APOE-Genotype and Insulin Modulate Estimated Effect of Dietary Macronutrients on Cognitive Performance: Panel Analyses in Nondiabetic Older Adults at Risk of Dementia.

BACKGROUND: The apolipoprotein E gene (APOE ε-2/3/4, combined as 6 different genotypes: ε-22/23/24/33/34/44) and insulin status modulate dementia risk and play a role in the metabolism of macronutrients. OBJECTIVES: We aimed to examine APOE-genotype and fasting insulin as effect modifiers of the slopes between dietary macronutrients and cognitive performance among older adults at risk of dementia. METHODS: Panel analyses-with diet and cognition measured at baseline and follow-up at years 1 and 2-were performed in a sub-sample from the FINGER (Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability) trial (n = 676, 60-77 y, 46% females, all nondiabetics). The associations between macronutrients (3-d food records, z-scores) and global cognition (modified Neuropsychological Test Battery, z-score) were analyzed in mixed regression models adjusted for confounders selected a priori. After a gradient was implied by the point estimates in categorical APOE analyses, we investigated a continuous APOE variable [APOE-gradient, coded -1 (for ε-23), -0.5 (ε-24), 0 (ε-33), 1 (ε-34), 2 (ε-44)] as an effect-modifier. RESULTS: At increasing levels of the APOE-gradient, a relatively more favorable slope between diet and cognition was observed for a lower carbohydrate/fat ratio [ß = -0.040, 95% confidence interval (CI): -0.074, -0.006; P = 0.020 for interaction diet × APOE-gradient), and higher protein (ß = 0.075, 95% CI: 0.042, 0.109; P = 9.4 × 10-6). Insulin concentration (log-linear) modulated the association between the carbohydrate/fat ratio and cognition by a quadratic interaction (ß = -0.016, P = 0.039). Coherent findings for exploratory predictors (fiber, fat subtypes, composite score, metabolic biomarkers) were compatible with published hypotheses of differential dietary adaptation by APOE, with cognition among ε-33 being relatively independent of dietary parameters-implying "metabolic flexibility." Antagonistic slopes to cognition for ε-23 (positive) compared with ε-34 and ε-44 (negative) were found for a Higher-carbohydrates-fiber-Lower-fat-protein composite score, even as within-subjects effects. CONCLUSIONS: APOE-based precision nutrition appears conceptually promising, but replications in wider samples are warranted, as well as support from trials. Both relative hyper- and hypoinsulinemia might modulate the effect of diet on cognition.

The Dietary Carbohydrate/Fat-Ratio and Cognitive Performance: Panel Analyses in Older Adults at Risk for Dementia.

Background: Roughly 80% of total energy intake (TEI) in most human diets originates from digestible carbohydrates (eCarb) and fat (eFat), but the impact of their proportions on cognitive performance is poorly understood. Objectives: Our primary aim was to investigate estimates of global cognition in relation to macronutrient intake, with the log-ratio eCarb/eFat (CFr) as the primary predictor variable of interest. Secondary predictors were protein and the saturated/total fat ratio. Exploratory comparisons of CFr with eCarb and eFat as separate predictors were an additional aim. Methods: The observations were made on panel data (years 0, 1, 2) from the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability, n = 1251; age 60-77 y; 47% females; selected by risk factors for dementia. Self-reported diet was assessed by 3-d food records. Global cognition was measured using a modified Neuropsychological Test Battery. A mixed linear regression model was used, adjusted for age, sex, education, body-mass index, cholesterol-lowering drugs, TEI, time, time × intervention/control group, with study site and subject as random factors. Estimates were standardized (mean = 0; SD = 1) with 95% CI. Results: CFr had a negative estimate to global cognition (ß = -0.022, CI: -0.039, -0.005; P = 0.011). The point estimate for protein was ß = 0.013 (P = 0.41), and for the saturated/total fat ratio, associations with cognition were nonlinear. CFr correlated highly with eCarb (Pearson's r = 0.92) and eFat (r = -0.94). The point estimate for CFr fell between eCarb (ß = -0.026, P < 0.001) and (inversely) eFat (ß = 0.017, P = 0.090). Conclusions: A lower CFr was associated with better global cognition among older adults at risk for dementia. Because this is an important target group for preventive interventions, clinical trials are warranted to further investigate the impact of macronutritional composition on cognitive health. The potential role of CFr as a predictor for cognitive health should be further studied.

Serum proBDNF Is Associated With Changes in the Ketone Body ß-Hydroxybutyrate and Shows Superior Repeatability Over Mature BDNF: Secondary Outcomes From a Cross-Over Trial in Healthy Older Adults.

Background: ß-hydroxybutyrate (BHB) can upregulate brain-derived neurotrophic factor (BDNF) in mice, but little is known about the associations between BHB and BDNF in humans. The primary aim here was to investigate whether ketosis (i.e., raised BHB levels), induced by a ketogenic supplement, influences serum levels of mature BDNF (mBDNF) and its precursor proBDNF in healthy older adults. A secondary aim was to determine the intra-individual stability (repeatability) of those biomarkers, measured as intra-class correlation coefficients (ICC). Method: Three of the arms in a 6-arm randomized cross-over trial were used for the current sub-study. Fifteen healthy volunteers, 65-75 y, 53% women, were tested once a week. Test oils, mixed in coffee and cream, were ingested after a 12-h fast. Labeled by their level of ketosis, the arms provided: sunflower oil (lowK); coconut oil (midK); caprylic acid + coconut oil (highK). Repeated blood samples were collected for 4 h after ingestion. Serum BDNF levels were analyzed for changes from baseline to 1, 2 and 4 h to compare the arms. Individual associations between BHB and BDNF were analyzed cross-sectionally and for a delayed response (changes in BHB 0-2 h to changes in BDNF at 0-4 h). ICC estimates were calculated from baseline levels from the three study days. Results: proBDNF increased more in highK vs. lowK between 0 and 4 h (z-score: ß = 0.25, 95% CI 0.07-0.44; p = 0.007). Individual change in BHB 0-2 h, predicted change in proBDNF 0-4 h, (ß = 0.40, CI 0.12-0.67; p = 0.006). Change in mBDNF was lower in highK vs. lowK at 0-2 h (ß = -0.88, CI -1.37 to -0.40; p < 0.001) and cumulatively 0-4 h (ß = -1.01, CI -1.75 to -0.27; p = 0.01), but this could not be predicted by BHB levels. ICC was 0.96 (95% CI 0.92-0.99) for proBDNF, and 0.72 (CI 0.47-0.89) for mBDNF. Conclusions: The findings support a link between changes in peripheral BHB and proBDNF in healthy older adults. For mBDNF, changes differed between arms but independent to BHB levels. Replication is warranted due to the small sample. Excellent repeatability encourages future investigations on proBDNF as a predictor of brain health. Clinical Trial Registration:ClinicalTrials.gov, NCT03904433.

Ketosis After Intake of Coconut Oil and Caprylic Acid-With and Without Glucose: A Cross-Over Study in Healthy Older Adults.

Introduction: Medium-chain-triglycerides (MCT), formed by fatty acids with a length of 6-12 carbon atoms (C6-C12), constitute about two thirds of coconut oil (Coc). MCT have specific metabolic properties which has led them to be described as ketogenic even in the absence of carbohydrate restriction. This effect has mainly been demonstrated for caprylic acid (C8), which constitutes about 6-8% of coconut oil. Our aim was to quantify ketosis and blood glucose after intake of Coc and C8, with and without glucose intake. Sunflower oil (Suf) was used as control, expected to not break fasting ketosis, nor induce supply-driven ketosis. Method: In a 6-arm cross-over design, 15 healthy volunteers-age 65-73, 53% women-were tested once a week. After a 12-h fast, ketones were measured during 4 h after intake of coffee with cream, in combination with each of the intervention arms in a randomized order: 1. Suf (30 g); 2. C8 (20 g) + Suf (10 g); 3. C8 (20 g) + Suf (10 g) + Glucose (50 g); 4. Coc (30 g); 5. Coc (30 g) + Glucose (50 g); 6. C8 (20 g) + Coc (30 g). The primary outcome was absolute blood levels of the ketone ß-hydroxybutyrate, area under the curve (AUC). ANOVA for repeated measures was performed to compare arms. Results: ß-hydroxybutyrate, AUC/time (mean ± SD), for arms were 1: 0.18 ± 0.11; 2: 0.45 ± 0.19; 3: 0.28 ± 0.12; 4: 0.22 ± 0.12; 5: 0.08 ± 0.04; 6: 0.45 ± 0.20 (mmol/L). Differences were significant (all p ≤ 0.02), except for arm 2 vs. 6, and 4 vs. 1 & 3. Blood glucose was stable in arm 1, 2, 4, & 6, at levels slightly below baseline (p ≤ 0.05) at all timepoints hours 1-4 after intake. Conclusions: C8 had a higher ketogenic effect than the other components. Coc was not significantly different from Suf, or C8 with glucose. In addition, we report that a 16-h non-carbohydrate window contributed to a mild ketosis, while blood glucose remained stable. Our results suggest that time-restricted feeding regarding carbohydrates may optimize ketosis from intake of MCT. Clinical Trial Registration: The study was registered as a clinical trial on ClinicalTrials.gov, NCT03904433.

Capillary blood tests may overestimate ketosis: triangulation between three different measures of ß-hydroxybutyrate.

The ketone body ß-hydroxybutyrate (BHB), assessed by a point-of-care meter in venous whole blood (BHBv), was used as the main outcome in a study on nutritional ketosis in healthy older adults. Two other BHB measures were also used in the study for validation and exploratory purposes, and here we report findings on correlation and agreement between those three methods. Ketosis in the range of 0-1.5 mmol/L was induced in 15 healthy volunteers by intake of medium-chain fatty acids after a 12-h fast. BHBv was assessed at 12 time points for 4 h. The same point-of-care meter was also used to test capillary blood (BHBc) at three time points, and a laboratory test determined total ketones (TK) in plasma (BHBp + acetoacetate) at four time points. A total of 180 cases included simultaneous data on BHBv, BHBc, BHBp, and TK. TK correlated with BHBp (Pearson's r = 0.99), BHBv (r = 0.91), and BHBc (r = 0.91), all P < 0.0001. BHBv and BHBp had good agreement in absolute values. However, the slope between BHBc and BHBv, measured with the same device, was in the range of 0.64-0.78 in different regression models, indicating substantially higher BHB concentrations in capillary versus venous blood. We conclude that all three methods are valid to detect relative changes in ketosis, but our results highlight the importance of method considerations and the possible need to adjust cutoffs, e.g., in the management of ketoacidosis and in the evaluation and comparison of dietary interventions.