed in other genotypes?
Findings
In this cohort study among 2157 older adults without dementia, higher total meat consumption was associated with slower cognitive decline and a reduced dementia risk among older adults with APOE ε3/ε4 and ε4/ε4 genotypes. Interactions by APOE genotype were observed for trajectories of global cognition and episodic memory.
Meaning
These findings suggest that higher meat consumption than conventionally recommended may be associated with benefits in a genetically defined subgroup comprising approximately one-quarter of the global population.
Abstract
Importance
The apolipoprotein E (APOE) ε4 allele increases Alzheimer disease risk. Understanding genotype-specific dietary needs could inform more personalized prevention strategies.
Objective
To test the hypothesis that higher meat consumption may be associated with cognitive health benefits in individuals with APOE genotypes ε3/ε4 and ε4/ε4 (APOE34/44) and to examine whether this association differs from that in other genotypes.
Design, Setting, and Participants
This population-based cohort study used panel data analyses conducted in January 2025 to January 2026 over 15 years of follow-up in the Swedish National Study on Aging and Care–Kungsholmen (SNAC-K), using strategies aligned with causal inference principles. Recruitment was done in 2001 to 2004 among adults without dementia aged 60 years or older.
Exposures
The primary exposure was total meat consumption in grams per total kilocalories assessed via validated food frequency questionnaires. The secondary exposure was the ratio of processed to total meat.
Main Outcomes and Measures
Global cognitive trajectory, measured as change in z score per 10 years, was analyzed by linear regression. Incident dementia was analyzed using Fine and Gray subdistribution hazard ratios (sHRs), treating nondementia death as a competing risk.
Results
Among 2157 older adults without dementia (mean [SD] age 71.2 [9.2] years; 1337 female [62.0%]), 1680 participants had longitudinal cognition data and 569 participants (26.4%) had APOE34/44 genotypes. During follow-up, 296 participants developed dementia and 690 died without dementia. Among participants with APOE34/44 genotypes, higher total meat consumption (top vs bottom quintile) was associated with better cognitive trajectories (β = 0.32; 95% CI, 0.07 to 0.56; P = .01) and reduced dementia risk (sHR, 0.45; 95% CI, 0.21 to 0.95; P = .04). No associations were found in participants with APOE22/23/24/33 genotypes (cognitive trajectory: β = –0.11; 95% CI, –0.27 to 0.06; P = .20; dementia: sHR, 0.95; 95% CI, 0.57 to 1.61; P = .86). P values for APOE interaction were .004 for cognition and .10 for dementia. In the top quintile of meat consumption, dementia risk and cognitive decline were similar between APOE strata. A higher ratio of processed to total meat was unfavorably associated with dementia (sHR, 1.14; 95% CI, 1.01 to 1.29; P = .04), showing no APOE interaction and no substantial difference between unprocessed red meat and poultry. Post hoc analyses suggested concordant APOE interaction for all-cause mortality (unprocessed meat exposure, APOE34/44: HR, 0.85; 95% CI, 0.73 to 0.99; P = 0.04; P for interaction = .03).
Conclusions and Relevance
In this study, higher meat consumption was associated with better cognitive trajectories and lower dementia risk among individuals with APOE34/44 genotypes. The expected cognitive disadvantage among individuals with APOE34/44 genotypes was not observed at high meat consumption, suggesting clinical and public health relevance.
Introduction
Apolipoprotein E (APOE) is the predominant genetic risk modifier for Alzheimer disease, with 3 variants (ε4/ε3/ε2 alleles), yielding 6 different genotypes.1 Empirical findings by us2 and others3,4 indicate that APOE modifies response to dietary factors, which may be explained by evolutionary aspects.5,6 APOE ε4 (APOE4) originated 1 million to 6 million years ago and represents the ancestral human form; APOE3 is thought to have emerged approximately 200 000 years ago, while APOE2 arose more recently.5 The most common genotype, ε3/ε3 (APOE33), accounts for more than 50% of APOE genotypes in most ethnic populations.7 APOE44 confers the highest Alzheimer risk, with odds ratios compared with APOE33 varying by ancestry. They approach 30 in East Asian, 13 in White, 6 in Black, and 4 in Hispanic populations, while corresponding odds ratios for APOE34 are 4, 3, 2, and 2, respectively.8
The plant-to-animal food ratio is heterogenous in ancestral human diets, but a general shift toward increased meat consumption occurred 2.5 million years ago.9 A hypothesis by Ben-Dor et al10 states that the shift toward a lower plant-based proportion was not linear but J shaped (see our interpretation in Figure 1).5,10,11,12,13 Those authors propose that a hypercarnivorous period occurred a few million years ago,10 which may correspond to APOE4 emergence,2 followed by a return toward more plant-based diets in the last hundreds of thousands of years, coinciding with APOE3 emergence. Conflicting hypotheses suggest that APOE4 may provide more5 or less11 adaptation to higher meat consumption. The former view is more compatible with the J-shaped hypothesis and implies that APOE3 may confer increased metabolic flexibility, potentially enabling evolutionary readaptation to a more plant-based, omnivorous diet.12
Figure 1. Overview of Hypotheses on Dietary Habits During Human Evolution and APOE-Specific Dietary Adaptations.
In contrast to the Finch et al11 suggestion that the APOE3 genotype may be meat adaptive, Huebbe et al5 argue that APOE4 may provide adaptation to meat consumption, citing resistance to parasitic infections, reduced detoxification capacity of plant compounds, and preferential binding to larger lipoproteins rather than high-density lipoprotein. Plant specialists (consuming a >70% plant-based diet), generalists (30%-70%), and hypercarnivores (<30%) are defined according to Ben-Dor et al,10 2021. Figure adapted from Norgren,12 2023. Curves represent ordinal approximations based on interpretation of referenced sources. The chimpanzee apolipoprotein E (ApoE) protein is monomorphic and believed to function more like human ApoE3 and ApoE2 than ApoE4.5,13
In our previous work among older adults at risk for dementia, a higher-carbohydrate-fiber-lower-fat-protein composite score (proxy for the plant-to-animal ratio) was negatively associated with global cognition among individuals with APOE34/44 genotypes. A lack of an association for APOE33 supported the notion of metabolic flexibility, whereas APOE2 carriers exhibited positive trends, suggesting opposing adaptation relative to APOE4. Building on these findings, we outlined our hypothesis on APOE-specific dietary adaptation, referencing the Bradford Hill viewpoints on decision-making from observational data.12,14 One supporting perspective is the lower prevalence of APOE4, suggesting adverse selection, in agricultural regions.15 In Europe, APOE4 allele frequency decreases gradually from up to 27% in the north to as low as 4% in the south.16
Reviews of studies on meat consumption in association with cognitive health outcomes indicate that findings and methodologies are inconsistent.17,18,19 Some studies reported APOE interaction analyses, although outcomes were not statistically significant.20,21 Our aim for this study was to estimate the effect of meat consumption on global cognition and dementia incidence among individuals with genotypes ε3/ε4 and ε4/ε4 (APOE34/44) compared with individuals with APOE22/23/24/33 (non-APOE34/44) genotypes in a population-based cohort of older adults. As stated prior to data access,12 we hypothesized that higher meat consumption would be associated with distinct benefits among individuals with APOE34/44 genotypes.
Methods
Study Population
The Swedish National Study on Aging and Care–Kungsholmen (SNAC-K) is an ongoing, longitudinal, population-based study targeting individuals aged 60 years or older in an urban area of Stockholm.22 In the first wave (2001-2004), 3363 of 5111 randomly selected individuals were enrolled; 2157 participants met the inclusion criteria for this analysis: baseline data on diet, cognition, and APOE status and no dementia at baseline (see flowchart in eFigure 1 in Supplement 1).
This cohort study was approved by the Swedish Ethical Review Authority. Written informed consent was obtained from all participants or proxies for those cognitively impaired. The study adhered to the Declaration of Helsinki and the Strengthening the Reporting of Observational Studies in Epidemiology—Nutritional Epidemiology (STROBE-Nut) reporting guideline.
Study Design
We used panel data with repeated exposure and outcome measures over a period of up to 15 years. Participants were assessed every 6 years until age 78 years and every 3 years thereafter. Cognitive performance and dementia were evaluated at each visit; dietary assessments were conducted at baseline and at 3- and 6-year follow-ups.
A triangulation approach was used to examine cognition, integrating longitudinal between-participant outcomes (years 0-15) as the primary analysis, sensitivity analyses on within-participant outcomes (years 0-6), and a cross-sectional baseline analysis to account for individuals lost before follow-up. Primary analyses were conceptualized as a parallel-group target trial,23 with mean dietary intake across follow-ups representing long-term exposure; within-participant analyses were framed as crossover trials, as detailed in a previous study24 and eMethods 1 in Supplement 1. Time-to-event analyses estimated the association of baseline diet (to preserve temporality) with incident dementia.
Exposure Variables
Dietary intake was assessed using validated, semiquantitative, 98-item food-frequency questionnaires capturing diet over the previous year.25 The primary exposure was total meat consumption (grams per total kilocalories, as explained in eFigure 2 in Supplement 1). Secondary exposures were the processed-to-total meat ratio and the log ratio of unprocessed red meat to poultry (correlation matrix for exposures shown in eTable 1 in Supplement 1). Processed meat was defined as meat transformed through salting, curing, fermentation, smoking, or other processes. Cognitive trajectories were analyzed from mean consumption levels across follow-ups, with sensitivity analyses using baseline and last measurement, respectively. Dementia analyses used baseline values. Data on dietary supplements were not available.
Outcome Variables
Global cognition was calculated as the mean z score across 4 domains: episodic memory (free recall and recognition), semantic memory (vocabulary), verbal fluency (animals and professions), and perceptual speed (digit cancellation and pattern comparison).26 Scores were standardized to baseline values of the study sample before creation of a composite score of global cognition for all individuals with data in at least 2 cognitive domains (2157 participants). For primary analyses, a longitudinal cognitive trajectory (change in z score per year) was precalculated for each participant and domain separately using linear regression, with linear time as the factor estimating log-transformed cognitive outcomes. Baseline data from all domains and at least 1 follow-up in each domain were specified as inclusion criteria (1680 participants).
Dementia was diagnosed per Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) (DSM-IV) criteria, based on medical and drug history, general and neurological exams, and cognitive tests (eg, Mini-Mental State Examination [MMSE] and clock drawing). Daily living independence was also assessed. Diagnoses followed a 3-step process: initial physician assessment, independent review of available information by a second physician, and adjudication by senior neurologists (including G.G.) if discrepant. For participants who died between follow-ups, clinical records and death certificates were reviewed.27
Covariates
Covariate selection followed recommendations for causal inference and used baseline values for adjustment assuming negligible exposure feedback mechanisms.23,28 For diet covariates, the mean across visits was used, when applicable, to align with the factor of interest. The model included age, sex, education, APOE status, living arrangements, lifelong occupation type, physical activity level, current smoking status, alcohol intake, total energy intake, and Alternative Healthy Eating Index (AHEI) score,29 excluding meat items, as a marker for adherence to dietary guidelines. Morbidity30 (<2, 2-5, or >5 chronic diseases) and baseline cognition (except on dementia analyses, where this was considered to be a mediating pathway) were also included. The composite morbidity variable was used for simplicity given that separate adjustment for multiple diagnoses (hypertension, dyslipidemia, diabetes, cardiovascular or cerebrovascular disease, obesity, anemia, and depression) yielded similar results. Because no covariate beyond age substantially modified estimates, reporting is restricted to our primary model. A sensitivity analysis explored 6 macronutrient parameters as potential mediators. For missing covariates (<1% for all), the mean or mode (for categorical variables) was imputed.
Effect Modifiers
APOE was determined by standard methods26 and primarily dichotomized as APOE34/44 vs non-APOE34/44 genotypes, consistent with our prior work.2 Ancillary stratifications explored a hypothesized interaction gradient (APOE22-23-24-33-34-44).12
Statistical Analysis
Between-participant analyses on cognition used linear regression, with quintile-based analyses in primary analyses to assess potential nonlinearity. A fixed-effects model was used for within-participant analyses.31 Robust standard errors were estimated. Dementia analyses used Fine and Gray models, treating nondementia death as a competing risk. Proportional subdistribution hazard assumptions were confirmed graphically using cumulative incidence curves. Dementia onset was defined as the midpoint between the last known dementia-free day and the day of diagnosis. Time zero was set at baseline, and exit was defined as the earliest occurrence of dementia onset, death, or, for survivors without dementia, the day after the last follow-up. Exit was truncated to 15.5 years from baseline for 37 participants, who were without dementia, who had their last follow-up beyond that time to mitigate potential selection bias. Given that attrition was not associated with APOE or meat consumption (Table; eFigure 3 in Supplement 1), meaningful bias from attrition was considered unlikely.
Table. Baseline Participant Characteristics by Meat Consumption Quintile and APOE Genotype.
| Characteristica | Participants, No. (%)b | |||||
|---|---|---|---|---|---|---|
| Q1 | Q2 | Q3 | Q4 | Q5 | Full sample | |
| Participants, No. | ||||||
| APOE34/44 | 108 | 131 | 102 | 100 | 128 | 2157 |
| Other genotype | 324 | 300 | 330 | 331 | 303 | |
| Age, mean (SD), y | ||||||
| APOE34/44 | 71.8 (9.3) | 72.1 (8.7) | 71.7 (8.9) | 68.9 (8.7) | 67.3 (7.6) | 71.2 (9.2) |
| Other genotype | 74.4 (9.7) | 73.6 (9.7) | 71.4 (8.9) | 69.8 (8.9) | 68.4 (8.3) | |
| Sex (females/males) | ||||||
| APOE34/44 | 77/31 (71.3/28.7) | 84/47 (64.1/35.9) | 58/44 (56.9/43.1) | 55/45 (55.0/45.0) | 80/48 (62.5/37.5) | 1337/820 (62.0/38.0) |
| Other genotype | 224/100 (69.1/30.9) | 185/115 (61.7/38.3) | 201/129 (60.9/39.1) | 204/127 (61.6/38.4) | 169/134 (55.8/44.2) | |
| Time to last follow-up visit, mean (SD), y | ||||||
| Overall | ||||||
| APOE34/44 | 8.6 (4.9) | 8.6 (4.9) | 8.0 (5.2) | 8.5 (4.7) | 9.3 (5.1) | 8.8 (5.1) |
| Other genotype | 8.5 (5.4) | 8.9 (5.2) | 9.0 (5.3) | 9.4 (4.9) | 8.7 (5.0) | |
| In participants included in cognition analyses | ||||||
| No. with data | 323 | 329 | 344 | 346 | 338 | 1680 |
| APOE34/44 | 10.4 (3.6) | 10.5 (3.6) | 10.4 (3.5) | 10.1 (3.3) | 11.3 (3.3) | 10.8 (3.6) |
| Other genotype | 10.8 (3.9) | 10.8 (3.8) | 11.0 (3.7) | 11.0 (3.4) | 10.5 (3.5) | |
| Global cognition, mean (SD), z score | ||||||
| APOE34/44 | –0.06 (1.03) | –0.07 (0.87) | –0.08 (0.93) | 0.10 (1.06) | 0.20 (0.98) | 0 (1) |
| Other genotype | –0.12 (1.08) | –0.17 (1.01) | 0.10 (0.94) | 0.01 (0.96) | 0.13 (1.00) | |
| Education, mean (SD), y | ||||||
| APOE34/44 | 12.7 (4.3) | 12.2 (3.9) | 12.7 (5.1) | 12.5 (4.1) | 12.9 (4.3) | 12.4 (4.2) |
| Other genotype | 12.6 (4.4) | 11.7 (3.9) | 12.3 (4.2) | 12.2 (3.9) | 12.6 (4.0) | |
| Living arrangements (not alone) | ||||||
| No. with data | 432 | 430 | 431 | 429 | 428 | 2150 |
| APOE34/44 | 186 (40.7) | 151 (46.6) | 163 (53.9) | 156 (65.7) | 128 (46.8) | 1066 (49.6) |
| Other genotype | 138 (42.6) | 148 (49.5) | 166 (50.5) | 174 (52.7) | 174 (57.6) | |
| Occupation (manual) | ||||||
| No. with data | 432 | 430 | 431 | 429 | 428 | 2150 |
| APOE34/44 | 19 (17.6) | 30 (22.9) | 15 (14.7) | 15 (15.2) | 23 (18.3) | 396 (18.4) |
| Other genotype | 69 (21.3) | 60 (20.1) | 63 (19.1) | 53 (16.1) | 49 (16.2) | |
| Physical activityc | ||||||
| Low | ||||||
| APOE34/44 | 18 (16.7) | 25 (19.1) | 19 (18.6) | 21 (21.0) | 28 (21.9) | 458 (21.2) |
| Other genotype | 70 (21.6) | 70 (23.3) | 62 (18.8) | 69 (20.8) | 76 (25.1) | |
| Mid | ||||||
| APOE34/44 | 66 (61.1) | 69 (52.7) | 51 (50.0) | 58 (58.0) | 67 (52.3) | 1141 (52.9) |
| Other genotype | 176 (54.3) | 148 (49.3) | 184 (55.8) | 173 (52.3) | 149 (49.2) | |
| High | ||||||
| APOE34/44 | 24 (22.2) | 37 (28.2) | 32 (31.4) | 21 (21.0) | 33 (25.8) | 558 (25.9) |
| Other genotype | 78 (24.1) | 82 (27.3) | 84 (25.5) | 89 (26.9) | 78 (25.7) | |
| Tobacco smoker (current) | ||||||
| No. with data | 429 | 429 | 430 | 430 | 426 | 2144 |
| APOE34/44 | 12 (11.1) | 16 (12.4) | 10 (9.9) | 14 (14.1) | 20 (15.9) | 279 (13.0) |
| Other genotype | 36 (11.2) | 30 (10.0) | 35 (10.6) | 47 (14.2) | 59 (19.7) | |
| BMI, mean (SD) | ||||||
| APOE34/44 | 24.9 (3.6) | 25.7 (3.5) | 25.7 (3.4) | 26.5 (4.3) | 26.6 (4.0) | 26.0 (4.0) |
| Other genotype | 25.2 (3.9) | 25.9 (4.2) | 25.7 (3.6) | 26.2 (4.3) | 26.9 (3.9) | |
| Systolic blood pressure, mean (SD), mm Hg | ||||||
| No. with data | 430 | 430 | 432 | 430 | 431 | 2153 |
| APOE34/44 | 143 (21) | 146 (19) | 143 (22) | 141 (17) | 144 (16) | 144 (19) |
| Other genotype | 145 (19) | 144 (19) | 144 (21) | 145 (19) | 144 (19) | |
| HbA1c level, mean (SD), % | ||||||
| No. with data | 427 | 430 | 425 | 423 | 421 | 2126 |
| APOE34/44 | 4.53 (0.43) | 4.53 (0.43) | 4.67 (0.86) | 4.66 (0.80) | 4.46 (0.47) | 4.57 (0.68) |
| Other genotype | 4.54 (0.57) | 4.56 (0.71) | 4.55 (0.75) | 4.61 (0.73) | 4.65 (0.75) | |
| Total cholesterol, mean (SD), mg/dL | ||||||
| No. with data | 426 | 429 | 426 | 423 | 420 | 2124 |
| APOE34/44 | 246 (43) | 243 (40) | 233 (40) | 241 (40) | 236 (41) | 234 (43) |
| Other genotype | 230 (43) | 233 (44) | 231 (43) | 231 (42) | 233 (44) | |
| Chronic diseases, mean (SD), No. | ||||||
| APOE34/44 | 3.8 (2.1) | 3.6 (2.0) | 3.4 (2.1) | 3.4 (2.1) | 3.4 (2.1) | 3.6 (2.2) |
| Other genotype | 3.7 (2.2) | 3.8 (2.3) | 3.5 (2.2) | 3.5 (2.4) | 3.5 (2.3) | |
| Diabetes | ||||||
| APOE34/44 | 5 (4.6) | 5 (3.8) | 10 (9.8) | 10 (10.0) | 7 (5.5) | 166 (7.7) |
| Other genotype | 15 (4.6) | 21 (7.0) | 24 (7.3) | 32 (9.7) | 37 (12.2) | |
| Total energy intake, mean (SD), kcal/d | ||||||
| APOE34/44 | 2028 (660) | 1972 (595) | 2037 (659) | 1993 (622) | 1783 (576) | 1970 (651) |
| Other genotype | 2022 (685) | 1984 (611) | 2029 (701) | 1973 (637) | 1864 (653) | |
| Carbohydrates, mean (SD), E% | ||||||
| APOE34/44 | 46.6 (6.4) | 45.6 (6.2) | 43.6 (6.3) | 43.3 (6.5) | 42.1 (5.8) | 44.2 (6.8) |
| Other genotype | 47.3 (7.7) | 44.8 (6.8) | 44.3 (6.3) | 43.7 (6.0) | 41.0 (6.3) | |
| Fat, mean (SD), E% | ||||||
| APOE34/44 | 33.3 (6.7) | 34.3 (7.0) | 35.2 (6.9) | 34.8 (7.1) | 34.8 (6.2) | 34.6 (6.9) |
| Other genotype | 33.2 (8.0) | 35.2 (7.1) | 34.4 (6.6) | 34.6 (6.3) | 35.8 (6.6) | |
| Protein, mean (SD), E% | ||||||
| APOE34/44 | 13.3 (2.5) | 13.5 (2.1) | 13.9 (2.1) | 14.4 (2.3) | 16.0 (2.4) | 14.2 (2.4) |
| Other genotype | 13.0 (2.3) | 13.3 (2.1) | 14.0 (2.1) | 14.6 (2.1) | 15.9 (2.3) | |
| Fiber, mean (SD), g | ||||||
| APOE34/44 | 27.2 (10.4) | 25.9 (10.1) | 25.2 (9.6) | 23.8 (8.9) | 22.5 (9.2) | 25.2 (10.3) |
| Other genotype | 28.0 (12.5) | 24.7 (9.8) | 26.2 (10.6) | 24.8 (9.1) | 22.4 (9.5) | |
| Alcohol, mean (SD), E% | ||||||
| APOE34/44 | 4.2 (4.3) | 4.1 (4.0) | 5.0 (5.1) | 5.2 (4.7) | 4.7 (4.4) | 4.5 (4.4) |
| Other genotype | 3.8 (5.0) | 4.3 (4.4) | 4.7 (3.9) | 4.7 (4.0) | 5.0 (4.4) | |
| SFA-to-PUFA ratio, mean (SD), log, z | ||||||
| APOE34/44 | 0.00 (1.24) | 0.03 (1.00) | –0.02 (0.90) | 0.00 (0.87) | –0.22 (0.83) | 0 (1) |
| Other genotype | 0.24 (1.25) | 0.05 (1.02) | –0.06 (0.94) | –0.05 (0.89) | –0.10 (0.83) | |
| Meat consumption, mean (SD), g/14 000 kcald | ||||||
| Total | ||||||
| APOE34/44 | 215 (97) | 398 (29) | 509 (32) | 642 (48) | 934 (214) | 539 (264) |
| Other genotype | 221 (98) | 395 (33) | 511 (33) | 642 (46) | 928 (220) | |
| Red meat, unprocessed | ||||||
| APOE34/44 | 114 (70) | 199 (63) | 255 (71) | 310 (87) | 490 (216) | 273 (165) |
| Other genotype | 109 (71) | 203 (63) | 260 (73) | 326 (96) | 468 (189) | |
| Poultry, unprocessed | ||||||
| APOE34/44 | 52 (47) | 74 (50) | 81 (51) | 115 (77) | 163 (124) | 98 (96) |
| Other genotype | 54 (47) | 72 (46) | 92 (58) | 109 (82) | 168 (170) | |
| Processed meat | ||||||
| APOE34/44 | 50 (51) | 124 (61) | 174 (81) | 217 (98) | 280 (170) | 168 (134) |
| Other genotype | 57 (56) | 121 (71) | 159 (84) | 207 (102) | 293 (192) | |
| Processed-to-total meat ratio, mean (SD), % | ||||||
| No. with data | 429 | 431 | 432 | 431 | 431 | 2154 |
| APOE34/44 | 21.0 (20.2) | 31.5 (15.6) | 34.0 (15.6) | 33.9 (15.1) | 30.0 (15.9) | 30 (18) |
| Other genotype | 23.9 (22.5) | 30.5 (17.5) | 31.1 (16.1) | 32.2 (15.6) | 31.6 (18.2) | |
| AHEI score (maximum = 110), mean (SD) | ||||||
| APOE34/44 | 66 (10) | 63 (10) | 61 (9) | 59 (11) | 61 (10) | 62 (10) |
| Other genotype | 64 (10) | 62 (9) | 63 (9) | 60 (9) | 58 (9) | |
Abbreviations: AHEI, Alternative Healthy Eating Index; BMI (body mass index; calculated as weight in kilograms divided by height in meters squared); SFA/PUFA, saturated/polyunsaturated fatty acids (log ratio).
SI conversion factors: To convert cholesterol to millimoles per liter, multiply by 0.0259; HbA1c to proportion of total hemoglobin, multiply by 0.01;
There were missing data for 16 participants for BMI, 4 participants for systolic blood pressure, 31 participants for HbA1c, 33 participants for total cholesterol, 7 participants for living arrangements, 7 participants for occupation, 13 participants for tobacco smoking, and 3 participants for the ratio of processed to total meat (due to 0 total meat consumption).
Sample size per quintile is indicated for APOE34/44 and other genotypes (APOE22/23/24/33).
Physical activity was categorized as low (light or moderate-to-intense exercise ≤2 to 3 times/month), high (moderate-to-intense exercise several times/week), or mid (all other levels).
Meat consumption is expressed as grams per 14 000 kcal (equivalent to weekly consumption for a 2000 kcal/d diet).
To address competing risk and health in broader terms, some post hoc analyses were conducted. All-cause mortality was examined by Cox regression (eMethods 2 in Supplement 1), and associations between meat consumption and some biomarkers were studied by linear regression. Replacement analyses for individual food groups applied log-ratio transformations.32
A 2-sided significance level of 5% was used for primary analyses, which were prespecified; therefore, no adjustment for multiple comparisons was performed. Ancillary analyses were exploratory and not intended for formal hypothesis testing; such reporting is primarily graphical. Analyses were performed in January 2025 to January 2026 using Stata statistical software version 18 (StataCorp).
Results
Description of Participants
Among 2157 participants (mean [SD] age 71.2 [9.2] years; 1337 female [62.0%]), the prevalence of APOE34/44 was 569 participants (26.4%). Baseline characteristics across quintiles of meat consumption by APOE34/44 genotype are presented in the Table and eFigures 4 to 5 in Supplement 1. Status at censoring, showing 296 patients with dementia and 924 deaths, is described in eTable 2 in Supplement 1. A total of 690 patients died without dementia. APOE distribution and other characteristics are detailed in eTable 3 in Supplement 1, including a separate column for the subsample of 1680 participants with cognitive trajectories, showing minimal differences from the full sample.
At baseline, 88 participants scored less than 27 on the MMSE and were excluded in sensitivity analyses, alongside participants with dementia within 3 years. Of participants with an MMSE score less than 27, a total of 87 scored in the range of 23 to 26, while 1 participant scored 20 and 22 participants had an APOE34/44 genotype. A total of 13 food records were collected after dementia diagnosis at follow-up. Excluding these records had a negligible effect on cognitive results, and given that dementia analyses used baseline diet, they were not affected.
Cognition Analyses
Linear analyses for total meat consumption and cognitive trajectories are shown in Figure 2A, with meat types analyzed in Figure 2B. Higher meat consumption was associated with favorable trajectories of global cognition and episodic memory for APOE34/44 but not for non-APOE34/44 genotypes, with interactions. Global cognitive change per 10 years by total meat quintile is illustrated in Figure 3A.33 For APOE34/44 in the top quintile (Q5), the trajectory was better than Q1 (β = 0.32; 95% CI, 0.07 to 0.56; P = .01) and similar to non-APOE34/44 genotypes regardless of quintile. In exploratory analyses, Q5 was also better compared with Q2 (P = .04), Q3 (P = .004), and Q4 (P = .03). The largest magnitude was observed for episodic memory (Figure 3B). No association was found with cognitive trajectory in participants with non-APOE34/44 genotypes (β = –0.11; 95% CI, –0.27 to 0.06; P = .20) for Q5 vs Q1. APOE interaction (P for interaction for cognitive decline = .004) was robust across sensitivity analyses and more pronounced among females, participants aged 72 years and younger, and individuals with higher hemoglobin A1c (HbA1c) levels, lower AHEI scores, and Cardiovascular Risk Factors, Aging, and Dementia (CAIDE) risk scores34 of 6 or greater (eFigure 6 in Supplement 1). Complete APOE stratification supported our primary dichotomization (eFigure 7 in Supplement 1). Mediation analyses on macronutrient parameters showed minimal changes (eFigure 8 in Supplement 1). Triangulation approaches (within-participant and cross-sectional baseline analyses) yielded results in the same direction as the primary analysis (eFigure 9 in Supplement 1). Comparing individuals who did vs did not develop dementia, APOE interactions remained similar, and dietary changes did not differ, collectively suggesting that reverse causation was unlikely (eFigure 10 in Supplement 1).
Figure 2. Forest Plots of Linear Associations of Meat Consumption With Cognitive Trajectories.
A, Estimates per composite cognitive outcome and cognitive subdomain as the outcome. B, Estimates by different meat variable as the exposure. Both analyses use linear regression between meat exposure (in grams per kilocalorie, z transformed) and cognitive trajectories (change in z score/y multiplied by 10) adjusted for age, sex, education, APOE status, living arrangements, occupation type, physical activity, smoking, alcohol intake, total energy intake, Alternative Healthy Eating Index score (calculated without meat items), number of chronic diseases, and baseline cognition. One SD equals the following consumption levels (standardized for 2000 kcal/d intake): 264 g/week for total meat consumption, 165 g/week for unprocessed red meat, 96 g/week for poultry, and 134 g/week for processed meat. P values are given for the interaction between meat variables and APOE status.
Figure 3. Line Graphs of Quintile-Based Associations of Meat Consumption With Cognitive Health Outcomes.
The figure displays 3 cognitive outcomes: global cognition (A), episodic memory (B), and dementia incidence (C), analyzed in the same subsample (1680 participants with ≥1 cognitive follow-up) by quintile (Q) of total meat consumption. Q assignment is based on weight per total energy intake. Linear regression (A and B) was adjusted for age, sex, education, APOE status, living arrangements, occupation type, physical activity level, smoking status, alcohol intake, total energy intake, Alternative Healthy Eating Index (AHEI; calculated without meat items when used as a covariate) score, baseline cognition, and number of chronic diseases. Similar adjustments, except for baseline cognition, were applied to subdistribution hazard ratios (sHRs) in panel C, whereas plotted values are crude to enhance scaling. D, Distributions of other dietary factors by meat quintile are illustrated among all 2157 participants. No factors changed results substantially when added as covariates (eFigure 8 in Supplement 1). As a reference, the shown consumption levels in Q3 to Q5 clearly exceed the Nordic Nutrition Recommendations,33 2023. P values are given for the interaction between exposure (Q5 vs Q1) and APOE genotype. E% indicates energy percentage; SFA/PUFA, saturated/polyunsaturated fatty acids.
A higher ratio of processed to total meat was associated with a worse cognitive trajectory in APOE34/44 only (β = –0.12; 95% CI, –0.23 to –0.01; P = .04), although without APOE interaction. The ratio of red meat to poultry was not associated with cognitive trajectory (Figure 2B).
Dementia Analyses
Subdistribution hazard ratios (sHRs) for dementia incidence aligned with findings on cognitive trajectories. For Q5 vs Q1 of total meat consumption among 2157 participant, sHRs were as follows: 0.72 (95% CI, 0.46-1.11; P = .14) in the full sample, 0.45 (95% CI, 0.21-0.95; P = .04) for APOE34/44 genotypes, and 0.95 (95% CI, 0.57-1.61; P = .86) for non-APOE34/44 genotypes (P for interaction = .10). For APOE34/44 compared with non-APOE34/44 genotypes, the sHR was 2.49 (95% CI, 1.47-4.22; P = .001) in Q1, decreasing to 1.17 (95% CI, 0.55-2.45; P = .69) in Q5 of total meat. Ancillary quintile analyses are reported in eTable 4 in Supplement 1, with a subsample shown in Figure 3C and cumulative incidence curves in eFigure 11 in Supplement 1. Linear estimates are shown in Figure 4A; these were more robust after excluding individuals with possible cognitive impairment (eFigure 12 in Supplement 1). Quintile-based analyses indicated that for APOE34/44, Q5 differed from Q1 through Q4, whereas a slight elevation observed in Q3 did not represent a statistically significant difference vs Q1, Q2, or Q4 (eTable 4 in Supplement 1).
Figure 4. Forest Plots of Dementia and All-Cause Mortality Analyses.
A, Dementia incidence was analyzed using Fine and Gray models, with nondementia death as a competing risk. One SD equals the following consumption levels (standardized for 2000 kcal/d intake): 264 g/week for total meat, 165 g/week for unprocessed red meat, 96 g/week for poultry, 134 g/week for processed meat. B, A post hoc analysis on all-cause mortality was conducted to guide interpretations of cognitive outcomes using Cox proportional hazards regression, excluding deaths occurring within the first year after baseline. Unprocessed red meat and poultry were grouped as unprocessed meat after concluding that the ratio between those meat types was not associated with the outcome (hazard ratio [HR], 1.01; 95% CI, 0.94-1.08; P = .87; P for APOE interaction = .89). Both analyses accounted for age, sex, education, APOE status, living arrangements, occupation type, physical activity level, smoking status, alcohol intake, total energy intake, Alternative Healthy Eating Index score (calculated without meat items), and number of chronic diseases. P values are given for the interaction between meat variables and APOE genotype. sHR indicates subdistribution hazard ratio.
A higher ratio of processed to total meat was associated with increased dementia risk (sHR, 1.14; 95% CI, 1.01-1.29; P = .04). This was not modified by APOE status (Figure 4A).
Post hoc Analyses
Findings appeared more robust among females. However, no interaction between meat consumption and sex, regardless of APOE genotype, was observed for global cognition (eFigure 6 in Supplement 1) or dementia (eTable 4 in Supplement 1).
Cox regression analyses on all-cause mortality revealed APOE interactions consistent with cognitive findings. Specifically, higher unprocessed meat consumption at baseline was associated with reduced mortality among participants with APOE34/44 genotypes (HR, 0.85; 95% CI, 0.73-0.99; P = .04; P for interaction = .03), with a trend in the opposite direction for non-APOE34/44 genotypes (Figure 4B; eFigure 13 in Supplement 1). Similar trends of APOE interaction were found for outcomes of cholesterol and HbA1c levels, but not for body mass index or blood pressure levels (eFigure 13 in Supplement 1).
We used the slope between dietary and circulating levels of vitamin B12 as a possible proxy for nutrient absorption to explore a mechanistic hypothesis. A 3-way interaction for dietary vitamin B12 × meat consumption × APOE status regressed on vitamin B12 in blood suggested that for participants with APOE33 genotypes, absorption did not vary across levels of meat consumption. However, for participants with APOE34/44 genotypes, absorption was greater with higher meat consumption, while the opposite pattern was observed for APOE2 carriers (eFigure 14 in Supplement 1).
Replacement analyses for individual food groups indicated more favorable cognitive associations in APOE34/44 genotypes when meat replaced cereals, dairy, oils, and legumes, with no associations for the replacement of eggs, fish, or tubers (eFigure 15 in Supplement 1). Total meat consumption was inversely associated with cereals, dairy, and fruits primarily (eFigure 16 in Supplement 1).
Discussion
In this population-based cohort study of older adults, higher total meat consumption was associated with favorable cognitive health outcomes for APOE34/44 genotypes. Findings were consistent for longitudinal analyses of global cognition, episodic memory, and dementia risk, with interactions by APOE status for global cognition and episodic memory. None of those estimates were statistically significant in non-APOE34/44 genotypes. A lower processed-to-total meat ratio was favorably associated with dementia, with no substantial difference between unprocessed red meat and poultry. When meat types were analyzed as fractions of the total diet, associations were observed for unprocessed red meat but not for processed meat; higher consumption of unprocessed red meat was associated with lower dementia risk regardless of APOE status. After exclusion of individuals with possible cognitive impairment, this association became more robust and extended to total meat. In post hoc analyses, mortality rates were lower with higher unprocessed meat consumption exclusively among participants with APOE34/44 genotypes, compatible with findings from a Chinese cohort study.35 While exposures that are associated with both dementia and death can yield paradoxical associations (eg, smoking may appear protective against dementia by reducing survival time),36 this explanation seems unlikely for APOE34/44 given the concurrent association with longer survival.
To our knowledge, this is the first study to demonstrate interactions between meat consumption and APOE status in the association with cognitive outcomes, in support of a prespecified hypothesis that higher meat consumption may be advantageous in APOE34/44 genotypes.12 However, our findings align with underappreciated patterns in 2 large cohorts. In the UK Biobank (493 888 participants), unprocessed red meat was inversely associated with dementia (P = .01), driven by APOE4 carriers (HR, 0.64 per 50-g/d increase; P < .001), with no associations in noncarriers (HR, 0.93; P = .59).21 In the Nurses’ Health Study (NHS) and Health Professionals Follow-up Study (133 771 participants), supplementary analyses revealed an APOE4 interaction (P < .001) for unprocessed red meat, showing favorable trends among carriers and adverse trends among noncarriers.37 These patterns were not emphasized in the original publications, possibly because APOE interaction was not the primary focus or because conventional significance thresholds were strictly applied, but they are consistent with our findings. For processed meat, which accounts for approximately one-third of total meat intake across the cohorts discussed and in our study (although measures are not directly comparable), prior studies have reported adverse associations with cognitive health outcomes,21,37 which we did not observe. However, a relative advantage of unprocessed over processed meat was shown for dementia.
Intriguingly, among APOE4 carriers in the NHS, women aged 70 years and older consuming 1 or more servings/d of unprocessed red meat compared with less than 0.50 servings/d had a cognitive advantage of the same magnitude as the disadvantage typically observed in APOE4 carriers vs noncarriers (approximately 3 years of cognitive aging).37 This parallels effect sizes observed in our cohort, where the expected excess risk among participants with APOE34/44 genotypes was absent in the highest quintile of meat consumption across global cognition, episodic memory (a hallmark feature of Alzheimer pathology38), and dementia outcomes. Given that these genotypes account for approximately 70% of Alzheimer dementia cases in Northern Europe and North America,39 the absolute number of potentially preventable cases is substantial.
Indirect evidence of a similar APOE interaction was implied by reports that the low-meat EAT-Lancet40 and Planetary Healthy diets41 were favorably associated with cognitive health outcomes only among non–APOE4 carriers. Notably, the 2025 EAT-Lancet 2.042 cites the NHS and UK Biobank studies discussed previously to support the statement that “red meat has been positively associated with…unhealthy ageing,” a conclusion that is contradicted by our interpretation.
Macronutrient parameters did not mediate our findings, prompting exploration of alternative explanations. Post hoc analyses using a proxy for vitamin B12 absorption suggested distinct APOE responses, with better nutrient uptake from meat than from other sources among participants with APOE34/44 genotypes. These findings raise the possibility that APOE-modified health associations may be influenced by the food matrix43 or by antinutritive factors in foods replacing meat (primarily cereals and dairy). Indeed, more favorable associations in APOE34/44 were observed when meat replaced relatively recent additions to the human diet, supporting an evolutionary perspective for future mechanistic research.
Contrary to the Finch et al,11 2004, hypothesis suggesting APOE3 as the “meat adaptive” allele, our findings suggest that this characteristic fits APOE4. However, their hypothesis was based on the assumption that “all direct human ancestors are believed to have been largely herbivorous.”11 This conflicts with the hypothesis by Ben-Dor et al,10 which implies that a temporary hypercarnivorous period may have coincided with the emergence of APOE4 in human evolution12 (Figure 1). While Finch et al11 associated high meat consumption with elevated cholesterol levels and chronic disease risk, our data showed that for participants with APOE34/44 genotypes, higher meat consumption aligned with lower blood cholesterol levels and a lower ratio of dietary saturated to polyunsaturated fat, a key dietary determinant of cholesterol levels.44 This is plausible given that unprocessed meat may contain more unsaturated than saturated fat.45
Strengths and Limitations
Our study’s strengths include a triangulation approach to cognition, showing consistency across between- and within-participants analyses, along with progression to dementia. We adjusted extensively for potential confounders and explored possible reversed causality in sensitivity analyses; however, residual bias cannot be ruled out. Furthermore, we validated our primary APOE dichotomization and provided supporting evidence for separating ε24 from ε34/44, in line with previous work,2 although results were similar to those of a conventional stratification by APOE4 carriers. Our study also includes several limitations. One limitation could be potential survival bias; that is, our selection may include particularly resilient individuals. Ethnic ancestry, although not explicitly measured, was homogeneous and predominantly Northern European, and this may limit generalizability. Self-reported dietary data may include errors, but we find that unlikely to have induced bias for APOE interactions.
Conclusions
In this cohort study, we found that the APOE34/44 group exhibited the anticipated excess risk of cognitive decline and dementia progression compared with participants with other genotypes when consuming meat at levels consistent with current dietary guideline targets. However, this disadvantageous association was absent at higher consumption levels, equivalent to more than twice the target.33 Viewed alongside reinterpreted evidence from NHS37 and UK Biobank21 focusing on unprocessed meat, these findings point to a consistent gene-diet interaction, with important implications for public health. Results reinforce the urgency of investing in precision nutrition research with a focus on APOE, which could ultimately inform future policy development.
eFigure 1. Flow-chart of recruitment and selection of participants
eFigure 2. Comparison of methods for measuring meat consumption
eFigure 3. Attrition by level of meat consumption
eFigure 4: Baseline characteristics of participants by meat consumption
eFigure 5: Health status at baseline by total meat consumption and APOE genotype
eFigure 6: Comparison of time-point(s) of exposure measurement and stratifications on various sub-groups
eFigure 7: Validation of our primary dichotomization (APOE34/44 vs others)
eFigure 8: Analysis of potential mediation or confounding by various diet parameters.
eFigure 9. Fixed-effects model adjusted for time and cross-sectional baseline analyses
eFigure 10: Analysis of cognitive trajectories and dietary changes by dementia status and APOE
eFigure 11: Cumulative incidence curves for dementia by APOE and total meat consumption
eFigure 12: Sensitivity analysis excluding individuals with possible cognitive impairment
eFigure 13. Noncognitive health outcomes
eFigure 14. Sensitivity analyses exploring the role of vitamin B12 status
eFigure 15. Analyses by individual food groups
eFigure 16. Distribution of consumption from other food groups by quintiles of meat consumption
eTable 1. Correlations between diet variables
eTable 2. Participant status
eTable 3. Baseline characteristics of participants without and with cognitive trajectories
eTable 4. Association of meat consumption at baseline with dementia risk over 15 years
eMethods 1: Triangulation: within-participant associations and participants with baseline data only
eMethods 2: Associations between meat consumption and all-cause mortality
eReferences.
Data Sharing Statement
References
- 1.Belloy ME, Napolioni V, Greicius MD. A quarter century of APOE and Alzheimer’s disease: progress to date and the path forward. Neuron. 2019;101(5):820-838. doi: 10.1016/j.neuron.2019.01.056 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Norgren J, Sindi S, Matton A, Kivipelto M, Kåreholt I. APOE-genotype and insulin modulate estimated effect of dietary macronutrients on cognitive performance: panel analyses in nondiabetic older adults at risk of dementia. J Nutr. 2023;153(12):3506-3520. doi: 10.1016/j.tjnut.2023.09.016 [DOI] [PubMed] [Google Scholar]
- 3.Egert S, Rimbach G, Huebbe P. ApoE genotype: from geographic distribution to function and responsiveness to dietary factors. Proc Nutr Soc. 2012;71(3):410-424. doi: 10.1017/S0029665112000249 [DOI] [PubMed] [Google Scholar]
- 4.Farmer BC, Johnson LA, Hanson AJ. Effects of apolipoprotein E on nutritional metabolism in dementia. Curr Opin Lipidol. 2019;30(1):10-15. doi: 10.1097/MOL.0000000000000566 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Huebbe P, Rimbach G. Evolution of human apolipoprotein E (APOE) isoforms: gene structure, protein function and interaction with dietary factors. Ageing Res Rev. 2017;37:146-161. doi: 10.1016/j.arr.2017.06.002 [DOI] [PubMed] [Google Scholar]
- 6.Yassine HN, Finch CE. APOE alleles and diet in brain aging and Alzheimer’s disease. Front Aging Neurosci. 2020;12:150. doi: 10.3389/fnagi.2020.00150 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Khajouei E, Ghisays V, Piras IS, et al. Phenome-wide association of APOE alleles in the All of Us Research Program. EBioMedicine. 2025;117:105768. doi: 10.1016/j.ebiom.2025.105768 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Belloy ME, Andrews SJ, Le Guen Y, et al. APOE genotype and Alzheimer disease risk across age, sex, and population ancestry. JAMA Neurol. 2023;80(12):1284-1294. doi: 10.1001/jamaneurol.2023.3599 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Pontzer H, Wood BM. Effects of evolution, ecology, and economy on human diet: insights from hunter-gatherers and other small-scale societies. Annu Rev Nutr. 2021;41:363-385. doi: 10.1146/annurev-nutr-111120-105520 [DOI] [PubMed] [Google Scholar]
- 10.Ben-Dor M, Sirtoli R, Barkai R. The evolution of the human trophic level during the Pleistocene. Am J Phys Anthropol. 2021;175(suppl 72):27-56. doi: 10.1002/ajpa.24247 [DOI] [PubMed] [Google Scholar]
- 11.Finch CE, Stanford CB. Meat-adaptive genes and the evolution of slower aging in humans. Q Rev Biol. 2004;79(1):3-50. doi: 10.1086/381662 [DOI] [PubMed] [Google Scholar]
- 12.Norgren J. The Impact of Macronutritional Composition and Ketosis on Cognitive Health: From Normal Aging to Alzheimer’s Disease: Doctoral Thesis. Karolinska Institutet; 2023. Accessed February 19, 2026. https://openarchive.ki.se/articles/thesis/The_impact_of_macronutritional_composition_and_ketosis_on_cognitive_health_from_normal_aging_to_Alzheimer_s_disease/26916826?file=48961099 [Google Scholar]
- 13.Zannis VI, Nicolosi RJ, Jensen E, Breslow JL, Hayes KC. Plasma and hepatic apoE isoproteins of nonhuman primates: differences in apoE among humans, apes, and New and Old World monkeys. J Lipid Res. 1985;26(12):1421-1430. doi: 10.1016/S0022-2275(20)34247-4 [DOI] [PubMed] [Google Scholar]
- 14.Phillips CV, Goodman KJ. The missed lessons of Sir Austin Bradford Hill. Epidemiol Perspect Innov. 2004;1(1):3. doi: 10.1186/1742-5573-1-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Henderson ST. High carbohydrate diets and Alzheimer’s disease. Med Hypotheses. 2004;62(5):689-700. doi: 10.1016/j.mehy.2003.11.028 [DOI] [PubMed] [Google Scholar]
- 16.Singh PP, Singh M, Mastana SS. APOE distribution in world populations with new data from India and the UK. Ann Hum Biol. 2006;33(3):279-308. doi: 10.1080/03014460600594513 [DOI] [PubMed] [Google Scholar]
- 17.Paul TL, Fleming SA. Dietary consumption of beef and red meat: a scoping review and evidence map on cognitive outcomes across the lifespan. Public Health Nutr. 2023;26(12):2912-2926. doi: 10.1017/S1368980023001933 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Quan W, Xu Y, Luo J, et al. Association of dietary meat consumption habits with neurodegenerative cognitive impairment: an updated systematic review and dose-response meta-analysis of 24 prospective cohort studies. Food Funct. 2022;13(24):12590-12601. doi: 10.1039/D2FO03168J [DOI] [PubMed] [Google Scholar]
- 19.Zhang H, Hardie L, Bawajeeh AO, Cade J. Meat consumption, cognitive function and disorders: a systematic review with narrative synthesis and meta-analysis. Nutrients. 2020;12(5):1528. doi: 10.3390/nu12051528 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Ylilauri MPT, Hantunen S, Lönnroos E, Salonen JT, Tuomainen TP, Virtanen JK. Associations of dairy, meat, and fish intakes with risk of incident dementia and with cognitive performance: the Kuopio Ischaemic Heart Disease Risk Factor Study (KIHD). Eur J Nutr. 2022;61(5):2531-2542. doi: 10.1007/s00394-022-02834-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Zhang H, Greenwood DC, Risch HA, Bunce D, Hardie LJ, Cade JE. Meat consumption and risk of incident dementia: cohort study of 493,888 UK Biobank participants. Am J Clin Nutr. 2021;114(1):175-184. doi: 10.1093/ajcn/nqab028 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Lagergren M, Fratiglioni L, Hallberg IR, et al. A longitudinal study integrating population, care and social services data: the Swedish National study on Aging and Care (SNAC). Aging Clin Exp Res. 2004;16(2):158-168. doi: 10.1007/BF03324546 [DOI] [PubMed] [Google Scholar]
- 23.Matthews AA, Danaei G, Islam N, Kurth T. Target trial emulation: applying principles of randomised trials to observational studies. BMJ. 2022;378:e071108. doi: 10.1136/bmj-2022-071108 [DOI] [PubMed] [Google Scholar]
- 24.Norgren J, Sindi S, Sandebring-Matton A, Ngandu T, Kivipelto M, Kåreholt I. The dietary carbohydrate/fat-ratio and cognitive performance: panel analyses in older adults at risk for dementia. Curr Dev Nutr. 2023;7(6):100096. doi: 10.1016/j.cdnut.2023.100096 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Carballo-Casla A, Avesani CM, Beridze G, et al. Protein intake and mortality in older adults with chronic kidney disease. JAMA Netw Open. 2024;7(8):e2426577. doi: 10.1001/jamanetworkopen.2024.26577 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Laukka EJ, Lövdén M, Herlitz A, et al. Genetic effects on old-age cognitive functioning: a population-based study. Psychol Aging. 2013;28(1):262-274. doi: 10.1037/a0030829 [DOI] [PubMed] [Google Scholar]
- 27.Grande G, Valletta M, Rizzuto D, et al. Blood-based biomarkers of Alzheimer’s disease and incident dementia in the community. Nat Med. 2025;31(6):2027-2035. doi: 10.1038/s41591-025-03605-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Franks DW, Ruxton GD, Sherratt T. Ecology needs a causal overhaul. Biol Rev Camb Philos Soc. 2025;100(5):1950-1969. doi: 10.1111/brv.70029 [DOI] [PubMed] [Google Scholar]
- 29.Chiuve SE, Fung TT, Rimm EB, et al. Alternative dietary indices both strongly predict risk of chronic disease. J Nutr. 2012;142(6):1009-1018. doi: 10.3945/jn.111.157222 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Calderón-Larrañaga A, Vetrano DL, Onder G, et al. Assessing and measuring chronic multimorbidity in the older population: a proposal for its operationalization. J Gerontol A Biol Sci Med Sci. 2017;72(10):1417-1423. doi: 10.1093/gerona/glw233 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Gunasekara FI, Richardson K, Carter K, Blakely T. Fixed effects analysis of repeated measures data. Int J Epidemiol. 2014;43(1):264-269. doi: 10.1093/ije/dyt221 [DOI] [PubMed] [Google Scholar]
- 32.Corrêa Leite ML. Log-ratio transformations for dietary compositions: numerical and conceptual questions. J Nutr Sci. 2021;10:e97. doi: 10.1017/jns.2021.93 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Nordic Council of Ministers . Nordic nutrition recommendations 2023: integrating environmental aspects. 2023. Accessed February 19,2026. https://pub.norden.org/nord2023-003/nord2023-003.pdf
- 34.Kivipelto M, Ngandu T. From heart health to brain health: legacy of the North Karelia Project for Dementia Research. Glob Heart. 2016;11(2):235-242. doi: 10.1016/j.gheart.2016.04.013 [DOI] [PubMed] [Google Scholar]
- 35.Jin X, Xiong S, Yuan C, et al. Apolipoprotein E genotype, meat, fish, and egg intake in relation to mortality among older adults: a longitudinal analysis in China. Front Med (Lausanne). 2021;8:697389. doi: 10.3389/fmed.2021.697389 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Rojas-Saunero LP, Young JG, Didelez V, Ikram MA, Swanson SA. Considering questions before methods in dementia research with competing events and causal goals. Am J Epidemiol. 2023;192(8):1415-1423. doi: 10.1093/aje/kwad090 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Li Y, Li Y, Gu X, et al. Long-term intake of red meat in relation to dementia risk and cognitive function in US adults. Neurology. 2025;104(3):e210286. doi: 10.1212/WNL.0000000000210286 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Weintraub S, Wicklund AH, Salmon DP. The neuropsychological profile of Alzheimer disease. Cold Spring Harb Perspect Med. 2012;2(4):a006171. doi: 10.1101/cshperspect.a006171 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Mattsson N, Groot C, Jansen WJ, et al. Prevalence of the apolipoprotein E ε4 allele in amyloid β positive subjects across the spectrum of Alzheimer’s disease. Alzheimers Dement. 2018;14(7):913-924. doi: 10.1016/j.jalz.2018.02.009 [DOI] [PubMed] [Google Scholar]
- 40.Samuelsson J, Glans I, Stubbendorff A, et al. Associations between the EAT-Lancet planetary health diet and incident dementia. J Prev Alzheimers Dis. 2025;12(6):100166. doi: 10.1016/j.tjpad.2025.100166 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Zhang JJ, Ye YX, Dorajoo R, et al. APOE genotype modifies the association between midlife adherence to the planetary healthy diet and cognitive function in later life among Chinese adults in Singapore. J Nutr. 2024;154(1):252-260. doi: 10.1016/j.tjnut.2023.11.024 [DOI] [PubMed] [Google Scholar]
- 42.Rockström J, Thilsted SH, Willett WC, et al. The EAT-Lancet Commission on healthy, sustainable, and just food systems. Lancet. 2025;406(10512):1625-1700. doi: 10.1016/S0140-6736(25)01201-2 [DOI] [PubMed] [Google Scholar]
- 43.Aguilera JM. The food matrix: implications in processing, nutrition and health. Crit Rev Food Sci Nutr. 2019;59(22):3612-3629. doi: 10.1080/10408398.2018.1502743 [DOI] [PubMed] [Google Scholar]
- 44.Hu FB, Manson JE, Willett WC. Types of dietary fat and risk of coronary heart disease: a critical review. J Am Coll Nutr. 2001;20(1):5-19. doi: 10.1080/07315724.2001.10719008 [DOI] [PubMed] [Google Scholar]
- 45.Geiker NRW, Bertram HC, Mejborn H, et al. Meat and human health-current knowledge and research gaps. Foods. 2021;10(7):1556. doi: 10.3390/foods10071556 [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
eFigure 1. Flow-chart of recruitment and selection of participants
eFigure 2. Comparison of methods for measuring meat consumption
eFigure 3. Attrition by level of meat consumption
eFigure 4: Baseline characteristics of participants by meat consumption
eFigure 5: Health status at baseline by total meat consumption and APOE genotype
eFigure 6: Comparison of time-point(s) of exposure measurement and stratifications on various sub-groups
eFigure 7: Validation of our primary dichotomization (APOE34/44 vs others)
eFigure 8: Analysis of potential mediation or confounding by various diet parameters.
eFigure 9. Fixed-effects model adjusted for time and cross-sectional baseline analyses
eFigure 10: Analysis of cognitive trajectories and dietary changes by dementia status and APOE
eFigure 11: Cumulative incidence curves for dementia by APOE and total meat consumption
eFigure 12: Sensitivity analysis excluding individuals with possible cognitive impairment
eFigure 13. Noncognitive health outcomes
eFigure 14. Sensitivity analyses exploring the role of vitamin B12 status
eFigure 15. Analyses by individual food groups
eFigure 16. Distribution of consumption from other food groups by quintiles of meat consumption
eTable 1. Correlations between diet variables
eTable 2. Participant status
eTable 3. Baseline characteristics of participants without and with cognitive trajectories
eTable 4. Association of meat consumption at baseline with dementia risk over 15 years
eMethods 1: Triangulation: within-participant associations and participants with baseline data only
eMethods 2: Associations between meat consumption and all-cause mortality
eReferences.
Data Sharing Statement
