Inject a peptide every week for a year and some people’s immune systems will notice. That is not a defect in these drugs; it is what happens with therapeutic peptides and proteins generally, and every approved GLP-1 label carries a section reporting how often it happened in the trials. What almost nobody does is read the second number in that section — the fraction of those antibodies that actually neutralize the drug — and without it the first number means nothing at all. Read the first number alone and you reach the wrong conclusion about which molecule to worry about.
The class table, and why you must not subtract across it
Here is what each current label reports, with the denominators, because the denominators are doing work. None of it maps onto how the molecules compare on weight loss, which the head-to-head article handles separately.
Tirzepatide sits at the top. Anti-tirzepatide antibodies were detected in 64.5% (1,591 of 2,467) of Zepbound-treated patients in the two weight reduction studies and 60.6% (137 of 226) in the sleep apnea studies.[1] Mounjaro reports 51% (2,570 of 5,025) across seven adult diabetes trials, and 49% (30 of 61) in pediatric patients.[2] Liraglutide is an order of magnitude lower: Victoza reports 9% (102 of 1,104) and Saxenda 2.8% (42 of 1,505).[3][4] Semaglutide is lower still. Ozempic reports 1% (32 of 3,150).[5] Wegovy reports 3% (50 of 1,709) at the 2.4 mg dose in one pair of studies and, under a newer assay, 11.2% (34 of 304) at the same dose and 15.4% (202 of 1,311) at 7.2 mg in a later pair.[6] Dulaglutide, outside the weight-loss labels but in the same class, came in at 1.6% of 4,006 patients.[7]
Every one of those labels opens its immunogenicity section with the same warning, and it should be taken literally: the observed incidence depends on the sensitivity and specificity of the assay, and differences in assay methods preclude meaningful comparison across studies.[6] The Wegovy figures make that concrete — same drug, same 2.4 mg dose, two different instruments, 3% and 11.2%. So the table above is a list of what each label says, not a ranking. Anyone publishing the difference between two of those numbers as a fact about the molecules is reporting an assay.
The number that actually matters is neutralization
An anti-drug antibody that binds the peptide somewhere harmless is a laboratory finding. One that blocks the part of the molecule that engages the receptor is a clinical problem. The labels report both, and the second column collapses.
On Zepbound, neutralizing antibodies against tirzepatide activity at the GIP and GLP-1 receptors were found in 2.8% and 2.7% of patients, and against the native hormones in 0.8% and 0.1%.[1] On Mounjaro the equivalent figures are 2% and 2%, with 0.9% and 0.4% against native GIP and GLP-1.[2] In the pediatric Mounjaro trial, where 49% of patients were antibody-positive, no neutralizing antibodies were detected at all.[2] Saxenda reports a neutralizing effect in vitro in 1.2% (18 of 1,505); Victoza in 1% (12 patients).[4][3] Dulaglutide, 0.9%.[7]
So the drug with 64.5% positivity has under 3% neutralization, and the pooled phase 3 analysis of tirzepatide is unambiguous about the consequence: treatment-emergent antibody status, antibody titer and neutralizing-antibody status had no effect on the pharmacokinetics or the efficacy of tirzepatide, across 5,025 evaluable patients, with maximum titers spanning 1:20 to 1:81,920.[8] The one thing that did differ was tolerability: more antibody-positive patients had hypersensitivity or injection-site reactions, most of them nonserious, which is the skin article’s subject rather than this one’s.
The reversal: the same rate meant the opposite thing for exenatide
If a high antibody rate were harmless in principle, this would be a short article. It is not, and the counterexample is inside the same drug class and the same analysis.
The LEAD program measured antibodies to liraglutide and to exenatide side by side. At 26 weeks, 8.7% (32 of 369) and 8.3% (49 of 587) of patients on liraglutide 1.2 mg and 1.8 mg had low-level antibodies, and glycemic efficacy was untouched — HbA1c reductions of 1.1 to 1.3% in antibody-positive patients against 1.2% in antibody-negative ones. In the LEAD-6 extension, 61% (113 of 185) of patients had anti-exenatide antibodies, and there high levels correlated with significantly smaller HbA1c reductions, P = 0.0022.[9]
Sixty-one percent and sixty-four percent are the same neighborhood. One cost patients glycemic control and one did not. Positivity is therefore not the variable; titer and neutralizing capacity are, and a headline that reports only the first is reporting the wrong one. The same analysis adds a detail worth keeping: patients switched from exenatide to liraglutide got a further 0.4% HbA1c reduction despite carrying anti-exenatide antibodies, so antibodies raised against one molecule did not travel to the next.[9]
What the labels admit they did not test
There is a question underneath all of this that the trials mostly left open, and the labels are unusually candid about it.
Some anti-drug antibodies cross-react with the body’s own hormone. On Wegovy, 28 of the 50 antibody-positive patients — 56% of them — developed antibodies that cross-reacted with native GLP-1, and the label states outright that the clinical consequences of that are unknown.[6] On Zepbound, 40% and 16.5% of treated patients in the weight studies developed antibodies cross-reactive to native GIP or native GLP-1.[1] The Victoza label goes further and names the gap: 56 patients (5%) had antibodies cross-reacting with native GLP-1, and those antibodies were not tested for a neutralizing effect against native GLP-1, so the potential for clinically significant neutralization of the native hormone was not assessed.[3] Ozempic records that the in vitro neutralizing activity of its antibodies is uncertain at this time.[5]
None of that is evidence of harm. It is an honest boundary on a reassuring dataset, and the reason to state it is that the dataset is otherwise so reassuring that it invites overreach in the other direction.
Why these particular peptides are quiet
There is a plausible structural reason the acylated GLP-1 analogs sit low on the table, and it is not an accident of manufacturing. Semaglutide and liraglutide both carry a fatty acid chain attached to a lysine residue, included to bind albumin and extend half-life.
A study testing three immunogenic model peptides and fourteen acylated analogs against 26 common HLA class II alleles found that acylation with fatty acids consistently decreased the peptides’ capacity to activate T cells, an effect most evident with longer fatty acids positioned inside the HLA class II binding core or near the C-terminus.[10] Binding-affinity changes were modest; the T-cell activation change was not. A modification made for pharmacokinetics appears to suppress the first step of an immune response as a side effect. That is a mechanism, on model peptides, rather than a measurement of these drugs — but it fits the labels.
Where compounding enters, and it is through the impurities
Everything above describes the approved products, made to a filed process with release specifications. A compounded preparation is the same amino acid sequence made by a different route from a different bulk substance. The immunological question is not about the sequence. It is about what else is in the vial, which is the subject of the contents article and the reason which powder went in is a regulated question at all.
Peptide-related impurities are a recognized immunogenicity risk because a deletion, an addition or a modified residue can create a novel T-cell epitope that the parent sequence does not contain — the reason the abbreviated pathway for generic peptides asks sponsors to characterize any new impurity against the originator.[11] A 2025 review of the field is blunt about how far that reasoning can currently be taken: regulatory guidance on impurity qualification thresholds is sparse, and the in vitro and in silico methods for assessing an impurity’s immunogenicity risk have real technical limits, which bites hardest exactly where clinical data does not exist — follow-on products.[12]
The measurements that do exist are pointed. An analysis of follow-on and compounded semaglutide and liraglutide ran mass spectrometry, photostability testing, a fibrillation assay and an MHC class II-associated peptide proteomics assay, and found potentially immunogenic peptides presented on dendritic cells stimulated with the impurities, in a distinct number and distribution from the originators; the compounded and follow-on materials carried different impurity profiles including amino acid deletions and additions, and compounded semaglutide showed significant disparity in strength, impurity sum and high-molecular-weight protein content against originator product on light exposure.[13] Separately, testing of compounded tirzepatide combined with vitamin B12 identified a widespread, previously unreported impurity formed by a reaction between the peptide and certain B12 analogs, at substantial levels, with clinical effects unknown.[14]
Two things have to be said about those studies at once. The chemistry is real and the methods are strong. And both were performed by the originators — the first by Novo Nordisk employees and shareholders, the second by Eli Lilly — which is a conflict of interest and also the reason the data exists, because nobody else has published comparable work.
The gap, stated plainly
No study has measured anti-drug antibody rates in patients taking a compounded GLP-1. Not a high rate, not a low rate — the measurement has not been made. What exists is an incidence table for the approved products, a mechanism connecting impurities to immune recognition, and analytical evidence that compounded material carries impurity profiles the approved products do not. That is a reason to care about who made the powder and what the lot was tested for, which is what a certificate of analysis is supposed to document, and what handling after the vial leaves the pharmacy does to the same molecule is a related question. It is not a reason to tell anyone their compounded vial has raised antibodies against them. Nobody knows that, in either direction, and saying otherwise would be inventing the study that does not exist.