GLP-1s

3 Bullet Point Summary 

• This is the most impactful class of medication released since I finished training. I say that as someone who was skeptical of it for a decade, for reasons that turned out to be wrong.

• The drug is not the plan. Roughly 85% of people started on one for weight loss are off it within two years, and the trial data say most of the weight comes back when they stop. The people who keep the result are the ones who built something underneath it.

• The right question is never "should I be on a GLP-1." It's what you're trying to accomplish — performance, quality of life, longevity, or disease management — and whether this drug, at this price, is the best available way to get there.

Let me start with the part that is easy to miss in the noise.

I have been in practice long enough to watch a lot of drugs get launched with a press release and quietly disappear. This class did not do that. In terms of effect size — how much the drug actually changes the thing you're measuring — GLP-1 receptor agonists are the most impactful medications I have seen released since I was in training. That is not a marketing statement. It is a comparison against every other class I prescribe.

It is also not an endorsement, because effect size is only half of the equation. The other half is what you do with it.

What a GLP-1 actually is

The class belongs to a family of hormones called incretins. Incretins are released from the gut in response to food, and their job is timing. They tell the pancreas to release insulin around a meal, when there is glucose arriving that needs somewhere to go.

Insulin is the most anabolic hormone in the body. Its function is storage. Glucose gets shuttled into the liver as glycogen, into fat as fat, or into muscle where it fuels contraction. When people say "insulin resistance," what they're describing is how efficiently that shuttling works. Resistance means the pancreas has to shout to get the same result.

Here is why I was skeptical of this class for years. In diabetes management, there are drugs that increase insulin release and drugs that don't, and the ones that increase insulin release reliably cause weight gain. There's a long-running argument in the diabetes literature about whether the complications of diabetes — the heart attacks, the strokes, the kidney disease — are driven more by the elevated blood sugar or by the elevated insulin that's been chasing it for twenty years. The honest answer is probably both. So when a new class arrived whose proposed mechanism was more insulin, my reaction was: I don't think I want that.

I was wrong, but not for the reason you'd assume. I was wrong because the insulin story is largely not why the drug works.

The order they arrived, and why the names are a mess

Exenatide came first, branded Byetta, back in the mid-2000s. The weight loss was real and unremarkable. Liraglutide followed around 2010 — modest, still nothing that would change how anyone practiced. Then semaglutide hit, branded Ozempic, and people started losing weight in a way nobody had seen from a pill or an injection before.

Early adopters did what early adopters do. Ozempic was approved for type 2 diabetes and only for type 2 diabetes, so the weight-loss use was off-label, which meant it belonged to whoever could pay cash for it. Hollywood found it before primary care did.

This is where the naming gets genuinely confusing, and it confuses clinicians too. The FDA approves a drug for an indication — a specific purpose. The same molecule can carry a different brand name for each indication. Semaglutide is Ozempic for diabetes and Wegovy for weight management. Tirzepatide is Mounjaro for diabetes and Zepbound for weight management. Same chemical, two labels, two marketing departments. It happens all over medicine: bupropion is Wellbutrin as an antidepressant and Zyban for smoking cessation. Medical students are mostly taught generic names to limit conflict of interest, which is good policy and leaves everyone slightly unequipped for a conversation with a patient who read about a brand on Instagram.

The pill version is its own case study in branding. Oral semaglutide launched as Rybelsus right around 2020, arrived with the pitch that nobody wants an injection, and landed in a market that had already decided it very much did want the injection. It never took. Novo eventually stopped fighting it and rebranded oral semaglutide under the name people already recognized — the Ozempic pill — and then brought an oral version through for weight management as well. The molecule didn't change. The name on the box did.

The newest entry isn't approved yet. Retatrutide is Lilly's triple agonist — it hits GIP, GLP-1, and glucagon receptors — and its first phase 3 results read out in May of this year at roughly 28% average weight loss over 80 weeks, with a subgroup reaching about 30% at two years. For context, 30% is the neighborhood people used to have to go to surgery for.

What the drug is probably actually doing

Officially, the mechanism is incretin biology: better-timed insulin, better partitioning of the energy you eat.

Anyone who has prescribed this class knows that isn't the main event. The main event is appetite. People eat less, and — more interestingly — they stop thinking about food. Patients describe it as the food noise going quiet. The craving still registers, but the payoff shrinks. You eat the thing you wanted, you notice it was fine, and you don't go back for four more.

That matters more than it sounds, because the failure mode it interrupts is a real one. Most people are not overeating at dinner. They are stressed all day, they finally sit down at ten at night, the sympathetic nervous system lets go, hunger arrives all at once, and they stand in the pantry. Then they wonder why the scale won't move. A drug that quiets that hour is doing something a nutrition handout has never once accomplished.

Why this is a metabolic conversation, not a weight conversation

The people who benefit most are the people who are the most metabolically unhealthy. That sentence covers almost everything I'd say about selection.

The problem is that we're bad at identifying them. The standard test is hemoglobin A1c, a three-month average of blood sugar, and an average is exactly the wrong tool for a disease of volatility. If you spike to 250 after a meal and drift low on the back end, your average can look fine. Worse, if your pancreas is compensating heroically — pumping out enough insulin to jam that sugar into fat and muscle — your blood sugar can look normal for years while the underlying problem gets steadily worse.

The numbers on this are worse than most clinicians appreciate. Against an oral glucose tolerance test, the sensitivity of A1c for diagnosing diabetes has been reported at around 41%. It misses more than half. It misses them unevenly across ethnic groups. A normal A1c is not a clean bill of metabolic health, and I have stopped treating it like one.

Which brings me to polycystic ovary syndrome, where I think this class is close to a no-brainer. PCOS presents as irregular cycles, androgenic features, difficulty losing weight, and trouble conceiving — and for decades the entire treatment menu was birth control and metformin. Young women were told, sometimes in one visit, that they would always be heavy and probably couldn't get pregnant. There's been a long-running push in the field to rename the condition around what actually drives it, because the underlying driver in most cases is insulin resistance and the ovary is downstream. The name hasn't formally changed. The understanding has. And when these women get metabolically healthier, a lot of things get better at once.

When I don't start it

My default is to do it without the drug, and that isn't philosophical stubbornness. It's a bet about durability: the slower the weight comes off, the more likely it stays off, because the thing that took it off was a set of behaviors rather than a prescription.

So if a patient is actively making changes and the changes are working, I generally don't add anything. Don't fix what isn't broken.

The trap is that most people can't tell whether it's working. Here's the version I see constantly: someone decides to get serious, starts training hard, cuts calories aggressively, and is now carrying increased physical stress plus a significant deficit on top of a life that was already stressful. Then every meal becomes a moral event — good food, bad food, guilt either way. Three months in, the scale hasn't moved and they conclude they're broken.

They're usually not. Body recomposition almost always precedes weight loss. Ten pounds of muscle on, ten pounds of fat off, and the scale reports nothing happened. This is why we measure body composition in the clinic. I can hold up an objective number and tell someone that what they're doing is working, which is a better intervention than a prescription and costs considerably less. The labs usually agree with the body composition before the scale does.

The part nobody plans for

Now the uncomfortable arithmetic.

About 85% of people who start a GLP-1 for weight loss are no longer taking it two years later. That is a real-world claims finding, not a trial artifact, and it's the single most important number in this article.

What happens after they stop is where I want to correct something I've said myself. The randomized trial extensions are stark: when the drug is withdrawn, most of the lost weight returns within a year, along with much of the metabolic improvement. I have summarized that in conversation as "90% of people regain all of it," and that's too clean. A 2026 Cleveland Clinic analysis of nearly 8,000 real-world patients found something more textured — a year after stopping, the obesity group had regained about half a percent of body weight on average, and 45% had maintained or continued losing. The catch is why: a fifth restarted the same drug, a quarter switched to another obesity medication, and others moved into supervised lifestyle programs. Real people don't stop and do nothing. They stop and do something else.

The honest synthesis is this. Stop the drug and change nothing, and the trial data are right — it comes back. Stop the drug into an actual plan, and it largely doesn't. The medication was never the intervention. It was the window.

That framing also explains why payers behave the way they do. If most people discontinue, and most of those regain, then from the perspective of whoever's paying for the eventual heart attack, the success rate looks thin. Meanwhile bariatric surgery — which for years cost about what two years of branded GLP-1 cost at launch prices — has decades of outcome data showing durable weight loss plus reductions in diabetes, cancer, and death. That comparison is less favorable to the drug than the internet thinks, and it deserves a real hearing rather than a reflexive dismissal.

How I actually dose it

By the book, you start at the introductory dose, escalate every four weeks, and keep climbing to the maximum.

I don't do that. If a patient is losing weight, I don't increase the dose. I keep people on the lowest dose that is still producing a result, for two reasons. My goal is never to keep someone on a medication for life — don't take what you don't need — and as a general rule, the harder you push on one biological pathway, the more you eventually pay for it in side effects.

Maintenance dosing, or what the internet calls microdosing, I use constantly. When people come all the way off, a lot of them regain. A low maintenance dose is frequently the difference between keeping the result and starting over.

Do we have long-term safety data for that? No. We have about a decade of use at scale and no thirty-year cohort, because the drug isn't thirty years old. So the comparison you're actually making is between a known risk and an unknown one. I know what carrying excess weight and elevated blood sugar for twenty years does — I watch it do it. I don't know what twenty years of a GLP-1 does. Reasonable people weigh those differently, and I'd rather say that plainly than pretend to a certainty I don't have.

There's a broader version of this worth naming. We have gotten used to being able to look anything up, and we've stopped tolerating the answer we don't know yet. That answer is not evasion. It's the actual state of the evidence on a lot of things, and learning to make decisions inside it is a skill.

Muscle, protein, and the hollowed-out look

What people call Ozempic face is mostly one thing: aggressive dosing, insufficient protein, and no training stimulus.

Skeletal muscle clears roughly 80% of insulin-mediated glucose. It is the largest and most effective metabolic organ you own. Resistance training upregulates the glucose transporters in muscle, which means a trained body preferentially refills muscle instead of fat with the same meal. This is the whole premise behind what Gabrielle Lyon has branded muscle-centric medicine, and the premise is right: if your muscle doesn't work, nothing downstream of it works either.

So the failure mode is predictable. Crank the dose, suppress appetite to near zero, eat nothing, train nothing, and the body spends muscle. I am now seeing protein-calorie malnutrition in American adults on these medications — dizziness, weakness, brain fog, labs that look like starvation, because that is what it is. The gaunt appearance isn't the drug. It's the depletion.

The fix is boring and it works. Eat protein like it's the job. Lift. If appetite is so suppressed that you cannot get protein in, the dose is too high, and that is a dosing error, not a personality failure.

Price, and why value is a ratio

Value is effect size divided by cost, and the denominator on this class has been moving fast.

At launch, branded GLP-1s ran about $1,200 a month cash. Then the manufacturers opened direct-to-patient channels, which brought it to roughly $500. Then the federal pricing deals took another bite. As of now, tirzepatide runs roughly $300 to $450 a month depending on dose through Lilly's direct channel, semaglutide is in the mid-$300s through the federal platform, the Medicare and Medicaid price is $245, and the oral agents are landing lower still. Medicare is now covering weight management in a way it explicitly wasn't two years ago. The effect size hasn't changed. The price fell by roughly two-thirds. So the value went up, materially.

That makes the clinical decision harder, not easier. Consider the alternative I'd have reached for two years ago: dapagliflozin, an SGLT2 inhibitor, now fully generic at around $10 a month. It makes you excrete glucose in your urine, protects the kidneys, protects the heart, and produces mild weight loss. Judged as effect per dollar, that used to be an easy win. At $10 versus $100 a month — which is where a GLP-1 lands for some people now — the gap has narrowed enough that I can no longer say one is obviously better value than the other. I consider that a good problem.

One note on the compounding era, since it shaped how a lot of people got access. When demand outran supply, the shortage designation let compounding pharmacies produce these molecules legally, at lower prices, at enormous scale. When the shortage was declared resolved, that door closed, and the manufacturers began enforcing against compounders, distributors, and prescribers. The peptide space changes legal status faster than almost anything I deal with — a product is available, then it isn't, then it is — and none of that movement has anything to do with biology.

Who's writing the prescription

Peter Attia has made a point worth repeating: in a perfect world, the person recommending an intervention has no financial stake in whether you take it. Talk to a surgeon about your back, and you are more likely to have surgery.

The margins in the peptide business are high enough that this bias is everywhere. A large share of weight-loss clinics make more money selling and distributing the drug than from the clinical encounter itself. There is a plain term for a business model whose revenue depends on you continuing to buy the substance, and drug dealers do not have your long-term interests at heart. Find someone whose compensation doesn't depend on your answer.

The corollary matters just as much, and it cuts against how a lot of physicians practice. If someone has decided they want something, they are going to get it. They can order it online this afternoon. The moment I refuse to have the conversation, I have not prevented anything — I've just removed myself from it, and sent them one step down a supply chain that gets less regulated at every stop. FDA-approved manufacturer, to compounding pharmacy, to research-chemical website: at each step you lose quality control and increase the odds of something in the vial that isn't on the label.

I would rather someone be doing a thing I have reservations about while I'm monitoring them than be doing it alone. That has been true of testosterone for years. 

The framework

When someone asks whether they should be on a GLP-1, the question I ask back is: what are we trying to accomplish? There are four answers, and they set different bars.

Performance — getting from fifteen percent body fat to eight. There's no health benefit there. It's a performance goal, and it's legitimate as long as we're both honest that it's a performance goal.

Quality of life. Longevity. Disease management. Each of those has its own evidence base and its own threshold for what's worth the tradeoff.

Where it's clearly indicated: type 2 diabetes, class III obesity, established cardiovascular disease, PCOS. Those are close to black and white, and the trials back them.

Where I push back: when paying for it would compromise your financial security. I've written before that financial security is the most underrated determinant of long-term health. Taking on medical debt for a drug is not a metabolic intervention. And in the elderly, or anyone already low on muscle — if losing weight means you can't get off the toilet, this is the wrong drug at any price.

Everyone else is in the middle, which is most people reading this. For them, my answer is that improving metabolic health is worth doing, we have precedent for treating risk factors early to lower long-term risk, and we don't yet know whether this particular version of that is cost-effective over a lifetime. So we decide together, using the literature, my clinical experience, and what you actually value.

I care about whether we made the right decision given the context — not about whether the answer was yes or no.

Which is the actual takeaway. Stop sorting medications into good and evil. Get educated, figure out what you're solving for, and decide whether the value is there for you.

Three Practical Take Aways 

1. Get a real metabolic picture before you decide anything. A normal A1c is not a clean bill of health — it misses more than half of dysglycemia when checked against a glucose tolerance test. Ask for fasting insulin, and ask what your post-meal numbers look like. You cannot evaluate whether a metabolic drug is worth it without knowing how metabolically healthy you actually are.

2. Build the exit before you start. The number that should worry you isn't the side-effect profile, it's that 85% of people are off this drug within two years. Decide now what the protein target is, what the training looks like, and whether the plan is a maintenance dose or a taper. If the answer is "I'll figure that out later," you have bought a window and no plan for what to do inside it.

3. Ask who profits from your answer. Before you take a recommendation, find out whether the person making it is paid based on what you decide. If most of that clinic's revenue comes from selling you the vial rather than from taking care of you, get a second opinion from someone with nothing riding on it.