AUTHORS
Lauren Fang, MD, Amina Akram, MBBS2, Buraq Ahmed3, Animesh Pardanani, PhD, MBBS4, Ayalew Tefferi, MD5, Naseema Gangat, MBBS6
AFFILIATIONS
1 Mayo Clinic, Rochester, Minnesota, United States
2 Mayo Clinic, Rochester, United States
3 University of Cambridge, Cambridge, United Kingdom
4 Mayo Clinic, Rochester, United States
5 Mayo Clinic, Rochester, United States
6 Mayo Clinic, Rochester, United States
ABSTRACT
Background
Polycythemia vera (PV) carries an increased risk of arterial and venous thrombosis. GLP-1 receptor agonists (GLP-1RA), increasingly prescribed for type 2 diabetes mellitus and obesity, promote weight loss and have anti-inflammatory and cardioprotective effects. This has prompted interest in GLP-1RA in myeloproliferative neoplasms (MPN), where preliminary registry analyses suggest more favorable outcomes in exposed patients (Chang et al., ISTH 2025; Cheema et al., Clin Lymphoma Myeloma Leuk 2025). However, such observations are susceptible to immortal time bias, in which longer-surviving patients are preferentially classified as exposed. We assessed the association of GLP-1RA exposure with overall survival (OS) and incidence of arterial and venous thrombosis in PV, using methods accounting for immortal time bias.
Methods
Within a Mayo Clinic PV cohort, GLP-1RA recipients were compared with non-recipients. OS was estimated by Kaplan-Meier analysis and Cox regression treating GLP-1RA exposure as a time-dependent covariate. Analyses were adjusted for age, sex, diabetes, prior thrombosis, and revised International Working Group (IWG) risk. Among recipients, thrombotic events were divided into pre-exposure, active exposure, and post-discontinuation periods. Incidence rates (events per 100 person-years) and incidence rate ratios (IRRs; exact Poisson tests) were compared between exposure periods and with non-recipients.
Results
Comparison of GLP-1RA users and non-users. Among 1,060 patients with PV, 93 (9%) received GLP-1RA, initiated a median of 59 months (range, -73 to 390) relative to PV diagnosis for a median of 23 months (range, 0-87). Semaglutide (59%) and tirzepatide (55%) were the most commonly prescribed GLP-1RAs. Compared with non-users, recipients were younger (median age, 56 vs 62 years; p<0.001) and more often had diabetes (49% vs 12%; p<0.001), hypertension (71% vs 53%; p<0.001), and a smoking history (35% vs 10%; p<0.001), with lower hemoglobin (15.5 vs 17.4 g/dL; p<0.001) and leukocytes (9.5 vs 10.8 x10^9/L; p=0.007). Sex (female, 51% vs 49%; p=0.84) and prior arterial (13% vs 14%) and venous (15% vs 14%) thrombosis were comparable. Recipients were less often high/very-high risk by the revised IWG survival model (16% vs 46%; p<0.001), a difference driven by younger age, with similar within-strata distributions. Aspirin (92% vs 94%) and cytoreductive therapy (85% vs 90%; p=0.18) exposure were similar.
Overall survival. After a median follow-up of 9.6 years, 1 (1%) recipient and 436 (45%) non-recipients had died. Without accounting for immortal time bias, GLP-1RA use was associated with superior OS (p<0.001), persisting after adjustment for age and sex (p=0.007), diabetes (p=0.004), and revised IWG risk (p=0.004). However, after modeling GLP-1RA as a time-dependent covariate, this association was markedly attenuated (p=0.046) and non-significant after multivariable adjustment (p=0.14-0.21), including among diabetics (p=0.28).
Arterial and Venous Thrombosis. During post-diagnosis follow-up, recipients experienced 15 arterial and 17 venous thrombotic events (152 and 153, respectively, among non-recipients). Among recipients, 24 events (11 arterial/13 venous) occurred before GLP-1RA initiation (662.0 person-years; 3.63 per 100 person-years), 4 (2 arterial/2 venous) during active exposure (126.4 person-years; 3.16 per 100 person-years), and 4 (2 arterial/2 venous) after discontinuation (41.1 person-years; 9.7 per 100 person-years). Non-recipients experienced 305 thrombotic events during 11,230 person-years of follow-up (2.7 per 100 person-years). The incidence rate during GLP-1RA exposure was similar to the pre-exposure period (IRR 0.87; 95% CI 0.22-2.54; p=1.00) and comparable to the rate in non-recipients (3.16 vs 2.72 per 100 person-years; IRR 1.17; 95% CI 0.32-3.01; p=0.59).
Conclusion
In PV, the apparent survival advantage of GLP-1RA therapy was fully explained by immortal time bias, with no independent survival benefit after time-dependent modeling. Thrombotic event rates during GLP-1RA exposure did not differ significantly from the recipients' pre-exposure period or from non-recipients. These findings should be interpreted cautiously given the limited number of on-treatment events and potential confounding by indication, as GLP-1RA recipients had a higher burden of diabetes and other cardiovascular risk factors than non-recipients.
DISCLOSURES
Lauren Fang, MD: No relevant conflicts of interest to declare.