There are two different liver questions, and a checkout page collapses them. One is whether these drugs treat fatty liver disease, which has an answer and a licensed product behind it — the histology is in the fatty liver article and the trial that produced it in the ESSENCE explainer. The other is whether a person whose liver already works badly — cirrhosis, portal hypertension, a Child-Pugh score on a chart — can safely take one. That question is answered by a single sentence on three labels, and by nothing at all on the fourth.
What the four labels say, side by side
Read on DailyMed on September 15, 2026, the Ozempic label carries a subsection numbered 8.7: “No dose adjustment of OZEMPIC is recommended for patients with hepatic impairment.” The Zepbound and Mounjaro labels carry the same subsection with the same instruction, both pointing at a clinical pharmacology study in which no change in tirzepatide exposure was observed.
The Wegovy label does not carry that subsection. Its Use in Specific Populations section runs 8.1 Pregnancy, 8.2 Lactation, 8.3 Females and Males of Reproductive Potential, 8.4 Pediatric Use, 8.5 Geriatric Use and 8.6 Type 2 Diabetes Mellitus, and then stops. There is no hepatic impairment subsection and no renal impairment subsection. The only hepatic statement anywhere in the document is a line in the pharmacokinetics section listing covariates found not to influence exposure, among them “mild, moderate, and severe hepatic impairment (Child-Pugh Class A-C) or fibrosis stage (F2 or F3 in patients with MASH).”
That is the asymmetry worth holding. The one product in this class with an indication to treat a liver disease is the one product whose label gives a prescriber no dedicated instruction about a damaged liver.
The reassurance rests on 25 people and one injection each
The semaglutide sentence traces to an open-label, parallel-group trial that enrolled four groups: 19 participants with normal hepatic function and eight with mild, ten with moderate and seven with severe impairment by Child-Pugh criteria. Each received a single subcutaneous 0.5 mg dose, with plasma concentrations followed for 35 days. No effect was to be declared if the 90% confidence interval for the exposure ratio fell inside 0.70 to 1.43. It did: AUC ratios were 0.95 (90% CI 0.77 to 1.16) for mild, 1.02 (0.93 to 1.12) for moderate and 0.97 (0.84 to 1.12) for severe.[1]
The tirzepatide sentence traces to a study of the same shape and smaller size: six with mild, six with moderate and seven with severe impairment against 13 controls, each given a single subcutaneous 5 mg dose. AUC ratios against control were 1.08 (90% CI 0.879 to 1.32), 0.960 (0.790 to 1.17) and 0.852 (0.699 to 1.04), and exposure showed no significant relationship to Child-Pugh score, serum albumin, total bilirubin or international normalized ratio.[2] The oral semaglutide study followed the same design over ten days of dosing in 32 participants with impairment, and reported half-lives of 142 to 156 hours across all groups.[3]
Two things follow, and only one of them is comforting. The first is that these are clean results: a liver that filters badly does not appear to change how much of either drug ends up in the blood, which makes pharmacokinetic sense for peptides cleared by proteolysis rather than by hepatic metabolism. The second is the dose. A 0.5 mg injection is the second rung of a ladder that ends at 2.4 mg, and a 5 mg injection is the bottom rung of one that ends at 15 mg — the schedules set out in the titration article. Nobody has given a maintenance dose to a person with Child-Pugh C cirrhosis in a controlled study and reported what happened over months. “No dose adjustment” is a statement about a concentration curve, not about whether the drug is a good idea.
The indication stops at one word
The Wegovy indication reads “for the treatment of noncirrhotic metabolic dysfunction-associated steatohepatitis (MASH) … with moderate to advanced liver fibrosis (consistent with stages F2 to F3 fibrosis) in adults,” approved under accelerated approval. Its Medication Guide translates that for patients as MASH “with moderate to advanced liver scarring (fibrosis), but not with cirrhosis of the liver.” The supporting trial enrolled patients with a biopsy showing fibrosis stage 2 or 3, and the week-72 analysis covered 800 F2 and F3 patients, 250 at F2 and 550 at F3.
Stage 4 is cirrhosis, and stage 4 was not enrolled. The one randomized trial that did enroll it tested 2.4 mg weekly for 48 weeks in biopsy-confirmed steatohepatitis-related compensated cirrhosis and did not improve fibrosis;[4] those figures belong to the MASH page and are not repeated here. The point for a reader with cirrhosis is simpler than the result: the population most likely to read a liver indication as permission is the population the indication explicitly excludes.
The observational record in cirrhosis is real and narrow
A Taiwanese national-insurance analysis matched 467 pairs of GLP-1 receptor agonist users and non-users among people with type 2 diabetes and cirrhosis, followed for a mean of 3.28 and 3.06 years. Death rates were 27.46 against 55.90 per 1,000 person-years, an adjusted hazard ratio of 0.47 (95% CI 0.32 to 0.69). Decompensated cirrhosis ran at aHR 0.70 (0.49 to 0.99), hepatic encephalopathy 0.59 (0.36 to 0.97) and liver failure 0.54 (0.34 to 0.85), with longer cumulative use associated with lower risk.[5] The cohort was compensated cirrhosis, the comparator was non-use rather than another drug, and the authors call for confirmation.
The largest active-comparator cohort is more measured and contains the reversal. Across Swedish, Danish and Norwegian registers, 91,479 initiators of a GLP-1 receptor agonist were compared against 244,004 initiators of a dipeptidyl peptidase-4 inhibitor. Serious liver events occurred at 16.9 against 19.2 per 10,000 person-years, adjusted hazard ratio 0.85 (95% CI 0.75 to 0.97) — but the rate difference was −2.1 per 10,000 person-years (95% CI −4.4 to 0.1), an interval that includes zero. Splitting the composite, cirrhosis carried the whole effect at 0.85 (0.75 to 0.97) while hepatocellular carcinoma did not move at all: 1.05 (0.80 to 1.39).[6]
Two limits travel with that number. The ratio is significant and the absolute difference is not, which is what a small effect on a rare outcome looks like. And the cohort excluded anyone with chronic liver disease other than fatty liver, so hepatitis B, hepatitis C, alcohol-related and autoimmune liver disease are absent from the estimate entirely. A reader whose cirrhosis came from something other than metabolic disease has not been counted in the reassuring study.
The risk that points the other way
Cirrhosis is a wasting illness, and the drug in question works by reducing intake. A meta-analysis of 22 cohort studies covering 6,965 patients with cirrhosis put the pooled prevalence of sarcopenia at 37.5% (95% CI 32.4 to 42.8), rising toward half in alcohol-associated liver disease and in Child-Pugh grade C disease. Sarcopenia carried an adjusted hazard ratio for death of 2.30 (95% CI 2.01 to 2.63), with no significant heterogeneity across any subgroup analysis.[7]
Set the two findings beside each other. In compensated cirrhosis with diabetes, treated patients died at roughly half the rate of untreated ones. In cirrhosis generally, more than a third of patients already carry a muscle-wasting diagnosis that doubles mortality, and lean mass is the tissue these drugs take alongside fat — the trial figures for that are in the muscle article. Neither finding cancels the other, and no study has been designed to say which one dominates in a person with a MELD score.
What has actually been dosed in decompensated disease
The only published program to escalate an incretin drug through a maintenance dose in Child-Pugh B and C cirrhosis used a different molecule. A phase 1 trial gave a single 0.3 mg dose of survodutide, a glucagon and GLP-1 dual agonist, to 41 participants across Child-Pugh A, B and C cirrhosis and matched healthy controls, finding exposure ratios whose 90% confidence intervals spanned 1. A second cohort of 41 escalated from 0.3 mg to 6.0 mg over 24 weeks and held it for four more. Drug-related treatment-emergent adverse events occurred in 82.4% of participants without cirrhosis and 87.5% of those with it.[8] Liver fat, liver stiffness, liver volume and body weight all fell.
That is 82 people, in a phase 1 trial, on a molecule approved for nothing, and it is the closest thing this field has to an answer. Survodutide’s wider program is covered in the pipeline article. Its authors note the reason the gap exists: people with decompensated cirrhosis are usually excluded from trials of investigational drugs.
What a cash-pay buyer is holding
Every label sentence above belongs to a branded, FDA-approved product. Most sellers in this market dispense compounded semaglutide or tirzepatide, which is not FDA-approved and which the FDA does not review for safety, efficacy or quality before it is dispensed — set out in what a compounded vial contains and priced on the compounded semaglutide board. A vial arrives without the prescribing information quoted here, which means the one document a prescriber would consult about a damaged liver is not in the box.
The eligibility question is also not a checkbox. Cirrhosis is not on either label’s contraindication list — those lists name medullary thyroid carcinoma, multiple endocrine neoplasia type 2 and serious hypersensitivity, and nothing else, as the eligibility article sets out. A liver problem therefore does not stop an online intake. It just means the prescriber needs information the intake may never ask for.
What none of this settles
No trial has randomized people with cirrhosis to a maintenance dose of an approved GLP-1 drug and measured decompensation, transplant or death. The mortality figures in cirrhosis come from a matched cohort of 467 pairs in one country, against non-use rather than an active comparator, in compensated disease with diabetes. The pharmacokinetic reassurance on three labels comes from single low doses in 44 people combined. And the one number in this literature that runs against the drug — a 37.5% background rate of sarcopenia carrying a hazard ratio of 2.30 — has never been measured in anyone taking it.
What is established is narrow. Hepatic impairment does not appear to change how much of either drug reaches the blood, the labels therefore ask for no dose change, and a liver indication exists for exactly one formulation in exactly one noncirrhotic stage range. Everything past that is a conversation with a hepatologist who can read an elastography result, not a field on an order form.