The impact of macronutritional composition and ketosis on cognitive health : from normal aging to Alzheimer’s disease

Abstract

Ketogenic diets (KD) are increasingly investigated for the prevention of cognitive decline and Alzheimer’s disease (AD). Without explicitly investigating a KD, this thesis disentangles two of its hallmarks: a reduced dietary carbohydrate/fat-ratio (CFr) and the metabolic state ketosis. Whether health effects from KD are primarily driven by ketosis or from other pathways related macronutritional changes, is not fully understood. Beyond CFr, KD may optionally be modified regarding protein, fat-subtypes, plant/animal-based food proportions, the timing of nutrient intake, and ketogenic supplements.

Strategies to induce ketosis in the absence of a carbohydrate restricted diet (Study I) and subsequent associations between induced ketosis and a biomarker essential for brain function (Study II) was investigated in a randomized clinical trial planned and performed within this doctoral project: In a 6-arm cross-over design, 15 healthy older adults (age 65-73, following their usual diet) were exposed to intake of oils with various composition of medium-chain triglycerides (MCT), with and without glucose. Blood levels of ketones (β-hydroxybutyrate, BHB) and brain-derived neurotrophic factor (BDNF) were thereafter monitored for 4 hours. Mature BDNF (mBDNF) and its precursor proBDNF are essential for brain plasticity, and their concentrations in serum have been associated with cognitive health. A methods comparison for measuring blood ketones (Study III) supports the internal validity of Study I and II.

The impact of self-reported CFr—in the non-ketogenic range—on cognitive performance (Study IV/V) was investigated by panel analyses on data (year 0, 1, and 2) from the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER). The sample (n=1259, age 60–77, 47% females) had no substantial cognitive impairment but had risk factors for developing dementia, and cognition at mean level or slightly lower than expected in screening test. Study V added stratified analyses based on genetics (APOE) and insulin status.

Study I: A 16-hour non-carbohydrate window and intake of 20 g caprylic acid (C8) contributed roughly equally to induce transient ketosis (0.45 mmol/L, AUC/time venous BHB hour 0-4, when combining the strategies). Coconut oil (which has a ≈7% fraction constituted by C8 and is dominated by lauric acid) did not share the ketogenic properties of purified C8 (difference –0.22 mmol/L, p<0.001).

Study II: Contrary to our expectations, change in mBDNF was lower (z-score: β=–0.88, p<0.001) after intake of C8 (higher ketosis) compared to sunflower oil (lower ketosis). Since associations between BHB and mBDNF appeared unrelated (p=0.43) on the individual level, alternative explanations to ketosis as a driver were discussed. In contrast, proBDNF increased more ( β=0.25, p=0.007) after intake of C8 compared to sunflower oil, and individual associations between BHB and proBDNF ( β=0.40, p=0.006) supported ketosis as a mechanistic link.

Study III: A handheld ketone meter correlated well with the laboratory method (r=0.91) and agreement was high when applied to venous whole blood (which was our primary outcome). However, absolute values were systematically higher in capillary blood, which should be considered in comparisons between studies.

Study IV: A lower CFr (log, z-score) estimated a higher composite z-score on a Neuropsychological Test Battery ( β=–0.022, p=0.011) in linear mixed regression. Methodological advantages of analyzing intake of carbohydrates and fat as a ratio compared to single variables were discussed. No significant associations were found for protein, and the saturated/total fat ratio had non-linear associations with cognitive performance.

Study V: APOE (𝜀-2/3/4), which is the most important AD risk gene, modified estimates between diet parameters (CFr, protein, saturated/total fat ratio, fiber, composite score) and cognitive performance in a sub-sample with insulin data, excluding diabetics (n=676). By increasing values of a continuous APOE-gradient [–1 (𝜀-23), –0.5 (𝜀-24), 0 (𝜀-33), 1 (𝜀-34), 2 (𝜀-44)], a less favorable estimate (p<0.0001 for interaction) was found for a Higher-carbohydrates-fiber-Lower-fat-protein composite score. Estimates for 𝜀-33 were relatively close to zero whereas 𝜀-44 (with some ambiguity for females) typically had an antagonistic estimate to 𝜀-23. Relative hypo- and hyper-insulinemia significantly magnified several estimates diet ->cognition in a dose dependent manner, primarily among 𝜀-34/44. The plant/animal-based proportion of macronutrients was discussed as a potential unmeasured confounder.

Conclusions: Macronutritional changes may be an alternative explanation to ketosis for what may drive potential cognitive effects from KD. Time-restricted carbohydrate intake may be considered as an alternative, or a complement, to C8-enriched MCT-oils for achieving mild ketosis. Signaling functions of ketones may be at work in transient mild/moderate ketosis, but whether our BDNF results have any cognitive implications requires further studies. To guide further research, our diet ->cognition analyses have strengthened the case for: (1) a precision nutrition approach based on APOE-genotype and insulin status; (2) not limiting interventions on carbohydrate restriction to the ketogenic range of CFr; (3) considering both ends of the insulin spectrum as representing distinct at-risk types susceptible to diet modifications. APOE-34/44 carriers may be optimal targets for studying potential benefits on brain health from CFr-reduction, and higher protein intake. The concept of universal macronutrient targets may be questioned, and stratified analyses may be encouraged in further studies.