Weighing the evidence on GLP1s and venous thromboembolism Open Access

An elevated body mass index (BMI) is a significant contributor to overall cardiovascular diseases and may contribute to one-quarter of the population’s attributable risk for venous thromboembolism (VTE).¹ Glucagon-like peptide 1 receptor agonists (GLP1-RAs) are a class of drugs that increase glucose-dependent insulin secretion, slow gastric emptying, and decrease appetite. Randomized control trials (RCTs) with these drugs have demonstrated improvements in cardiovascular outcomes² among those with diabetes and impressive weight loss, even in patients without diabetes.³ Perhaps more than any class of drugs in recent history, GLP1-RAs have created a deluge of public interest,⁴ capturing media attention and receiving celebrity endorsements, with some economists predicting that they will move the needle not only on the bathroom scale but ultimately on the national gross domestic product.⁵ 

In this issue of Blood Advances, Chiang et al⁶ build on existing studies by showing that GLP1-RAs, when compared with dipeptidyl peptidase 4 inhibitors (DPP4is), specifically reduce the VTE risk in patients with diabetes mellitus, type 2, and that the mechanism seems to be independent of the initial BMI. A target trial emulation (TTE) methodology was used based on the availability of detailed electronic health records from 140 health care systems and 21 countries (TriNetX), which led to the identification of a large population of patients who initiated either GLP1-RAs or DPP4is and enabled the evaluation of the subsequent VTE risk over a 1-year timeframe. In a propensity-score matched cohort of 540 258 patients with diabetes and a mean BMI of 33.8 kg/m², ∼20% VTE reduction was found with GLP1-RAs when compared with DPP4is. The rate of VTE reduction with GLP1-RAs was similar in patients with obesity (hazard ratio [HR], 0.78; 95% confidence interval [CI], 0.71-0.87) and those without obesity (HR, 0.68; 95% CI, 0.58-0.79). The reduction in VTE was also seen when GLP1-RAs were compared with other diabetes medications in sensitivity analyses.

Several choices around the study design should be highlighted to better understand the strength of the study conclusions. TTE is a technique that uses observational data to carry out a mock RCT and, in doing so, supports more causal interpretations than those that can be concluded from observational data while being more pragmatic and expeditious.⁷ With TTE, there are 2 key features, namely (1) the design is set up so that key elements used to describe the treatment effect are specified (population eligibility criteria, treatment strategies, treatment assignment from time zero, end points, outcome measures, etc), and (2) baseline confounders are accounted for to mimic randomization (herein, they used propensity score matching). Although this study provided a large population to help address the clinical questions, it also has inherent limitations related to the TriNetX database, namely data validity, coding practices, and completeness can vary across institutions and patients. For example, only 60% of patients in this study had available hemoglobin A1c and BMI data, and follow-up BMI data were not available to understand how the degree of weight loss might have influenced the findings. Despite propensity score matching, there may still be unmeasured variables that may explain why prescribers select certain medications over others, which, in turn, would affect the study outcomes.

With the important limitations of the study design noted, if the findings are replicated, this represents a potentially substantial off-target effect of the GLP1-RAs. Although it remains to be seen whether a similar VTE risk reduction might be seen in a nondiabetic population, the authors demonstrated a significant VTE risk reduction in both the nonobese and obese subgroups with risk reduced across all initial BMI categories. This study’s results bring attention to older findings from the HOPE-3 and JUPITER RCTs that demonstrated a 47% reduction in VTE with rosuvastatin⁸; it is intriguing to consider future possibilities of long-term nonanticoagulant treatments to reduce the primary or secondary risk of VTE. Important questions remain regarding the underlying mechanism of VTE reduction, namely is the reduction in VTE events related merely to the benefits of weight loss or something specific about VTE risk in diabetes or are there other mechanisms at play and could they be synergistic with the VTE reductions seen with statins? Could the solution to decreasing bleeding while preventing VTE lie outside of the coagulation cascade altogether? The broader impact of potentially reducing the incidence of VTE at a larger population level is also exciting to consider as GLP1-RAs are gaining widespread use for weight loss in people with obesity.

The authors showed a reduced VTE risk in a population of people with diabetes who were treated with GLP1-RAs when compared with those treated with DPP4is. Although this conclusion should be interpreted with some caution because of the study design limitations, it should be considered a robust initial evaluation. Future research is needed to confirm this finding and should explore whether this treatment strategy effectively reduces VTE risk in a general population and should investigate the underlying mechanisms further.

Conflict-of-interest disclosure: The authors declare no competing financial interests.