The GLP-1 receptor agonist class: biology, approved drugs, pipeline, and what we still do not know

Glucagon-like peptide-1 is a 30-amino-acid peptide cleaved from proglucagon in the L-cells of the distal small intestine and colon. It's released in response to nutrient ingestion, particularly carbohydrates and fats reaching the gut, and it does its physiologic work and is then gone, fast. Native GLP-1 has a circulating half-life of roughly 1 to 2 minutes (Holst, Physiological Reviews 2007) ^[1] because dipeptidyl peptidase-4 cleaves it at the second amino acid almost the moment it enters plasma. This is why the original drug development problem was not "find a GLP-1 mimic" but "find one that survives long enough to dose."

The hormone's receptor, GLP-1R, is a Class B G-protein-coupled receptor, and its tissue distribution is the reason the class behaves the way it does. Pancreatic beta cells express it; activation couples through Gαs to raise cAMP and trigger glucose-dependent insulin secretion, meaning the drug only drives insulin release when glucose is already elevated. Alpha cells express it; activation suppresses glucagon secretion, also glucose-dependent. This is the incretin axis, and it's the cleanest part of the pharmacology.

The rest of the receptor distribution is what made the class interesting. GLP-1R is expressed in the hypothalamic arcuate nucleus and the brainstem nucleus tractus solitarius, where it suppresses the orexigenic NPY/AgRP neurons that drive food-seeking and amplifies satiety signaling. It sits on vagal afferents and enteric neurons, which is how it slows gastric emptying - the so-called ileal brake. It's found on endothelial cells, where it appears to support nitric oxide production and dampen vascular inflammation. It's expressed in the kidney, where it drives natriuresis and modulates inflammatory tone. And it sits in the mesolimbic dopamine system - the ventral tegmental area and nucleus accumbens - which is the receptor footprint that has powered the substance-use and food-addiction work over the last three years ^[2].

This is a broad receptor distribution for what was originally framed as a gut hormone. Most of the class's therapeutic expansions are explained by it.

The mechanism of a GLP-1 receptor agonist is straightforward in principle and complicated in detail. The principle: occupy GLP-1R, signal through it, and do so for far longer than native GLP-1 ever could. The detail is in the engineering - fatty acid side chains for albumin binding (liraglutide, semaglutide), Fc fusion (dulaglutide), encapsulation in slow-release microspheres (exenatide ER), or, in the case of orforglipron, a non-peptide small molecule that binds the same receptor without DPP-4 vulnerability at all.

At the beta cell, the agonist drives glucose-dependent insulin secretion, which is the reason these drugs don't cause hypoglycemia as monotherapy. At the alpha cell, glucagon release is suppressed, again glucose-dependently. At the hypothalamus and brainstem, the drug reduces meal size and meal frequency through direct receptor activation. At the gut, gastric emptying slows, which contributes both to glycemic smoothing and to the nausea that limits dose escalation. At the endothelium and the kidney, the receptor signals appear to be directly anti-inflammatory, and these are the effects we now think underlie the cardiovascular and renal trial wins.

The mesolimbic effects are more interesting and less settled. Preclinical work has shown that GLP-1R activation in the nucleus accumbens attenuates dopamine release in response to alcohol, cocaine, nicotine, and palatable food. Whether the same attenuation is doing meaningful work in human substance use disorder is the open question of the moment, and we come back to it below.

Two structural variants matter. Tirzepatide is a dual GIP/GLP-1 receptor agonist; the added GIP receptor activity amplifies insulin secretion and (the working hypothesis) modulates adipose tissue handling in ways that drive deeper weight loss than pure GLP-1 single agonism. Retatrutide adds a third arm - glucagon receptor co-agonism - which raises basal energy expenditure on top of the appetite suppression. CagriSema pairs semaglutide with cagrilintide, an amylin analogue, on the theory that amylin's central satiety signal is additive to GLP-1's. The class is no longer pure. The mono-agonist era is ending.

The first agent, exenatide, was approved in 2005. It's a synthetic version of exendin-4, isolated from the saliva of Heloderma suspectum, the Gila monster, where the peptide had evolved to survive proteolytic degradation in ways native GLP-1 had not. Twice-daily injection, modest weight loss, modest HbA1c reduction. A useful drug, not a transformative one.

Liraglutide followed in 2010 for type 2 diabetes (Victoza) and in 2014 for obesity at the higher 3.0 mg dose (Saxenda). The 2014 obesity approval mattered: it was the first acknowledgement from FDA that this mechanism produced clinically meaningful weight loss in a non-diabetic population, even if the magnitude (5-8% body weight, on average) now looks small. Albiglutide came and went the same year. Dulaglutide arrived in 2014 and added once-weekly dosing, which mattered more for adherence than the trial literature usually captures.

The inflection point was semaglutide. Approved as Ozempic in 2017 for type 2 diabetes, the drug delivered HbA1c reductions and weight loss that exceeded liraglutide enough to reshape prescribing patterns within two years. Oral semaglutide (Rybelsus) followed in 2019 - a peptide engineered to survive gastric pH through co-formulation with the absorption enhancer SNAC, dosed daily on an empty stomach with rigid fasting requirements. In 2021 the higher-dose 2.4 mg formulation reached the obesity market as Wegovy, and the STEP-1 readout (-14.9% body weight at 68 weeks) ^[3] was the first time the class produced weight loss in the range that bariatric surgery had previously owned alone.

Tirzepatide changed the ceiling again. Approved as Mounjaro in 2022 for type 2 diabetes and as Zepbound in 2023 for obesity, the dual GIP/GLP-1 agonist produced -20.2% mean body weight at 72 weeks in SURMOUNT-5 ^[4], the head-to-head against semaglutide. It's the first time we have seen an obesity drug clear 20% mean loss in a randomized population.

Two regulatory expansions in 2024 mattered structurally. FDA approved Wegovy for cardiovascular risk reduction in overweight and obese adults without diabetes, the first weight-loss indication ever tied to a hard cardiovascular endpoint. And FDA approved Zepbound for obstructive sleep apnea in patients with obesity, a sign that downstream consequences of adiposity are starting to be treated as direct indications rather than as bonuses on a weight-loss label.

The trajectory, read honestly: a glucose-lowering drug class developed slowly, gained an obesity indication almost reluctantly, then over the last five years started delivering outcome data - cardiovascular, renal, sleep, hepatic - that have repositioned it as a metabolic-disease drug rather than a diabetes drug. Most of the surprises have favored the class.

The class is now broad enough that comparison matters. The table below summarizes the major approved agents on the dimensions that drive clinical choice: receptor scope, half-life, route, and dosing cadence.

A few patterns are worth flagging. Half-life has gotten longer over time, and that trend is the single biggest driver of patient adherence; weekly dosing is the floor now, and the orforglipron filing pushes toward once-daily oral as the next benchmark. Receptor scope has gotten broader, and the dual and triple agonists in development suggest mono-agonism is no longer the frontier. And the head-to-head efficacy data - tirzepatide outperforming semaglutide in SURPASS-2 on glycemic control and in SURMOUNT-5 on weight loss ^[5] - establishes that within the class, the dual agonist is currently the deeper-acting drug.

The interesting question is no longer "is there a better GLP-1 agonist coming." It's "what shape will the class take when it includes oral small molecules, triple agonists, and amylin combinations." Five agents are far enough along to matter.

Orforglipron is a non-peptide small molecule GLP-1 receptor agonist, oral, with no food or water restrictions - a meaningful contrast to Rybelsus, which requires fasting and tight timing. The Phase 3 ATTAIN-1 trial in obesity without diabetes (N≈3,127, 72 weeks) reported -12.4% mean body weight at the 36 mg dose, with 59.6% of patients losing at least 10%. ATTAIN-2 in obesity with diabetes hit real HbA1c and weight reductions and showed a 50.6% drop in high-sensitivity CRP at the highest dose, which is the inflammation-marker readout worth watching ^[6]. Lilly filed for FDA approval, and the regulatory decision has been delayed into 2026. If approved, this would be the first oral GLP-1 agonist with a meaningful obesity label, and the implications for compounding economics and prescribing patterns are large.

Retatrutide is the triple agonist (GLP-1, GIP, glucagon receptor). The Phase 2 readout reported -24.2% body weight at 12 mg over 48 weeks, with 36-48% of patients in the 8-12 mg arms losing at least 25% of body weight. No placebo patient reached that threshold ^[7]. This is the largest mean weight loss any agent in this class has produced in a controlled trial. The side-effect burden was also the highest among pipeline agents in head-to-head meta-analysis. The glucagon receptor arm is doing meaningful work, and it's the arm we understand the least at the tissue level.

CagriSema combines semaglutide 2.4 mg with cagrilintide 2.4 mg, an amylin analogue. Novo Nordisk has filed for Phase 3 readout-based approval. The mechanism here is mechanistically clean - two satiety signals from two receptor systems - and the trial efficacy data have been positive, though the headline weight loss did not exceed retatrutide's.

Survodutide (GLP-1/glucagon dual agonist, Boehringer Ingelheim) and mazdutide (GLP-1/glucagon dual agonist, Innovent) are the other glucagon co-agonists in Phase 3. Survodutide's Phase 2 MASH data have been particularly strong, with histological resolution and fibrosis improvement that exceeds what pure GLP-1 monotherapy has shown ^[8].

The pipeline read: weight loss is climbing toward the surgical range, oral formulation is solving the adherence problem, and glucagon co-agonism is emerging as the mechanistic addition that does the most work in hepatic disease. Tolerability remains the limiting factor, particularly for the triple agonists.

The most distinctive feature of this class is that the secondary indications keep arriving on actual outcome data, not on mechanism speculation. The cardiovascular reduction in SUSTAIN-6, then SELECT. The renal protection in FLOW. The OSA approval for tirzepatide. The MASH data. The sleep, liver, and kidney signals are no longer surprising - they're now part of how the class is positioned.

What follows in the next several sections is the honest read on the indications that are either approved on outcome data, approved on regulatory pathway, or generating signal strong enough to be discussed at scientific meetings. The Alzheimer's null result gets its own section because the field needs to absorb it.

The EVOKE and EVOKE+ trials enrolled 3,808 patients with mild cognitive impairment or early Alzheimer's dementia and randomized them to semaglutide versus placebo for three years. The trial read out in late 2024. Semaglutide did not outperform placebo on the primary cognitive endpoints. The safety profile was acceptable; the efficacy was null ^[9].

This matters more than a typical Alzheimer's trial failure because the prior probability had been pushed unusually high. The target-trial-emulation analyses out of electronic health records had reported 40-70% reductions in first-time Alzheimer's diagnoses in T2DM patients on semaglutide versus comparators, with hazard ratios as low as 0.33 against insulin. The preclinical work in murine and primate models had shown neuroprotective effects through mechanisms that looked plausible. A subset of the metabolic-psychiatric community had treated the question as nearly settled in semaglutide's favor.

EVOKE settled it the other way for early AD. The right inference is not that GLP-1 signaling does nothing in the brain - the receptor distribution in mesolimbic and hypothalamic circuits is well established and the substance-use signals are real. The right inference is that the observational signal was probably confounded by indication, by survival, and by the metabolic-health differential between patients who do and don't end up on a GLP-1 agonist. Healthy users get healthier diagnoses, and electronic health records don't see all of the reasons.

What EVOKE doesn't rule out is a role earlier in the disease trajectory, in patients with metabolic risk factors but no cognitive impairment yet, or as part of a combination regimen with amyloid-targeting agents. Those trials would take a decade. The honest read for now: the class doesn't slow established early AD at semaglutide-on-label doses. Stop quoting the observational hazard ratios as if they were trial data.

FLOW was a Phase 3b randomized trial of semaglutide 1.0 mg weekly in 3,533 patients with chronic kidney disease and type 2 diabetes, median follow-up 3.4 years, stopped early for efficacy. The composite primary endpoint - kidney failure, sustained 50% eGFR decline, or kidney/cardiovascular death - was reduced by 24%. Major adverse cardiovascular events fell 18%. Cardiovascular death fell 29%. All-cause mortality fell 20% ^[10].

These are large effects in a population with a high event rate, and the trial earned its early stop. In the hierarchy of GLP-1 outcome trials, FLOW sits alongside LEADER and SUSTAIN-6 in changing prescribing patterns - and arguably above them, because the renal effect is mechanistically novel and because it adds a population (CKD patients) where prescribers had been more cautious about glycemic agents in general.

The pre-specified renal analysis of SELECT extended the signal to a different population: overweight and obese patients without diabetes, with established cardiovascular disease. Major kidney events dropped 22%. This is the first evidence of renoprotection in a non-diabetic, non-CKD population, and it's the reason renal protection is now framed as a class effect rather than a diabetes-specific finding.

What we don't yet know: whether tirzepatide and the dual agonists produce comparable renal benefit, whether the effect is durable past 3-4 years, and whether it extends to non-diabetic CKD without obesity. The trials to answer these are running.

Metabolic dysfunction-associated steatohepatitis is the indication where the dual and triple agonists are showing the most differentiation from pure GLP-1 monotherapy. Liraglutide and semaglutide both produce histological MASH resolution at meaningful rates, but fibrosis improvement - the harder endpoint, and the one that tracks long-term liver outcomes - has been limited with monotherapy.

Tirzepatide's SYNERGY-NASH Phase 2 trial reported higher rates of both MASH resolution and fibrosis improvement than the GLP-1 monotherapy literature, which is the first signal that the GIP arm contributes specifically to hepatic outcomes. Survodutide, the GLP-1/glucagon dual agonist, has produced what may be the strongest Phase 2 MASH data in the class - both resolution and fibrosis improvement - and the glucagon receptor activity is the leading mechanistic explanation, given glucagon's known effects on hepatic lipid handling ^[8].

The regulatory context: resmetirom's approval in 2024 established that FDA will accept histological surrogate endpoints for MASH approval in advance of long-term outcome data. This opens a pathway that GLP-1 and incretin-based therapies are now actively pursuing. The likely shape of the next two years: at least one Phase 3 readout in MASH from a GLP-1-based regimen, probably positive on resolution, possibly positive on fibrosis, leading to a label.

The honest framing: MASH is where the class moves from "treats metabolic disease" to "treats organ damage from metabolic disease," which is a different and more consequential clinical position.

The substance use disorder data are where editorial caution matters most, because the patient-facing enthusiasm is well ahead of the trial literature. Here is what actually exists.

A 12-week randomized trial of dulaglutide 1.5 mg weekly in alcohol use disorder reported a real reduction in alcohol intake versus placebo (N=76/75, p=0.04). This is the cleanest piece of evidence in the file. It's one trial, modest in size, with a behavioral endpoint that's difficult to measure rigorously.

A 4-month prospective observational study in 113 patients on semaglutide reported food addiction prevalence (measured by the Yale Food Addiction Scale) falling from 57.5% to 4.2%. This is a striking effect size in an unblinded design where weight loss and food-reward changes are confounded with the measurement instrument itself.

Retrospective cohort work in over 500,000 patients with opioid use disorder and over 800,000 with alcohol use disorder reported lower rates of opioid overdose and alcohol intoxication in those prescribed GIP/GLP-1 agonists versus comparators ^[11]. These analyses are subject to the same confounding issues as the early Alzheimer's observational work - healthier patients are prescribed metabolic drugs and have lower event rates for many reasons.

The preclinical case is genuinely strong: rodent models reliably show that semaglutide and exendin-4 reduce alcohol consumption, relapse-like drinking, cocaine reward, nicotine intake, and amphetamine-seeking, and the mesolimbic dopamine attenuation is a mechanistically clean explanation. The receptor is in the right place. The signal is consistent across drugs and species.

What is missing: adequately powered Phase 3 randomized trials with hard substance-use endpoints. NCT05520775 and similar trials are running. Until they read out, the appropriate framing is that GLP-1 agonists have a biologically plausible mechanism and consistent preclinical effect on reward circuitry, that early human data are suggestive, and that none of this constitutes evidence to prescribe these drugs for substance use disorder. The EVOKE experience should be a permanent caution against assuming the observational signal will replicate in randomized trials.

Liraglutide 3.0 mg (Saxenda) was approved for adolescents aged 12 and older with obesity in 2020. Semaglutide 2.4 mg (Wegovy) received the same indication in 2022 based on STEP TEENS, which reported a -16.1% mean BMI reduction at 68 weeks in adolescents - efficacy comparable in magnitude to the adult STEP-1 readout. The adolescent safety profile has tracked the adult profile, with the same gastrointestinal dominance and the same lack of confirmed thyroid signal.

The pediatric data raise questions the class is not yet equipped to answer. We don't know what 30 years of GLP-1 receptor signaling does to a body that started the drug at 14. We don't know whether the appetite and reward-system effects shape developing eating behavior in ways that persist or rebound after discontinuation. We don't have long-term bone density, fertility, or growth-trajectory data. The clinical case for adolescent obesity pharmacotherapy is strong on near-term morbidity grounds - adolescent obesity tracks into adult disease - but the long-tail uncertainty is real and should be named.

Tirzepatide pediatric trials are ongoing. The pediatric label will follow if efficacy and safety hold.

The most interesting mechanistic claim in the field is that GLP-1 agonists don't simply suppress appetite - they lower the body's defended weight, the homeostatic set point around which energy balance is regulated. The hypothesis predicts that patients on these drugs settle at a new, lower steady-state weight; that the defense against further loss kicks in at that lower weight rather than at the original one; and that the rebound after discontinuation reflects the set point reverting, not the drug effect simply ending.

The mechanistic plausibility is real. Hypothalamic GLP-1R signaling demonstrably modifies NPY/AgRP and POMC neuron activity, and these are the circuits classically implicated in set-point defense. The leptin and ghrelin axes interact with GLP-1 signaling in ways that could shift the homeostatic equilibrium rather than merely override it. STEP and SURMOUNT extension data show that patients maintain weight loss as long as they remain on therapy, with weight plateaus rather than continued loss, which is what a reset hypothesis would predict.

Here is where the hypothesis breaks down. The STEP-1 extension showed that patients who discontinued semaglutide regained roughly two-thirds of lost weight within a year. SURMOUNT-4 showed comparable rebound after tirzepatide withdrawal. If the set point had truly reset, you would expect the new steady state to persist after the drug was removed - that's the definition of a defended weight. What the discontinuation data actually show is that the drug holds appetite suppressed, holds gastric emptying delayed, holds satiety signaling amplified, and when you remove it, the homeostatic system returns to where it was. The defended weight doesn't appear to have moved. The drug is doing the defending.

The careful reading, which is where most thoughtful people in the field have landed: GLP-1 signaling produces a state that functions like a lower set point for as long as the drug is on board, but the underlying homeostatic architecture is unchanged. This has prescribing implications. It means these are chronic therapies for a chronic condition, in the same sense that antihypertensives are. The "reset" framing oversells what the drugs do, even if the day-to-day clinical experience genuinely feels like a reset to patients taking them.

This section exists because the confusion is real and consequential. Several peptides are routinely marketed alongside GLP-1 agonists, or framed in the same conversation, and they don't bind the GLP-1 receptor.

Tesamorelin is a growth hormone-releasing hormone (GHRH) analogue. It binds the GHRH receptor on pituitary somatotrophs and drives endogenous growth hormone release. Its FDA-approved indication is HIV-associated lipodystrophy, and its mechanism has nothing to do with the incretin axis. The visceral fat reduction it produces comes from GH-driven lipolysis, not from appetite suppression or insulin sensitization through GLP-1 signaling.

AOD-9604 is a synthetic fragment of human growth hormone (amino acids 176-191), originally developed as an anti-obesity agent and now sold as a research peptide. The mechanism claim is enhanced fat oxidation through a growth-hormone-related pathway. The evidence base in humans is thin, the original Phase 2 trials were negative on the primary weight-loss endpoint, and it's not a GLP-1 receptor ligand by any reading.

MOTS-c is a mitochondrial-derived peptide that signals through AMPK and influences metabolic homeostasis at the cellular level. It doesn't bind GLP-1R. Its physiology is interesting and largely preclinical. Conflating it with the incretin class is a category error.

The source of the recurring confusion is mostly commercial. Compounded-peptide retailers tend to market across categories, and the marketing copy collapses distinctions that the pharmacology preserves. A useful test: if a peptide is described as "GLP-1-like" or "with similar effects to semaglutide" without naming the receptor it binds, it almost certainly doesn't bind GLP-1R, and the marketing is the data you are being shown.

The approved-indication footprint of the class has expanded faster than the regulatory framework around it has stabilized. As of mid-2026, the approvals include type 2 diabetes (every agent), obesity (liraglutide 3.0, semaglutide 2.4, tirzepatide), cardiovascular risk reduction in established CVD (semaglutide for both diabetic and non-diabetic populations, dulaglutide for diabetes), obstructive sleep apnea with obesity (tirzepatide), and chronic kidney disease with diabetes (semaglutide, on the FLOW data). Adolescent obesity indications cover liraglutide and semaglutide.

The compounding question has been the more contentious regulatory thread. During the 2023-2024 semaglutide and tirzepatide shortages, 503A and 503B compounding pharmacies produced large volumes of compounded incretin-class products, often in formulations and combinations that bypassed brand-name pricing and label restrictions. The FDA removed semaglutide and tirzepatide from the drug shortage list in 2024 and 2025 respectively, which triggered restrictions on compounding under the relevant statutory authorities. In December 2024 the American Diabetes Association issued a formal statement recommending against use of non-FDA-approved compounded GLP-1 and GIP/GLP-1 products, citing content, quality, safety, and effectiveness concerns ^[12]. Enforcement against outsourcing facilities has tightened through 2025-2026.

The orforglipron filing sits in the middle of this. Lilly submitted for FDA approval in late 2025, and the regulatory decision has been delayed into 2026. Approval would produce the first oral small-molecule GLP-1 agonist with a meaningful obesity label, changing both prescribing patterns and the economics of the compounding market.

The class has a remarkably consistent adverse event profile. Gastrointestinal effects - nausea, vomiting, diarrhea, constipation - dominate, are dose-dependent, peak during titration, and account for most discontinuations. In head-to-head trials, tirzepatide and semaglutide produce GI adverse events at broadly similar rates (~17-22%), with discontinuation rates in the same range. Orforglipron's Phase 3 GI burden was somewhat higher at the top doses, with nausea up to 36.4% and vomiting up to 23.1% versus single-digit placebo rates.

The shared serious-event flags are well established. Pancreatitis carries a class warning, though the absolute risk in trials has been low and the original signal from observational data has not fully replicated in RCTs. Medullary thyroid carcinoma carries a boxed warning across the class on the basis of a rodent C-cell tumor signal that has not been confirmed in humans; the contraindication for patients with MEN2 or personal/family history of medullary thyroid cancer is standard. Gallbladder disease - cholelithiasis and cholecystitis - is elevated, plausibly mediated by both direct biliary effects and rapid weight loss. Acute kidney injury occurs and tracks dehydration from severe GI losses. Heart rate increases of 2-4 bpm are consistent across agents ^[13].

Drug-specific concerns: tirzepatide's hypoglycemia rate sits between 0.2% and 1.7% across populations, slightly elevated above semaglutide's 0.4%, attributable to the GIP arm. The retatrutide Phase 2 safety profile was the highest-burden among pipeline agents in meta-analysis, consistent with triple-receptor activation amplifying both efficacy and side effects.

What we don't know, honestly: long-term effects of multi-year GLP-1R activation on lean body mass, bone density, gastrointestinal motility after eventual discontinuation, and pancreatic exocrine function. The trial follow-up periods are mostly 1-3 years; the prescribing patterns suggest patients will be on these drugs for decades. The thyroid signal in humans remains unresolved despite repeated reassuring analyses. The cancer epidemiology - the 2024 retrospective work showed reduced risk for most GI and reproductive cancers but an increased kidney cancer signal versus metformin - needs RCT-derived data the class is unlikely to generate ^[14].

The honest read on safety: this is a well-tolerated class with a manageable adverse event profile in trials of 1-3 years' duration, and the longer-term unknowns are real and should be acknowledged in patient counseling.

A short list of what the field doesn't yet know, ordered by how much I think it matters.

Whether the substance use disorder signal replicates in adequately powered randomized trials with hard endpoints, and whether the effect is durable or dependent on continued dosing. The mechanistic case is strong; the human trial case is suggestive; the evidence is not there yet.

Whether the renal protection seen in FLOW and SELECT extends to non-diabetic CKD populations and to tirzepatide and the dual agonists. The class effect framing is plausible but currently rests largely on semaglutide data.

Whether the next generation of triple agonists can deliver retatrutide-level weight loss with tolerability profiles that allow chronic use. The Phase 3 readouts over the next two years will answer this.

Whether the MASH approvals materialize on histological surrogates, and whether long-term hepatic outcome data confirm fibrosis benefit.

Whether oral small-molecule GLP-1 agonism (orforglipron and successors) shifts prescribing patterns enough to change the public-health footprint of the class. The injection barrier is real for a meaningful fraction of patients, and an effective daily oral with no food restrictions would change adoption curves.

Whether the set-point reset framing survives longer-term discontinuation data, or whether it continues to look like persistent appetite suppression rather than altered homeostasis. This matters for how we frame these drugs to patients and to payers.

What 20 years of GLP-1R agonism does, in any tissue. We don't have the data. We won't have it for a long time. Prescribers and patients are making a long-tail bet that the early safety signal will hold, and so far it has, but "so far" is a 20-year-old class with most of its prescribing volume concentrated in the last five years.

The class has done more than most of us expected when exenatide was approved in 2005. It has also failed in places (EVOKE) where the prior probability had been pushed too high by observational data. The reasonable posture toward the next five years is the same as the reasonable posture toward the last five: take the outcome trial data seriously, hold the observational data lightly, and assume the indications will keep expanding faster than the long-term safety database can keep up with them.