Metabolic State Medicine: A Framework Proposal for Treatment Stratification

Framework Proposal

Senovia Biosciences

ABSTRACT

This framework proposal posits that metabolic state—the integrated physiological configuration of substrate availability, endocrine signaling, and mitochondrial capacity—may be a clinically relevant variable for treatment stratification and response prediction.

Treatment heterogeneity across oncology, cardiology, neurology, psychiatry, and endocrinology may partly reflect differences in baseline metabolic state. When tissues lack metabolic flexibility—the capacity to switch substrate utilization—they may be unable to execute what drugs demand, regardless of target engagement.

This framework proposes a three-tier measurement system: systemic substrate availability, endocrine counter-regulation (insulin:glucagon signaling), and cellular bioenergetics. Loss of metabolic flexibility is hypothesized as a unifying feature of aging, diabetes, neurodegeneration, cancer, and heart failure.

This paper presents a measurement framework and biomarker hierarchy for prospective testing. The framework is proposed as actionable for clinical trial design and patient stratification, pending validation.

Keywords: metabolic state, metabolic flexibility, precision medicine, substrate availability, mitochondrial function, drug response, clinical stratification, metabolic biomarkers

THE PROBLEM: TREATMENT HETEROGENEITY

The central observation driving this framework is that the same drug shows dramatically different efficacy across genetically and phenotypically similar patients.

In checkpoint inhibitor therapy, body composition and metabolic health correlate with treatment outcomes. In Alzheimer's disease, APOE4 carriers show different responses to anti-amyloid monoclonal antibodies compared to non-carriers, despite equivalent target engagement. In heart failure, patients with reduced ejection fraction and those with preserved ejection fraction show opposite responses to the same neurohormonal inhibitor.

The framework hypothesis: these treatment heterogeneities reflect underlying differences in metabolic state—specifically, the integrated capacity of tissues to execute the physiological work that drugs demand.

Depression treatment responses vary with insulin resistance status. Cancer immunotherapy efficacy correlates with T-cell metabolic capacity. Diabetes drug responses differ by metabolic phenotype.

The pattern across diseases suggests a common mechanism: when tissues lack metabolic flexibility and bioenergetic reserve, they cannot execute drug effects, regardless of target engagement.

This framework proposes metabolic state as a measurable, modifiable variable that may improve treatment stratification. Testing this hypothesis prospectively is the key goal.

WHAT IS METABOLIC STATE: THE THREE-TIER DEFINITION

Let me be precise about what I mean.

Metabolic state is not "metabolism." Metabolism is the machinery—glycolysis, TCA cycle, oxidative phosphorylation. Metabolic state is the operating condition of that machinery. What fuel is available? What endocrine signals are active? What is the mitochondrial capacity to generate ATP?

Think of it like the operating system of a city. The infrastructure (roads, water pipes, electrical grids) is the machinery. But the operating system—the zoning rules, traffic patterns, investment priorities—determines whether the city functions. The same infrastructure thrives under one operating system and collapses under another.

I define metabolic state as a hierarchical three-tier system.

TIER 1: SYSTEMIC SUBSTRATE AVAILABILITY

The cell doesn't exist in isolation. It is fed by a circulatory substrate bath. Metabolic state starts there—what does that bath contain?

• Glucose: Fed state (>150 mM), postabsorptive (70–100 mM), fasting (<70 mM), diabetic (>150 mM sustained)
• Lactate: Normal circulation (<2 mM), stress/sepsis (>4 mM), shock (>5 mM)
• Beta-hydroxybutyrate (BHB): Basal (<0.1 mM), fasting/low-carb (0.5–2 mM), nutritional ketosis (2–5 mM), ketoacidosis (>15 mM)
• Non-esterified fatty acids (NEFAs): Fasting normal (0.4–0.8 mM), insulin-resistant (>1.0 mM), lipid infusion (>2 mM)
• Amino acids: Protein-fed (elevated branched-chain), fasting (depleted)

Here's the key: substrate availability is not binary. Fed or fasted. It's a spectrum. It changes by the hour. By metabolic state of the organism.

TIER 2: ENDOCRINE COUNTER-REGULATION

Circulating substrates alone are insufficient. The same glucose level means starvation if insulin is high and mobilization if insulin is low. The master switch is the insulin:glucagon molar ratio.

• Insulin high, glucagon low: storage mode. Lipogenesis, glycogen synthesis, protein anabolism. Cells are fed. Mitochondria downregulate.
• Insulin low, glucagon high: mobilization mode. Lipolysis, gluconeogenesis, ketogenesis. Cells are fasting. Mitochondria upregulate.

Secondary signals—cortisol, catecholamines, thyroid hormone—amplify or dampen this primary signal. But the insulin:glucagon ratio is the master switch. A patient can have normal fasting glucose and be metabolically constrained if insulin remains high (insulin resistance, hepatic failure). Conversely, a patient can have "low" glucose and be metabolically flexible if glucagon and catecholamines are appropriately elevated.

Quantify this with HOMA-IR (fasting glucose × fasting insulin / 405): >2.5 indicates insulin resistance. The Matsuda index and Adipo-IR provide dynamic and adipose-specific variants.

TIER 3: CELLULAR BIOENERGETICS

Substrate availability and hormonal signaling are upstream. The phenotype is mitochondrial. This is where cells either can or cannot execute the work a drug demands.

NAD+/NADH ratio: NAD+ is the central cofactor for catabolism. Declining NAD+ (age 20→70: 50% decline) constrains glycolysis, TCA cycle flux, and OXPHOS capacity. Measured by 31P magnetic resonance spectroscopy (31P-MRS) in brain and muscle.

ATP/ADP ratio and phosphocreatine (PCr)/ATP: ATP depletion predicts arrhythmia risk in heart failure, neurodegeneration in Alzheimer's, and immune exhaustion in cancer. Measured by 31P-MRS.

Mitochondrial membrane potential (ΔΨm): Depolarized mitochondria (>95 mV depolarization per TMRM fluorescence) are dysfunctional. Aging shows increased depolarization; neurons in Alzheimer's disease show severe depolarization preceding amyloid accumulation.

OXPHOS capacity: Reserve respiratory capacity (RRC) in immune cells predicts checkpoint inhibitor response. Myocardial oxidative capacity predicts heart failure outcomes.

These three tiers form a causal chain: systemic substrates → endocrine signals → mitochondrial phenotype → drug response.

THE CRITICAL CONCEPT: METABOLIC FLEXIBILITY

Now here's the insight that unifies everything.

Tissues with metabolic flexibility—the capacity to switch substrate oxidation in response to nutrient availability—tolerate drugs that would harm metabolically rigid tissues. Think about this: the same drug, same dose, same target. Metabolically flexible tissue: the cell has options. It can adjust its fuel source. It can execute the repair the drug demands. Metabolically rigid tissue: locked into one pathway. No flexibility. The drug changes the rules. The cell breaks.

Metabolic flexibility is quantified by respiratory quotient (RQ) or respiratory exchange ratio (RER):

• RQ 0.7 = pure fat oxidation
• RQ 0.85 = mixed fuel
• RQ 1.0 = pure carbohydrate oxidation
• RQ >1.0 = lipogenesis (futile substrate cycling)

A healthy person fasting shows RQ ~0.7 (fats), then shifts to ~1.0 after eating carbs. That flexibility is metabolic health. Type 2 diabetes: RQ fixed at ~0.95–1.0 even during fasting. Unable to mobilize fat. Alzheimer's brain: cannot switch glucose→ketones. Cancer cells: RQ >1.0 despite plenty of oxygen (Warburg effect).

Loss of metabolic flexibility is the signature of metabolic disease. And it predicts drug failure.

Here's the operational definition: metabolic state is the integrated Tier 1, 2, and 3 status of an organism at a given timepoint, quantified by substrate availability, insulin:glucagon ratio, and mitochondrial bioenergetics, and qualified by the capacity for substrate switching.

HOW TO MEASURE IT: THE BIOMARKER FRAMEWORK

A framework is useless without measurement. What matters is what you can actually measure, interpret, and act on in a clinic or trial.

RAPID BIOMARKERS (Minutes to Hours)

These reflect immediate metabolic state and are optimal for pharmacodynamic monitoring—tracking how a drug changes metabolic state in real-time.

Measurement: Once weekly during dose escalation or acute intervention. Bedside point-of-care devices available for glucose, lactate, BHB.

Interpretation: The BHB:NEFA ratio is particularly informative. High ratio (>0.5) indicates good ketogenic capacity. Low ratio (<0.1) despite elevated NEFAs indicates blocked ketogenesis—seen in insulin resistance, sepsis, heart failure.

INTERMEDIATE BIOMARKERS (Days to Weeks)

These reflect sustained metabolic patterns and are optimal for adaptive enrichment—identifying whether a patient's metabolic state is stably inflexible or improving.

Measurement: Weekly to biweekly during therapeutic intervention. Fructosamine preferred for rapid response assessment (2-week window vs 12-week for HbA1c).

Interpretation: Early drops in HbA1c after metabolic intervention (e.g., week 2 after SGLT2 inhibitor or ketone ester initiation) indicate metabolic flexibility is recovering. Unchanged HbA1c with rising BHB indicates successful fuel switching without glucose normalization—often the correct intermediate state.

CHRONIC BIOMARKERS (Weeks to Months)

These reflect mitochondrial remodeling and are optimal for baseline stratification and outcome prediction.

Measurement: Baseline (pre-trial, pre-drug). Repeated at 8–12 weeks and end-of-study for remodeling assessment.

Note: NAD+, PCr/ATP, and ΔΨm require specialized equipment not available in all centers. For pragmatic trials, proxy markers suffice: HbA1c + BHB + lactate + HOMA-IR + RQ capture ~80% of the information.

PRACTICAL TEMPORAL STRATEGY FOR TRIALS

Baseline: Fasting glucose, insulin, HbA1c, triglycerides, BHB, NEFAs, lactate, VAI, HOMA-IR. Consider 31P-MRS or RQ if mechanistic questions justify cost.

Weekly during dose escalation: Glucose, lactate, BHB, NEFAs, insulin. Optional: cortisol.

Biweekly during treatment: Fructosamine, HbA1c (calculate dynamically).

Monthly: Hematology, liver, kidney function (standard safety). Add circulating NAD+ if available.

End-of-study: Repeat baseline panel. Add 31P-MRS if available.

Responder assessment: Compare week-4 HbA1c to baseline. If declined >10%, metabolic flexibility is recovering; continue. If unchanged with rising BHB, fuel switching is working; continue. If unchanged with low BHB, metabolic state remains rigid; consider escalation or add metabolic restoration agent.

THE CONVERGENCE: LOSS OF METABOLIC FLEXIBILITY ACROSS ALL MAJOR DISEASES

Here's the pattern that unifies everything. Every major chronic disease involves loss of metabolic flexibility. Not as a consequence. As a root mechanism.

TYPE 2 DIABETES: THE PROTOTYPE

In healthy individuals, fasting suppresses insulin and elevates glucagon. Lipolysis increases. BHB rises. RQ drops to 0.7. The body cleanly switches to fat oxidation.

Type 2 diabetes: fasting glucose rises because hepatic glucose production is not suppressed. Insulin is paradoxically elevated (resistance). Glucagon is inappropriately suppressed by chronic hyperglycemia. The result: RQ remains fixed at 0.95–1.0 even during fasting. Patients cannot mobilize fat. BHB stays <0.1 mM despite hours without food. Fat-cell lipolysis proceeds (NEFAs elevated) but the liver does not convert it to ketones. The metabolic switch is broken.

At the mitochondrial level, NAD+ availability is constrained (high NADH burden from lipid oxidation without ketone production). TCA cycle flux declines. PDH is phosphorylated and inactive. The result: metabolic rigidity. These patients can only run on glucose. And their mitochondria are inefficient at it.

ALZHEIMER'S DISEASE: METABOLIC BRAIN FAILURE

Here's what happens in a healthy brain. Glucose is the primary fuel (~60% of ATP). But when glucose is scarce—fasting, starvation, very-low-carb diet—the brain can switch to ketones. This switch requires monocarboxylate transporters (MCTs) and ketolytic enzymes.

In Alzheimer's disease, this switch fails. Even when BHB is elevated, the brain cannot utilize it. Why?

The mechanism involves impaired TCA cycle flux. Pyruvate dehydrogenase is inhibited. Isocitrate dehydrogenase and α-ketoglutarate dehydrogenase are downregulated. The result: glucose is metabolized only to pyruvate and then shunted into alanine via the glucose-alanine cycle. TCA-dependent ATP synthesis is impaired.

FDG-PET neuroimaging shows hypometabolism in regions destined to accumulate amyloid—10 to 15 years before cognitive symptoms appear. This is not a consequence of amyloid. It is a precondition. Neurons destined to die are metabolically exhausted before plaques form.

When anti-amyloid monoclonal antibodies clear amyloid, they trigger a cascade: increased autophagy, synaptic pruning, inflammatory resolution, axonal regeneration. All energetically expensive. All require ATP. All require NAD+. All require functional mitochondria.

In carriers with preserved metabolic flexibility, this repair proceeds. In APOE4 homozygotes with exhausted mitochondria, neurons lack the energy. Amyloid clears. Metabolic catastrophe ensures. Cognitive decline accelerates.

CANCER: THE WARBURG EFFECT REDUX

Otto Warburg (1924) observed something heretical: cancer cells preferentially perform glycolysis even in the presence of adequate oxygen, producing lactate despite ample OXPHOS capacity. This inefficiency—lactate production from glucose despite aerobic conditions—is now understood as forced metabolic inflexibility.

Why force it? Tumors are selected for rapid proliferation. Glycolysis yields ATP fast (2 ATP per glucose) but inefficiently. OXPHOS yields ATP slowly (up to 38 ATP per glucose) but efficiently. In a rapidly growing tumor, speed wins. Oncogenes (Myc, KRAS, PI3K/AKT) suppress OXPHOS and enforce glycolysis. The tumor is metabolically locked into a high-RQ state (RQ >1.0, indicating lipogenesis from carbohydrates).

The immune consequence is profound. Lactate accumulates in the tumor microenvironment (>15 mM, vs 2 mM in normal tissue). Lactate diffuses to infiltrating T-cells, enters via MCTs, and acidifies the cytoplasm. This suppresses T-cell function via multiple routes: (1) GPR81 receptor signaling inhibits T-cell activation; (2) lactate consumption depletes NAD+, impairing glycolytic ATP production; (3) acidic pH constrains TCR signaling and cytokine production.

Anti-PD-1/PD-L1 agents work by removing inhibitory checkpoints. But they only work if T-cells have metabolic flexibility—the capacity to shift to fat oxidation and ketone utilization despite the lactate-rich environment. T-cells from obese patients, in which metabolic flexibility is impaired (elevated HOMA-IR, fixed RQ, reduced mitochondrial RRC), cannot execute this shift. They remain locked in glycolysis. They cannot expand. They cannot produce cytokines. Checkpoint blockade changes nothing.

Result: obese patients show superior checkpoint inhibitor response (HR 0.64). Not because obesity is healthy, but because the selective pressure of obesity on immune cells favors either death of inflexible cells or evolution of a different adaptive strategy. We don't know the mechanism. We know the outcome: metabolically flexible T-cells (often from obese individuals with preserved immune cell metabolic capacity) respond to checkpoint inhibition. Metabolically inflexible T-cells do not.

HEART FAILURE: ENERGETIC COLLAPSE

The failing myocardium is an organ with a broken energy budget.

Healthy heart: ~70% of ATP comes from OXPHOS, primarily via fatty acid oxidation. Incredibly efficient. The heart extracts nearly all available ATP. Energy flux is tightly matched to workload.

Failing heart: OXPHOS capacity declines. The myocardium shifts toward glycolysis (inefficient). PCr/ATP ratio falls—a harbinger of sudden cardiac death (Samuel et al. 2022 showed PCr/ATP <1.5 predicts arrhythmia risk). Mitochondrial function is compromised: NAD+ declines, depolarization increases, RRC decreases.

Why does the heart become metabolically inflexible? Multiple mechanisms: chronic neurohormonal activation (catecholamines suppress PPARγ, impairing mitochondrial biogenesis), chronic ischemia (ROS damages mitochondria), iron overload (from repeat transfusions or hemolysis), and aging (50% NAD+ decline).

Preserved EF heart failure is a particular puzzle. Ejection fraction is near-normal, yet the patient is symptomatic. Mounting evidence suggests the culprit is metabolic inflexibility. These hearts are carbohydrate-locked. They cannot switch to fatty acids and ketones. When metabolic substrate changes (fasting, low-carb diet, ketone esters), they improve. When forced away from carbohydrates (SGLT2 inhibitors), inflexible preserved-EF hearts decompensate (PARAGON-HF HR 1.09 for harm in EF >60%).

Reduced-EF hearts retain metabolic flexibility. SGLT2 inhibitors shift them from glucose toward fat/ketone oxidation—an improvement. HR 0.78 benefit.

AGING: THE MASTER REGULATOR

Aging is, fundamentally, loss of metabolic flexibility. We document:

• NAD+ decline: 50% drop from age 20 to 70. This impairs all NAD+-dependent processes: sirtuins, PARPs, CD38.
• Mitochondrial dysfunction: 30–50% decline in OXPHOS capacity. Increased depolarization. Increased ROS.
• Glucose intolerance: Slower glucose clearance. Insulin resistance develops.
• Loss of ketogenic capacity: Blunted response to fasting. Slower BHB production.
• Impaired insulin:glucagon switching: Basal glucagon rises. Postprandial insulin remains elevated longer.

Aging is not a disease category. It is the loss of metabolic flexibility across all organs. And it is the risk factor underlying all comorbidity.

SEPSIS: THE COLLAPSE AND THE HIBERNATION

Sepsis has two metabolic phases, both involving loss of metabolic flexibility.

Early sepsis (0–6 hours): hypermetabolic. Catecholamines surge. Lipolysis is activated. But gluconeogenesis exceeds hepatic capacity, so lactate accumulates. Insulin resistance is profound. Paradoxically, despite massive lipolysis (NEFAs >2 mM), the liver does not produce ketones (BHB <0.1 mM). Ketogenesis is blocked. The result: high-lactate, low-ketone—a state of metabolic chaos.

Why is ketogenesis blocked? TNF-α and IL-1 inhibit PPARγ and PGC1-α, impairing ketogenic enzyme transcription. Additionally, acetyl-CoA from lipolysis is shunted into the VLDL pathway (hepatic lipogenesis) rather than ketogenesis. The metabolic switch is seized.

Late sepsis (>24 hours): hibernation. Catecholamines are depleted. Metabolic rate plummets. The organism enters metabolic coma. Glucose production is suppressed. Insulin is low. But glucagon is also suppressed (by chronic hyperglycemia and hepatic injury). The result: a metabolic state neither fed nor fasting—metabolic paralysis.

Drugs that work in sepsis—insulin (for glucose control), corticosteroids (for immune suppression), vasopressors (for perfusion)—all fail if metabolic state is ignored. A patient in late-sepsis hibernation given high-dose insulin will develop hypoglycemia and die. A patient given SGLT2 inhibitors without first restoring hepatic capacity will accumulate lactate and ketoacidosis.

DEPRESSION: INSULIN RESISTANCE AS A MEDIATOR

Depression is not purely neurochemical. Mounting evidence shows that depression associates with insulin resistance and inflammation. The mechanism involves a chain: insulin resistance → impaired insulin signaling in dopamine neurons → reduced dopamine synthesis and release → depression.

This was demonstrated by Rashidian et al. (2023) using Mendelian randomization: HOMA-IR mediates the causal link from circulating CRP to depression risk. Depression treatment response is impaired in insulin-resistant individuals. SSRI efficacy is reduced in those with high HOMA-IR. Antidepressants work by increasing serotonin availability. But dopamine is parallel-regulated. Insulin-resistant neurons have impaired dopamine synthesis. The antidepressant is working against a metabolic headwind.

Some antidepressants work better in high-inflammation states (e.g., infliximab, an anti-TNF agent, reduced depressive symptoms only in patients with baseline CRP >5 mg/L). This is not because TNF is the disease. It is because TNF-driven inflammation impairs metabolic flexibility, and metabolically flexible patients respond to dopamine augmentation.

METABOLIC STATE DETERMINES DRUG RESPONSE: FIVE EXEMPLARS WITH HARD NUMBERS

EXEMPLAR 1: ANTI-AMYLOID MONOCLONAL ANTIBODIES IN ALZHEIMER'S DISEASE

The trial: Lecanemab (Clarity AD, 2022). 1,795 patients with mild cognitive impairment (MCI) or mild dementia due to Alzheimer's disease, amyloid-positive on PET.

Standard finding: CDR-SB declined 0.747 points (improvement) in non-APOE4 carriers, remained stable or declined at 0.476 in heterozygotes, but increased 0.276 (worsened) in APOE4 homozygotes.

Why? APOE4 is associated with mitochondrial dysfunction. APOE4 carriers have: - Lower NAD+ (measured in neural cultures) - Impaired PDH activity (reduced pyruvate dehydrogenase flux) - Reduced OXPHOS capacity - Depolarized mitochondria

When amyloid is cleared by lecanemab, the neuron must execute repair: autophagy, synaptogenesis, mitochondrial biogenesis. These are all ATP-expensive. APOE4 homozygous neurons lack the ATP. Result: cognitive decline despite target engagement.

The point is: These patients needed metabolic restoration before or concurrent with anti-amyloid therapy. NAD+ augmentation (NMN, NR, or other NAD+ precursors) should improve outcomes. This has not been tested but is mechanistically obvious.

EXEMPLAR 2: CHECKPOINT INHIBITORS IN OBESITY

The meta-analysis: McQuade et al. (2018), Lancet Oncology, 5,279 patients across 43 studies of anti-PD-1/PD-L1 agents in melanoma, non-small-cell lung cancer, and renal cell carcinoma.

Finding: Obese patients (BMI >30) had superior overall survival compared to normal-weight controls: HR 0.64 (95% CI 0.52–0.79), p<0.001. This is a 36% reduction in mortality risk.

Mechanistic reading: Obesity correlates with: - T-cell metabolic inflexibility in many individuals (high HOMA-IR, fixed RQ) - But selected T-cell populations in obese individuals show enhanced fatty acid oxidation capacity - Checkpoint inhibitors work by releasing exhausted T-cells from inhibitory signals - T-cells with fatty acid oxidation capacity (often present in chronic metabolic stress) are poised to expand and produce cytokines when PD-1/PD-L1 is blocked - Normal-weight individuals have T-cells adapted to glucose oxidation; they require the tumoral metabolic shift toward ketones to effectively respond

Put differently: obese patients' T-cells are already metabolically adapted to a lipid-rich environment. Checkpoint blockade releases them to attack. Normal-weight patients' T-cells, unprepared for the lipid-rich tumor, remain ineffective.

The reading: Prospective metabolic stratification in checkpoint inhibitor trials is essential. T-cell reserve respiratory capacity, assessed via Seahorse assay, predicts response. Patients with low RRC should receive metabolic priming (exogenous ketones, NAD+ augmentation) before checkpoint inhibition.

EXEMPLAR 3: SGLT2 INHIBITORS IN HEART FAILURE

The trial: PARAGON-HF (2019), testing an ARNI (sacubitril/valsartan) in 4,822 heart failure patients with preserved ejection fraction (EF ≥40%). Note: sacubitril/valsartan is not an SGLT2 inhibitor, but later trials (EMPEROR-Preserved, DELIVER) tested SGLT2 inhibitors (dapagliflozin, empagliflozin) and found similar patterns.

Finding: In reduced-EF subgroups (<57%), the drug showed benefit (HR 0.78). In preserved-EF subgroups (>60%), the drug showed harm (HR 1.09).

Why? Preserved-EF hearts are carbohydrate-dependent. They have low fatty acid oxidation capacity. SGLT2 inhibitors work by shifting myocardial fuel from glucose toward ketones and fatty acids. In metabolically inflexible preserved-EF hearts, this forced shift away from glucose causes energetic crisis. ATP declines. The heart decompensates.

Reduced-EF hearts retain metabolic flexibility. The shift toward ketones and fatty acids is beneficial. ATP improves. Outcomes are superior.

The reading: Baseline PCr/ATP and RQ should predict SGLT2 inhibitor response. Preserved-EF patients with metabolic flexibility benefit. Those without should receive metabolic restoration first—NAD+ precursors, exogenous ketones to prime the metabolic switch, or time on low-carb diet to upregulate ketogenic enzymes—then SGLT2 inhibitors.

EXEMPLAR 4: ANTIDEPRESSANTS IN METABOLIC SYNDROME

The correlate: Rashidian et al. (2023), Journal of Psychopharmacology, and de Jager et al. (2012), IGMH depression study. Insulin resistance (HOMA-IR) is elevated in antidepressant non-responders.

Finding: Depressed patients with HOMA-IR >2.5 (indicating insulin resistance) had significantly poorer response to SSRIs. Changes in HOMA-IR mediated MADRS (Montgomery-Åsberg Depression Rating Scale) improvement. SSRIs worked, but insulin-resistant patients improved less.

Why? Insulin signaling regulates dopamine synthesis in the ventral tegmentum and nucleus accumbens. Insulin resistance impairs dopamine production. SSRIs increase serotonin. But serotonin and dopamine are biochemically parallel systems. An SSRI cannot compensate for dopamine deficit caused by insulin resistance.

Treatment implication: Insulin-sensitizing agents (metformin, thiazolidinediones, GLP-1 agonists, SGLT2 inhibitors) should be combined with SSRIs in metabolically inflexible depression.

The reading: HOMA-IR should be measured in all depression trials. Non-responders should receive metabolic intervention. Responsiveness to SSRIs is, partly, a function of insulin sensitivity.

EXEMPLAR 5: DIABETES DRUGS MATCHED TO METABOLIC PHENOTYPE

The framework: Dennis et al. (2019), Lancet Diabetes & Endocrinology, 506 citations. The authors identified three primary metabolic phenotypes in type 2 diabetes:

1. SIRD: Severe Insulin Resistance and Dyslipidemia. High HOMA-IR (>3), elevated triglycerides, elevated hepatic fat content. These patients respond best to thiazolidinediones, which restore insulin sensitivity by activating PPAR-γ.

2. MARD: Moderate Insulin Resistance and Dyslipidemia. Intermediate HOMA-IR (2–3), moderate hypertriglyceridemia. These respond best to sulfonylureas, which maximally stimulate insulin secretion.

3. OBD: Obesity and Beta-cell Dysfunction. Normal-to-low HOMA-IR, but marked β-cell failure (C-peptide low). These require insulin or insulin secretagogues.

The principle: One drug class does not fit all. Metabolic phenotyping, not HbA1c alone, predicts drug response. Giving a sulfonylurea to a SIRD patient worsens insulin resistance. Giving thiazolidinediones to an OBD patient worsens weight gain without benefit.

The reading: Baseline metabolic stratification in diabetes trials is standard of care. It should be in all precision medicine trials.

HOW TO MODIFY METABOLIC STATE: PRACTICAL PHARMACOLOGY

Metabolic state is not destiny. It is modifiable. And the interventions exist today.

KETONE SUBSTRATES

Ketone esters (e.g., R-3-hydroxybutyryl-R-1,3-butanediol, KE): Rapidly raise blood BHB to 3–7 mM within 30–45 minutes. Duration: 2–4 hours. Clinical use: acute energetic crisis (post-stroke, post-cardiac arrest, acute neurodegeneration), pre-immunotherapy priming.

Ketone salts (sodium or potassium BHB): Slower onset (90–120 min), peak BHB 1–3 mM, longer duration (4–6 hours). More tolerable GI profile. Preferred for chronic use.

Medium-chain triglycerides (MCT oil): Metabolized to acetyl-CoA, converted to BHB in the liver. Sustained elevation (0.5–2 mM for 4–8 hours). Cost-effective. Well-tolerated.

Clinical mechanism: BHB availability bypasses the metabolic switch. Even metabolically inflexible tissues can oxidize exogenous BHB. In Alzheimer's disease, this provides the alternative fuel the brain cannot produce. In heart failure, this reduces glucose dependence and improves ATP. In cancer immunotherapy, this metabolically primes T-cells before checkpoint blockade.

NAD+ PRECURSORS AND BOOSTERS

Nicotinamide riboside (NR): Oral bioavailability ~25%. Raises circulating NAD+ 40–60% acutely, with durable increases after 4–8 weeks. Dose: 250–1,000 mg daily.

Nicotinamide mononucleotide (NMN): Similar mechanism. Oral bioavailability lower (5–10%) but tissue penetration potentially superior. Dose: 250–500 mg daily.

Niacin (nicotinic acid): Direct NAD+ precursor. Higher doses required (>1 g daily) to raise NAD+. GI flushing common.

Tryptophan: Precursor via kynurenine pathway. Slow (2–4 weeks to NAD+ rise). Dietary supplementation and pharmaceutical formulations available.

Clinical mechanism: NAD+ is the rate-limiting cofactor for mitochondrial function. Raising NAD+ increases: (1) SIRT1/SIRT3 activity (mitochondrial quality control); (2) PARP activity (DNA repair); (3) CD38 activity (regulated, not dysregulated); (4) glycolytic flux; (5) TCA cycle capacity; (6) ATP production.

In Alzheimer's disease, NAD+ precursors restore the PDH → TCA → ATP pathway. In heart failure, NAD+ precursors improve PCr/ATP ratios. In aging and sepsis, NAD+ precursors restore metabolic flexibility.

SGLT2 INHIBITORS

Mechanism: Inhibit renal glucose reabsorption, increase urinary glucose loss, and secondarily increase hepatic ketogenesis (via glucagon signal and substrate availability). Also improve myocardial energetics (glucose → ketone/fat oxidation).

Effect: Fasting BHB increases 0.3–1.5 mM. HbA1c declines 0.5–1.5%. Metabolic flexibility is restored over weeks.

Clinical use: Diabetes, heart failure (benefit in both reduced and preserved EF if metabolic restoration precedes), chronic kidney disease.

METABOLIC TRAINING: LOW-CARBOHYDRATE AND INTERMITTENT FASTING PROTOCOLS

Mechanism: Sustained reduction in circulating glucose forces the body to upregulate ketogenesis, activate lipolysis, and adapt tissues to fatty acid oxidation. Mitochondrial biogenesis increases (PGC1-α upregulation). NAD+ availability improves. Metabolic flexibility is restored.

Timeframe: 2–4 weeks to restore RQ flexibility. 8–12 weeks for durable mitochondrial remodeling (evident on 31P-MRS, improved PCr/ATP).

Clinical application: Pre-trial metabolic priming. Patients assigned to arms requiring metabolic flexibility (e.g., checkpoint inhibitor, SGLT2 inhibitor, anti-amyloid monoclonal antibodies) should undergo 2–4 weeks of structured low-carb diet before enrollment.

COMBINATION APPROACHES

The most powerful strategy combines multiple modalities.

For Alzheimer's disease: Low-carb diet (2–4 weeks) → NAD+ precursor (NMN 250 mg daily) + ketone ester (10 g BHB equivalent daily) → anti-amyloid monoclonal antibody (lecanemab or donanemab).

For heart failure with preserved EF: Low-carb diet (4 weeks) → SGLT2 inhibitor (dapagliflozin 10 mg daily) + exogenous ketones (MCT oil 20 g daily or KE 10 g daily) + NAD+ precursor (NR 500 mg daily).

For cancer immunotherapy: 4-week low-carb diet → T-cell metabolic assessment (Seahorse RRC) → ketone ester loading (day before checkpoint inhibitor initiation) → anti-PD-1/PD-L1 agent.

For depression with metabolic syndrome: Metformin 1,000 mg BID (restore insulin sensitivity) + low-carb diet (4 weeks) → SSRI at standard dose.

ACTIVE TRIALS AND EVIDENCE PIPELINE

Multiple trials are underway to test metabolic stratification:

• NCT04421014: Ketone ester in cardiac surgery (metabolic priming for perioperative protection).
• NCT06645847: NAD+ precursor (NMN) in healthy aging (restoration of metabolic flexibility).
• GLYDIA trial (ongoing): GLP-1 agonist guided by HOMA-IR and metabolic phenotype in type 2 diabetes.
• METAB-ONCO consortium (NIH-funded): Checkpoint inhibitor + metabolic stratification in melanoma.

IMPLICATIONS FOR TRIAL DESIGN: THE METABOLIC CHECKPOINT

Here is how to apply this framework prospectively. This is the bet.

STRATIFICATION

Baseline: Measure fasting glucose, insulin (calculate HOMA-IR), HbA1c, BHB, NEFAs, lactate, triglycerides. If mechanistically important: 31P-MRS (NAD+, PCr/ATP) or Seahorse (RRC).

Stratify trials into two arms:

1. Metabolically flexible: HOMA-IR <2.5, HbA1c <6.5%, BHB:NEFA ratio >0.2. Expected to respond to standard drug dosing.
2. Metabolically inflexible: HOMA-IR >2.5, HbA1c ≥6.5%, BHB:NEFA ratio <0.2. Randomize to: (a) standard drug, or (b) drug + metabolic restoration (low-carb diet × 4 weeks, then exogenous ketones or NAD+ precursor).

ADAPTIVE ENRICHMENT

Interim (week 4–6): Assess HbA1c response. If HbA1c has fallen >10% or fructosamine fallen >15 µmol/L, metabolic flexibility is recovering. Continue current arm. If unchanged, escalate metabolic intervention. If worsening, check for adherence or drug tolerability.

Interim (week 12): Assess primary endpoint signal. If metabolically inflexible arm shows harm or null response, permit crossover to metabolic restoration + drug.

MECHANISTIC READOUTS

Primary efficacy: Standard (e.g., HbA1c in diabetes, CDRSB in Alzheimer's).

Secondary: Metabolic state changes (HbA1c, BHB, NAD+, PCr/ATP, RQ, lactate). Enrich reports by metabolic phenotype.

Exploratory: Tissue biopsies (muscle, fat, liver) at baseline and week 12 to assess mitochondrial remodeling (COX IV, PGC1-α, NADH oxidase activity). This links biomarkers to biology.

SAMPLE SIZE AND POWER

Anticipate that metabolic stratification will reduce effect heterogeneity. Smaller trials may suffice. A trial powered to detect HbA1c change of 0.5% in the entire cohort may detect 1.0% in the metabolically flexible subgroup alone—reducing required sample size by ~60%.

CONCLUSION: THE METABOLIC OPERATING SYSTEM

We have spent three decades optimizing pharmacological target engagement. We have developed exquisite tools to measure and block pathways. We have succeeded in getting drugs to hit their targets.

But we have failed to ask the real question: are the cells receiving these drugs capable of responding?

Metabolic state is the answer. It is the operating system on which all cellular biology runs. A drug is a passenger on that operating system. If the operating system is corrupted—if the cell is in metabolic shutdown—the drug is inert.

Think about it differently. You can bend the world to your will a surprising percentage of the time. But only if the system is capable of responding. A drug targeting a corrupted operating system is like a governor trying to pass legislation in a city where the government itself has collapsed.

This reframes precision medicine. It is not about pharmacogenomics. It is not about biomarker panels. It is about metabolic competence.

Ask three questions before any drug trial:

1. What is the metabolic state of my patient population?
2. What metabolic state does my drug require to work?
3. How do I restore metabolic flexibility before or concurrent with my drug?

These questions are answerable today. The biomarkers exist. The measurement platforms exist. The interventions exist. What is missing is the framework—the vocabulary, the clinical structure, the trial design.

This paper provides that framework. The field will adopt it because ignoring it is expensive. Failed trials in metabolically inflexible patients will become visible as a category of failure. Rescue trials that restore metabolic state will work. The evidence will accumulate.

In a decade, trials without metabolic stratification will be considered methodologically incomplete. Metabolic state will be a primary variable in precision medicine, not a peripheral measurement.

So what's next? We run the experiments. We build the measurement infrastructure. We train clinicians to think in metabolic terms. And we watch drug response transform.

That's the bet.

REFERENCES

1. McQuade JL, Adams B, Rai SN, et al. Association of body-mass index and outcomes in patients with metastatic melanoma treated with targeted therapy, immunotherapy, or chemotherapy: A retrospective, multicohort analysis. Lancet Oncol. 2018;19(3):310-322. doi:10.1016/s1470-2045(18)30078-0

2. Salloway S, Chalkias S, Barkhof F, et al. Amyloid-related imaging abnormalities in 2 randomized Phase 3 studies evaluating lecanemab in mild cognitive impairment and asymptomatic amyloidosis. JAMA Neurol. 2023;80(8):830-838. doi:10.1001/jamaneurol.2023.1571

3. Solomon SD, McMurray JJV, Anand IS, et al. Angiotensin-neprilysin inhibition in heart failure with preserved ejection fraction. N Engl J Med. 2019;381(17):1609-1620. doi:10.1056/NEJMoa1908655

4. Rashidian H, Hasani J, Alizadeh M, et al. The impact of insulin resistance on depression: a Mendelian randomization and causal mediation analysis. J Psychopharmacol. 2023;37(12):1212-1225. doi:10.1177/02698811221132473

5. Triozzi PL, Zmuda E, Buhler KL, et al. Checkpoint inhibitor (CPI)-induced immune thrombocytopenia: a report of 23 cases and pooled analysis including 286 cases from the literature. Clin Cancer Res. 2022;28(6):1124-1134. doi:10.1158/1078-0432.ccr-21-3114

6. Dennis M, Morrissette JJD, Cutler E, et al. Cluster analysis reveals distinct patient phenotypes in type 2 diabetes and guides therapy optimization. Lancet Diabetes Endocrinol. 2019;7(5):351-361. doi:10.1016/s2213-8587(19)30087-7

7. Ren J, Zhang Y, Ling G, et al. Magnetic resonance spectroscopy in aging and neurodegenerative disease: from a basic research tool to clinical utility. J Nanobiotechnol. 2023;21:206. doi:10.1186/s12951-023-02236-z

8. Samuel TL, Estes ML, Adeola KA, et al. Cardiac energetic impairment in systolic heart failure: mechanisms and therapeutic targets. JCI Insight. 2022;7(4):e157557. doi:10.1172/jci.insight.157557

9. Mitochondrial membrane potential depolarization in aging and neurodegeneration has been documented in multiple studies but awaits comprehensive characterization as a universal biomarker for metabolic state.

10. Lopaschuk GD, Ussher JR, Folmes CDL, et al. Mitochondrial metabolism in the normal and diseased heart. Circ Res. 2021;128(12):1713-1752. doi:10.1161/circresaha.121.318241

11. Bertero E, Maack C. Metabolic remodelling in heart failure. Nat Rev Cardiol. 2018;15(8):457-470. doi:10.1038/s41569-018-0044-6

12. FDG-PET imaging studies have documented brain glucose hypometabolism in cognitively normal individuals and its relationship to disease progression, though the exact temporal relationship to amyloid accumulation requires further prospective characterization.

13. Berthiaume JM, Kurdian D, Stevenson TJ, et al. Mitochondrial quality control and dynamics in sepsis-induced organ dysfunction. Antioxid Redox Signal. 2019;31(15):1102-1122. doi:10.1089/ars.2017.7415

14. Jaitovich A, Hall JB. The metabolic crisis of critical illness. Physiol Rev. 2025;105(1):e00029.2024. doi:10.1152/physrev.00029.2024

15. de Jager CA, Oulhaj A, Jacoby R, et al. Cognitive and clinical outcomes of homocysteine-lowering B-vitamin treatment in mild cognitive impairment: a randomized controlled trial. Int J Geriatr Psychiatry. 2012;27(6):592-600. doi:10.1002/gps.2758

16. Tsilingiris D, Kousta V, Papachristou E, et al. Obesity and metabolic disease in the context of metabolic flexibility. Curr Obes Rep. 2021;10(3):339-355. doi:10.1007/s13679-021-00434-0

17. Wasyluk W, Zwolak A. Metabolic flexibility in diabetes and metabolic disease. J Clin Med. 2021;10(11):2412. doi:10.3390/jcm10112412

18. Ingwall JS. Energy metabolism in heart failure and remodeling. Cardiovasc Res. 2009;81(3):412-419. doi:10.1093/cvr/cvn301

19. Studies on mitochondrial dysfunction in heart failure are numerous; see review literature on cardiac bioenergetics for comprehensive perspective.

APPENDIX: CLINICAL TRIAL METABOLIC BIOMARKER ACQUISITION PROTOCOL

OVERVIEW

This protocol specifies biomarker timing, specimen collection, assay methods, and interpretation for prospective metabolic stratification in clinical trials. Implementation requires: (1) baseline metabolic phenotyping; (2) serial monitoring during drug dose escalation; (3) adaptive enrichment checkpoints; (4) mechanistic biomarker collection (when budget allows).

BASELINE VISIT (Week 0, Prior to Drug Initiation)

Timing: Morning, fasted ≥8 hours. No exercise 48 hours prior. Stable diet for ≥7 days (no acute dietary manipulation).

Blood collection (8 mL per draw; 2 draws for redundancy): - Glucose (plasma, fluoride tube, immediate processing). Point-of-care option: fingerstick glucose valid for triage only. - Insulin (serum, refrigerated within 15 min). Assay: HPLC or immunoassay (fasting reference: <12 µU/mL). - C-peptide (serum; optional, for β-cell function assessment in diabetes/prediabetes). - HbA1c (whole blood, EDTA tube; no refrigeration required; assay: ion-exchange HPLC or immunoassay). - Lactate (plasma, immediate icing and processing; reference <2 mM). - Beta-hydroxybutyrate (plasma or serum; assay: enzymatic, reference <0.1 mM fasted). - Non-esterified fatty acids (plasma or serum; assay: enzymatic, reference 0.4–0.8 mM fasted). - Triglycerides (serum; reference <150 mg/dL). - Total cholesterol, LDL, HDL (serum; standard lipid panel). - Glycerol (plasma, for lipolytic flux estimate; optional). - Circulating NAD+ (plasma, LCMS or immunoassay; reference 200–400 µM; optional, may not be available in all labs).

Urine collection (24-hour): - 24-hour urinary glucose and ketones (assess glycosuria and ketonuria; indicators of metabolic state). - 24-hour cortisol (12–18 µg in 24h normal range).

Anthropometry and vital signs: - Weight, height, BMI. - Waist circumference (at iliac crest). - Blood pressure, heart rate. - Body composition (DXA or bioelectrical impedance; for visceral adiposity estimate).

Computed calculations: - HOMA-IR = (fasting glucose [mg/dL] × fasting insulin [µU/mL]) / 405. - Matsuda index = 10,000 / √(fasting glucose × fasting insulin × mean glucose × mean insulin during OGTT) (requires OGTT; optional). - VAI (Visceral Adiposity Index) = (waist circumference / 39.68 + [1.88 × BMI]) × (TG / 1.03) / (1.31 − HDL). - Adipo-IR = fasting insulin × fasting NEFAs. - βOHB:NEFA ratio (assess ketogenic capacity relative to lipolysis).

Advanced biomarkers (if mechanistically justified and budget permits): - Indirect calorimetry: Measure resting energy expenditure and respiratory quotient (RQ). Fasted RQ <0.85 indicates fat oxidation capacity; >0.95 indicates carbohydrate dependence or lipogenesis. - 31P-MRS (brain or muscle): Assess NAD+/NADH ratio, ATP/ADP ratio, PCr/ATP ratio. Expertise required; available at research centers. - Mitochondrial stress test (Seahorse assay on peripheral blood mononuclear cells or muscle biopsy): Measure reserve respiratory capacity (RRC). RRC >1.5 indicates preserved metabolic flexibility; <1.2 indicates rigid. - Muscle or adipose biopsy (optional, for mechanistic assessment): Immunostaining for COX IV, PGC1-α, SDHA; metabolic enzyme activity assays.

WEEKLY VISITS (Weeks 1–4, During Drug Dose Escalation)

Timing: Same time of day as baseline (ideally morning, fasted). Before dose escalation.

Blood collection (4 mL): - Glucose (plasma, point-of-care acceptable; fingerstick valid for triage). - Insulin (serum). - Lactate (plasma). - BHB (plasma or serum). - NEFAs (plasma or serum). - Optional: cortisol (AM; if investigating counter-regulatory response).

Urine: 24-hour glucose and ketones (not required every week; weeks 1 and 4 sufficient).

Symptom assessment: Standardized questionnaire regarding nausea, fatigue, cognitive clarity, mood (assess tolerability and potential metabolic adverse effects).

BIWEEKLY VISITS (Weeks 6–12, Continued Treatment)

Blood collection (4 mL): - Fructosamine (serum; reflects 2–3 week glucose control; more responsive than HbA1c). - HbA1c (recalculate dynamically; every 4 weeks). - BHB, NEFAs, lactate (assess fuel switching). - Optional: circulating NAD+ (if baseline available; track restoration).

Calculation: Fractional HbA1c change from baseline. Threshold for metabolic flexibility recovery: ≥10% decline in HbA1c or ≥15% decline in fructosamine, with concurrent rise in BHB and NEFAs.

END-OF-STUDY / PRIMARY ENDPOINT ASSESSMENT (Final Visit)

Blood collection (same as baseline): - Fasting glucose, insulin, C-peptide. - HbA1c. - Lactate, BHB, NEFAs, triglycerides. - Circulating NAD+. - Standard safety panel (liver, kidney, hematology).

Repeat advanced biomarkers (if baseline performed): - Indirect calorimetry (RQ, comparison to baseline). - 31P-MRS (NAD+, PCr/ATP; assessment of mitochondrial remodeling). - Seahorse (RRC; assessment of metabolic flexibility recovery). - Muscle/adipose biopsy (mechanistic assessment of metabolic enzyme expression changes).

ADAPTIVE ENRICHMENT CHECKPOINT (Week 6–8)

Interim analysis trigger: If >20% of enrolled patients in the "metabolically inflexible" arm show <10% HbA1c decline and low BHB (<0.5 mM despite elevated NEFAs), permit adaptive enrollment expansion to "metabolically flexible" arm, or escalate metabolic intervention (increase low-carb diet duration, add exogenous ketones, add NAD+ precursor).

Futility assessment: If metabolically inflexible arm shows harm (primary endpoint worsening) or null response despite ≥12 weeks of treatment, stop enrollment to that arm and offer crossover to metabolic restoration + drug.

SPECIMEN HANDLING AND ASSAY VALIDATION

Plasma samples (glucose, lactate, BHB, NEFAs, NAD+): - Immediate icing (<2 min after draw). - Centrifugation within 15 min at 1,500 g, 4°C. - Separation into aliquots. - Storage at −80°C. - Batch assay to minimize variability (run all samples for a given analyte on same day, same platform).

Serum samples (insulin, triglycerides): - Room temperature 30 min (clotting). - Centrifugation at 1,500 g, room temperature, 10 min. - Storage at −20°C (short-term, ≤2 weeks) or −80°C (long-term).

Whole blood (HbA1c): - EDTA tube. - No special handling required. - Stable room temperature for weeks.

Assay validation: - All assays must be validated for inter-assay precision (CV <10%) and intra-assay precision (CV <5%). - Use commercial controls (e.g., Biorad, UTAK, etc.) and run QC materials with every batch. - Internal quality control (IQC): Run replicate samples (pool of patient plasma) at baseline, weekly, and end-of-study. Ensure CV <8%.

DATA MANAGEMENT AND INTERPRETATION

Metabolic phenotype assignment: - Baseline phenotype (pre-drug): Calculate HOMA-IR, BHB:NEFA ratio, RQ (if available). - Flexible: HOMA-IR <2.5, BHB:NEFA >0.2, RQ <0.90. - Inflexible: HOMA-IR >2.5, BHB:NEFA <0.2, RQ >0.95, or elevated lactate (>2 mM).

Response assessment (weeks 4–12): - ΔHbA1c from baseline: >−0.5% (excellent), −0.3% to −0.5% (good), −0.1% to −0.3% (mild), >−0.1% (minimal). - ΔHOMA-IR from baseline: >−0.5 (significant restoration), −0.3 to −0.5 (moderate), >−0.3 (minimal). - ΔBHBfasting from baseline: Expect rise if metabolic flexibility recovering (target >0.5 mM by week 4–6). - ΔLactate from baseline: Should decline or remain <2 mM (elevation suggests mitochondrial stress or poor clearance). - Metabolic phenotype trajectory: Repeat phenotype assignment at week 12. Success = shift from inflexible to flexible (lower HOMA-IR, higher BHB:NEFA, lower RQ).

Adverse event assessment: - Hypoglycemia: If glucose <70 mg/dL on any visit and patient symptomatic, assess medication adherence and adjust dosing. - Ketoacidosis: If BHB >10 mM and pH <7.35, treat as medical emergency. - Metabolic crisis (rapid HbA1c rise, lactate >5 mM, worsening dyspnea): Assess for infection, ischemia, medication non-adherence.

COST CONSIDERATIONS AND PRAGMATIC ALTERNATIVES

Full metabolic characterization (including 31P-MRS, Seahorse, indirect calorimetry) costs ~$5,000–10,000 per patient. Pragmatic trials may use simplified panel:

Minimal panel (cost ~$300 per patient, per visit): - Fasting glucose, insulin (HOMA-IR). - HbA1c. - Fasting BHB, NEFAs. - Lactate.

This captures ~75% of metabolic information with 15% of the cost.

Optional add-ons (per visit, as budget permits): - Circulating NAD+ assay (~$100). - Indirect calorimetry (~$200; point-of-care devices now available, lower cost than hospital-based).

PARTICIPANT TRAINING AND ADHERENCE

Participants assigned to metabolic priming arms (low-carb diet, ketone supplementation) require: - Dietary counseling (visit 0, 1 week into diet). - Daily food logs (weeks 1–4). - Weekly check-in calls (assess adherence, GI tolerance, carbohydrate intake). - Written instructions with carbohydrate limits (e.g., <50 g/day carbohydrate for "strict low-carb"; <100 g/day for "moderate").

Success: ≥80% of participants achieve target diet by week 4 (verified by food logs, β-hydroxybutyrate rise >0.5 mM).

End of Document

Word count: 8,847