β-hydroxybutyrate (BHB) was quantified in the posteromedial cortex (PMC) using ¹H-MRS (PRESS). Acute and acute-on-chronic responses were defined as the within-visit ratios of ~75 minutes post-drink to pre-drink (post/pre) at Weeks 0 and 4, respectively. The chronic effect was defined as the ratio of pre-drink BHB at Week 4 to pre-drink at Week 0 (Week 4 pre / Week 0 pre). BHB values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS PRESS assessment. BHB estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 40% were excluded from analysis.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

Glucose was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a glucose-to-creatine (Glucose/Cr) ratio. The chronic response was defined as the ratio of pre-drink Glucose/Cr at Week 4 to pre-drink Glucose/Cr at Week 0 (Week 4 pre / Week 0 pre). Glucose/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Glucose/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

N-acetyl-aspartate was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a N-acetyl-aspartate-to-creatine (N-acetyl-aspartate/Cr) ratio. The chronic response was defined as the ratio of pre-drink N-acetyl-aspartate/Cr at Week 4 to pre-drink N-acetyl-aspartate/Cr at Week 0 (Week 4 pre / Week 0 pre). N-acetyl-aspartate/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. N-acetyl-aspartate/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Lactate was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Lactate-to-creatine (Lactate/Cr) ratio. The chronic response was defined as the ratio of pre-drink Lactate/Cr at Week 4 to pre-drink Lactate/Cr at Week 0 (Week 4 pre / Week 0 pre). Lactate/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Lactate/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Glycine was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Glycine-to-creatine (Glycine/Cr) ratio. The chronic response was defined as the ratio of pre-drink Glycine/Cr at Week 4 to pre-drink Glycine/Cr at Week 0 (Week 4 pre / Week 0 pre). Glycine/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Glycine/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Glutamate was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Glutamate-to-creatine (Glutamate/Cr) ratio. The chronic response was defined as the ratio of pre-drink Glutamate/Cr at Week 4 to pre-drink Glutamate/Cr at Week 0 (Week 4 pre / Week 0 pre). Glutamate/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Glutamate/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Gamma-aminobutyric acid was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Gamma-aminobutyric acid-to-creatine (Gamma-aminobutyric acid/Cr) ratio. The chronic response was defined as the ratio of pre-drink Gamma-aminobutyric acid/Cr at Week 4 to pre-drink Gamma-aminobutyric acid/Cr at Week 0 (Week 4 pre / Week 0 pre). Gamma-aminobutyric acid/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Gamma-aminobutyric acid/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Glutamine was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Glutamine-to-creatine (Glutamine/Cr) ratio. The chronic response was defined as the ratio of pre-drink Glutamine/Cr at Week 4 to pre-drink Glutamine/Cr at Week 0 (Week 4 pre / Week 0 pre). Glutamine/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Glutamine/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Glutathione was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Glutathione-to-creatine (Glutathione/Cr) ratio. The chronic response was defined as the ratio of pre-drink Glutathione/Cr at Week 4 to pre-drink Glutathione/Cr at Week 0 (Week 4 pre / Week 0 pre). Glutathione/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Glutathione/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Myo-inositol was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Myo-inositol-to-creatine (Myo-inositol/Cr) ratio. The chronic response was defined as the ratio of pre-drink Myo-inositol/Cr at Week 4 to pre-drink Myo-inositol/Cr at Week 0 (Week 4 pre / Week 0 pre). Myo-inositol/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Myo-inositol/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

N-acetyl-aspartyl-glutamate was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a N-acetyl-aspartyl-glutamate-to-creatine (N-acetyl-aspartyl-glutamate/Cr) ratio. The chronic response was defined as the ratio of pre-drink N-acetyl-aspartyl-glutamate/Cr at Week 4 to pre-drink N-acetyl-aspartyl-glutamate/Cr at Week 0 (Week 4 pre / Week 0 pre). N-acetyl-aspartyl-glutamate/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. N-acetyl-aspartyl-glutamate/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Aspartate was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as an Aspartate-to-creatine (Aspartate/Cr) ratio. The chronic response was defined as the ratio of pre-drink Aspartate/Cr at Week 4 to pre-drink Aspartate/Cr at Week 0 (Week 4 pre / Week 0 pre). Aspartate/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Aspartate/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Ascorbate was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as an Ascorbate-to-creatine (Ascorbate/Cr) ratio. The chronic response was defined as the ratio of pre-drink Ascorbate/Cr at Week 4 to pre-drink Ascorbate/Cr at Week 0 (Week 4 pre / Week 0 pre). Ascorbate/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Ascorbate/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Alanine was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as an Alanine-to-creatine (Alanine/Cr) ratio. The chronic response was defined as the ratio of pre-drink Alanine/Cr at Week 4 to pre-drink Alanine/Cr at Week 0 (Week 4 pre / Week 0 pre). Alanine/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Alanine/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Scyllo-inositol was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Scyllo-inositol-to-creatine (Scyllo-inositol/Cr) ratio. The chronic response was defined as the ratio of pre-drink Scyllo-inositol/Cr at Week 4 to pre-drink Scyllo-inositol/Cr at Week 0 (Week 4 pre / Week 0 pre). Scyllo-inositol/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Scyllo-inositol/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Phosphocholine was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Phosphocholine-to-creatine (Phosphocholine/Cr) ratio. The chronic response was defined as the ratio of pre-drink Phosphocholine/Cr at Week 4 to pre-drink Phosphocholine/Cr at Week 0 (Week 4 pre / Week 0 pre). Phosphocholine/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Phosphocholine/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Glycerophosphocholine was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Glycerophosphocholine-to-creatine (Glycerophosphocholine/Cr) ratio. The chronic response was defined as the ratio of pre-drink Glycerophosphocholine/Cr at Week 4 to pre-drink Glycerophosphocholine/Cr at Week 0 (Week 4 pre / Week 0 pre). Glycerophosphocholine/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Glycerophosphocholine/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Phosphoethanolamine was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Phosphoethanolamine-to-creatine (Phosphoethanolamine/Cr) ratio. The chronic response was defined as the ratio of pre-drink Phosphoethanolamine/Cr at Week 4 to pre-drink Phosphoethanolamine/Cr at Week 0 (Week 4 pre / Week 0 pre). Phosphoethanolamine/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Phosphoethanolamine/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Taurine was quantified in the posteromedial cortex (PMC) using ¹H-MRS (J-PRESS) at pre-drink assessments and expressed as a Taurine-to-creatine (Taurine/Cr) ratio. The chronic response was defined as the ratio of pre-drink Taurine/Cr at Week 4 to pre-drink Taurine/Cr at Week 0 (Week 4 pre / Week 0 pre). Taurine/Cr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline ¹H-MRS J-PRESS assessment. Taurine/Cr estimates with Cramér-Rao lower bounds (CRLB, %SD) ≥ 20% were excluded from analysis.

Post-exercise phosphocreatine (PCr) recovery time constant (τPCr) was measured by thigh ³¹P-MRS at pre-drink assessments. The chronic response was defined as the ratio of pre-drink τPCr at Week 4 to pre-drink τPCr at Week 0 (Week 4 pre / Week 0 pre). τPCr values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline thigh ³¹P-MRS assessment. Assessments were excluded from analysis if the τPCr relative fit error (SD/estimate) ≥ 0.20, if PCr drop area < 30%, or if intramuscular minimum pH < 6.8.

Phosphocreatine (PCr) depletion area during exercise (PCr drop area) was measured by thigh ³¹P-MRS at pre-drink assessments. The chronic response was defined as the ratio of pre-drink PCr drop area at Week 4 to pre-drink PCr drop area at Week 0 (Week 4 pre / Week 0 pre). PCr drop area values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline thigh ³¹P-MRS assessment. Assessments were excluded from analysis if the τPCr relative fit error (SD/estimate) ≥ 0.20, if PCr drop area < 30%, or if intramuscular minimum pH < 6.8.

Intramuscular minimum pH during exercise was measured by thigh ³¹P-MRS at pre-drink assessments. The chronic response was defined as the difference between pre-drink PCr drop area at Week 4 and pre-drink PCr drop area at Week 0 (Week 4 pre - Week 0 pre). Intramuscular minimum pH values were derived from spectral fitting with predefined quality-control criteria.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline thigh ³¹P-MRS assessment. Assessments were excluded from analysis if the τPCr relative fit error (SD/estimate) ≥ 0.20, if PCr drop area < 30%, or if intramuscular minimum pH < 6.8.

G-BrainAGE was derived from pre-drink 3D T1-weighted MP-RAGE brain MRI using FreeSurfer (longitudinal pipeline) features processed through the CentileBrain platform with sex-specific pretrained models. G-BrainAGE represents the difference between the model-predicted brain age and chronological age (predicted age - chronological age), with higher values indicating an "older-appearing" brain relative to age and lower values indicating a "younger-appearing" brain. The chronic effect was assessed as the change in G-BrainAGE from Week 0 to Week 4.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline brain MRI (T1-weighted MP-RAGE) assessment.

The Logical Memory test assesses verbal episodic memory for short stories. Participants recall two stories immediately and after a 20-30 minute delay. Outcomes include immediate and delayed recall subscores, each ranging from 0 to 50 (25 units/story × 2 stories), with higher scores indicating better performance. Verbatim subscores reflect exact recall; gist subscores reflect recall of essential elements.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

MoCA is a brief assessment of global cognitive function across multiple domains (e.g., attention, executive function, memory, language, visuospatial abilities, and orientation). Total scores range from 0 to 30, with higher scores indicating better cognitive performance.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

The Eriksen flanker task assesses selective attention and inhibitory control by requiring responses to a target stimulus while ignoring distracting flankers. The outcome is a composite score integrating response accuracy and response time into a single performance metric. Composite scores range from 0 to 10, with higher scores indicating better performance.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

The Dimensional Set Shifting task assesses cognitive flexibility (ability to shift between rules or stimulus dimensions). The outcome is a task-derived composite performance score; scores range from 0 to 10, with higher scores indicating better performance.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

The Digit Symbol Substitution Test (DSST) assesses processing speed, attention, and visuomotor coordination by requiring participants to match symbols to digits using a provided key within a fixed time. The outcome is the total number of correct symbol-digit matches. Total scores range from 0 to 93, with higher scores indicating better performance.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.

The Free and Cued Selective Reminding Test (FCSRT) assesses verbal episodic memory using controlled learning and recall of a 16-word list with semantic cues. Total Free Recall (TFR) is the sum of freely recalled words across three learning trials (range 0 to 48). Total Delayed Free Recall (TDFR) is the number of freely recalled words after a delay (range 0 to 16). Higher scores indicate better memory performance.

Analysis was performed in a modified intention-to-treat population, which included all randomized participants who received ≥ 1 dose of the intervention and had ≥ 1 post-baseline efficacy assessment.