In January 1975, Roger Unger and Lelio Orci made an unfashionable argument in The Lancet: diabetes was not simply a disease of missing insulin. They argued that glucagon is relatively or absolutely elevated in every form of endogenous hyperglycemia and proposed a two-hormone model: lack of insulin leaves glucose underused, while excess glucagon drives its overproduction.¹ By 2012, Unger was calling glucagon excess the sine qua non of diabetes, a view that remains debated.²

So industry built pills to block the glucagon receptor. Merck’s MK-0893 lowered A1C by up to 1.5 percentage points at its 80-mg dose over 12 weeks and raised LDL cholesterol dose-dependently.³ Lilly’s LY2409021 raised liver aminotransferases, modestly and reversibly.⁴ Neither reached the market.

Today the most potent obesity drug in late-stage testing activates that same receptor on purpose. Retatrutide, which also activates the GIP and GLP-1 receptors, produced 25.0% mean weight loss at 80 weeks in TRIUMPH-1 by the treatment-regimen estimand,⁵ which I covered on September 30 in Everyone Will Quote 28%. Read Page Four.

Amylin took a quieter road. It was isolated from the amyloid that clogs islets in type 2 diabetes,⁶ and the only approved amylin drug, pramlintide, is injected before each major meal and moved weight by about 1 to 1.5 kg in its label trials.⁷ Now cagrilintide paired with semaglutide is under FDA review,⁸ and amylin agonists from Lilly and from Zealand with Roche are in Phase 3.^9,10

The myth worth retiring: that the next generation of obesity drugs is GLP-1 with the volume turned up. Glucagon and amylin act through distinct receptor systems, mostly in different places: glucagon in the liver, amylin in a patch of brainstem outside the blood-brain barrier. This issue is the mechanism, with the trials as supporting evidence, and it labels what comes from physiology, what comes from animals, and what has been shown in patients.

If you have paid time left on an individual Granola plan, we'll match it with a Wispr Flow subscription that includes Notetaker and dictation, and add bonus time, up to 12 months total. Sign in or create a Wispr account and submit proof of your plan to check eligibility.

Try Notetaker Free

Glucagon and GLP-1 come from the same gene. In the pancreatic α-cell, prohormone convertase 2 cuts proglucagon into glucagon; in the intestinal L-cell, prohormone convertase 1/3 cuts it into GLP-1, GLP-2, glicentin, and oxyntomodulin.¹¹ One gene, two hormones with what look like opposite agendas: glucagon raises glucose, GLP-1 lowers it.

Oxyntomodulin is the tell: glucagon with a short tail that activates both receptors. Nature had already built a co-agonist. In a 4-week trial, people injecting it three times a day before meals lost 2.3 kg against 0.5 kg on saline.¹² The biology worked in people. A peptide that lasts minutes in plasma could not be a weekly drug.

Glucagon’s receptor sits mostly in the liver, and one signaling chain carries its instructions there. It is a class B G protein-coupled receptor that couples to G_s, raises cyclic AMP, and switches on protein kinase A.¹¹ From that one kinase come three well-established instructions and a fourth that is proposed.

The first is the famous one: break down glycogen within minutes and, through the transcription factor CREB and its coactivator CRTC2, switch on gluconeogenic genes over hours. That is the half of glucagon that made it the villain.

The second is the one obesity medicine cares about: the hepatocyte stops building fat and starts burning it. Protein kinase A, with AMPK contributing, inhibits acetyl-CoA carboxylase, so malonyl-CoA falls. Malonyl-CoA is the brake on CPT1, the gate that admits long-chain fatty acids into mitochondria. Release the brake and fatty acids go to β-oxidation and ketone production.¹¹

The third reaches beyond glucose. Glucagon promotes hepatic amino-acid breakdown and disposal of their nitrogen as urea.¹³

The fourth is a proposed pathway, and it shows why acute physiology cannot simply be carried forward into chronic treatment. In healthy volunteers, a glucagon injection raised circulating FGF21 within hours, and in mice lacking FGF21, or lacking FXR in the liver, a glucagon agonist lost its weight effect.^14,15 Yet in retatrutide’s MASLD study, circulating FGF21 was lower at higher doses, and the authors considered increased circulating FGF21 an unlikely explanation for the liver benefit.¹⁶

Energy expenditure is another acute finding. In 11 healthy men, a glucagon infusion raised energy expenditure 15% without detectably activating brown fat.¹⁷ Whether that persists through months of co-agonist dosing in people with obesity has not been shown.

Amino acids stimulate the α-cell to release glucagon, and glucagon tells the liver to clear them. That closed loop is the liver–α-cell axis.¹³ Block the receptor and the loop opens: fasting glucose falls, but amino acid turnover slows and lipid handling shifts. MK-0893’s LDL rise was linked to increased cholesterol absorption alongside higher GLP-2, and ezetimibe abolished it in mice.³ Glucagon is not only a glucose hormone. It is the liver’s fuel-clearing signal.

That reframes fatty liver. Richter and colleagues propose glucagon resistance in MASLD: the glucose-raising arm keeps working while the amino acid arm is impaired, so the α-cell keeps secreting and the liver keeps making glucose.¹³ They argue this may help explain the high glucagon levels of type 2 diabetes. If they are right, some hyperglucagonemia starts in a fatty liver, not only in a broken islet.

At weight-loss doses, glucagon alone would push glucose up. In a co-agonist, the GLP-1 arm releases insulin in a glucose-dependent way and covers that side. In 2009, a glucagon and GLP-1 co-agonist normalized adiposity in obese mice through both lower food intake and higher energy expenditure.¹⁸ Since then the design question has been the ratio: retatrutide has roughly balanced glucagon and GLP-1 activity with stronger GIP activity,¹⁹ and survodutide is a glucagon and GLP-1 dual agonist.²⁰

A 2026 paper from Boehringer Ingelheim’s team asked where survodutide acts in the brain. In human and mouse tissue, they measured receptor expression in two appetite-related regions, the area postrema and the arcuate nucleus. The glucagon receptor was virtually undetectable; the GLP-1 receptor was present.²¹ The functional experiments were in mice: a long-acting pure glucagon agonist did not activate satiety-associated neurons or reduce acute food intake, yet still lowered body weight.

These findings support a predominantly GLP-1-mediated appetite effect for survodutide in the systems studied, alongside peripheral contributions from glucagon. Low receptor detection in selected regions does not establish the absence of all central or indirect appetite effects, and the functional work has not been repeated in people.

The liver is where glucagon-containing drugs stand out. In retatrutide’s Phase 2 MRI substudy, liver fat fell 82.4% on 12 mg by 24 weeks, and 86% reached a normal liver fat fraction, against 0% on placebo.¹⁶ In survodutide’s Phase 2 biopsy trial, MASH improved without fibrosis worsening in 62% on 4.8 mg against 14% on placebo; 6.0 mg reached 43% in that primary analysis, which counted participants without a week-48 biopsy as non-responders.²² In Phase 3 SYNCHRONIZE-MASLD, 84.2% on survodutide 6.0 mg cut liver fat by at least 30% at 48 weeks, against 24.3% on placebo (efficacy estimand).²³

What these trials cannot do is isolate glucagon. Each drug also activates the GLP-1 receptor, retatrutide activates GIP as well, and weight loss by itself lowers liver fat. None randomized people to the same molecule with and without glucagon activity.

The weight results sit in the table further down: survodutide 6.0 mg produced 13.0% at 76 weeks against 5.4% on placebo, and mazdutide 9 mg produced 16.65% at 60 weeks against 1.50%.^24,25 Separate trials cannot tell us how much each receptor contributes. Population, exposure, titration, discontinuation, duration, and estimand all differ, so I do not rank these results against each other or against retatrutide.

Heart rate rises with these drugs. Glucagon raises heart rate on its own, which is why emergency physicians use it in beta-blocker overdose, but GLP-1 agonists raise it too, so the drug-level effect cannot be assigned to one receptor. In retatrutide’s Phase 2, the rise peaked at 24 weeks and then declined;²⁶ survodutide raised heart rate 3.2 to 3.5 beats per minute at week 76.²⁷ SYNCHRONIZE-CVOT, survodutide’s cardiovascular outcomes trial, is scheduled as an American Heart Association late-breaking presentation on November 9.²⁸

Dysesthesia, an abnormal skin sensation, was reported in 12.5% on retatrutide 12 mg against 0.9% on placebo in TRIUMPH-1, according to Lilly’s release; its mechanism is unknown.²⁹ Gastrointestinal events affected 89.7% on survodutide 6.0 mg against 47.9% on placebo.²⁴ These are findings for the combination drugs, not evidence that glucagon causes them.

The theoretical muscle cost of faster amino acid breakdown has limited data behind it. In a retatrutide DXA substudy in people with type 2 diabetes, fat mass fell 23.2% on 12 mg over 36 weeks, and the lean share of weight lost was similar to other obesity treatments.³⁰ Only 103 participants completed treatment and both scans. That is limited reassurance, not an answer to the long-term muscle question.

Amylin was found by studying damage. Two groups identified it in islet amyloid in 1986 and 1987, and Garth Cooper’s group sequenced a 37-amino-acid peptide 46% identical to calcitonin gene-related peptide.⁶ It is co-secreted with insulin by the β-cell after a meal, at roughly 1 molecule for every 100 of insulin.³¹ People with type 1 diabetes make essentially none, and advancing type 2 diabetes can lose it along with β-cell function. Insulin replacement does not replace amylin.

Amylin’s receptor is a hybrid: the calcitonin receptor, another class B GPCR, paired with a receptor activity-modifying protein. Add RAMP1, RAMP2, or RAMP3 and the complex becomes AMY1, AMY2, or AMY3, each with its own peptide preferences.³²

That gives drug designers a dial the GLP-1 field never had. Cagrilintide is non-selective. Eloralintide prefers AMY1 and, in rats, caused less conditioned taste avoidance than cagrilintide, the rodent version of “that food made me sick.”³³ Selectivity is, in effect, a hypothesis about nausea, and it is untested in people: in eloralintide’s Phase 2, nausea reached 64% on 6 mg and 25% with a gradual 3-to-9 mg escalation, against 14% on placebo.³⁴

Amylin’s main target is the area postrema, a circumventricular organ in the caudal hindbrain that samples blood directly and relays through the nucleus of the solitary tract. In rats, an intact area postrema is required for amylin to end meals.³¹ In people, the same signal slows gastric emptying and trims the post-meal glucagon surge, and the pramlintide label notes it does not alter the counter-regulatory response to insulin-induced hypoglycemia.⁷

The area postrema also carries GLP-1 receptors.²¹ Two hormone systems converge on one small structure through distinct receptor systems. That is the case for stacking them, and the reason to watch whether their nausea stacks too.

In leptin-resistant obese rats, amylin pretreatment partially restored hypothalamic leptin signaling, and the pair produced synergistic, fat-specific weight loss.³⁵ The human trial was smaller and selected. Everyone first took pramlintide with a reduced-calorie diet for 4 weeks, and only those who lost 2% to 8% were randomized to pramlintide, metreleptin, or both. At 24 weeks, the combination group had lost 12.7% from enrollment, a figure that includes the lead-in, against about 8% with either drug alone. No one has tested the idea with today’s long-acting analogs.

Human amylin aggregates into the same fibrils Cooper isolated. Rat amylin does not, because prolines break up the β-sheets fibrils need, and pramlintide borrowed them at positions 25, 28, and 29.³² That fixed stability, not dosing. Pramlintide is injected before each major meal, requires an initial 50% reduction in mealtime insulin followed by individualized adjustment, carries a boxed warning for severe hypoglycemia when used with insulin, and moved weight 0.8 to 1.6 kg in its label trials.⁷ The new analogs are weekly.

The cleanest amylin evidence comes from a four-arm randomized trial. REDEFINE-1 randomized 3,417 adults to cagrilintide plus semaglutide, either drug alone, or placebo for 68 weeks. Weight loss was 20.4% with the combination, 14.9% with semaglutide, 11.5% with cagrilintide, and 3.0% with placebo (treatment-policy estimand).³⁶ Compared with the semaglutide-alone arm under a flexible dosing protocol, adding cagrilintide produced about 5.5 percentage points more weight loss. Flexible meant investigators could slow escalation or lower the dose and participants could continue on a lower dose, so not everyone received the maximum.

Alone, eloralintide 9 mg produced 20% at 48 weeks against 0.4% on placebo (efficacy estimand).³⁴ Petrelintide produced 9.8% mean weight loss at the 28-week primary endpoint in the 5-mg group, against 1.7% on placebo; the 7-mg and 9-mg groups reached 9.3% and 9.4%, so higher doses added little by that point.³⁷ Reported reductions reached up to 10.7% at week 42, against 1.7% on placebo, under the efficacy estimand.³⁸ Across petrelintide doses, nausea occurred in 20% against 6% on placebo and vomiting in 3% against 6%.³⁷ Those rates come from one placebo-controlled trial; they do not show better tolerability than GLP-1 drugs, which were not tested alongside it.

Lean mass. In the REDEFINE-1 DXA substudy (n = 252), fat made up 66.9% of the weight lost on CagriSema, 69.7% on semaglutide, and 62.9% on cagrilintide alone.³⁹ These findings do not demonstrate a preferential lean-tissue-sparing effect from adding cagrilintide. DXA lean soft tissue includes water and organs; it is not a measure of skeletal muscle, or of how well muscle works. On function, the presenters reported improved patient-reported physical function with CagriSema versus placebo and sit-to-stand performance that improved similarly with CagriSema and placebo.^39,40

Tolerability. GI events affected 79.6% on CagriSema, 73.8% on semaglutide, 54.0% on cagrilintide, and 39.9% on placebo; 5.9% of CagriSema participants stopped treatment because of adverse events, against 3.6%, 2.6%, and 3.5%.³⁶ Amylin alone may be gentler; stacked on a GLP-1, it is not automatically so.

Superiority. In open-label REDEFINE-4, reported by press release, CagriSema produced 23.0% at 84 weeks against 25.5% for tirzepatide 15 mg (efficacy estimand) and missed non-inferiority.⁴¹ Lilly’s different answer, eloralintide added to tirzepatide, is the subject of 23.3% Weight Loss in Type 2 Diabetes. Can Patients Stay on Treatment?

Glucagon was the villain because we judged it by the glucose it raises. Drugs that include it now produce large reductions in liver fat, though the share owed to glucagon itself is still unmeasured. Amylin was a footnote because the native peptide clumps and the first analog needed mealtime shots. Chemistry fixed both problems, and REDEFINE-1 showed that cagrilintide adds to semaglutide. Neither class has shown long-term outcomes yet.