Should You Consider a GLP-1? Metabolic Health, Microdosing, and What Semaglutide, Tirzepatide and Retatrutide Actually Do

Why we are writing this now

Two years ago the conversation was about weight: whether these drugs worked, whether using them was cheating, whether the side effects were worth it. That is largely settled. They work, they are not cheating, and the side effects are real but manageable in the right hands.

The conversation that replaced it is far more useful. It is about metabolic health — the observation that when you improve metabolic function, a remarkable number of problems people were told to live with begin to resolve, and that a meaningful share of that benefit shows up whether or not the scale moves much.

We are not primarily interested in whether you weigh less. We are interested in whether your fasting insulin comes down, your inflammation falls, your visceral fat shrinks, your joints stop aching, and your ten-year risk of the diseases that actually kill people goes down. Sometimes a GLP-1 is the right tool for that. Sometimes it is emphatically the wrong one. And often the right dose is a fraction of what an online clinic will sell you.

It is also worth naming the scale of the problem these medications arrived into. A generation of prediabetes was on course to become a generation of diabetes, dialysis and cardiovascular disease, and there were never going to be enough clinicians or enough money to care for all of it. Hospital medicine is full of people who cycle in and out having their symptoms managed while the disease underneath is never actually addressed. Whatever else is true about these drugs, they are the first intervention in decades with a realistic chance of reaching that underlying problem at a population level. That is worth acknowledging even while we argue about how they should be used.

And a word about the people taking them. In an ideal world everyone would get morning light, eat unprocessed food, sleep well and never become metabolically ill. In the real world people end up in a hole — and you cannot always climb out of a hole the way you fell into it. Telling someone who has gained sixty pounds over fifteen years to simply reverse the process is often not useful advice. Sometimes a different route out is required. That is not a moral failure on their part, and we do not treat it as one.

The question is not “do I want to lose weight?” The question is “is my metabolism driving me toward the diseases I am most afraid of — and what is the least invasive thing that changes that trajectory?”

How we decide whether to prescribe one at all

We use these medications thoughtfully and selectively, as one tool inside a larger plan rather than as a default first move.

Where we reach for a GLP-1 is when there is a metabolic component that is not being well managed by the things that should manage it first:

• Diet — real nutritional work, not a vague instruction to eat better. Protein, meal sequencing, removing the foods that are driving the glucose curve.
• Exercise — resistance training in particular, because muscle is the largest glucose sink you own.
• Circadian rhythm and sleep — light exposure, consistent timing, adequate duration. Short sleep manufactures insulin resistance on its own.
• Gut function — dysbiosis, intestinal permeability and SIBO all degrade the cells that make your own GLP-1. Fixing this resolves the problem outright for some patients.
• Stress load — because cortisol drives the same insulin resistance we are trying to treat.
• Natural GLP-1 support — the dietary and microbiome strategies in section 14, which have real if modest evidence behind them.

The three kinds of patient who come to us about this

It may help to say plainly who actually walks through our door asking about GLP-1s, because the right answer is different for each of them.

When the foundational work has been done properly and the markers still will not move, that is when we look at low-dose GLP-1 therapy — and “low-dose” is meant literally. We routinely use a fraction of the labeled doses and hold there, rather than climbing a ladder built for a different population.

That is also the practical reason we compound. A vial and a syringe let us dose precisely where a patient needs to be and titrate in small increments. A fixed-dose pen cannot do that.

Look at what these drugs are being tested for

Here is the argument we find most persuasive, and it does not come from any single trial. It comes from stepping back and looking at the shape of the entire research program.

As of August 2026, ClinicalTrials.gov lists 1,668 registered interventional studies of the six established GLP-1-based medications, with a further 226 for the newer incretin agents. 450 of them are still running. Semaglutide alone accounts for 606.³

The interesting part is not the number. It is the range. When we enumerated all 606 semaglutide trials and sorted them by condition, they fell into 41 distinct disease domains, and 37.5% of them named a condition outside the obesity, diabetes and metabolic-syndrome core altogether.³ Cardiovascular disease. Kidney disease. Fatty liver. Heart failure. Sleep apnea. Alcohol use disorder. Osteoarthritis. Peripheral artery disease. Polycystic ovary syndrome. Psoriasis. Atrial fibrillation. Cancer. Bone. Neuropathy. Sarcopenia. Glaucoma. Systemic sclerosis. Even trials registered under headings like aging and longevity.

Ask yourself what it would take for one molecule to be a plausible candidate in that many unrelated-looking diseases. Either the drug is a miracle, or those diseases are less unrelated than the way we file them suggests.

We think it is the second thing. And we think the common thread is metabolic.

And these are not speculative associations. The conditions that keep showing benefit are ones in which metabolic dysfunction is an established driver: cardiovascular disease, chronic kidney disease, heart failure with preserved ejection fraction, metabolic fatty liver, obstructive sleep apnea, peripheral artery disease, weight-bearing joint disease. In each of those, large randomized trials have now shown improvement on hard endpoints — not just weight.^{4,5,6,7,8,9,10} Section 4 gives the headline numbers.

Not everything has worked. The trials in Alzheimer’s and Parkinson’s disease were negative, and we cover that honestly further down.^11,12 But the pattern of what works and what does not is itself informative: the benefit shows up along one axis, and that axis is metabolic.

The most parsimonious reading of the evidence is that a large share of what we call separate chronic diseases are downstream expressions of the same metabolic dysfunction — which is why one intervention aimed at that dysfunction keeps producing benefit in organs that look unrelated on paper. That has been the working assumption of functional medicine for decades. What is new is that we can now point at hard endpoints in tens of thousands of randomized patients.

It is also why the rest of this article spends so much time on your metabolism — your fasting insulin, your visceral fat, your inflammation, your muscle — rather than on the drug. The drug is one way to move that axis. It is not the axis.

1. Metabolic health is the root, and almost nobody has it

Start with the number that reframes everything. When researchers took a nationally representative sample of American adults and asked a simple question — how many meet all five criteria for optimal metabolic health, without medication — the answer was 12.2%.¹³ About one in eight.

The criteria were not exotic — waist circumference, fasting glucose and A1c, blood pressure, triglycerides and HDL, all without medication. Among adults with obesity, 0.5% met all five. Among adults at a normal weight, fewer than a third did.¹³

A stricter analysis, adding the absence of established cardiovascular disease, put the figure at 6.8% in 2017–2018 and falling. The component that collapsed hardest was glucose regulation, from 59.4% of adults to 36.9%.¹⁴

One honesty note, since the figure gets stretched: this measures optimal health on a strict, medication-free threshold. Someone whose blood pressure is well controlled on one medication counts as not optimal. The correct reading is that one in eight Americans is metabolically excellent — alarming enough without embellishment.

Where that leaves us

The CDC’s most recent count puts 40.1 million Americans — 12.0% of the population — in the diabetes category, with more than a quarter of them undiagnosed. Another 115.2 million adults, more than two in five, have prediabetes, and roughly eight in ten do not know it.¹⁵

Now narrow that to the group this article is addressed to: among adults 65 and older, 52.1% have prediabetes.¹⁵ A coin flip. That is the strongest argument there is for why metabolic screening should be routine after fifty, and is not.

The diseases you are most afraid of run through the same door

Of the ten leading causes of death in the United States, four are conditions in which insulin resistance, visceral adiposity and metabolic inflammation play a major driving role — heart disease, stroke, diabetes and kidney disease — and three more, cancer, Alzheimer’s and chronic liver disease, are conditions where they contribute.¹⁶

The associations are not subtle. In a meta-analysis of 65 studies covering 516,325 people without diabetes, those in the highest insulin-resistance category had a 64% higher risk of cardiovascular events.¹⁷ In a national cohort, a fasting insulin at or above 10 µU/mL was associated with roughly double the risk of cancer death — and that held in people who were not obese.¹⁸

That last point is the recurring theme of metabolic medicine: the problem is not the number on the scale, it is what the tissue is doing. One correction worth making, since fasting insulin gets promoted as the cardiovascular marker: in that same meta-analysis, fasting insulin alone did not predict coronary disease.¹⁷ HOMA-IR, which combines it with fasting glucose, is the number that carries the outcome data. We still order fasting insulin — it is the input HOMA-IR needs, and it moves early.

The problem starts long before the diagnosis

The Whitehall II cohort followed 6,538 people and reconstructed the trajectory of the 505 who developed type 2 diabetes. Insulin sensitivity was already falling steeply about five years before diagnosis. Beta-cell function first rose — the pancreas compensating — then collapsed. Fasting glucose only turned sharply upward in the final three years.¹⁹

So there is a window, measured in years, when the compensation is visible in insulin while the glucose numbers your doctor is watching still look fine. Measure only glucose and A1c and you find out late. That is the entire clinical argument for measuring insulin.

Muscle insulin resistance runs earlier still — detectable decades before overt hyperglycemia.²⁰ Muscle matters here because under experimental conditions it accounts for roughly 85% of insulin-stimulated glucose disposal.²¹ It is the largest glucose sink you own.

Which is why the single most important thing you can do for your metabolic health has nothing to do with a prescription pad. It is building and keeping muscle.

Why the scale is a poor instrument

In a national sample, 23.5% of normal-weight American adults were metabolically abnormal, and 31.7% of adults with obesity were metabolically healthy.²²

The clinical implication is direct. A slim 60-year-old with a fasting insulin of 14, triglycerides of 180 and central fat is at real risk, and will usually be told her labs are fine. A 250-pound man with a fasting insulin of 4 and clean lipids is in better metabolic shape than she is. If the only screen anyone runs is a scale and a fasting glucose, both get missed.

Which brings us to the reason this article spends so much of its length on small doses. The problem we are usually treating is not visible obesity. It is a hidden metabolic problem in someone who looks well — and the dose required to move a hidden problem is not the dose required to produce 20% weight loss.

2. What a GLP-1 actually is

Glucagon-like peptide-1 is not a drug. It is a hormone your own body makes. It is produced by specialized L cells lining the lower small intestine and colon, and also in the brainstem, and it is released in response to nutrients arriving in the gut.

A peptide is simply a short chain of amino acids — the same building blocks as protein, assembled into a signaling molecule rather than structural tissue. Your native GLP-1 is 30-odd amino acids long, and it is destroyed within a couple of minutes by an enzyme called DPP-4. That brief pulse is the point: it is a meal signal, not a standing instruction.

What that signal does

• It amplifies insulin release — but only when glucose is elevated. This glucose-dependence is why GLP-1 medications, used alone, rarely cause hypoglycemia. We will return to this in the sleep section, because it matters more than people realize.
• It suppresses glucagon after a meal, reducing the liver’s output of glucose at exactly the moment you do not need more.
• It slows gastric emptying, flattening the post-meal glucose curve and prolonging fullness.
• It acts in the brain on regions governing satiety, reward and food-seeking — which is why appetite and, for many people, “food noise” change so dramatically.
• It has effects far outside the gut. GLP-1 receptors have been documented on macrophages, microglia, monocytes, neutrophils, eosinophils and several T-cell populations.²³ This is the basis for the anti-inflammatory effects seen in trials, and for a great deal of speculation that has run well ahead of the data.

How the drugs differ from the hormone

Pharmaceutical chemists made two changes to native GLP-1: they altered the sequence so DPP-4 cannot chew it up, and they attached a fatty acid chain that binds tightly to albumin in the blood, creating a slow-release depot. Semaglutide shares roughly 94% structural homology with human GLP-1 and has a half-life of about one week, with drug still circulating for around five weeks after the last dose.²⁴

That last point is worth sitting with. When you inject once weekly, you are not producing a pulse. You are producing continuous receptor occupancy at concentrations far above anything your body generates. Native GLP-1 operates in the picomolar range; steady-state semaglutide sits in the nanomolar range — roughly a thousandfold higher in molar terms, though the comparison is imperfect because semaglutide is more than 99% albumin-bound and is a modified analogue with different receptor kinetics.²⁵

This is the single most important fact for understanding the microdosing argument. These medications are not hormone replacement. At standard doses they are frankly pharmacologic. Whether every patient needs a frankly pharmacologic dose is the open question.

Two things that complicate the picture

Not everyone responds the same way, and some of that is genetic. A genome-wide analysis pooling observational data and large randomized trials found that variants in the GLP-1 receptor gene itself and in ARRB1 meaningfully shift response. Combining the two genes identified roughly 4% of the population with a 30% greater A1c reduction than the 9% with the poorest response.²⁶ That study measured glycemic response in type 2 diabetes rather than weight loss, so it should not be transferred wholesale — but it establishes the principle. Non-responders are real, and someone who fails on semaglutide may do well on tirzepatide.

Some common medications appear to suppress your own GLP-1. A 2024 study in Cell Metabolism followed 30 patients starting atorvastatin plus 10 controls and found that active GLP-1 concentrations fell significantly within four weeks. The proposed mechanism runs through the gut microbiome: statins reduced Clostridium species, lowering ursodeoxycholic acid, and both Clostridium transplantation and UDCA supplementation reversed the glucose intolerance in animal models.²⁷

3. Semaglutide, tirzepatide, retatrutide — what actually differs

These are not three versions of the same drug. They hit different receptors, and the differences are large enough to change who should take which.

Semaglutide — the single agonist

Semaglutide acts at the GLP-1 receptor and nothing else. It is sold as Ozempic (weekly injection, approved for type 2 diabetes), Wegovy (weekly injection, approved for weight management and for cardiovascular risk reduction), and Rybelsus (daily tablet for diabetes). In December 2025 the FDA approved a 25 mg daily oral tablet for weight management — the first oral GLP-1 for obesity, and a genuinely different product from the 2.4 mg injection despite sharing a molecule.²⁸

A newer development worth knowing about: orforglipron, marketed as Foundayo, is the first non-peptide GLP-1 receptor agonist available as an ordinary daily pill — no injection, no refrigeration, no food-timing restrictions. Its weight-loss effect looks broadly comparable to the first-generation injectables rather than to tirzepatide. The significance is access rather than potency, and for patients who will not inject, that matters a great deal.

Tirzepatide — the dual agonist, and a widely repeated error

Tirzepatide, sold as Mounjaro (diabetes) and Zepbound (weight management and, since 2024, moderate-to-severe obstructive sleep apnea with obesity), activates both the GLP-1 receptor and the GIP receptor. GIP is the other major incretin hormone, and adding it appears to improve fat handling. It also appears better tolerated: in the head-to-head trial, discontinuation for gastrointestinal reasons was 2.7% on tirzepatide versus 5.6% on semaglutide, despite substantially greater weight loss.²⁹

You will constantly read that tirzepatide has a “5:1 GIP-to-GLP-1 ratio.” That is a misreading of the source paper. In the original characterization, tirzepatide bound the GIP receptor with affinity comparable to native GIP, but bound the GLP-1 receptor roughly fivefold more weakly than native GLP-1 does — and was about thirteenfold weaker in signaling.³⁰ The “5” is a comparison of the drug against a natural hormone, not a ratio between two receptors. Taking the published binding constants at face value, tirzepatide’s own imbalance toward GIP is closer to thirtyfold.

The practical statement is simpler and correct: tirzepatide is a GIP-weighted dual agonist, not a balanced one. That is likely part of why it outperforms semaglutide.

Retatrutide — the triple agonist, not yet available

Retatrutide adds a third target: the glucagon receptor. That sounds paradoxical — glucagon raises blood sugar — but glucagon receptor agonism also increases energy expenditure and drives hepatic fat oxidation. In practice the combination produces the largest weight reductions yet recorded with a drug.

The phase 3 numbers, from Lilly’s own releases: TRIUMPH-1 (n=2,339, 80 weeks) produced mean weight reductions of 19.0% at 4 mg, 25.9% at 9 mg and 28.3% at 12 mg versus 2.2% for placebo.³¹ Two further phase 3 trials, in diabetes with obesity and in severe obesity with cardiovascular disease, reported 20.8% and 22.6% at the top dose.³²

Two distinctive signals from the retatrutide program: dose-dependent heart rate increases that peaked around 24 weeks then declined³⁵, and dysesthesia — abnormal skin sensation — in up to 12.5% of treated participants versus 0.9% on placebo.³¹ Neither appears with semaglutide or tirzepatide to the same degree.

What is genuinely interesting about retatrutide

The lipid effect is real and larger than people expect. Across the phase 2 programme retatrutide lowered LDL cholesterol dose-dependently — roughly 22% at the top dose — cut triglycerides by about 40%, and reduced apolipoprotein B by 21–24%, with large reductions in small LDL particles and triglyceride-rich lipoproteins. Systolic blood pressure fell 5–8 mmHg.^35,36 There is a plausible mechanism beyond weight loss: mouse work shows glucagon-receptor signaling accelerates the breakdown of PCSK9, the same protein targeted by the most potent cholesterol drugs we have.³⁷ That is mouse data, PCSK9 has never been measured in a person on retatrutide, and the lipid changes in trials cannot be separated from losing a quarter of your body weight. But it is a genuinely intriguing lead.

The muscle question — where the hope is running ahead of the data. There is real excitement in the training and biohacking world that retatrutide might spare muscle in a way the other molecules do not. We would love that to be true. It is not established. The only DXA data come from a substudy of the diabetes trial in which 103 people completed scans: total fat mass fell 26.1% at 8 mg versus 4.5% on placebo, and the investigators reported that the proportion of weight lost as lean mass was similar to other obesity treatments — neither better nor worse.³⁸ The phase 3 programme has reported no body composition data at all, and no head-to-head trial has compared retatrutide with semaglutide or tirzepatide on lean mass.

There is a mechanistic story behind the hope — the glucagon component stimulates FGF-21, and obesity appears to be an FGF-21-resistant state, so lean people might in theory extract a muscle-preserving effect that trial populations cannot. We find it interesting. We would also point out that the rodent evidence on FGF-21 and muscle mostly runs the other way, and that no one has published this chain. Treat it as a hypothesis worth testing, not a reason to choose the drug.

Side by side

The only true head-to-head

SURMOUNT-5 randomized 751 adults with obesity and without diabetes to maximum tolerated tirzepatide or maximum tolerated semaglutide for 72 weeks. Tirzepatide produced 20.2% weight reduction versus 13.7% for semaglutide — a 6.5 percentage point difference — with waist circumference falling 18.4 cm versus 13.0 cm.²⁹

Two caveats. The trial was open label, which is a real limitation when the primary endpoint is body weight. And note the comparison is at maximum tolerated doses, which tells you nothing about how the two molecules compare at low doses — a question nobody has studied.

Retatrutide’s apparent superiority over both is a cross-trial inference, not a head-to-head result. Different populations, different eras, different protocols. It is very probably the most potent of the three. That is not the same as having been demonstrated.

The dose ladders, for reference

Two things about these ladders that patients are almost never told. First, the starting dose of tirzepatide — 2.5 mg — is explicitly an initiation dose, not a maintenance dose. It was never studied as a stopping point. Second, the ladder is a protocol, not a destination. There is nothing in the pharmacology that requires you to climb it merely because four weeks have elapsed.

4. What the outcome trials show — beyond the scale

The intro gave you the scoreboard. This section is the short version of what sits behind it, plus the results we think are most worth understanding.

Weight loss trials tell you about weight. Outcome trials tell you whether people have fewer heart attacks, fewer hospitalizations, less kidney failure and less inflammation. Over the past three years these medications have accumulated a body of hard-endpoint evidence that very few interventions in medicine can match — and a few instructive failures.

The headline results

Cardiovascular. SELECT randomized 17,604 adults with established cardiovascular disease and no diabetes. Major cardiovascular events occurred in 6.5% versus 8.0%, hazard ratio 0.80.⁴ In absolute terms that is 1.5 percentage points over about four years — roughly 67 people treated for four years to prevent one event. A genuinely good preventive result, comparable to statins in similar populations, and the first time a weight medication has demonstrated one. Not a miracle. Both framings belong in the same sentence.

Kidney. FLOW enrolled 3,533 people with diabetes and chronic kidney disease and was stopped early for efficacy: the composite of kidney failure, major eGFR decline or kidney/cardiovascular death fell 24%, with death from cardiovascular causes down 29% and from any cause down 20%.⁵

Heart failure with preserved ejection fraction — a condition that has resisted almost everything. SUMMIT: events in 9.9% versus 15.3%.⁶ STEP-HFpEF: symptom scores up sharply, six-minute walk distance up 20 metres, and the number we find most interesting — C-reactive protein fell 43.5% versus 7.3% on placebo.⁴⁴

That CRP figure is not a weight-loss effect in different clothing. It is systemic inflammation measurably falling, and it is the best published evidence for what patients keep telling us: less pain, less swelling, better mornings.

Fatty liver — the one we think is most underappreciated

Metabolic fatty liver disease is now a leading cause of liver failure and a major risk factor for liver cancer, and the sequence is well understood: fat accumulates, fat drives inflammation, inflammation drives fibrosis, fibrosis becomes cirrhosis. It is also almost entirely silent until late, which is why we calculate a FIB-4 on everyone.

ESSENCE is the phase 3 trial: 1,197 people with biopsy-confirmed steatohepatitis and fibrosis. At the 72-week interim analysis, steatohepatitis resolved without worsening fibrosis in 62.9% of the semaglutide group versus 34.3% on placebo, with fibrosis improving in 36.8% versus 22.4%.⁷ In the phase 2 tirzepatide trial, resolution rates were 44%, 56% and 62% at 5, 10 and 15 mg versus 10% on placebo.⁴⁵

Those resolution figures are robust. The fibrosis numbers deserve more caution than they usually get — in the tirzepatide trial, fibrosis improvement was a secondary endpoint, was not adjusted for multiple comparisons, and did not reach conventional significance at the two higher doses in the registered analysis. And ESSENCE reports an interim look at part of its population; the outcomes that actually matter — cirrhosis, decompensation, death — do not read out until week 240.

Even with those caveats, this is a disease with essentially no drug treatment a few years ago, reversing on biopsy. If you have fatty liver, this is the section to bring to your next appointment.

Addiction — real, and smaller than the enthusiasm

Patients tell us they are drinking less without trying, or that the “food noise” has gone quiet. There is now controlled data behind the alcohol part: two randomized trials found meaningful reductions in heavy drinking, with the larger trial showing a drug-attributable difference of about 13.7 percentage points in heavy drinking days over placebo — noting that both arms also received therapy and both improved substantially.^46,47

One distinction worth having, because patients worry about it: this appears to act mainly on wanting rather than liking — two separate circuits in the brain. The compulsive pull quiets down; the capacity to enjoy a good meal or a glass of something on an occasion does not disappear. Most people describe it as relief rather than loss.

The effect is not universal across substances. A randomized trial of dulaglutide added to standard smoking-cessation care found no difference in abstinence (63% versus 65%), though it did prevent the usual post-quit weight gain.⁴⁸ No GLP-1 is approved for any addiction.

The failure that matters most: Alzheimer’s disease

Observational data had been encouraging. Then the trials read out. evoke and evoke+ randomized 3,808 people with confirmed early Alzheimer’s disease to oral semaglutide or placebo. Both were null on the primary cognitive endpoint, and the extension phases were discontinued.¹¹ A more rigorous observational analysis had already returned a null result before the trials reported.⁴⁹

The result that best captures our position

In STEP 3, every participant received intensive behavioral therapy — a low-calorie diet and 30 counseling visits. Placebo plus that program produced 5.7% weight loss. Semaglutide plus the same program produced 16.0%.⁵¹

Read it in both directions. Intensive lifestyle work alone delivered a real, clinically meaningful result. And the medication roughly tripled it. That is an argument for combination, not substitution — and it is exactly how we use these drugs.

5. The dose question — and what “microdosing” really means

Why we are talking about small doses at all

Ten thousand years ago we were hunter-gatherers. Food was not available on every corner. If the hunt went badly we went without for days, sometimes weeks. Our bodies handled that well — we dropped into a fasted state, shifted into ketosis, and ran the systemic clean-out that biologists call autophagy. That second half of our metabolism was not exotic. It was the ordinary rhythm of being alive.

Now there is food on every street corner, and a great deal of it is processed, seed-oil-laden and engineered to be eaten quickly. Add chronic stress and poor sleep on top, and there is almost no window in a modern day when the body is not digesting something. The fasted state has effectively disappeared from most people’s lives, and with it a good deal of the repair that used to happen there.

Some people can fast deliberately and do it well. It takes real discipline, and most patients cannot sustain it. In our experience a low-dose GLP-1 is the most reliable way we have found to make a fasted state accessible again — to quiet the constant food signal enough that a person can go the hours their metabolism was built to go. It simulates, in our view, something close to the ancestral pattern. We are having to use a drug to do it. That is simply where we are.

Why the standard doses do not fit everyone

The dose ladders for these medications were built in trials of people with a mean BMI around 38, many with type 2 diabetes, whose goal was substantial weight reduction. The doses were escalated until weight loss was maximized or the patient could not tolerate more.

That is an entirely reasonable way to develop an obesity drug. It is a poor way to determine the dose for a 68-year-old with a BMI of 26, a fasting insulin of 15, stubborn visceral fat, aching joints and a CRP of 4 — who does not need to lose thirty pounds and cannot afford to lose muscle.

The dose that produces maximum weight loss and the dose that produces maximum metabolic benefit are not necessarily the same dose.

What microdosing was supposed to mean

As originally described, a microdose is roughly one-fifth to one-tenth of the standard starting dose — not one-tenth of the maximum dose, which is a different and much larger number. The concept borrows directly from low-dose naltrexone: give the system a small nudge and let it work better, rather than overriding it pharmacologically.

That distinction is being abused commercially. A telehealth company that starts you on semaglutide 0.25 mg and calls it a microdose because it is a tenth of 2.4 mg is describing the ordinary first rung of the standard ladder. That is not a microdose. It is onboarding.

What the evidence actually supports — and where it doesn’t

We want to be very clear here, because this is the section where enthusiasm most often outruns data.

There is no randomized controlled trial of GLP-1 microdosing. None. What exists is dose-response data from phase 2 trials, and it is genuinely informative — but it points in different directions for different molecules.

The honest synthesis: low-dose efficacy is molecule-specific. Semaglutide clearly does something meaningful at doses at the bottom of the ladder. Retatrutide demonstrably does not at 0.5 mg. Tirzepatide is untested in that range. Generalizing “microdosing works” across the whole class is not supportable, and we do not do it.

The inflammatory and immune signal

The most interesting reports from low-dose use have nothing to do with weight: joint pain, psoriasis, eczema, mast cell reactivity, chronic inflammatory symptoms. GLP-1 receptors are documented on multiple immune cell populations, which makes the biology plausible.²³ The one published clinical report is a 2025 case series describing benefit in mast cell activation syndrome, much of it at doses well below conventional.⁵⁴

We take it seriously as a signal, and we are clear about what it is: an uncontrolled, retrospective case series with subjective endpoints and no comparison group. Hypothesis-generating, not evidence of efficacy. Also worth knowing — you will see it claimed that GLP-1 receptors sit on mast cells. We went looking for the primary source and could not find one. Any benefit there may well be indirect.

We are not alone in noticing this. Other clinicians using small doses report improvement in rheumatoid arthritis, psoriatic disease, inflammatory bowel disease and endometriosis, and some have begun arguing that microdosed GLP-1s should become routine in autoimmune care. We understand the enthusiasm and we share some of it. But we have to be straight about where that stands: there is not one completed randomized trial in any of those conditions. Not in rheumatoid arthritis, not in inflammatory bowel disease, not in psoriatic arthritis, not in endometriosis — the trials that exist are still recruiting. Everything published is retrospective cohorts and case series.

So this belongs in the category of promising clinical observation, and we treat it that way: worth trying in a patient who has exhausted better-evidenced options, monitored closely, and never described to you as established.

So what actually counts as a microdose?

Roughly a tenth of the standard starting dose — sometimes a fifth. Not a tenth of the maximum dose, which is a much larger and quite different number, and the source of most of the confusion in this space.

A telehealth company that starts you on the ordinary first rung of the ladder and calls it a microdose because it is a fraction of the top dose is describing standard onboarding. That is not what we mean.

We are not going to publish a dose conversion table here, because working out milligrams from syringe units is exactly where errors happen — and the errors are not small. The FDA has documented patients self-administering 5 to 20 times the intended dose because of confusion between milligrams, millilitres and syringe “units,” with some cases requiring hospitalization.⁵⁵

If you are on a compounded vial, the concentration printed on your label determines everything, and the same number of units can mean very different doses from two different pharmacies. Never carry a unit number from one vial to the next. Have your prescriber do the arithmetic with you and write the dose in both units and milligrams before your first injection. We do this in the room, every time.

Finding the dose, and knowing when you have passed it

The goal is the minimum effective dose — the smallest exposure that moves the marker you are actually treating. That marker might be A1c. It might be fasting insulin, CRP, waist circumference, liver enzymes, joint pain or the volume of the food noise. It is rarely the scale alone.

Signs the dose has gone past useful and into counterproductive:

• Flattened mood and lost drive. Losing interest in things you enjoy. This is the most common sign of overshoot and it usually resolves within a dose reduction or two.
• Moving less all day — not by choice, just less.
• Protein intake falling because eating has become a chore.
• Strength declining — fewer reps at the same load, longer recovery.
• Weight loss stalling at a higher dose that resumes when the dose comes down. Clinically common, and it surprises people.
• Persistent nausea, reflux or constipation that has not settled after two or three weeks at a dose.

One free thing to try before escalating: rotate the injection site. Absorption differs between abdomen, thigh, arm and buttock, and patients who have settled into one spot sometimes respond to a change.

If you feel terrible, check the obvious things first

A lot of what gets reported online as “this drug made me depressed” turns out, on examination, to be something more mundane and more fixable.

People on a large dose sometimes eat one meal in a day. That is not a recipe for feeling well, whatever the pharmacology. Blood pressure drops when intake falls and electrolytes are not replaced. Blood sugar runs low. Protein and micronutrients fall off a cliff. Any one of those will make a person feel flat, foggy and unmotivated, and none of them is a reason to conclude the medication is doing something to your brain.

There is a social dimension too, and it is underrated. Eating is not only fuel — it is how most people see each other. Patients who stop showing up to meals often lose more than calories. If you have quietly withdrawn from the table, that is worth noticing, and it is worth keeping the ritual even when the appetite is gone.

So before anyone decides the drug is the problem: check blood pressure, check electrolytes and hydration, count the actual protein, and ask honestly whether you are eating with people. Then reduce the dose. In our experience that sequence resolves most of it.

What the first ninety days actually look like

Patients ask this more than any other question, and the marketing answers it badly. Here is the honest version of a course that is going well.

Week one. The commonest experience is that very little happens, which surprises people. Some notice appetite quieting within a day or two; many notice nothing for a fortnight. Mild nausea, if it appears, usually shows up in the 24 to 48 hours after the injection and settles. Eat smaller and slower, keep the fat content of meals moderate, and drink more water than feels necessary.

Weeks two to four. This is when most people notice the thing they did not expect: the mental chatter about food goes quiet. Patients describe walking past the kitchen without the negotiation. It is often more striking than the appetite change itself, and for people who have spent decades in that negotiation it can be genuinely emotional. Constipation is the effect most likely to bother you here — get ahead of it with water, fiber and magnesium rather than waiting.

Weeks four to eight. The decision point, and the one most clinics get wrong. The default script says escalate at four weeks. We ask a different question: is this dose doing the job? If the markers we are treating are moving, if you feel well and if strength is holding, we hold. This is where a lot of unnecessary side effects get avoided.

Around week eight to twelve. Repeat labs — fasting insulin, glucose, A1c, hs-CRP, liver and kidney function. This is the appointment that tells us whether the dose is right, and it is usually the first objective confirmation that something real is happening. Frequently the labs move before the scale does, which is worth knowing in advance so you do not read a slow week as failure.

Throughout. Protein at every meal from day one, not once the weight starts moving. Resistance training from week one, not later. These are the two things that determine whether you end up smaller and stronger or just smaller, and they are hardest to start once appetite has already fallen away.

6. Muscle, bone, and the 40% myth

The most frightening claim about these medications is that they eat your muscle. It has been repeated so often it now functions as common knowledge. It is largely wrong — but there is a real problem hiding underneath it, and the real problem is arguably more actionable than the fake one.

What DXA “lean mass” actually measures

Start here, because everything else follows. On a DXA scan the body is split into fat, bone mineral, and everything else. That third compartment is called lean mass — and it is a subtraction, not a measurement of muscle. It includes organs, connective tissue, extracellular fluid, and glycogen with the water bound to it.⁵⁷

When someone loses weight, the liver shrinks, glycogen falls, fatty infiltration reverses and fluid drops. All of that registers as “lean mass loss.” In one controlled animal comparison, loss of liver mass exceeded the change in muscle mass entirely.⁵⁸ So any lean-mass figure is an upper bound on muscle loss, and usually a generous one.

The actual numbers

The best body-composition data comes from the DXA substudy of SURMOUNT-1, in which 160 participants were scanned at baseline and at 72 weeks:⁵⁹

Read the last row twice. The ratio of fat to lean tissue lost was identical in the drug arm and the placebo arm. The medication changed how much people lost. It did not change what they lost. That is the single most important body-composition finding in this entire field, and it directly contradicts the claim that GLP-1s cause disproportionate muscle loss.

Where did “40%” come from? Across the literature, lean mass as a proportion of total weight lost ranges from about 15% to 60% depending on the trial, the population and the rate of loss.⁵⁷ Forty percent sits in the upper half of a wide range, not at its centre — and well above what the largest DXA substudy actually found. In the largest DXA substudy it was 25% — the same as dieting.

Meanwhile, in the semaglutide body-composition substudy, lean mass as a proportion of the body actually rose by about three percentage points, and the lean-to-fat ratio improved — more so in those who lost the most weight.⁶⁰ Body composition got better, not worse.

Function is what matters, and function held

A 2026 study examined this in both mice and humans. In calorie-matched animals, pharmacological and dietary weight loss produced comparable, modest effects on muscle; relative muscle mass and strength improved, and running time to exhaustion rose robustly. In the human arm, thigh muscle cross-sectional area fell significantly — yet neither knee extensor strength nor grip strength declined.⁵⁸

The same paper settles a mechanistic question people have been speculating about: skeletal muscle does not appear to have a functional GLP-1 receptor.⁵⁸ So any claim that GLP-1 agonism directly protects muscle is unsupported — and so is any claim that it directly attacks muscle. The drug is not talking to your muscle at all. Weight loss is.

Now the real problem

If muscle loss on a GLP-1 is proportionate to weight loss, and weight loss is very large, then muscle loss is also very large in absolute terms. That is the actual risk, and it is entirely a function of two things you control: how fast you lose, and what you do while losing.

It matters most in exactly the group most likely to be handed a prescription casually. In older adults with type 2 diabetes, GLP-1 initiation was associated with a modest increase in fragility fractures, HR 1.11 (1.01–1.21).⁶¹ And a 2026 review noted that while short and medium-term trials show grip strength preserved despite falling lean mass, longitudinal work in older adults has documented accelerated sarcopenia with prolonged semaglutide use — concluding that lean mass is an unreliable predictor of strength, and that strength should be measured directly.⁶²

Bone

The bone data are modest but real. In a 52-week randomized trial in 64 adults at increased fracture risk, semaglutide 1.0 mg produced greater declines in bone mineral density at the lumbar spine (−0.020 g/cm², P=0.001) and total hip (−0.018 g/cm², P=0.007) than placebo, with higher bone resorption markers and no increase in bone formation.⁶³ The authors attributed the decline primarily to concurrent weight loss.

The mechanism is unglamorous and worth understanding: bone remodels in response to load. Lose thirty pounds and every step, every stair, every rise from a chair applies less force to the skeleton. If nothing replaces that load, bone density follows the body weight down. This is not a drug effect. It is a physics effect. And it is preventable.

There is a more optimistic possibility worth mentioning, clearly labeled as unproven. GLP-1 receptors are present on bone cells, and the signaling appears to stimulate osteoblasts — the cells that build bone — while inhibiting the osteoclasts that break it down. If that holds in humans, the net effect once a patient is weight-stable could be favourable rather than harmful. That is mechanistic and largely preclinical reasoning, not a finding. What we would take from it practically: the bone risk sits in the losing phase, which is another argument for losing slowly and loading the skeleton while you do.

What is coming for muscle

The most interesting frontier in this field is combining a GLP-1 with a compound that releases the body’s brakes on muscle growth — myostatin and activin A. A phase 2 trial pairing semaglutide with one such antibody produced 17.8 kg of weight loss with only 0.8–2.3% lean mass loss, against 14.2 kg and 4.7–6.9% lean loss for semaglutide alone.⁶⁶ Animal work has gone further, with large fat reductions alongside actual lean mass gain.

If these approaches prove safe and scalable they could genuinely change body-recomposition medicine — losing fat while building muscle, rather than trading one for some of the other. They are phase 2, discontinuation rates in the antibody arms were higher, and none of it is available. For now the muscle-preserving intervention that works is the one that has always worked, and it involves picking heavy things up.

7. Fat loss mobilizes what was stored in the fat

This is a short section, but it matters to the patients we see — people who have already been through mold illness, heavy metal exposure, chemical sensitivity, or years of accumulated environmental burden.

Adipose tissue is not inert storage. It is where the body puts fat-soluble compounds it cannot easily excrete: polychlorinated biphenyls, organochlorine pesticides, dioxins, flame retardants. People with obesity carry two to three times the total adipose content of these compounds that lean individuals do.⁶⁷ When that fat compartment shrinks, they have to go somewhere.

The measurements bear this out. Blood concentrations of persistent organic pollutants rise by roughly 2 to 4% per kilogram of weight lost, and most remain elevated a year later.⁶⁸ At the same time — and this is the reassuring half — total body burden falls, by about 15% within six to twelve months in patients followed through major weight loss.⁶⁷ The rise in the blood is redistribution out of a shrinking reservoir, not accumulation. On balance you end up carrying less.

One correction worth making, since it is widely repeated: visceral fat is not a privileged toxin reservoir. These compounds distribute according to lipid content rather than depot identity, and concentrations in visceral and subcutaneous fat correlate closely.⁶⁹ Visceral fat is still the metabolically dangerous compartment — just not for this reason.

The one group the literature genuinely flags for caution is women undergoing major weight reduction shortly before pregnancy or breastfeeding, because mobilized compounds partition into the fetus and into breast milk.⁶⁸ That caution we endorse without reservation. For everyone else, the authors of the largest review put it plainly: the benefits of weight loss substantially outweigh the risks.

Why we insist on amino acids

The reason is simpler than the detox argument and better supported: most patients on these medications are not eating enough protein. Fewer than half reach 1.2 g/kg⁶⁴, and protein intake during an energy deficit is the difference between losing fat and losing tissue.⁶⁵

When appetite is pharmacologically suppressed, a free-form essential amino acid supplement solves a practical problem: it delivers the building blocks without requiring the patient to face down another meal. We use PerfectAmino by BodyHealth → for exactly this. Glycine is worth adding separately for its role in glutathione synthesis and its calming effect at night. Glycine Powder →

8. Stress matters — and it should be addressed first

This is short, but we would not leave it out, because it changes what we do in the clinic more often than almost anything else in this article.

We do not start a GLP-1 in a patient whose stress load is clearly elevated. Not because the medication is dangerous in that setting, but because the timing is wrong.

If you are in the middle of a divorce, a job that is eating you alive, a long post-viral tail, an active mold or Lyme picture, or months of poor sleep — that is not the moment to add another demand to the system.

Why we take this position

Cortisol drives the exact problem we are trying to fix. This is not a soft argument. In classic clamp studies, cortisol required roughly 2.6 times more insulin to suppress the liver’s glucose output, and the defect was postreceptor — meaning it is not something you overcome by making more insulin.⁷⁰ Glucocorticoid exposure at the tissue level is on its own sufficient to produce visceral obesity, insulin resistance and hypertension.⁷¹ A chronically stressed patient is being pushed toward metabolic dysfunction by an upstream signal. Adding a medication without touching that signal is asking one intervention to overcome another that is still running.

Poor sleep does the same thing, faster. One week at five hours a night reduced insulin sensitivity by 11–20% in healthy men.⁷² Short sleep manufactures the condition we are treating.

And clinically, these patients simply do worse in the first weeks. We see lower heart rate variability, higher resting heart rate and reduced exercise tolerance early on. In most people that settles. In someone who arrives already depleted, it often does not — and they conclude the medication failed them and will not try again. There is no trial evidence on this; nobody has studied baseline autonomic state as a predictor of GLP-1 tolerance. It is our clinical observation, and we are labeling it as such.

What we look at

• Resting heart rate and HRV trend from a wearable — compared to that person’s own baseline, not a population chart. Low HRV and elevated resting heart rate both carry independent metabolic risk: in 8,185 adults followed over eight years, the lowest HRV quartile had roughly 20% higher risk of developing type 2 diabetes.⁷³
• Sleep — duration, consistency, and whether it is restorative.
• Exercise tolerance — same session, same load, higher effort and slower recovery is the earliest signal and the one patients notice first.
• Cortisol rhythm where the history warrants it — a four-point salivary panel showing the shape of the day, not a single serum value. A flattened daily slope predicts inflammation, adiposity and mortality.⁷⁴

What we do first

1. Sleep, before anything else. Duration, consistency, a dark cool room.
2. Slow-paced breathing. Ten minutes a day. Meta-analysis of HRV biofeedback found a large effect on stress and anxiety.⁷⁵
3. The right exercise dose — often less than the patient is doing. Exercise is the best-evidenced way to improve HRV⁷⁶, but a depleted patient hammering high-intensity work is adding to the problem.
4. Remove the obvious inputs. Alcohol, late caffeine, aggressive fasting and chronic under-eating are all cortisol-raising and all optional.
5. Treat the underlying driver. Active mold exposure, an untreated infection, unmanaged thyroid disease, sleep apnea, or a genuinely unsustainable situation. Sometimes the intervention needed is not medical.
6. Targeted support. Ashwagandha has the best evidence among the adaptogens for lowering cortisol specifically — a modest, low-certainty effect rather than a dramatic one.⁷⁷ Adaptogen support → Magnesium →
7. Reassess in 8–12 weeks. Then we have the conversation about a GLP-1 — and when we do, it usually goes better.

9. The sleep side effect nobody warned you about

We have seen enough of this in our own patients to write about it: people who were sleeping normally start waking at three or four in the morning and cannot get back down. Sometimes it is fragmented sleep throughout the night. It tends to appear within the first few weeks, and it is more common after a dose increase.

You will not find it in the prescribing information. Insomnia is not a listed adverse reaction for either Wegovy or Zepbound — it does not appear in the adverse reaction tables or the postmarketing sections.^78,43 It does show up in spontaneous adverse event reporting, where it was the single most frequently reported psychiatric event associated with these medications, though the statistical signal is weaker than the raw frequency suggests.⁷⁹

And the strongest controlled sleep data run in the opposite direction entirely. In two 52-week randomized trials, tirzepatide dramatically improved obstructive sleep apnea and significantly improved patient-reported sleep disturbance.⁸ For a great many people these medications make sleep better, not worse. Both things are true, and if you are one of the people who sleeps better, this section is not about you.

What is probably going on

Two contributors look most likely, and neither is what patients usually assume.

It is probably not low blood sugar. GLP-1’s insulin-releasing effect switches off before glucose reaches a low range, and the body’s counter-regulatory response stays fully intact — which is why hypoglycemia is rare on these drugs unless insulin or a sulfonylurea is also on board.⁸⁰

It is probably not the drug raising cortisol. In a randomized crossover trial, three weeks of GLP-1 therapy produced no change in cortisol at all compared with placebo.⁸¹

What does hold up: these medications modestly raise heart rate — 1 to 4 bpm for semaglutide and 1 to 3 for tirzepatide per their labels^78,43 — and in one detailed monitoring study the increase was concentrated at night.⁸² A faster, less variable heartbeat overnight is a plausible route to early-morning waking. The effect is reversible; heart rate returns to baseline after stopping.⁸³ Separately, a steep calorie deficit can raise evening cortisol, transiently, in the early weeks.⁸⁴ That is a direct argument against stacking aggressive fasting on top of the medication.

What we do about it

1. Reduce the dose before doing anything else. If it started after a dose increase, this is the answer more often than not, and it is free.
2. Slow the rate of loss. A steep deficit is doing more here than the drug is.
3. Stop stacking aggressive fasting on top. A 16- to 18-hour window plus a GLP-1 plus a training program is three appetite-suppressing inputs at once. Adequate protein spread across the day matters far more than extra fasted hours.
4. Move the injection. Some patients do better shifting the injection day so peak-effect days fall where sleep matters least. No trial for this; it costs nothing to try.
5. Check the heart rate. Resting and overnight, plus HRV if there is a tracker. A persistent rise is a reason to hold or reduce, independent of sleep.
6. Do not skip dinner protein. Patients whose appetite dies in the evening often eat almost nothing after mid-afternoon.
7. Support the wind-down. Magnesium and glycine at night are low-risk and worth trying. Glycine Powder →

If the sleep disruption comes with low HRV, a rising resting heart rate and a chronically stressed history, read the previous section again — that is a signal to step back and address the underlying state rather than keep adjusting around the edges.

One last thing worth knowing: the relationship runs both ways. A single night of sleep deprivation delayed the GLP-1 response to breakfast by about 90 minutes in healthy men.⁸⁵ Poor sleep degrades incretin signaling, which degrades glucose control, which degrades sleep. Breaking that loop anywhere helps.

10. The real risks, in proportion

Every one of these medications carries genuine risk. Several of the risks you have read about are much smaller than the headlines suggested, and a couple are larger. Here is our attempt to put each one at its correct size.

Gastrointestinal effects — common, and mostly a dosing problem

Nausea, vomiting, diarrhea, constipation, reflux and early satiety are the dominant adverse effects and the dominant reason people stop. They are dose-dependent and titration-dependent. Most of what patients experience as “I can’t tolerate this drug” is actually “I was escalated too fast.” Slower titration, smaller increments, and a willingness to sit at a lower dose indefinitely solve most of it.

What is not just an inconvenience: persistent vomiting or a near-total collapse in intake is a medical problem, not a sign the drug is working. Reduced appetite reduces micronutrients, not just calories — among adults on GLP-1 therapy the proportion falling below recommended intake was 98.6% for vitamin D, 94.2% for choline, 89.9% for magnesium and 88.4% for iron.⁶⁴ Eat, take a good B-complex and multivitamin, and tell us if you have stopped being able to.

Gallbladder disease — real, and higher in weight-loss dosing

A meta-analysis of 76 randomized trials in 103,371 participants found a relative risk of 1.37 for gallbladder or biliary disease overall — but with an important split. In trials using these drugs for weight loss, the relative risk was 2.29 (1.64–3.18), versus 1.27 for diabetes trials. Higher doses and longer duration both increased risk.⁸⁶ In absolute terms, the Wegovy label reports cholelithiasis in 1.6% versus 0.7% on placebo.⁷⁸

This is partly a drug effect and partly the well-known consequence of rapid weight loss itself. It is another argument for losing slowly, and for not skipping dietary fat entirely — a gallbladder that never contracts becomes a gallbladder that forms stones.

Pancreatitis — the alarm has not replicated

The 2023 analysis behind the headlines reported a hazard ratio of 9.09 for pancreatitis. Read the confidence interval: 1.25 to 66.00. That spans two orders of magnitude, which tells you the estimate rests on a handful of events in a small cohort.⁸⁷

A far larger analysis subsequently matched 20,459 GLP-1 users to 20,459 controls and found no increased risk of acute pancreatitis. Among those who did develop it, GLP-1 users had significantly lower rates of complicated disease and lower mortality.⁸⁸ Pancreatitis remains in every label and remains a reason to stop immediately if severe abdominal pain develops — but the nine-fold figure should not be quoted as an established risk.

So should we be checking lipase and amylase?

This is the question we get most often, and the answer is more specific than yes or no.

We do not run surveillance lipase in a patient who feels well. These medications raise pancreatic enzymes in a large fraction of people who never develop pancreatitis, asymptomatic elevations correlate poorly with clinical disease, and no label or society recommends serial monitoring. A mildly raised lipase in someone who feels fine mostly generates imaging, anxiety and the occasional unnecessary discontinuation.

What we do instead: a baseline lipase, amylase and triglycerides before starting, so that a later value can be interpreted; a careful history for prior pancreatitis, gallstones, heavy alcohol use and hypertriglyceridemia, which identifies far more risk than any lab draw; triglycerides on every panel, since severe elevation is itself a pancreatitis cause and usually improves on therapy; lipase immediately if symptoms appear — severe epigastric pain boring through to the back, with nausea or vomiting (lipase is the more specific test; amylase adds little); and a low threshold for gallbladder imaging with right upper quadrant pain or fatty-food intolerance.

Thyroid cancer — a rodent finding that has not appeared in humans

Every label carries a boxed warning about thyroid C-cell tumors. Read the wording: in rodents, these drugs cause C-cell tumors, and “it is unknown whether” the same happens in humans.^78,43 The contraindication in personal or family history of medullary thyroid carcinoma or MEN2 is precautionary.

The largest human cohort — 145,410 GLP-1 users versus 291,667 comparators — found no increased thyroid cancer risk (HR 0.93), including for medullary thyroid cancer specifically.⁸⁹ A French case-control study is the outlier.⁹⁰ On current evidence the human signal is absent. The contraindication stands regardless.

Delayed gastric emptying and anesthesia

It is worth separating two things that get conflated. Slowing the stomach is the intended mechanism — it is part of how these drugs work, and it is not a complication. True gastroparesis, meaning severe dysfunction with persistent nausea and vomiting, is a different and relatively uncommon thing. The risk of crossing from one to the other rises with aggressive dose escalation, with high doses, and with eating as though the stomach were emptying normally when it is not. Start low, go up slowly, and do not eat a large meal out of habit on a dose that has slowed you down.

These drugs slow the stomach by design. At endoscopy, retained gastric contents were found in 13.6% of GLP-1 users versus 2.3% of non-users — though only about 0.2% of those showed evidence of aspiration, and a 24-hour liquid diet plus a 12-hour fast produced no retained solids at all.⁹¹

Current multi-society guidance favours shared decision-making over blanket drug-holding: hold daily formulations the day of surgery and weekly formulations a week prior where appropriate, use a preoperative liquid diet for at least 24 hours where delayed emptying is a concern, and consider rapid sequence induction rather than cancelling.⁹²

Tell every anesthesiologist, endoscopist, dentist and surgeon that you are on a GLP-1. Do not assume it is in the chart.

SIBO and gut motility

Two figures circulate here and both are misleading. “GLP-1s increase SIBO by 45%” has no traceable source. “76% of users have SIBO” comes from 99 symptomatic patients referred for breath testing with no control group — a test-positivity rate, not a prevalence.⁹³ What the best data actually show is a doubling of a very small number: 0.177 versus 0.083 diagnoses per 1,000 patient-years.⁹⁴

But the mechanism is entirely plausible — slowed motility is how SIBO gets established in the first place — and in a practice like ours, where a large share of patients arrive with pre-existing gut dysfunction, the population-level statistics are beside the point. We are not treating a population. We are treating people who already have it.

Alcohol — we would say no

Our position is that you should not drink at all while taking a GLP-1, and we say this more firmly here than in most contexts.

The stomach is already slowed, so alcohol sits in it longer and your usual tolerance is no longer a reliable guide. You are eating far less, so a larger share of what remains carries no nutrition at a point when protein and micronutrient intake are already falling short in most patients on these medications.⁶⁴ Alcohol is a leading cause of pancreatitis and a driver of gallbladder and liver disease — the three areas where these drugs carry their own signal, so stacking them is unnecessary. It dismantles sleep architecture at exactly the time when many patients are already sleeping poorly. And it suppresses fat oxidation and drives visceral fat — the compartment we are specifically trying to shrink.

The good news is that this is one of the easier asks, because many patients find the desire simply goes away. Reduced drinking is one of the most consistently reported effects of the whole class.^47,46 Treat that as an opportunity rather than a curiosity, and let it stay gone. Our full argument is in the alcohol and optimal wellness article.

Contraception and pregnancy

Tirzepatide reduces the absorption of oral contraceptives — peak concentrations fell 55–66% in pharmacokinetic studies. The label instructs patients using oral hormonal contraception to switch to a non-oral method or add a barrier method for four weeks after starting and for four weeks after every dose escalation.⁴³ This warning is specific to tirzepatide; the semaglutide label carries no equivalent statement.

On pregnancy: inadvertent first-trimester exposure has been reassuring. A multinational study found major congenital malformation risk of 0.95 (0.72–1.26) relative to insulin⁹⁵, and a 2026 target trial emulation in over 3,500 pregnancies found no significant difference in malformations or fetal growth — while explicitly cautioning that the confidence intervals were wide enough that the results “shouldn’t be interpreted as proof of safety.”⁹⁶

Consensus guidance published in July 2026 states that incretin therapy should not be deliberately continued through pregnancy, that contraception should be discussed at initiation and at every dose escalation, and that women with inadvertent exposure should be reassured that current data show no increased risk of congenital anomalies.⁹⁷ The semaglutide label advises discontinuing at least two months before a planned pregnancy because of the long half-life; the tirzepatide label, with its shorter half-life, specifies no pre-conception washout.^78,43

Combine that with the restored fertility discussed earlier⁹⁸ and the contraceptive interaction above, and the conclusion writes itself: if you can become pregnant, contraception is part of this prescription.

11. Who should not take these medications

We spend most of this article explaining why these drugs are more useful than their critics allow. Here is the other side, and we hold it just as firmly.

Absolute and near-absolute contraindications

• Personal or family history of medullary thyroid carcinoma, or MEN2 syndrome. The evidence in humans is reassuring, but this contraindication is in every label and we honor it.
• Pregnancy, or planning pregnancy. Discontinue at least two months before conception for semaglutide.⁷⁸ Not to be continued deliberately through pregnancy.⁹⁷
• Prior pancreatitis — not an absolute bar, but a conversation that requires a real risk-benefit discussion and closer monitoring.
• Type 1 diabetes, outside of specialist management.
• Established gastroparesis or significant bowel dysmotility. You are adding a motility-slowing drug to a motility problem.
• Active SIBO, or significant irritable bowel syndrome. Not a permanent bar — a sequencing decision. Treat the gut, then revisit.

The group we worry about most

Frail or near-frail older adults with no meaningful excess weight. We see clinics doing this and it troubles us. An 80-year-old with a BMI of 23 who does not strength train has limited reserve, and a modest amount of extra weight in that population is arguably protective. Fragility fracture risk is measurably elevated in older adults on these medications⁶¹, and longitudinal data suggest accelerated sarcopenia with prolonged use.⁶²

This is not an argument that nobody over 70 should take one — we have patients in their seventies doing very well on low doses with rising strength. It is an argument that in this age group, resistance training and adequate protein are a precondition for the prescription, not something to discuss later.

The other group we worry about

Young, lean people using these drugs cosmetically. They are not vanity drugs. The risk profile that is entirely reasonable for someone with a BMI of 36 and prediabetes is not reasonable for someone with a BMI of 22 who wants to look different in photographs.

There is a second reason to be careful in the young that gets very little airtime. These medications produce GLP-1 signaling far above anything the body generates on its own — roughly a thousandfold higher in molar terms than native hormone²⁵ — and GLP-1 receptors are distributed throughout the brain, including in the circuits that govern appetite, reward and motivation. What years of that exposure does to a brain that is still developing is simply unknown. It may be fine. Nobody has the data, and we are not comfortable being the ones who find out.

Related and more delicate: a history of restrictive eating or an eating disorder. There are reports of these medications quieting the noise for people with binge-type patterns and we do not dismiss them — but a drug that pharmacologically enforces reduced intake in someone with a history of restriction is a dangerous combination, and it needs a clinician who treats eating disorders involved. Notably, only about half of the online prescribing platforms studied even asked about it.⁹⁹

And the one nobody wants to hear

If you are not willing to strength train and eat protein, we would rather not prescribe this. That is not moralizing — it is the direct implication of the body composition data. A GLP-1 without resistance training and adequate protein produces a smaller person with proportionally less muscle and less bone, who will regain fat when the drug stops and be worse off metabolically than when they started. The medication amplifies whatever you are already doing. If you are not doing anything, it amplifies that too.

12. Where your medication comes from

The public conversation lumps together things that are not alike. The question that matters is not brand versus compounded. It is supervised versus unsupervised.

A brand product from a licensed pharmacy, prescribed by a clinician who examined you and ran labs, is appropriate. So is a compounded prescription written for you specifically by a clinician who knows your history, dispensed by a licensed compounding pharmacy, as part of a monitored program. That is how most of our own patients receive these medications.

Why we use compounded preparations

Two reasons, and both are clinical rather than economic.

First, dose flexibility. A vial and a syringe let us dose precisely where a patient needs to be. Everything in the dosing section of this article — the minimum effective dose, holding at a fraction of the labeled starting dose, titrating on data rather than on the calendar — is far easier to deliver than with a pen that steps in fixed increments. It is also why we have been glad to see both manufacturers add single-dose vials alongside their pens.

Second, and less well known: we can add things. A compounded preparation can include adjuncts the standard product does not offer. The most common in our practice is vitamin B6 (pyridoxine) compounded into the vial for nausea — a small addition that meaningfully improves tolerability for patients who would otherwise struggle through the first weeks or quit. Other supportive ingredients can go in where a patient’s picture calls for it. You cannot do that with a pen.

Compounding does carry one specific risk that we manage rather than pretend away: dose calculation errors. The FDA has documented patients self-administering 5 to 20 times the intended dose, driven by confusion between milligrams, millilitres and syringe “units.”⁵⁵ That is exactly why we do the arithmetic with the patient in the room, write the dose in both units and milligrams, and confirm the concentration on the actual vial before the first injection.

What we do advise against

Prescribing mills. A web form, a five-minute approval, no labs, no examination and an automatic dose escalation every four weeks regardless of how you are doing. In a 2026 study of 49 online GLP-1 prescribing websites, only two required any blood work and roughly two-thirds required no live contact with a clinician at all.⁹⁹ The product may be perfectly good. The care around it is absent — which means nobody is checking your insulin, your kidney function, your gallbladder history, your SIBO or your body composition, and nobody is there to say hold.

Gray-market and “research” peptides. Vials sold online labeled “for research use only, not for human consumption” meet no pharmaceutical standard for identity, purity, sterility or stated concentration — one analysis of product bought this way found purity as low as 7.7% against a claimed 99%.¹⁰⁰ If a product cannot legally say it is for human use, take that at face value.

The clearest current example is retatrutide, which is not approved in any country and which the FDA has said cannot lawfully be compounded at all.³³ In August 2026 Eli Lilly filed six lawsuits against peptide vendors, a compounding pharmacy and a med spa selling it.³⁴ We do not prescribe it, we cannot obtain it, and we would discourage anyone from sourcing it independently — not because the molecule is uninteresting, but because nothing sold under that name today has a verified identity or dose behind it.

One pattern worth naming, because we see it: someone tries one peptide, gets a good result, and the website suggests three more. Peptides get stacked on peptides — often alongside stimulants or ADHD medication — with no clinician watching the combination and no way to tell which one is responsible for anything. That is a business model, not a protocol. If you are taking more than one of these, tell us what they are.

13. The labs we run — before, and during

If you take one practical thing from this article, take this: you cannot manage what you have not measured, and the panel your annual physical runs will not measure most of it.

A standard comprehensive metabolic panel and lipid panel do not include fasting insulin. Neither the American Diabetes Association nor the US Preventive Services Task Force recommends insulin for screening. So the marker that moves first — years before glucose does¹⁹ — is the one nobody orders.

Before starting

What the numbers should look like

We want to be careful here, because this is an area where confident-sounding targets circulate without support.

The best population-derived HOMA-IR cut-points come from a study of 2,459 adults, which found thresholds of 1.85 in men and 2.07 in women against metabolic syndrome — corresponding to roughly the 70th to 75th percentile of that population.¹⁰² Note what that means: these are statistical cut-points from one Spanish cohort, not biologically derived thresholds, and insulin immunoassays are not standardized between laboratories, so a HOMA-IR from one lab is not strictly interchangeable with another.

The commonly promoted targets of “fasting insulin under 5” and “HOMA-IR under 1.0” are wellness-industry conventions. We could not find primary literature supporting them and we are not going to present them as evidence-based. What is citable: fasting insulin at or above 10 µU/mL was associated with roughly doubled cancer mortality in a national cohort — including in non-obese people.¹⁸

Our practical position: we want HOMA-IR comfortably below the 1.85–2.07 range, we treat a fasting insulin approaching 10 as a signal that deserves attention rather than reassurance, and we care far more about the direction of travel across serial measurements than about any single value.

During therapy

• Every visit: weight, waist circumference, blood pressure, resting heart rate, protein intake review, training log, sleep and recovery, and how exercise is actually feeling.
• At 8–12 weeks: fasting insulin, glucose, A1c, hs-CRP, comprehensive metabolic panel with eGFR, liver enzymes. This is where you find out whether the dose is doing the job.
• Every 3–6 months: body composition. This is the measurement that tells you whether the weight coming off is the weight you wanted off.
• Every 6–12 months: full nutrient panel, lipids with ApoB, urine ACR.
• Bone density: at baseline in anyone post-menopausal or over 65, and repeated per risk.

14. Supporting your own GLP-1 — what works and what does not

Many patients would prefer not to inject anything, and many should not need to. We have written separately about natural GLP-1 agonists in more detail. Here we want to do something that article and most others in this space do not do: separate what has human trial data from what is marketing.

Start with the honest comparison

Native GLP-1 circulates in the picomolar range — single digits fasting, tens of pmol/L after a meal — in brief pulses that are cleared within minutes. Semaglutide at steady state sits in the nanomolar range, roughly a thousandfold higher in molar terms, with continuous receptor occupancy for a week at a time.²⁵

No food, herb or probiotic closes that gap. Anything that claims to be “nature’s Ozempic” is selling you something. What dietary strategies can do is meaningfully amplify your own post-meal incretin response, and that is genuinely worth doing — especially in people whose metabolic dysfunction has not yet reached the point of needing pharmacology.

What has real human evidence

Protein before carbohydrate — the strongest effect on this list

In a randomized trial in people with type 2 diabetes, 50 g of whey protein in water taken before a high-glycemic breakfast raised total GLP-1 by 141% and intact GLP-1 by 298%, and reduced the postprandial glucose excursion by 28%.¹⁰³ That is a large effect from a food.

It was a single-meal study in 15 people and 50 g is a substantial dose, so do not over-read the exact numbers. But it supports a principle we use constantly: protein first, then vegetables, then starch. Sequencing the same meal differently changes the glucose curve.

Fermentable fiber and short-chain fatty acids

Colonic fermentation of fiber produces short-chain fatty acids, and propionate in particular stimulates L cells to release GLP-1 and PYY. In a controlled study, targeted delivery of 10 g of propionate to the colon raised postprandial GLP-1 and PYY and reduced energy intake acutely; over 24 weeks it prevented weight gain and adverse intra-abdominal fat distribution in overweight adults.¹⁰⁴

One qualifier: that study used an engineered ester designed to deliver propionate specifically to the colon, not ordinary dietary fiber. Fiber is still among the best things you can eat — just do not attribute those exact numbers to a bowl of oats.

Polyphenol-rich spices, and fat

A randomized dose-response crossover in healthy men found that a polyphenol-rich curry made with mixed spices raised postprandial GLP-1 by 17.3% at the lower dose and 31.6% at the higher dose, with everything else in the meal held constant.¹⁰⁵ The spice blend contained turmeric, ginger and capsaicin-containing peppers among others, so the effect belongs to the blend rather than to any single spice.

Separately, 19 g of olive oil taken with a vegetable meal significantly raised GLP-1 secretion in overweight people with type 2 diabetes, via the 2-monoacylglycerol/GPR119 pathway.¹⁰⁶

What is genuinely useful but not for the reason usually claimed

Akkermansia muciniphila

In a randomized, double-blind, placebo-controlled trial, pasteurized Akkermansia at 10¹⁰ CFU daily for three months improved insulin sensitivity by 28.6%, reduced insulinemia by 34.1% and lowered total cholesterol by 8.7%. Weight and fat mass changes did not reach significance.¹⁰⁷

Those are excellent results and we use this organism — it is what we stock for this purpose. GLP-1 Probiotic by Pendulum → One honest note when we recommend it: GLP-1 itself was not measured in that trial. Use it for the insulin-sensitivity and gut-barrier evidence, which is genuinely good, and read the GLP-1 in the product name as the mechanism being targeted rather than a demonstrated human effect.

What does not hold up

The strategy we actually use

Before reaching for a peptide — and alongside it, if we do — this is the sequence:

1. Protein at every meal, front-loaded. 30–40 g at breakfast changes the whole day’s glucose and appetite pattern. PerfectAmino →
2. Meal sequencing. Protein and vegetables before starch. Free, immediate, and reproducible on a continuous glucose monitor.
3. Fermentable fiber, increased gradually. Legumes, alliums, cooled starches, artichoke, oats. If you have SIBO, this order gets reversed — treat the overgrowth first.
4. Support the organisms that make your GLP-1. GLP-1 Probiotic by Pendulum →
5. Fix the gut. This is where our practice differs most from a conventional one. Endogenous GLP-1 is made by cells in a gut lining that has to be healthy to make it. Dysbiosis, intestinal permeability and SIBO all degrade that machinery. See our articles on intestinal permeability, SIBO and the microbiome. MegaSporeBiotic → Tributyrin-X →
6. Resistance training. Not for GLP-1 — for the glucose sink. This is the intervention with the largest effect on insulin sensitivity of anything on this list.
7. Sleep and circadian rhythm. One night of sleep deprivation delayed the GLP-1 response to breakfast by roughly 90 minutes.⁸⁵ Chronic disruption blunts the whole incretin rhythm.
8. Remove alcohol. For the reasons covered above and elsewhere on this site.
9. Reassess in 12 weeks with labs. A meaningful proportion of people never need step nine.
10. Then, if the markers have not moved, consider a peptide — at the lowest dose that works.

Products not stocked in our shop are available through our practitioner dispensary: Fullscript →

15. Two cases from our practice

Both are de-identified and details have been altered where they do not change the clinical point. We have chosen them deliberately to span the range — a lean older woman on a fraction of a starting dose, and a patient with long-standing obesity on full therapeutic dosing. The same principles apply to both; the doses do not.

Case one: the 73-year-old who was told she was too old for this

A woman in her early seventies, 5’4”, whose BMI had drifted from 23.7 to 24.7 over about six months. Not remotely obese, and not sedentary either: daily walking, resistance training three times a week, golf. She came in frustrated that she was gaining weight despite doing everything right.

Her labs told a story the scale did not. A1c 5.7% — newly in the prediabetes range — and a mildly reduced eGFR of 51. Her lipid panel looked unremarkable, but ApoB was 76 and Lp(a) low, which is why we run those rather than relying on LDL-C alone.

We started her on compounded tirzepatide at a dose that worked out to roughly half to two-thirds of the labeled starting dose — about a tenth of the maximum.

She tolerated it well — no significant nausea, improved energy, clearer thinking, better appetite control. She lost about nine pounds and kept training.

Her internist raised a sarcopenia concern. That concern is legitimate and we want to say so plainly — age, a near-lean BMI and incretin therapy together absolutely warrant attention to muscle. But we think the disagreement was framed wrong. It is not GLP-1 versus muscle preservation. The question is: can we obtain the metabolic benefit at the minimum effective dose while objectively demonstrating that muscle and function are preserved?

Two decisions followed. First, we did not escalate. The pharmacy’s printed instruction was to double the dose within a month and double it again after that. She was benefiting where she was, with no adverse effects, and there was no clinical reason to chase a schedule written for a different patient. As she approached a lean BMI, the goal changed entirely — not another 15% of body weight, but the lowest exposure that maintains the benefit without continued weight or tissue loss.

Second, we made the monitoring match the actual concern: body composition, her training log, documented protein intake, waist circumference, and repeat A1c, fasting insulin and kidney function. Her A1c became the endpoint that mattered — does it move to 5.4 while weight stabilizes and strength holds? That is a far more interesting question than whether she loses another five pounds.

One thing we changed: we had been discussing pushing her fasting window toward 16–18 hours. We stopped. Stacking aggressive fasting on top of an incretin in a lean 73-year-old is exactly the combination that produces the sarcopenia her internist was worried about.

The teaching point: “she’s 73 and not obese, so she shouldn’t be on this” and “she has prediabetes and metabolic drift, so she should” are both reasonable positions. What resolves them is not argument — it is measurement. Dose to the minimum that works, prove muscle and function are holding, and be willing to stop if they are not.

Case two: sixty-five pounds, and the first time it ever stayed off

A woman who had struggled with obesity for most of her adult life. Every previous attempt had followed the same arc — loss, plateau, regain, and a little more than she started with. She had done this enough times to have stopped expecting a different result.

She started tirzepatide with us as part of a full program, not as a standalone prescription. She has now lost sixty-five pounds and sits at a BMI of about 22 — the first time in her adult life she has reached and held a normal weight.

What we want to draw attention to is not the number. It is everything that went alongside it.

• Resistance training, consistently. Not walking, not “staying active” — actual progressive strength work, from early on and throughout.
• Protein defended deliberately. Her intake settled around 1,300 calories a day, which is exactly where protein normally collapses. It did not, because she uses free-form essential amino acids to guarantee the floor rather than hoping food gets her there. PerfectAmino →
• A full nutritional program underneath it. Gut support, micronutrient repletion, and the rest of what we would be doing for her metabolically with or without a peptide.
• Monitoring every six months, coordinated with her primary care physician. Not us operating in a silo — her PCP sees the same numbers we do.
• Zero side effects, across the whole course. No nausea, no gallbladder trouble, no vision symptoms, no sleep disruption. We attribute a good part of that to unhurried titration and to never chasing a dose she did not need.
• She is now titrating down. Slowly, deliberately, with the plan being to come off entirely and hold the result with the lifestyle architecture she has built.

16. What happens when you stop — and how we plan the exit

This question deserves a straight answer, because the answer is uncomfortable and the marketing avoids it.

The data are consistent and not flattering. In the STEP 1 extension, participants regained about two-thirds of the weight they had lost within a year of stopping, and among those with baseline prediabetes the proportion back in the normal glucose range fell from 93.6% at the end of treatment to 43.3% a year later.¹¹⁰ SURMOUNT-4 tested it directly: those who continued lost a further 5.5%, while those switched to placebo gained 14.0%.¹¹¹

A meta-analysis of 37 studies found mean regain of about 0.4 kg per month after stopping, with cardiometabolic markers projected back to pre-treatment levels in roughly 1.4 years — and regain running about four times faster than after a diet-and-exercise intervention.¹¹² It typically begins around eight weeks after discontinuation.¹¹³

Our long-term position: the drug is a bridge, not a destination

Here is how we frame this with patients at the very first visit, before anyone injects anything.

We use a GLP-1 to reach defined health goals. Once those goals are reached, our job is to get you off it — and onto the natural GLP-1 support and lean-mass strategies that hold the result.

One clinician we like describes it as a training-wheels approach: the lowest dose that has an effect, alongside the lifestyle work, with the intention of peeling the training wheels away — or leaving them on if that is what a particular person needs. That is exactly how we think about it.

That means the goals have to be defined in advance, and they have to be real. Not “lose weight.” A fasting insulin below a stated number. A HOMA-IR below a stated number. An hs-CRP that has normalized. An A1c that has moved from 5.9 to 5.4. Liver enzymes that have come down. Albuminuria that has resolved. A waist circumference target. Lean mass held or increased on serial body composition. When those are met, the question changes from how do we improve this to how do we hold this without the drug.

This matters most for the low-dose patients, because the temptation is greatest there. A small dose feels benign, side effects are minimal, the labs look good, and it is easy to keep going indefinitely. If a low dose has done its work — inflammation down, insulin down, visceral fat gone — the goal has been achieved and the exit conversation should start.

What we transition onto is not a supplement that replaces the drug, because no such thing exists. It is the architecture the drug bought time to build:

• Protein at every meal, front-loaded — the single most reliable way to raise your own post-meal GLP-1¹⁰³, and the foundation of lean mass.
• Fermentable fiber and a repaired gut, so the L cells that make your GLP-1 are working.
• Resistance training, permanently. This is the non-negotiable that holds body composition and the glucose sink.
• Meal sequencing, sleep, alcohol removal, and the stress work from earlier in this article.
• A slow taper rather than a stop, with labs and body composition tracked through it and a willingness to pause the taper if markers move.
• An honest agreement to revisit if the numbers drift back. Restarting is not a failure. Silently regaining and not telling anyone is.

Some patients will not be able to come off, and we said so above — that is a legitimate clinical outcome, not a character verdict. But we think the default should be the exit, planned from day one, rather than an open-ended prescription that nobody ever revisits. Section 14 above is the map for what comes after.

17. How we do this at Tree of Light

18. If you want to go further

These are the sources we most often point patients toward. We do not agree with every claim in every one of them — that is rather the point of reading more than one.

Listen

• “The Ozempic Expert: Ozempic Transforms Your Gut Microbiome! People Are Being Overdosed On Ozempic!” — the best long-form treatment of the two themes at the centre of this article: that standard doses are too high for many people, and that these medications reshape the gut microbiome in ways we are only starting to map.
• Dr. Mark Hyman and Dr. Tyna Moore, “What We Got Wrong About GLP-1s (And What’s Actually Right)” — two clinicians revisiting their own earlier positions in public, and the clearest available explanation of why the standard dose is not right for everyone.
• The Dr. Tyna Show (drtyna.com) — her series on getting these medications right is the most detailed patient-facing material on low-dose strategy that we know of.
• Huberman Lab — the GLP-1 and metabolic health episodes are unusually careful about separating what the drug does from what weight loss does, and about the confounders behind reported side effects. Good on the open questions, including what supraphysiologic GLP-1 signaling might mean for a developing brain.

Read

• Benjamin Bikman, Why We Get Sick — the clearest book-length case that insulin resistance sits underneath most modern chronic disease. If the argument in section 1 interested you, start here.
• Peter Attia, Outlive — the best general framework we know of for thinking about the diseases that actually kill people and what moves the odds decades in advance.
• Robert Lustig, Metabolical — on the food environment that produced the problem in the first place.
• Casey Means, Good Energy — a very accessible entry point, particularly on continuous glucose monitoring and what your own data can tell you.

On this site

• Natural GLP-1 agonists: probiotics, peptides and herbals
• Low-carb, high-fat eating and insulin resistance
• SIBO and the breath testing protocol
• Intestinal permeability and why the microbiome matters
• Protein and amino acids, and alcohol and optimal wellness

19. Questions we get asked

References

Sources are listed in order of first citation. Where a widely repeated claim was not supported by its stated source, that is noted in the text rather than omitted.