375 - The ketogenic diet, ketosis, and hyperbaric oxygen: weight loss, cognition, cancer, and more

KEYPOINTS

• 01

  Dom D'Agostino came to ketosis through Navy oxygen-toxicity research

  Navy SEALs use closed-circuit rebreathers with near-100% oxygen to avoid bubble trails, but below about 50 feet oxygen toxicity becomes possible and at ~100 feet a diver may seize within minutes. Oxygen acts as a stimulant: it drives glutamate release, inhibits the enzyme converting glutamate to GABA and bursts reactive oxygen species. Because ketogenic diets control seizures when drugs fail, D'Agostino pivoted to ketones; Duke trials now push volunteers in chambers to an EEG seizure and compare latency under dietary, supplemental or combined ketosis.

• 02

  Nutritional ketosis is defined by blood BHB above 0.5 mM

  Fasting suppresses insulin, mobilizing fatty acids that the liver turns into beta-hydroxybutyrate and acetoacetate, which fuel the brain. The original epilepsy diet was ~90% fat; modern versions are 60–70% fat with more protein, because restricting protein stunted kids' growth. D'Agostino calls it the most researched diet with an objective biomarker, with 'pleiotropic' effects on metabolism, neuropharmacology and epigenetics.

• 03

  Ketogenic therapy works where seizure drugs fail

  In drug-resistant pediatric epilepsy about two-thirds respond; of those roughly a third get 95–100% control, and 10–15% are 'super responders' who stay seizure-free after one or two years on the diet. Adults respond less (~30–40%), partly adherence. In brain slices, sodium BHB silenced evoked seizure activity and made neurons more resilient to toxins.

• 04

  For weight loss: high protein, fiber-carbs, and count calories

  D'Agostino's mistakes list starts with not tracking. He recommends 20–30% protein (he eats up to ~220 g at 220 lb), up to 50–100 g carbs if about a third is fiber (broccoli, cauliflower, avocado, cucumber, asparagus — not carrots, tomatoes or most peppers), and a 10–20% calorie deficit. Tracking apps showed him eating ~4,000 kcal when he believed 3,200; 'calories don't matter on keto' is false.

• 05

  Watch lipids and certain genotypes

  A strict keto diet sent his LDL and apoB soaring; he has an NPC1L1 variant, so a small dose of ezetimibe controls it. He's seen LDL and triglycerides rise and HDL fall in some people, especially APOE4 homozygotes, and urges tracking ApoB, LDL, triglycerides, HDL, A1c and insulin. Early on, cutting protein from ~250 g to 70–80 g cost him strength.

• 06

  Esters, salts and 1,3-butanediol are not the same thing

  1,3-butanediol, a NASA-era food preservative, is metabolized by alcohol dehydrogenase like ethanol; ~150–200 mL/day for therapeutic ketosis doubled his liver enzymes over two weeks and can make people 'drunk stupid'. Esters bind BHB or acetoacetate to butanediol for a fast double peak. Salts ionically bind BHB to sodium, potassium, calcium and magnesium; he argues hypertension risk is tied to sodium chloride, not sodium BHB.

• 07

  Rapid ketone spikes trigger insulin and crashes

  Rising above about 1.5–2 mM quickly prompts a counter-regulatory insulin release; two hours after a big ester dose people can be both hypoglycemic and low in ketones. Salts don't produce that effect in his experience. He says higher isn't better: above 2–3 mM from supplements approaches 'energy toxicity' and acidosis, and animals above 6–7 mM on some esters died. 1 mM BHB supplies roughly 10% more brain energy.

• 08

  D- and L-BHB behave differently

  Meters and continuous ketone monitors measure only D-BHB. D is burned quickly for fuel; L is metabolized three to four times slower, accumulates in the brain and retains signaling effects such as NLRP3 inflammasome suppression. Pure D-BHB monoester didn't help seizures in his models; a 1:1 BHB-acetoacetate mix (redox balance) did.

• 09

  Cancer: a metabolic adjunct, still lacking trials

  A 49-author framework led by Thomas Seyfried proposes a glucose-ketone index of 1–4 plus drugs targeting glucose and glutamine for glioblastoma. Pressed by Peter, D'Agostino concedes it won't cure GBM, the goal is extending survival, and there are no RCTs; funding now favors adjuncts to CAR-T and checkpoint inhibitors. Peter suggests pancreatic cancer may be a better first model.

• 10

  Alzheimer's: a fuel-rescue rationale with patient selection

  Brain glucose use declines with age while ketone uptake is preserved (Stephen Cunnane's dual PET work). Acute ketone elevation improves cognition in impaired people, but long-term randomized amyloid outcomes are pending; D'Agostino would select patients with marked glucose hypometabolism.