Using data collected from nationwide population-based electronic health records in the United Kingdom, we found that metformin use was associated with a statistically significantly lower risk of incident VTE, as well as PE and DVT individually, in comparison with sulfonylurea use among persons with type 2 diabetes. The findings remained consistent across various sensitivity analyses.
Comparison With Previous Studies
To date, there is a paucity of data regarding the VTE risk of metformin use. Several in vivo studies have shown a potentially protective effect of metformin use on the risk of venous thrombosis.[14,17,20,44] Some studies demonstrated that the protective effect of metformin may operate through its inhibiting the activation, adhesion, and aggregation of platelets.[14,17] Another study showed that metformin use decreased the size of a venous thrombus and thus could prolong the average time to occlusive thrombosis in the inferior vena cava. To our knowledge, only 1 cohort study investigated the association between metformin and risk of incident DVT among patients with type 2 diabetes; the authors reported a decreased risk of DVT (HR = 0.43, 95% CI: 0.24, 0.76) among metformin users. In that study, however, the risk of DVT among metformin users was compared with that among nonusers of metformin. As the result, the findings might have been susceptible to channeling bias. For example, the nonusers of metformin may have had more severe diabetes than metformin users, which may have led them to receive more intensive antidiabetic treatment for their illness. As a result, disease severity may have confounded the association reported in the study. In addition, patients in the nonuser group had higher prevalences of CVD, history of fracture, and major surgery; thus, these risk factors may have made the nonusers of metformin more susceptible to the risk of DVT. Although several new medications are now available in preference to sulfonylureas as first-line medications for type 2 diabetes when patients have contraindications for metformin use, sulfonylureas remain the most commonly prescribed antidiabetic drugs after metformin.[45,46] Previous studies have found that there is no difference in the effect on HbA1c levels between metformin and sulfonylureas; thus, we chose sulfonylurea initiators as an active comparator for metformin initiators to minimize confounding by indication.[30,48,49] Compared with initiators of sulfonylureas, we confirmed previous findings that metformin use was associated with lower risks of both DVT and PE, a more severe subtype of VTE.
Several potential biological mechanisms might account for our findings. First, metformin has a favorable impact on endothelial cell function, especially in improving the integrity and adherence of endothelial cells, as well as preventing cellular apoptosis and atherosclerosis.[50,51] Second, previous in vitro and in vivo studies showed that the function of platelets would be suppressed after administration of metformin by inhibiting mitochondrial DNA release and decreasing diphosphate-induced activation, adhesiveness, and aggregation.[18,50,52] Third, metformin was reported to be associated with a decreased fibrinogen level, reduced activity of coagulation factors such as Factor VII and Factor XIII, and improved lysability by changing the clot structure.[17,53] These vascular protection benefits of metformin might potentially contribute to a lower risk of VTE.
Strengths and Limitations
This study had several strengths. First, we implemented a new-user design to assess the risk of incident VTE for metformin use. This design can minimize the potential for selection bias. Second, the use of an active comparator rather than a nonuser comparison group might have minimized confounding by indication. Third, we started our follow-up after the second dispensed prescription, because patients with a second prescription are more likely to adhere to their medication regimen. Fourth, the observed associations remained consistent in several sensitivity analyses independent of potential confounders, suggesting robustness of our findings. Finally, since numerous studies have demonstrated that metformin use is associated with a reduced risk of all-cause mortality and incident CVD,[41,42] the metformin findings related to these 2 conditions as the positive control outcomes provided further evidence of validity for our main study findings.
The limitations of this study should be acknowledged as well. First, although every effort was made to control for the measured confounders, including inverse probability of treatment weighting, residual confounding cannot be fully removed in an observational epidemiologic study. For example, in the THIN database, information on severity for most diseases was not collected. If the severity of other comorbid conditions in the metformin group differed from that in the sulfonylurea group, this may have led to potential bias. Second, we could only use data and variables that are recorded in the THIN database. Consequently, individuals who received out-of-system prescriptions were at an increased risk for differential misclassification. Third, HbA1c and hematocrit testing are not routine in general practice in the United Kingdom; therefore, serum HbA1c level and hematocrit were not included in the inverse probability of treatment weighting model, although we conducted sensitivity analyses with adjustment for the level of HbA1c or hematocrit to control the potential confounding effect. Fourth, the current study was conducted among patients with type 2 diabetes (a high-risk population of VTE patients); thus, our findings may not be generalizable to other populations whose pathophysiology may modify the role of metformin in VTE risk. Fifth, we could not determine from these analyses whether metformin is associated with a reduced risk of VTE outcomes or sulfonylureas are associated with an increased risk.
Because VTE is a leading cause of disability and death, efforts to elucidate its mechanisms and improve its prevention are vitally important, especially among persons with type 2 diabetes and other high-risk populations. Our study found that metformin use was associated with a lower risk of VTE, suggesting a role in vascular protection. If these findings can be confirmed by other investigators, it would provide a potentially preventive measure in controlling for this common but serious disease in the high-risk population. In addition, to date, the cause of one-third to one-half of VTE cases is still unknown. A better understanding of the association between metformin and VTE would raise hypotheses on the pathological mechanisms of VTE. Furthermore, previous studies were mainly conducted in animals to illustrate the potential role of metformin in modulating the molecular pathways involved in thrombosis; thus, studies conducted among human beings may be needed. In this population-based cohort study, we generally found a potentially antithrombotic impact of metformin. If our findings are confirmed by other observational studies, they may provide empirical evidence for conducting a clinical trial of metformin use in a population with a high risk of VTE, including patients with type 2 diabetes. Additionally, considering that sulfonylureas are mostly prescribed as a clinical alternative for persons with contraindications to metformin, and considering that VTE is a leading cause of morbidity and mortality around the world, precautions should be taken when prescribing sulfonylureas to patients with type 2 diabetes.
This study provides empirical evidence that metformin use is associated with a lower risk of VTE, as well as PE and DVT separately, as compared with sulfonylurea use in patients with type 2 diabetes. Physicians and patients should be aware of the potential risk of VTE when initiating sulfonylurea use for the treatment of type 2 diabetes.
CI, confidence interval; CVD, cardiovascular disease; DVT, deep vein thrombosis; HbA1c, hemoglobin A1c; HR, hazard ratio; PE, pulmonary embolism; SD, standard deviation; THIN, The Health Improvement Network; VTE, venous thromboembolism.
Data from this study are available for purchase from The Health Improvement Network (THIN) (email@example.com).
This study was approved by the THIN Scientific Review Committee.
THIN is a registered trademark of Cegedim SA (Boulogne-Billancourt, France) in the United Kingdom and other countries. Reference made to the THIN database is intended to be descriptive of the data asset licensed by IQVIA (IQVIA Solutions UK Ltd., London, United Kingdom). This work used deidentified data provided by patients as a part of their routine primary care.
The views expressed in this article are solely those of the authors.
Am J Epidemiol. 2022;191(5):856-866. © 2022 Oxford University Press