Risk of Gynecologic Tumors Linked to Using GLP-1 Receptor Agonists and SGLT2 Inhibitors: A Comparison of 91 Clinical Trials

Glucagon-like peptide-1 (GLP-1) receptor agonists and sodium–glucose co-transporter 2 (SGLT2) inhibitors are modern antidiabetic agents that offer novel mechanisms of glycemic regulation, distinct from traditional therapies [1]. Nonetheless, recent investigations have raised safety concerns, with emerging evidence implicating a possible link between these agents and tumorigenesis [2, 3]. A growing number of case series, clinical investigations, and real-world data analyses have pointed to a potential association between their use and cancer development [4 –6]. Laboratory-based research has proposed several plausible biological mechanisms for this effect. For instance, SGLT2 inhibitors may exert anti-apoptotic activity via AMPK/SIRT1 signaling [7, 8] or ERK1/2 activation through calcium-mediated pathways [9], both of which may follow a dose-response pattern [10, 11]. Similarly, GLP-1 receptor agonists have been shown to interfere with apoptotic regulation via BCL2/BAX/caspase-3 pathways [12] or influence gene expression through epigenetic modifications such as DNA demethylation of tumor suppressor loci [13].

Gynecologic tumors—which include malignancies of the uterus, cervix, endometrium, ovary, breast, vagina, and vulva—are prevalent among individuals who meet the indications for GLP-1 receptor agonists or SGLT2 inhibitors [14 –16]. For example, in the meta-analysis by Friberg et al. [17], the authors summarized that diabetes was statistically significantly associated with an increased risk of endometrial cancer with a risk ratio of 2.10. Similarly, in another meta-analysis by Qin et al. [18], they noticed a positive association between obesity and the risk/prevalence of uterine tumors with an odds ratio (OR) of 1.19. Notably, several clinical trials have hinted at an increased incidence of gynecologic neoplasms among patients using these agents, although statistical significance was not consistently achieved [19 –21]. In scenarios where trial data yield inconclusive findings due to design or power limitations, meta-analyses serve as critical tools to uncover patterns across heterogeneous datasets. By integrating results from large-scale studies and clinical databases, meta-analyses are regarded as robust indicators of real-world therapeutic outcomes and have frequently informed subsequent research priorities [22, 23].

However, existing meta-analyses in this field offer conflicting interpretations [24 –27]. Many prior efforts employed conventional pairwise comparisons and were limited by several methodological constraints. For example, some pooled drugs of different classes or doses into broad categories, potentially obscuring drug-specific carcinogenic signals—especially those linked to agents with unique pharmacokinetic or pharmacodynamic profiles such as tirzepatide. Others stratified analyses by compound but failed to account for dose-dependent variability [24 –26]. This limitation is notable given evidence from experimental models where glucose-dependent insulinotropic peptide (GIP) agonists, such as tirzepatide, have been shown to promote tumor proliferation in a dose-sensitive manner, particularly in colorectal cancer cell lines [28].

To address these gaps, we designed a comprehensive network meta-analysis (NMA) that stratifies GLP-1 receptor agonists and SGLT2 inhibitors by both agent and dosage. This approach enables a refined comparison of gynecologic tumor risks attributable to individual regimens. Our current NMA is conceptually aligned with our prior investigations on the neurodegenerative [29], metastatic cancer prevention [30], and auditory [31] risks associated with these agents. To our knowledge, no prior NMA has systematically evaluated site-specific gynecologic tumor risks across different agents and dosages. Therefore, this study aims to clarify the comparative risk profiles of these treatments in relation to gynecologic tumor development.

This network meta-analysis (NMA) was performed using a confirmatory strategy based on Cochrane recommendations [32], targeting predefined adverse outcomes—specifically, the incidence of gynecologic tumors. In brief, we listed targeted adverse events as outcomes of interest (i.e. incidence of gynecologic tumors based on biological plausibility here) in our review protocol as our main interest. As notified in Cochrane manuals "review authors should assume adverse events to be measured regularly and consistently in studies" [32], we strictly focused on randomized controlled trials (RCTs) with systematically report of adverse events or explicitly evaluating target outcomes [33]. The reporting adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, including the extension for network meta-analyses (PRISMA NMA) [34] (see eTable 1). This study protocol was registered with PROSPERO (registration number CRD420251003016) and received ethics approval from the Institutional Review Board of the Tri-Service General Hospital, National Defense Medical Center (TSGHIRB E202516007).

To our knowledge, this is the first network meta-analysis (NMA) to comprehensively evaluate gynecologic tumor risk from GLP-1 receptor agonists and SGLT2 inhibitors, stratified by both agent and dosage, with exclusive focus on female participants. Our findings show that tirzepatide at 15mg/week—a dosage approved in clinical practice [140] —was significantly associated with increased risk of both overall gynecologic tumors and intra-uterus tumor. Besides, the average cumulated dosage tirzepatide received before developing overall gynecologic tumors or intra-uterus tumor was 1067.0 mg and 1206.3 mg, respectively, which meant that subjects receiving tirzepatide at 15mg/week would need 71.1 weeks and 80.4 weeks to achieve it. These results support the hypothesis that carcinogenic potential may depend not only on drug class but also on specific regimen and dosages.

The most clinically significant observation in this study is that only tirzepatide, and only at its highest dose (15mg/week), was associated with a significant elevation in gynecologic tumor risk, specifically intra-uterus tumor, which differed from the results in the recent retrospective cohort study by Dai et al. [141]. In that report by Dai et al., the authors pooled various GLP-1 receptor agonists with different pharmacology properties into one group and found that GLP-1 receptor agonists were associated with a reduced risk of endometrial cancer and ovarian cancer [141]. This approach of pooling various medications with different pharmacologic properties without accounting for differences in pharmacologic characteristics or dosages (i.e. grouping pure GLP-1 agonists with GLP-1/GIP agonists) risks inconclusive results due to underlying heterogeneity within the comparison group. Specifically, in the report by Dai et al., although the authors demonstrated reduced risks of endometrial cancer (hazard ratio [HR], 0.75 [95% CI, 0.57–0.99.57.99]; p = 0.05) and ovarian cancer (HR, 0.53 [95% CI, 0.29–0.96.29.96]; p = 0.04) by GLP-1 receptor agonist prescription, the wide confidence interval and borderline p value might diminish the confidence of their findings [141]. For example, Dicembrini et al. found no significant association between SGLT2 inhibitors and breast cancer [24], while subsequent meta-analyses by Spiazzi et al. and Xu et al. extended the analysis to other tumor types (endometrial and uterine) and still observed no significant risks [25, 26]. On the GLP-1 side, Piccoli et al. similarly reported no significant risk of breast cancer across agents [27]. Silverii et al. demonstrated that overall GLP-1 receptor agonists might reduce uterine cancer risk in obese patients [142]. Although a few studies stratified results by compound [24 –26], most did not incorporate dose-specific effects—potentially overlooking subtle but meaningful differences in safety. Finally, Kamrul-Hasan et al. demonstrated the insignificant association between different dosages of tirzepatide and risks of breast cancers, ovarian cancers, and uterine cancers [143]. However, the heterogeneity in the control arms might limit the statistical significance in the meta-analysis by Kamrul-Hasan [143]. This is important because drug safety profiles may be highly dose-sensitive, with certain adverse effects manifesting only at higher thresholds [144]. Indeed, Prabakaran et al. found that GLP-1 receptor agonists promoted cell proliferation in vitro in a dose-dependent fashion [28].

Our study advances this field by leveraging NMA methodology to dissect compound- and dose-specific risks. The association we found between high-dose tirzepatide and gynecologic tumors implies that not all GLP-1 receptor agonists are equally safe in this context, and that dose escalation may shift the risk–benefit balance unfavorably. These insights could aid clinicians in selecting safer regimens, particularly for patients with underlying risk factors for gynecologic malignancy.

Tirzepatide represents a unique agent within this therapeutic class, often referred to as a "twin-cretin" because of its dual activity as both GLP-1 and glucose-dependent insulinotropic polypeptide (GIP) receptor agonist [145]. While this dual mechanism enhances glycemic and weight-loss benefits, GIP signaling may also pose oncologic concerns. Prior studies have indicated that GIP receptor activation can stimulate growth in various tumor types, particularly solid tumors [146]. In a preclinical model, Prabakaran et al. demonstrated that GIP agonism enhanced colorectal cancer cell proliferation in a dose-responsive manner [28]. This raises legitimate concerns regarding the carcinogenic potential of high-dose tirzepatide [147].

In particular, tirzepatide may increase the synthesis of insulin, IGF-1, IGF-2, and their binding proteins—factors that have been linked to elevated cancer risk, especially in patients with prior pancreatic injury [147]. Moreover, in genetically predisposed populations, such as those with a family history of medullary thyroid carcinoma or multiple endocrine neoplasia type 2, tirzepatide might further elevate thyroid cancer risk through GIP-mediated pathways [148]. Interestingly, uterine tissues appear to express minimal GIP receptor activity [149], suggesting that alternate mechanisms may drive the observed association with uterine tumors. One hypothesis is that tirzepatide may act synergistically with circulating glucose or insulin levels to activate growth factor–like pathways and stimulate cell proliferation through pleiotropic mechanisms [150]. Besides, the diverse result of tirzepatide in uterine tumor and endometrial tumor might reflect the fact that tirzepatide would exert diverse effects on uterine and endometrial cells. Nevertheless, there is currently no direct clinical or mechanistic study linking tirzepatide to uterine tumorigenesis. Therefore, prospective clinical and translational studies are needed to further investigate this association.

This study possesses several methodological strengths. First, by utilizing an NMA framework, we were able to compare multiple agents simultaneously while maintaining internal consistency across studies. Second, the strict inclusion of randomized controlled trials (RCTs) enhanced the validity of our findings by minimizing confounding. Third, our approach excluded trials involving patients with baseline gynecologic tumors, allowing us to focus on potential causality rather than recurrence or progression. Fourth, we conducted detailed subgroup analyses by tumor site, enabling more precise interpretation of risk across gynecologic tumor subtypes. Finally, robustness was confirmed through Bayesian-based sensitivity analyses, which corroborated our primary findings.

Nonetheless, certain limitations should be acknowledged. The average trial duration in our dataset was 122.0 weeks. While this may seem adequate, tumor development—particularly for solid tumors—often requires longer latency periods [151]. Therefore, longer-term follow-up studies would be valuable to confirm our observations. Second, although our RCT-focused design enhanced internal validity, it inherently excluded observational data that might offer insights into late-onset events. However, RCTs are less prone to confounding and reporting bias. Third, variation in diagnostic criteria across countries and trials may have introduced heterogeneity. Unlike standardized tumor surveillance protocols, adverse event reporting is typically less rigorous, which may lead to underreporting or misclassification of tumor events. Specifically, many of the included RCTs were designed to assess cardiometabolic endpoints—not tumor risk—so relevant variables such as tumor family history or hormonal status were not adjusted for in the original designs. To overcome these inherent limitations (diagnostic criteria variation and potentially missing report of target adverse events), we adopted confirmatory approach based on Cochrane recommendations [32] and strictly focused on RCTs with systematically report of adverse events or explicitly evaluating target outcomes [33]. These adaptation strategies had been widely used in various high impact articles, which discovered new insight regarding medication adverse events from RCTs not designed for target adverse events [152, 153]. While this may affect effect size precision, it does not negate the utility of our findings as hypothesis-generating evidence. Moreover, our use of a Cochrane-endorsed confirmatory strategy strengthens the interpretability of tumor signals derived from adverse event data.

This NMA highlights a potential association between tirzepatide and gynecologic tumor risk, especially intra-uterus tumors, at the commonly used 15mg/week dose. While most GLP-1 receptor agonists and SGLT2 inhibitors appear oncologically safe, our results underscore that regimen-specific risks must not be overlooked. Given that many candidates for these agents already have predisposing risk factors for cancer, treatment decisions should be individualized. Besides, clinicians should carefully monitor the possibility of gynecologic tumors, such as intra-uterus tumors, when their patients receiving 15mg/week tirzepatide for about 71.1 to 80.4 weeks (about 1.3 years). However, more data are needed to confirm whether this association is causal and not due to chance/bias or not. Therefore, our findings call for further long-term trials and mechanistic studies to clarify the oncogenic potential of tirzepatide and to guide safer clinical use.

Additional file 1

Supplementary Material 2