What does it do to the liver?

What this archive is for. Trials of the compound come first, ranked by how many people were actually given it. Studies that only observed people are below them and labelled.

1. 01SemaglutidePeptideStrong. At least two finished studies in people, and at least 1,000 people who took it. 12 completed trials, 42,363 people given it.10 reported

   1,197 people given it1 trial1 found nothing

   • found nothingSemaglutide healed liver inflammation in a serious fatty liver disease, on a look taken partway through. Liver samples were taken from the first 800 of 1,197 patients at 72 weeks. The inflammation cleared without the scarring getting worse in 62.9% on the drug against 34.3% on the dummy, and scarring improved in 36.8% against 22.4%. Bodily pain scores were no different. This is a look taken partway through, so the full trial has not reported yet.

     The numbers ESSENCE interim analysis at week 72 in the first 800 of 1,197 patients with biopsy-defined MASH: steatohepatitis resolution in 62.9% versus 34.3% and fibrosis improvement in 36.8% versus 22.4%. Bodily pain scores did not differ.

     New England Journal of Medicine, 2025· 1,197 people

2. 02SurvodutidePeptideStrong. At least two finished studies in people, and at least 1,000 people who took it. 5 completed trials, 2,034 people given it.7 reported

   293 people given it2 trials

   • Survodutide healed liver damage in a lot of the people who had it. Two hundred and ninety-three adults with liver inflammation confirmed by a tissue sample were treated for eleven months. Between 47 and 62 out of every 100 on the drug, depending on dose, had their liver damage improve without their scarring getting worse. On a dummy injection, 14 out of 100 did. This is a mid stage trial. It measured what a pathologist sees down a microscope, not how long anyone lived.

     The numbers 48-week trial in 293 adults with biopsy-confirmed MASH: histologic improvement without worsening fibrosis in 47 to 62% on survodutide versus 14% on placebo.

     New England Journal of Medicine, 2024· 293 peoplepoints the wrong way

   • This trial is running and has published nothing, because it is not due to finish until 2031. It is giving survodutide or a dummy injection to 1,800 adults whose fatty liver disease has already scarred the liver. It is built to show two things: that the disease clears without the scarring worsening, and that the scarring itself improves. Nothing can be said about the answer yet. This is the test that would settle the liver claims.

     The numbers A Phase 3 trial in 1,800 adults with MASH, with resolution of MASH without worsening fibrosis and at least a 1 point fibrosis improvement as co-primary endpoints, recruiting with estimated completion in December 2031.

     ClinicalTrials.gov, Boehringer Ingelheim, 2026· number of people not givenpoints the wrong way

3. 03TesamorelinPeptideStrong. At least two finished studies in people, and at least 1,000 people who took it. 8 completed trials, 1,729 people given it.10 reported

   149 people given it3 trials1 found nothing

   • found nothingIt worked on liver fat, in a small trial. 61 people with HIV and a fatty liver were split evenly and given a daily injection of tesamorelin or a dummy version for a year. Nobody involved knew which. Liver fat came down by about a third more on tesamorelin than on the dummy. By the end 35% of the tesamorelin group had liver fat back under the normal threshold, against 4% on the dummy. With 61 people this is a first answer, not a settled one.

     The numbers 61 people with HIV and NAFLD (NCT02196831) randomized 30 to 30 to 12 months of tesamorelin 2 mg daily or placebo, with hepatic fat fraction as the primary endpoint. Hepatic fat fraction fell further on tesamorelin, absolute effect minus 4.1 percent (95% CI minus 7.6 to minus 0.7, p = 0.018), a minus 37 percent relative reduction (95% CI minus 67 to minus 7, p = 0.016). At 12 months 35 percent on tesamorelin and 4 percent on placebo had a hepatic fat fraction under 5 percent (p = 0.0069). Fasting glucose and glycated haemoglobin did not differ between groups. Injection site complaints were more common on tesamorelin, none judged serious.

     The Lancet HIV, 2019· 61 people

   • Tesamorelin shrank both deep belly fat and liver fat, in a small trial. 50 people with HIV and belly fat were split by chance, and 48 were treated for six months. Deep fat around the organs fell 42 cm2 more than on the dummy shot, and liver fat fell about 3 percentage points more. Fasting blood sugar rose 7 mg/dL more than the dummy in the first two weeks. By six months that blood sugar difference had gone.

     The numbers 50 randomized, 48 treated over 6 months: visceral fat treatment effect -42 cm2 (95% CI -71 to -14) and net liver fat -2.9% lipid-to-water. Fasting glucose rose 7 mg/dL more than placebo at 2 weeks.

     JAMA, 2014· 50 people

   • The drug still worked in people taking the newer HIV medicines that tend to add weight. 38 of the 61 people in this trial were on integrase inhibitors, and 31 finished the year. Their deep belly fat fell about 25 cm2 while it rose about 14 cm2 on the dummy shot, and liver fat fell 4.2% against 0.5%. Blood sugar control did not get worse. This is a slice of a small trial, not a trial of its own.

     The numbers 38 participants on integrase-inhibitor regimens within a 61-person randomized trial, 31 completing 12 months: median visceral fat -25 cm2 versus +14 cm2 and hepatic fat -4.2% versus -0.5%, without worsening glycemic control.

     AIDS, 2024· 38 peoplepoints the wrong way

4. 04EfinopegdutidePeptideModerate. At least one finished study in people, and at least 100 people who took it. 1 completed trial, 145 people given it.

   145 people given it1 trial, 1 watch-only1 found nothing

   • found nothingA dual-acting drug cleared far more liver fat than semaglutide in a head to head test. 145 people with fatty liver disease were split by chance between the two drugs for about six months. Everybody knew which drug they were taking. Liver fat fell by roughly three quarters on efinopegdutide against a little over two fifths on semaglutide. Weight loss was similar. Liver fat on a scan is a marker. Nobody here was followed long enough to see scarring, liver failure or death.

     The numbers 145 participants with liver fat content of 10 percent or more, randomised 1:1 to efinopegdutide 10 mg or semaglutide 1 mg weekly for 24 weeks, open-label. Mean baseline BMI 34.3 and liver fat 20.3 percent; 33.1 percent had type 2 diabetes. Relative reduction in liver fat 72.7 percent (90% CI 66.8 to 78.7) against 42.3 percent (90% CI 36.5 to 48.1), p < 0.001. Body weight reduction 8.5 percent against 7.1 percent, p = 0.085, not significant. Slightly higher adverse events on efinopegdutide, mainly gastrointestinal.

     Journal of Hepatology, 2023· 145 people

   • The follow-on trial is running in the disease stage that actually matters, and it has reported nothing. It enrols people with steatohepatitis, which is the point where liver fat has turned into inflammation and scarring. Enrolment is closed and the study continues. No results have been posted. Liver fat on a scan is a stand-in. Scarring getting worse, and what happens to people because of it, is the real question and it is still open.

     The numbers The follow-on programme in steatohepatitis, which is the disease stage where liver fat matters clinically. Liver fat is a biomarker; fibrosis progression and liver outcomes are not.

     ClinicalTrials.gov, 2026· number of people not givenobserved only, nobody was given it

5. 05CotadutidePeptideStrong. At least two finished studies in people, and at least 1,000 people who took it. 6 completed trials, 1,221 people given it.

   74 people given it1 trial

   • Cotadutide cut liver fat in people with confirmed liver inflammation, but never looked at the liver itself again. 74 people with biopsy-confirmed disease were split by chance between two doses and a dummy, for about five months. At the top dose, liver fat fell by five percentage points more than on the dummy. Side effects hit nine in ten at the top dose against fewer than four in ten on the dummy. No second biopsy was taken, so nothing is known about scarring. The sponsor paid.

     The numbers PROXYMO: 74 people with biopsy-proven non-cirrhotic MASH with fibrosis, randomised to cotadutide 600 mcg, 300 mcg or placebo for 19 weeks. At 600 mcg, absolute liver fat fraction fell 5.0 percentage points against placebo (95 percent CI 1.5 to 8.5), ALT fell 23.5 U per L and AST 16.8 U per L. Any adverse event in 91.7, 76.9 and 37.5 percent; discontinuation for adverse events 16.7, 7.7 and 4.2 percent. No histological endpoint. Sponsor funded.

     Clinical Gastroenterology and Hepatology, 2024· 74 people

6. 06GHRP-6PeptideLimited. People have taken it in a published study, but fewer than 100 of them. 5 completed trials, 42 people given it.8 reported

   10 people given it1 trial

   • Block the growth hormone receptor and this peptide turns out to push blood sugar directly. 10 healthy lean men were studied, and nobody involved knew who got what. With the receptor blocked by a drug, and the men not fasting, GHRP-6 sent insulin up roughly eightfold and blood sugar up noticeably. Free fatty acids fell. The effect vanished when the men were fasting and when the blocker was absent. The authors read that as the peptide acting on the gut, liver and pancreas directly.

     The numbers 10 healthy non-obese men, double-blind and placebo-controlled. With the GH receptor blocked by pegvisomant in the non-fasting state, GHRP-6 at 1 microgram per kilogram raised insulin from 10.3 to 81.3 mU/L (P<0.001) and glucose from 4.2 to 6.0 mmol/L (P<0.05), with free fatty acids falling. The effect did not appear during fasting or without the blocker, which the authors read as a direct non-pituitary action on the gastro-entero-hepatic axis.

     Journal of Clinical Endocrinology and Metabolism, 2001· 10 people

7. 07RetatrutidePeptideStrong. At least two finished studies in people, and at least 1,000 people who took it. 5 completed trials, 1,345 people given it.8 reported

   0 people given it1 trial

   • Retatrutide cleared fat out of the liver in a small side study. People with obesity and a fatty liver took the drug or a dummy injection for six months, inside the larger trial. By the end, liver fat had come back into the normal range for 86% of those on the strongest dose, 79% on the middle dose and 52% on the lowest. Nobody on the dummy injection got there. This measured fat on a scan, not whether anyone lived longer or felt better.

     The numbers Phase 2a substudy (NCT04881760): at 24 weeks, normal liver fat (below 5%) was reached by 86% of participants on 12 mg, 79% on 8 mg and 52% on 4 mg versus 0% on placebo.

     Nature Medicine, 2024· number of people not given

8. 08PemvidutidePeptideLimited. People have taken it in a published study, but fewer than 100 of them. 2 completed trials, none stating how many people.1 reported

   0 people given it1 trial, 1 watch-only

   • A weekly shot cut the amount of fat sitting in people's livers, compared with a dummy shot. One half of the molecule acts on the liver directly, which is why it was expected to do this. Note what was measured. The amount of fat in the liver, not damage to the liver, and not how anyone felt.

     The numbers Once-weekly pemvidutide reduced liver fat content in MASLD, with the glucagon component acting directly on the liver.

     Journal of Hepatology, 2025· number of people not givenobserved only, nobody was given it

   • The drug cleared up liver inflammation in six months without letting scarring get worse, which is what it was asked to do. It did not improve the scarring itself by that point. Those are two different things, and scarring is the one that predicts what happens to a person over years. Six months is early to expect scar tissue to change.

     The numbers Met MASH resolution without worsening of fibrosis at 24 weeks; did not meet fibrosis improvement at that timepoint.

     The Lancet, 2025· number of people not givenpoints the wrong way

9. 09OsteocalcinPeptidePreliminary. Studied in people, but no published study has had people take it. 2 human studies where nobody was given it.

   nobody was given it1 watch-only study

   • The arrow points the other way. Osteocalcin was measured in 5,169 people in China, and lower levels went with higher blood sugar, insulin resistance, triglycerides, liver fat and weight. That is exactly what the bone hormone story predicts. Then the researchers used genetics to test which way the causation runs, and the genes say type 2 diabetes lowers osteocalcin rather than low osteocalcin causing diabetes. Nobody was given anything. This is the most important caveat on the page.

     The numbers 5,169 Chinese participants with measured osteocalcin, plus bidirectional two-sample Mendelian randomisation against Biobank Japan and East Asian diabetes GWAS. Osteocalcin correlated with lower glucose, insulin resistance, triglycerides, liver fat and BMI, but the causal analysis supported type 2 diabetes lowering osteocalcin (causal effect -0.03, -0.05 to -0.01, p = 0.006 and p = 0.001 in two datasets), not the reverse.

     Journal of Bone and Mineral Research, 2021· 5,169 peopleobserved only, nobody was given it

10. 10MOTS-cPeptideLimited. People have taken it in a published study, but fewer than 100 of them. 1 completed trial, 30 people given it.6 reported

    nobody was given it1 watch-only study

    • This study points the other way, and the MOTS-c story does not hold together. A hundred and fifty-nine adults without diabetes or heart disease had their blood measured. Nobody was given anything. Those with metabolic syndrome carried more MOTS-c, not less, and levels rose with belly fat and with fat in the liver. That is the reverse of what the diabetes and heart studies found. When studies disagree this sharply, the honest reading is that nobody knows what a blood MOTS-c level means.

      The numbers 125 Chinese adults plus 34 additional Caucasian women without overt type 2 diabetes or cardiovascular disease: plasma MOTS-c was higher, not lower, in those with metabolic syndrome and rose with android and liver fat, the opposite direction from studies in diabetes and coronary disease.

      Biochimica et Biophysica Acta, General Subjects, 2021· 125 peopleobserved only, nobody was given it

11. 11SHLP2PeptidePreliminary. Studied in people, but no published study has had people take it. 2 human studies where nobody was given it.6 reported

    nobody was given it1 watch-only study

    • Nobody was given anything in this study. Researchers measured two peptides in the blood of people without diabetes and compared them with how much fat those people carried. The more belly fat and liver fat someone had, the higher their peptide levels ran. That is the opposite direction to the mouse work, where giving the peptide protects against weight gain. Neither finding disproves the other. Neither one can be used to predict the other.

      The numbers Human plasma measurement in people without diabetes. Both MOTS-c and SHLP2 associated positively with android fat and liver fat. Worth reading carefully: the direction is positive, meaning higher peptide with more fat, which does not sit neatly with the mouse story in which giving the peptide protects against diet-induced obesity. An association in one direction and an intervention in the other are not in conflict, but neither can be used to predict the other.

      Biochimica et Biophysica Acta, General Subjects, 2021· number of people not givenobserved only, nobody was given it

    Also measured in animals or in the lab

    • This paper found SHLP2 and five sister peptides. In people, it only measured blood levels, and they fall with age. Nobody was given anything. In rodents, SHLP2 put into the brain helped the body handle blood sugar. In a dish, it cut cell death and cell damage, and helped mitochondria run.

      The numbers The naming paper. An in silico search of the 16S rRNA region of mitochondrial DNA that encodes humanin found six more short open reading frames, SHLP1 to SHLP6. SHLP2 and SHLP3 reduced apoptosis and reactive oxygen species and improved mitochondrial metabolism in vitro and enhanced 3T3-L1 preadipocyte differentiation in culture. Rodent clamp studies showed intracerebrally infused SHLP2 increasing glucose uptake and suppressing hepatic glucose production. It also reports that circulating SHLP2 declines with age in human plasma, which is a measurement, not an intervention.

      Aging (Albany NY), 2016· lab (cells in a dish)

Vitamins, minerals, amino acids and botanicals sold over the counter, held to the same standard.

1. 01Green tea extractSupplementModerate. At least one finished study in people, and at least 100 people who took it. 1 completed trial, 1,075 people given it.4 reported

   1,075 people given it1 trial, 1 watch-only

   • A year of high dose green tea extract harmed the liver in a small but real share of women. 1,075 women past menopause took an extract with 843 mg of EGCG a day, or a dummy pill, for one year. Liver enzymes rose in 36 women on the extract against 4 on the dummy pill. Just over 1% had a liver problem serious enough to be recorded as such. Nausea and skin reactions were also more common on the extract. The National Cancer Institute paid for it.

     The numbers 1,075 postmenopausal women randomised to extract with 843 mg EGCG a day or placebo for one year. Adverse events 75.6 versus 72.8 percent; serious 2.2 versus 1.5 percent; more nausea and skin events on extract. ALT elevation in 36 (6.7 percent) versus 4 (0.7 percent), p < 0.001; 1.3 percent had ALT-related serious adverse events. US National Cancer Institute funded.

     Food and Chemical Toxicology, 2015· 1,075 people

   • Who gets liver injury from green tea extract may be partly in their genes. This is a second look at the Minnesota trial, covering the 400 women who took the extract. Women carrying one version of a liver-processing gene saw liver enzymes rise about 80% above their starting level at six and nine months. Women with the other version rose about 30%. That is a big difference between people taking the same supplement. It does not tell you which version you carry.

     The numbers Secondary analysis of the Minnesota trial (400 women on extract, 98 percent white, mean age 59.8). Mean ALT rose 78 to 82 percent from baseline at months 6 and 9 in UGT1A4 rs6755571 A/C carriers against 28 to 30 percent in C/C, p < 0.001 and p = 0.004. Points to a genetic component in who gets liver injury.

     Journal of Dietary Supplements, 2023· 400 peopleobserved only, nobody was given it

2. 02EGCGSupplementStrong. At least two finished studies in people, and at least 1,000 people who took it. 3 completed trials, 1,172 people given it.5 reported

   1,075 people given it1 trial

   • Green tea extract damaged some livers, at ten times the dummy group's rate. 1,075 women past menopause took an extract delivering 843 mg of EGCG a day, or a dummy, for a year. A liver enzyme rose in 36 women on the extract, 6.7 percent, against 4 on the dummy, 0.7 percent. 1.3 percent had a liver problem recorded as serious. Overall side effect rates were close, 75.6 against 72.8 percent. Funded by the US National Cancer Institute.

     The numbers 1,075 postmenopausal women randomised to an extract delivering 843 mg EGCG a day or placebo for one year. Alanine aminotransferase elevation in 36 women on extract (6.7 percent) against 4 on placebo (0.7 percent), p < 0.001, with 1.3 percent having an ALT-related serious adverse event. Overall adverse event rates were similar, 75.6 against 72.8 percent, with more nausea and skin events on extract. US National Cancer Institute funded.

     Food and Chemical Toxicology, 2015· 1,075 people

3. 03TUDCASupplementLimited. People have taken it in a published study, but fewer than 100 of them. 3 completed trials, 54 people given it.7 reported

   20 people given it1 trial1 found nothing

   • found nothingFour weeks of TUDCA improved insulin sensitivity in liver and muscle, and not in fat. 20 obese adults took 1,750 mg a day or a dummy for four weeks, with detailed infusion tests and samples from muscle and fat. Liver and muscle insulin sensitivity rose by roughly 30 percent while the dummy group did not change. The cell stress markers that were meant to explain all this did not move in either tissue. NIH funded. Twenty people over four weeks is a mechanism study.

     The numbers 20 obese adults, mean age 48 and BMI 37, randomised to 1,750 mg a day of TUDCA or placebo for 4 weeks, with two-stage hyperinsulinaemic-euglycaemic clamps, tracer infusions and muscle and adipose biopsies. Hepatic and muscle insulin sensitivity rose approximately 30 percent (p < 0.05) with no change on placebo, and muscle insulin signalling improved. Markers of endoplasmic reticulum stress in muscle and adipose tissue did not change after either treatment, despite that being the hypothesised mechanism. NIH funded.

     Diabetes, 2010· 20 people

   What people using it report

   • Two separate populations discuss it and they do not agree. In steroid and testosterone forums it is taken as liver insurance, commonly at 250 to 750 mg a day and sometimes 1 to 2 g alongside oral compounds, and people report liver enzyme values falling quickly. In longevity and harm reduction discussion the same doses are treated as unproven for anyone whose liver is healthy.
4. 04ResveratrolSupplementModerate. At least one finished study in people, and at least 100 people who took it. 14 completed trials, 403 people given it.7 reported

   17 people given it3 trials2 found nothing

   • This is the strongest good human result resveratrol has, and it rests on eleven men. Obese but otherwise healthy men took 150 mg a day for a month, and also spent a month on a dummy pill. Their resting metabolism slowed the way it does when people eat less. Liver fat, blood sugar, triglycerides, a liver enzyme and inflammation markers all fell. Blood pressure improved. Muscle samples showed the cell's energy machinery switched on.

     The numbers Eleven healthy obese men, randomised double-blind crossover, 150 mg per day for 30 days. Sleeping and resting metabolic rate fell; muscle AMPK was activated with higher SIRT1 and PGC-1alpha protein and improved mitochondrial respiration on a fatty-acid substrate; intrahepatic lipid, circulating glucose, triglycerides, ALT and inflammation markers fell; systolic blood pressure and HOMA improved. The strongest positive mechanistic human result in the literature, in eleven men.

     Cell Metabolism, 2011· 11 people

   • found nothingThis is the study that explains why so many resveratrol trials come back empty. Six volunteers took a labelled dose, and their guts absorbed at least 70 percent of it. Almost none of it stayed as resveratrol. The gut wall and the liver tag it chemically almost immediately, so what ends up circulating is a set of breakdown products. You can swallow it, absorb it, and still have barely any of the actual compound in your blood.

     The numbers The finding that reframes the null trials. Using 14C-resveratrol in six volunteers, absorption of a 25 mg oral dose was at least 70% and peak plasma resveratrol plus metabolites reached about 2 micromolar with a 9.2 hour half-life, but only trace amounts of unchanged resveratrol were detectable. Rapid intestinal and hepatic sulfation is the rate-limiting step; glucuronidation and microbial hydrogenation of the double bond are the other routes.

     Drug Metabolism and Disposition, 2004· 6 people

   • found nothingResveratrol did not help fatty liver, and it may have irritated it. Twenty overweight men took 3,000 mg daily for eight weeks. Insulin resistance, liver fat, belly fat, blood fats and antioxidant activity were all unchanged, and so were the genes it was meant to act on. Two liver enzymes rose against the dummy, which the authors read as the liver being put under strain. A later six month trial with tissue samples also found no improvement.

     The numbers Twenty overweight or obese men with NAFLD, 3,000 mg daily for eight weeks. No improvement in insulin resistance, steatosis, abdominal fat distribution, plasma lipids or antioxidant activity, and no change in NQO1, PTP1B, IL6 or HO1 transcription. ALT and AST rose significantly versus placebo through week 6, which the authors read as increased hepatic stress. A 2016 trial of 1.5 g daily for six months with paired biopsies also found no histological improvement, while a 2014 trial of 500 mg for 12 weeks alongside lifestyle advice did report improved inflammatory markers.

     Clinical Gastroenterology and Hepatology, 2014· number of people not given

5. 05SulforaphaneSupplementModerate. At least one finished study in people, and at least 100 people who took it. 7 completed trials, 423 people given it.6 reported

   0 people given it1 trial

   • The human part of this much-cited paper is the smallest part of it. Sulforaphane was picked out by computer, by matching a diabetic liver pattern against 3,800 drug patterns. It cut sugar production in liver cells. In diabetic animals it improved glucose handling about as much as metformin. Finally, in a small human group, obese patients with poorly controlled type 2 diabetes took concentrated sprout extract and their fasting and average blood sugar came down. A later prediabetes trial could not confirm that human piece.

     The numbers A multi-part paper. Sulforaphane was identified computationally by matching a type 2 diabetes liver disease signature against 3,800 drug signatures; it suppressed glucose production in hepatic cells via NRF2 nuclear translocation, attenuated glucose intolerance in diabetic animals by a magnitude similar to metformin, and in the final human component, given as concentrated broccoli sprout extract, reduced fasting blood glucose and HbA1c in obese patients with dysregulated type 2 diabetes. The human arm is the smallest part of the paper and is the part that the later prediabetes trial did not confirm on its primary endpoint.

     Science Translational Medicine, 2017· number of people not given

6. 06MIB-626SupplementModerate. At least one finished study in people, and at least 100 people who took it. 4 completed trials, 107 people given it.1 reported

   0 people given it1 trial1 found nothing

   • found nothingThe tablet moved some body measurements and left every measure of physical function untouched. 30 overweight or obese adults aged 45 and over took MIB-626 or a dummy tablet for four weeks. Weight fell by about two kilograms. Blood pressure and cholesterol both came down. Strength, stamina, how quickly muscles tired and stair climbing power did not change. Insulin sensitivity, liver fat and belly fat did not change either. The manufacturer helped fund it. This study looked at many things at once, so treat any single result as a hint.

     The numbers Randomised 2:1, 30 overweight or obese adults aged 45 and over, MIB-626 two 500 mg tablets twice daily for 28 days. Body weight fell 1.9 kg (95 percent CI -3.3 to -0.5, P=.008), diastolic blood pressure fell 7.01 mmHg (-13.44 to -0.59, P=.034), total cholesterol fell 26.89 mg/dL (P=.004) and LDL fell 18.73 mg/dL (P=.007). Muscle strength, muscle fatigability, aerobic capacity and stair-climbing power did not change. Insulin sensitivity, hepatic fat and intra-abdominal fat did not change in either group. Metro International Biotech, the manufacturer, is a listed funder. This is a physiologic study with multiple outcomes rather than a trial powered on one of them.

     Journal of Clinical Endocrinology and Metabolism, 2023· number of people not given

7. 07PterostilbeneSupplementLimited. People have taken it in a published study, but fewer than 100 of them. 1 completed trial, 80 people given it.4 reported

   nobody was given it1 watch-only study

   • This is a safety report and it makes no claim that the compound works. It covers the same 80 patients, of whom 91.3 percent finished, averaging 54 years old and 71 percent women. Liver, kidney and blood sugar markers showed no drug reactions, and no meaningful self-reported or major reactions turned up. The authors conclude pterostilbene is generally safe up to 250 mg a day. Read it beside the trial it came from. Safe is not the same as useful, and that same trial found bad cholesterol rising.

     The numbers The safety report from the same 80-patient trial, 91.3 percent completion, average age 54, 71 percent female. No adverse drug reactions on hepatic, renal or glucose markers and no statistically significant self-reported or major adverse reactions. The stated conclusion is that pterostilbene is generally safe for use in humans up to 250 mg a day. It reports safety only and makes no efficacy claim.

     Journal of Toxicology, 2013· 80 peopleobserved only, nobody was given it

8. 08Nicotinamide RibosideSupplementLimited. People have taken it in a published study, but fewer than 100 of them. 4 completed trials, 90 people given it.5 reported

   nobody was given it1 watch-only study

   • Swallowing this form of vitamin B3 really does raise NAD in the blood. In a small human study the level rose by up to nearly three times. That single fact is what the entire NAD supplement industry is built on. Raising a level in the blood is the start of an argument, not the end of it. Nobody in this study got healthier.

     The numbers Oral NR raised blood NAD+ up to 2.7-fold in a human pilot and showed superior hepatic NAD+ kinetics in mice.

     Nature Communications, 2016· number of people not givenobserved only, nobody was given it

Approved drugs, biologics and hormones that have been measured for this. They are here because leaving them out would misrepresent the evidence, and they are below the peptides because that is not what most people come here to read.

1. 01OrforglipronSmall moleculeStrong. At least two finished studies in people, and at least 1,000 people who took it. 8 completed trials, 17,383 people given it.8 reported

   11,220 people given it1 trial

   • Pooled liver data across the whole late-stage programme show no liver signal. 11,220 people were combined from seven trials, about 6,900 on orforglipron and 4,300 on something else, followed for up to two years. Six people in each half hit the liver enzyme threshold that triggers concern, and every orforglipron case had another explanation. None met the standard that marks true drug-caused liver injury. Liver enzymes fell on average. This is the manufacturer analysing its own trials.

     The numbers 11,220 participants pooled across seven phase 3 trials, 6,920 on orforglipron and 4,300 on placebo, oral semaglutide, dapagliflozin or insulin glargine, up to 104 weeks including safety follow-up. Six orforglipron and six comparator participants (0.1 percent each) reached ALT or AST at three times the upper limit of normal with bilirubin at twice it; alternative causes explained all orforglipron cases and none met Hy's Law. Aminotransferases fell on average.

     Diabetes, Obesity and Metabolism, 2026· 11,220 people

2. 02MethotrexateSmall moleculeStrong. At least two finished studies in people, and at least 1,000 people who took it. 2 completed trials, 5,022 people given it.4 reported

   4,786 people given it1 trial1 found nothing

   • found nothingLow-dose methotrexate did not prevent heart attacks or strokes in people who already had heart disease, and it did not lower the inflammation markers it was supposed to. 4,786 people took it or a dummy pill for about two years, and all took folate too. This is the trial that separates methotrexate from the inflammation story it was borrowed for. It was stopped early. It carried real side effects, including liver enzyme rises, lower blood counts and more skin cancers.

     The numbers 4,786 patients with previous myocardial infarction or multivessel coronary disease plus type 2 diabetes or metabolic syndrome, randomised double-blind to methotrexate 15 to 20 mg weekly or placebo, all on 1 mg folate daily, median follow-up 2.3 years, stopped early. Methotrexate did not reduce cardiovascular events: primary endpoint in 201 versus 207 patients (4.13 versus 4.31 per 100 person-years, HR 0.96, 95% CI 0.79 to 1.16), and it did not reduce IL-1 beta, IL-6 or CRP. More liver enzyme elevations, lower leukocyte counts and haematocrit, and more non-basal-cell skin cancers on methotrexate.

     The New England Journal of Medicine, 2019· 4,786 people

3. 03FGF21BiologicModerate. At least one finished study in people, and at least 100 people who took it. 2 completed trials, 350 people given it.4 reported

   350 people given it2 trials

   • A lab-built version of FGF21 healed liver scarring in a mid-sized study. 222 people with fatty liver disease and moderate to advanced scarring got one of three doses or a dummy shot, for 24 weeks. Scarring improved without the disease getting worse in 7% of the dummy group, against 22% to 27% on the drug. The disease cleared in 2% of the dummy group and 23% to 37% on the drug. Nausea and diarrhoea were common. 89bio paid for it.

     The numbers ENLIVEN phase 2b, 222 randomised and 219 treated, biopsy-confirmed NASH with F2 or F3 fibrosis, 24 weeks. Fibrosis improvement without NASH worsening: 7 percent placebo, 22 percent on 15 mg weekly, 26 percent on 30 mg weekly (p 0.009), 27 percent on 44 mg every 2 weeks (p 0.008). NASH resolution: 2 percent placebo versus 37, 23 and 26 percent. Most common adverse events nausea and diarrhoea. Funded by 89bio.

     New England Journal of Medicine, 2023· 222 peoplepoints the wrong way

   • The higher dose worked at two years and the lower dose did not. 128 people with fatty liver disease and moderate to advanced scarring got one of two doses or a dummy shot. After 96 weeks, scarring improved by at least one stage in 19% of the dummy group, 30% on the lower dose and 49% on the higher. Only the higher dose clearly beat the dummy. Almost everyone in every group reported side effects, dummy group included. Akero Therapeutics paid for it.

     The numbers 128 randomised at 41 US centres, MASH with F2 or F3 fibrosis, efruxifermin 28 mg or 50 mg weekly or placebo. At 96 weeks, fibrosis improvement of at least one stage without MASH worsening: 19 percent placebo, 30 percent on 28 mg (difference 12 points, p 0.19), 49 percent on 50 mg (difference 31 points, 95 percent CI 12 to 49, p 0.003). Adverse events in 95 to 100 percent of every group including placebo. Sponsor Akero Therapeutics.

     Lancet, 2025· 128 peoplepoints the wrong way

   Also measured in animals or in the lab

   • In mice bred to make extra FGF21. They lived much longer without eating less. It seems to work by turning down growth hormone signals in the liver. The same mice had stunted growth and bone loss. There is no human lifespan data.

     The numbers Transgenic overexpression of FGF21 markedly extended lifespan in mice without reducing food intake, apparently by blunting hepatic growth hormone and IGF-1 signalling. The same paper records that FGF21 blocks somatic growth and causes bone loss in mice. There is no human lifespan data.

     eLife, 2012· animal study

4. 04EfruxiferminBiologicModerate. At least one finished study in people, and at least 100 people who took it. 1 completed trial, 181 people given it.2 reported

   181 people given it1 trial1 found nothing

   • found nothingThe trial failed the test it was built to pass, then looked better later. 181 people with liver scarring from fatty liver disease got efruxifermin or a dummy injection. At 36 weeks, when the answer was due, 19% on the 50 mg dose had less scarring against 13% on the dummy, which is not a gap you can lean on. By 96 weeks it was 29% against 11%. A result that arrives after the deadline is a reason to run another trial, not a finding.

     The numbers SYMMETRY, a Phase 2b trial in 181 patients with biopsy-confirmed compensated cirrhosis. The primary endpoint was missed: at 36 weeks fibrosis improved without worsening MASH in 19% on 50 mg against 13% on placebo. At 96 weeks the figures were 29% on 50 mg against 11% on placebo, a 16 percentage point difference (95% CI 2 to 30).

     New England Journal of Medicine, 2025· 181 people

5. 05Efimosfermin alfaBiologicModerate. At least one finished study in people, and at least 100 people who took it. 2 completed trials, 186 people given it.2 reported

   102 people given it1 trial

   • Liver fat fell sharply. 102 people with fatty liver disease got efimosfermin or a dummy shot for 12 weeks. By week 12, 89 percent of those on the drug had cut their liver fat by at least 30 percent, against 7 percent on the dummy. Side effects were mostly mild nausea and diarrhoea. Liver fat on a scan is not scarring on a biopsy, and 12 weeks is short. Boston Pharmaceuticals paid for the study.

     The numbers 102 participants, 12 weeks, phase 2a with safety and tolerability as the stated primary measure and efficacy only as exploratory biomarkers. At week 12, 47 of 53 evaluable participants on efimosfermin (89 percent) had at least a 30 percent reduction in liver fat fraction against 2 of 30 (7 percent) on placebo. Gastrointestinal events were the commonest adverse event, 40 percent against 24 percent, with no treatment related deaths. Funded by Boston Pharmaceuticals.

     The Lancet Gastroenterology and Hepatology, 2025· 102 people

6. 06MibavademabBiologicLimited. People have taken it in a published study, but fewer than 100 of them. 2 completed trials, 1 person given it.

   1 people given it1 trial1 found nothing

   • found nothingThe antibody only worked in people who were already short of leptin. In an early study in people, compared with a dummy shot, it was well tolerated. In the authors' words, it reduced body weight over twelve weeks in people with low leptin in the blood, and had no effect on weight in people whose leptin was higher. One patient given it outside a trial had better triglycerides and less liver fat. In mice bred without leptin, it brought weight, eating and blood sugar back to normal.

     The numbers A fully human antibody activating the leptin receptor with or without leptin. In obese leptin knockout mice it normalised body weight, food intake, blood glucose and insulin sensitivity. In a randomised double-blind placebo-controlled two-part phase 1 it was well tolerated. Verbatim: treatment of individuals with overweight or obesity decreased body weight over 12 weeks in those with low circulating leptin concentrations, under 8 ng/mL, but had no effect on body weight in individuals with higher baseline leptin. Compassionate use in one patient with atypical partial lipodystrophy and neutralising antibodies to metreleptin was associated with improvements in triglycerides and hepatic steatosis.

     Science Translational Medicine, 2023· 1 people

7. 07PegozaferminBiologicLimited. People have taken it in a published study, but fewer than 100 of them. 1 completed trial, none stating how many people.2 reported

   0 people given it1 trial

   • Liver scarring improved on the drug in a mid-stage study. 219 people with moderate or severe scarring from fatty liver disease got pegozafermin or a dummy shot for 24 weeks. Scarring improved in 22 to 27 percent of those on the drug, against 7 percent on the dummy. Nausea and diarrhoea were the most common side effects. Twenty four weeks is a short window for liver scarring, and 89bio, which made the drug, paid for the study.

     The numbers Fibrosis improvement without worsening of MASH at 24 weeks.

     New England Journal of Medicine, 2023· number of people not givenpoints the wrong way

Other compounds

1. 01ATX-304Small moleculePreliminary. Studied in people, but no published study has had people take it. 1 human study where nobody was given it.2 reported

   1 animal or lab study

   Measured in animals or in the lab

   • In mice. It lowered liver fat and oxidative damage in fatty liver disease.

     The numbers Reduced liver fat and oxidative stress in mice.

     JCI Insight, 2025· animal study

2. 02SupaglutideBiologicModerate. At least one finished study in people, and at least 100 people who took it. 3 completed trials, 190 people given it.

   1 animal or lab study

   Measured in animals or in the lab

   • In monkeys. Fifteen rhesus monkeys with fatty liver disease got weekly shots of the drug or a dummy shot for three months. Liver fat on a scan fell 40 percent against the dummy group. Biopsies showed less fat, and scarring did not get worse. Nobody has tested this on the liver in people.

     The numbers Fifteen rhesus monkeys with biopsy-confirmed metabolic dysfunction-associated steatohepatitis, given weekly subcutaneous supaglutide at 50 or 150 micrograms per kg or placebo for three months. Liver fat measured by MRI proton density fat fraction fell 40 percent from baseline against placebo, and histological steatosis improved without worsening of fibrosis on ultrasound-guided biopsy. Fifteen monkeys, three months, no human liver endpoint has been tested.

     Diabetology and Metabolic Syndrome, 2024· animal studypoints the wrong way

3. 03EmpagliflozinSmall moleculeStrong. At least two finished studies in people, and at least 1,000 people who took it. 3 completed trials, 16,738 people given it.4 reported

   1 animal or lab study

   Measured in animals or in the lab

   • In mice. Naturally old male mice given empagliflozin lived 5.9 percent longer on average. They also did better on learning, memory and balance tests, weighed less, and showed fewer signs of worn-out cells in the liver. Their gut bacteria changed too. This is one lab's study, and nobody else has repeated it.

     The numbers The entire mouse lifespan case, from one laboratory. Empagliflozin extended the median survival of male mice by 5.9 percent, improved learning, memory and motor balance, lowered body weight, reduced hepatic P21 and P16, altered gut flora composition and raised short-chain fatty acids. A single-site study, not part of any multi-site replication programme.

     GeroScience, 2024· animal study