Semaglutide and Papillary Thyroid Carcinoma: Current Evidence on Risk, Progression, and Mechanisms

1. Introduction

Glucagon-like peptide 1 receptor agonists (GLP-1RAs), including semaglutide, liraglutide, dulaglutide, exenatide, and the dual GIP/GLP-1 agonist tirzepatide, have changed the pharmacologic management of type 2 diabetes and obesity. Semaglutide, marketed as Ozempic for diabetes and Wegovy for weight management, has reached tens of millions of prescriptions annually in the United States, with continued growth projected through the mid-2030s. This scale of exposure has increased scrutiny of rare adverse events that were difficult to detect in trials conducted before approval.

Concern about thyroid malignancy originates in preclinical toxicology. Rodent studies found increases in thyroid C-cell hyperplasia, C-cell adenomas, and medullary thyroid carcinoma (MTC) that depended on dose and duration of exposure, following lifetime exposure to plasma concentrations relevant to clinical use [1]. Because rodent C cells express the GLP-1 receptor at much higher density than human C cells, and because no comparable signal has been clearly established in long-term human data, regulators required a boxed warning restricted to patients with a personal or family history of MTC or MEN2, while noting explicitly that the relevance of the rodent finding to humans is unknown.

Papillary thyroid carcinoma (PTC) arises from follicular epithelial cells, a lineage distinct from the parafollicular C cells implicated in the rodent signal, and accounts for the majority of thyroid cancer diagnoses. A prior narrative review addressed GLP-1RAs and thyroid cancer broadly [2]. This review narrows the question further by examining what the literature specifically shows about semaglutide and PTC, including new PTC incidence, progression in patients with existing disease, receptor-level mechanistic plausibility, and the evidentiary weight of each study design.

2. Materials and Methods

This is a narrative literature review. No protocol was registered, and no PRISMA flow diagram was generated. PubMed, Google Scholar, and journal-specific databases were searched between May and August 2026 using combinations of the terms semaglutide, Ozempic, GLP-1 receptor agonist, thyroid cancer, papillary thyroid carcinoma, differentiated thyroid cancer, medullary thyroid carcinoma, GLP-1 receptor expression, and FAERS. Reference lists of retrieved reviews and meta-analyses were hand-searched for additional primary studies. Articles were included if they reported original quantitative data on thyroid cancer incidence, progression, or receptor biology in relation to semaglutide or GLP-1RAs as a class, or if they were case reports or mechanistic studies addressing PTC specifically. No date restriction was applied, although emphasis was placed on studies published between 2023 and 2026 given the pace of change in this literature. A 2024 narrative review addressing GLP-1RAs and thyroid cancer broadly [2] is acknowledged as prior work that this review builds on and updates with more recent, PTC- and progression-specific data.

3. Results

3.1. Biological Rationale and Receptor Expression Evidence

The regulatory warning on GLP-1RAs originates from rodent studies showing C-cell hyperplasia, adenomas, and carcinomas that depended on dose and duration of exposure after lifetime treatment, driven by dense GLP-1 receptor expression on rodent C cells. Human C cells express the receptor at substantially lower density, which is the main reason regulators have cautioned against direct extrapolation to humans.

Whether follicular cells, the cell of origin for PTC, express clinically meaningful levels of the GLP-1 receptor remains unsettled. Gier et al. [3] reported GLP-1 receptor immunoreactivity in approximately 18% of PTC specimens, and Jung and Kwon [4] reported expression in 32.1% of PTC samples, with expression correlating with certain clinicopathologic features. He et al. [5] found GLP-1R and IGF-1R expression more pronounced in PTC than in normal thyroid tissue, yet GLP-1R agonist exposure did not increase PTC cell proliferation or measurably alter Akt/Erk signaling or cellular energy metabolism in their model. Waser et al. [6] reported the opposite finding, that papillary thyroid cancers do not express GLP-1R at all, and concluded that PTC is not a plausible target tissue for GLP-1 analogs in humans. A polymorphism study linked a specific GLP-1R genetic variant and expression pattern to PTC risk in an Egyptian population [7], suggesting that genetic variation across populations may partly explain the differing expression findings. Kennedy et al. [8] described GLP-1R overexpression as defining a distinct immunogenetic subset within primary and metastatic thyroid cancer, raising the possibility that only a molecularly defined subgroup of PTC, rather than PTC as a whole, might be biologically responsive to GLP-1R agonism.

This inconsistency is relevant to the interpretation of downstream functional and clinical data. If only a minority of PTCs express the receptor, and expression is inconsistently detected across studies and populations, then population-level studies that do not stratify by receptor status may dilute or obscure a subgroup-specific effect in either direction.

3.2. Functional and Mechanistic Studies

Wang et al. [9] found that in PTC xenograft and macrophage coculture models, semaglutide did not directly increase PTC cell proliferation but reduced tumor size by reprogramming tumor-associated macrophages from a protumor M2 phenotype toward an antitumor M1 phenotype, mediated through downregulation of PPARG via the GLP-1R/PPARG/ACSL1 pathway. This is a tumor-suppressive mechanism operating in the tumor microenvironment rather than directly on tumor cells. Zhang et al. [10] reported a related finding at the cellular level: liraglutide inhibited proliferation and migration of GLP-1R-overexpressing papillary and medullary thyroid cancer cell lines through changes in PI3K/Akt/mTOR-related signaling. Taken together, the available functional data, though limited to preclinical models, consistently point away from a proliferative effect of GLP-1RA exposure on PTC cells themselves, even in models selected for receptor expression.

3.3. Evidence from Randomized Controlled Trials and Large Pooled Trial Analyses

Vilsbøll et al. [11] pooled data from 93 liraglutide or semaglutide phase 2 and 3 clinical trials comprising 101,732 participants, together with postmarketing surveillance data, and found low absolute thyroid cancer event rates and hazard ratios of 1.70 versus placebo and 1.83 versus active comparators, neither of which reached statistical significance. The authors noted that a trend toward more thyroid cancer diagnoses in GLP-1RA arms was plausibly explained by detection bias from more frequent clinical monitoring rather than a true excess in disease. Postmarketing surveillance in the same analysis found a thyroid cancer reporting rate of approximately 0.001 cases per 100 patient-years, which did not support an association between liraglutide or semaglutide and thyroid cancer.

Nagendra et al. [12], in a semaglutide-specific meta-analysis of 37 RCTs and 19 real-world studies, found no increased risk of thyroid cancer versus placebo (OR 2.04, 95% CI 0.33–12.61) or active comparators (OR 1.19, 95% CI 0.15–9.66). Feier et al. [1] systematically reviewed ten RCTs (14,550 participants, 7830 on semaglutide) and similarly reported thyroid cancer incidence under 1% with no significant excess risk. A broader meta-analysis by Ko et al. [13], published in Annals of Internal Medicine, examined cancer risk across GLP-1RAs and dual agonists as a drug class rather than thyroid cancer alone, and is among the most methodologically current syntheses addressing oncologic risk for this drug class overall.

A recurring point, raised explicitly by Correra et al. [14] in an analysis of semaglutide's oncogenic potential, is that most RCTs were not designed or powered to detect cancer endpoints. A null finding in this context indicates an absence of detectable excess risk within the trials' statistical power, not definitive proof of oncologic neutrality.

3.4. Evidence from Real-World Cohort and Case-Control Studies

Bezin et al. [15] conducted a French nested case-control study using the national SNDS database and reported that one to three years of GLP-1RA use was associated with increased risk of all thyroid cancers (adjusted HR 1.58, 95% CI 1.27–1.95) and medullary thyroid cancer specifically (adjusted HR 1.78, 95% CI 1.04–3.05). This finding generated published discussion concerning residual confounding and detection bias [16,17,18], which the original authors addressed in a formal response [19].

Several larger and more recent cohort studies have not replicated this signal. Morales et al. [20] conducted a multi-database retrospective cohort study using propensity score adjustment across United States and European claims databases and did not find a consistent increase in thyroid tumor risk associated with GLP-1RA use once only diagnostically sound cohorts were retained. A nationwide Korean cohort study by Bea et al. [21] compared GLP-1RA users with DPP-4 inhibitor users and found no increased risk of thyroid cancer, with a weighted hazard ratio of 0.98. A Scandinavian cohort study spanning Denmark, Norway, and Sweden, with a mean follow-up of 3.9 years, similarly found no substantial increase in thyroid cancer risk, with the upper confidence bound consistent with no more than a 31% relative increase [22].

A target trial emulation using United States claims data by Brito et al. [23] found no significant overall increase in thyroid cancer risk with GLP-1RA use, but did observe a higher diagnosis rate in the first year after initiation, a pattern the authors attributed to increased clinical contact and earlier detection rather than a true acceleration of disease. This finding offers a direct empirical illustration of the detection bias mechanism proposed on theoretical grounds by critics of the Bezin et al. findings [15]. Two additional recent cohort studies examined patients with thyroid nodules or broader comparator drug classes. Balachandra et al. [24] found no significant difference in thyroid cancer risk between GLP-1 analog users and metformin users among patients with known thyroid nodules (risk ratio 0.99), and Sciscent et al. [25], using the TriNetX database, found no increased five-year thyroid cancer risk with GLP-1RA use compared with SGLT2 inhibitors, metformin, or DPP-4 inhibitors. An earlier cohort study by Funch et al. [26], examining a United States commercially insured population initiating liraglutide, similarly found no significant elevation in thyroid cancer risk after adjustment for latency, while noting the same underlying concern about surveillance-related detection bias.

3.5. Progression in Patients with Existing Differentiated Thyroid Cancer

Most of the evidence above addresses new thyroid cancer incidence in populations not previously exposed to the drug, which is a different clinical question from whether GLP-1RA exposure affects the course of disease in a patient who already has PTC, a question directly relevant to patients and clinicians managing known disease. Patrizio et al. [27] provide the most directly relevant data identified for this review: a matched cohort of 1072 patients with differentiated thyroid cancer, the vast majority with papillary histology, followed for a median of 69 months, in which GLP-1RA exposure was not significantly associated with disease recurrence or structural progression. Because this is a progression-specific, histologically relevant, multi-year cohort, it provides direct information on the progression question specifically, although it remains a single study and would benefit from replication.

3.6. Evidence from Pharmacovigilance and Spontaneous Reporting Databases

FAERS-based disproportionality analyses produced a signal that differed from clinical trial and cohort data, with the caveat that such databases cannot establish incidence or causation. Abi Zeid Daou et al. [28] found statistically significantly elevated reporting odds ratios for thyroid cancer for several GLP-1RAs, including semaglutide (ROR 7.61, 95% CI 6.37–9.08). Yang et al. [29], in an independent FAERS analysis spanning 2004 to 2021, similarly identified tumor-related adverse event signals associated with GLP-1RA use as a drug class. Darcy et al. [30] integrated the Surveillance, Epidemiology, and End Results (SEER) cancer registry with FAERS data to model GLP-1RA-associated medullary thyroid cancer risk, an approach that helps address the lack of a denominator in FAERS but was designed around MTC rather than PTC specifically. As with other FAERS-based findings discussed in this review, reporting patterns following publications and media coverage, along with the inability to confirm histologic subtype in many reports, limit how directly these signals can be applied to a PTC-specific question.

3.7. Case Reports

Two recent case reports describe different individual presentations and are discussed here because they illustrate the range of findings in the literature. Haddadin et al. [31] described a patient in whom PTC was discovered during endocrine evaluation after semaglutide use for metabolic syndrome. The case was notable for an unremarkable thyroid exam and an indolent presentation, and the authors framed it within the broader body of literature that does not support an association, rather than as evidence of causation. Beshay et al. [32] separately reported a case of mixed medullary and papillary thyroid carcinoma discovered in a patient on tirzepatide, a related dual GIP/GLP-1 agonist, again noting that current data do not support a confirmed association with differentiated thyroid cancer specifically.

Stiewig Rapp and Basheer [33] reported a 35-year-old woman who developed a PTC measuring 6.8 cm within five weeks of initiating low-dose semaglutide (0.25 mg weekly) for weight loss, with no prior thyroid abnormalities documented. This interval and tumor size are notable, and the authors were appropriately cautious that a single case cannot establish causation. A five-week interval is also a short period for a solid tumor of this size to arise without any prior growth, which raises the alternative possibility of a pre-existing, previously undetected lesion that became clinically apparent around the time of presentation. This case represents a data point that differs from the pattern observed in the population and in the progression-level evidence reviewed above, and it is best read as an individual report that supports continued case-level monitoring and reporting rather than as evidence to be weighed equally against the cohort and progression-level data.

3.8. Sex-Based Differences

Papillary thyroid carcinoma has a well-established female predominance, with incidence roughly three times higher in women than men, and none of the established risk factors fully explain this pattern. Despite the lower incidence in men, male sex is recognized as an unfavorable prognostic factor in PTC, and men who develop the disease tend to have more aggressive tumor behavior. In vitro studies indicate that thyroid cancer cells express estrogen receptors and that estrogen has a proliferative effect on PTC cells, yet epidemiologic studies have not established a clear association between thyroid cancer and female hormonal factors, and the mechanism behind the sex disparity remains unresolved.

Whether GLP-1 receptor expression in PTC tissue differs by sex has been examined directly in only one of the studies reviewed here. Jung and Kwon [4] tested sex alongside age, tumor size, extrathyroidal extension, lymph node metastasis, and multifocality as clinicopathologic correlates of GLP-1R expression and found that only tumor multifocality showed a significant association; sex was not a significant predictor of receptor expression in their cohort.

Sex-specific data on GLP-1RA-associated thyroid cancer risk specifically remain limited. A large propensity-matched cohort study using the TriNetX network, examining 1.1 million patients with obesity, found that GLP-1RA use was associated with reduced risk of thyroid and other endocrine gland cancers overall, and that this reduction was present in both sexes when the analysis was stratified [34]. This finding is reassuring but comes from a broad cancer panel study comparing GLP-1RA-treated patients with obesity to untreated patients with obesity, rather than a study specifically focused on thyroid cancer mechanisms or comparator antidiabetic drugs, and it does not isolate PTC from other thyroid cancer subtypes.

The progression cohort that provides the most direct evidence in this review, Patrizio et al. [27], enrolled a population that was 71% female, consistent with the underlying demographics of differentiated thyroid cancer, but the published results do not report a sex-stratified subgroup analysis of the association between GLP-1RA exposure and progression. Similarly, the French case-control study by Bezin et al. [15] matched cases and controls by sex as part of the study design but did not report a sex-stratified risk estimate for thyroid cancer specifically. A related, smaller cohort from the same research group examined tumor kinetics in patients with low-risk PTC under active surveillance, comparing 18 patients exposed to GLP-1RA with 37 unexposed patients, but this study was not designed to test sex as an effect modifier [35].

Taken together, sex is not currently established as a factor that modifies the relationship between GLP-1RA exposure and PTC incidence, progression, or receptor biology, but this reflects an absence of dedicated study rather than evidence of no effect. Given the pronounced and largely unexplained sex disparity in PTC epidemiology generally, and the female-predominant population that receives GLP-1RA therapy for both diabetes and obesity indications, sex-stratified analysis represents a specific and achievable next step for future cohort and mechanistic studies in this area.

4. Discussion

Reconciling these study types requires attention to what each design can and cannot show. Randomized controlled trial and pooled trial-level evidence, now anchored by a pooled analysis of more than 100,000 participants, does not show a statistically robust increase in thyroid cancer risk with semaglutide [1,11,12]. The largest and longest follow-up cohort studies are similarly consistent with no increased risk for new-onset disease [20,21,22], and, for the progression question specifically, the available progression-focused cohort in patients with existing differentiated thyroid cancer found no signal over nearly six years of follow-up [27]. However, the generally indolent natural history of differentiated thyroid cancer makes duration of drug exposure an important limitation when interpreting these findings. Across the randomized and observational literature, median GLP-1RA exposure has generally not exceeded approximately four to five years, and even the Patrizio et al. cohort [27] paired approximately 5.5 years of surveillance with substantially shorter cumulative on-drug exposure. In the pooled analysis by Vilsbøll et al. [11], hazard ratios for thyroid cancer were directionally above unity across comparisons, including 1.70 versus placebo, with the lower confidence bound approaching the null, although none reached statistical significance. These estimates do not establish increased risk, but their direction and imprecision underscore the need for adequately powered studies with semaglutide exposure beyond five years and longer follow-up to determine whether a long-latency effect exists.

By contrast, the French case-control study and FAERS-based analyses report elevated risk signals [15,28,29,30], and both designs are subject to detection bias and reporting patterns influenced by publicity, a mechanism illustrated empirically by the first-year diagnosis pattern observed in the Brito et al. target trial emulation [23]. Obesity itself is an independent risk factor for thyroid cancer, and because GLP-1RA users are disproportionately drawn from a population with obesity, some portion of any observed association may reflect confounding by indication.

The receptor expression literature adds a layer of biological uncertainty rather than resolving it. Studies disagree on whether and how often PTC tissue expresses the GLP-1 receptor at all [3,4,6], and the studies that did find expression did not find that receptor agonism drove proliferation [5,9,10]. The case report describing rapid enlargement of a PTC following semaglutide initiation [33] differs from the pattern seen in most of the population-level and progression-specific evidence, and a formal evidence gap analysis by Salama et al. [36] identified GLP-1RA effects on thyroid function and thyroid disease outcomes, including PTC, as an area warranting dedicated prospective study. Sex-specific effect modification is a related and largely unexamined dimension of this evidence gap. Given the pronounced female predominance of PTC and the female-predominant populations exposed to GLP-1RA therapy, the absence of sex-stratified reporting across most of the cohort and case-control literature reviewed here limits the ability to determine whether the relationship between GLP-1RA exposure and PTC outcomes differs between men and women.

4.1. Limitations

Several limitations recur across the body of evidence reviewed here. Follow-up duration and, importantly, cumulative GLP-1RA exposure remain short relative to the indolent natural history of most differentiated thyroid cancers. Adequately powered studies with semaglutide exposure beyond five years and longer follow-up will therefore be important to determine whether a long-latency effect exists. Observational and pharmacovigilance studies remain subject to detection bias. Obesity is a shared risk factor for both semaglutide use and thyroid cancer, which complicates adjustment in claims-based studies. Receptor expression studies use varied immunohistochemical methods and relatively small PTC samples, which may explain some of the disagreement among them. FAERS and other spontaneous reporting systems cannot generate incidence estimates. Sex-stratified analysis is reported in only a minority of the studies reviewed here, limiting the ability to draw conclusions about whether GLP-1RA-associated thyroid outcomes differ between men and women. This narrative review is also subject to the selection and interpretation limitations inherent to non-systematic methodology. A formal systematic review or meta-analysis restricted specifically to PTC, rather than thyroid cancer broadly, has not yet been published and would be a natural next step given the evidence gap identified by Salama et al. [36].

5. Conclusions

Current evidence, taken as a whole, does not support semaglutide as a driver of papillary thyroid carcinoma incidence or progression. Large pooled trial data, multiple national cohort studies, and a multi-year progression cohort in patients with differentiated thyroid cancer found no significant association. Receptor expression and mechanistic studies show inconsistent results but, on balance, do not support a proliferative effect of GLP-1 receptor agonism on PTC cells. Signals from a French case-control study and FAERS disproportionality analyses, along with a single case report describing rapid tumor enlargement, indicate that some uncertainty remains, and the field has recognized GLP-1RA effects on thyroid disease as an evidence gap. Available data provide reassurance for most patients, while continued attention to and reporting of atypical presentations, such as rapid tumor growth, remain appropriate. Prospective, receptor-status-stratified, progression-focused studies specific to PTC, rather than thyroid cancer as an aggregate outcome, represent a clear path toward resolving the remaining uncertainty. Sex-stratified analysis, largely absent from the current literature despite the pronounced female predominance of PTC, represents an additional and achievable direction for future studies in this area.