The process of selecting individuals for the study is shown in Web Figure 1. We identified a total of 117,472 initiators of metformin monotherapy and 13,835 initiators of sulfonylurea monotherapy during the study period. The mean duration of follow-up was 3.60 (standard deviation (SD), 3.30) years for the metformin group and 2.63 (SD = 2.78) years for the sulfonylurea group. As is shown in Table 1, individuals in the metformin group were younger (mean age = 62.3 (SD, 11.1) years vs. 67.1 (SD, 11.5) years), had a higher body mass index (32.5 (SD, 6.3) vs. 27.5 (SD, 5.1)) and had a lower prevalence of several comorbid conditions than the sulfonylurea group. The baseline characteristics of included individuals were well balanced between the 2 groups in the weighted sample, with all standardized differences less than 0.1.
The risk of VTE was lower in the metformin group than in the sulfonylurea group (Figure 1A). As Table 2 shows, 555 new cases of VTE occurred among the 117,472 metformin monotherapy initiators (1.3/1,000 person-years) and 75 new cases occurred among the 13,835 sulfonylurea monotherapy initiators (2.1/1,000 person-years) during the entire follow-up period. The multivariable-adjusted rate difference for VTE between the 2 groups was −0.8 (95% CI: −1.3, −0.3) per 1,000 person-years, and the corresponding adjusted HR was 0.65 (95% CI: 0.51, 0.84). Since the proportional hazards assumption was violated (P < 0.05), follow-up time was divided into 2, 4, 6, and 10 years. The adjusted HR was 0.49 (95% CI: 0.35, 0.70) over 2 years, 0.51 (95% CI: 0.38, 0.68) over 4 years, 0.53 (95% CI: 0.40, 0.70) over 6 years, and 0.64 (95% CI: 0.49, 0.82) over 10 years. The risk of PE or DVT in the metformin group was also lower than that in the sulfonylurea group (Figure 1B and 1C). During total follow-up, 266 new cases of PE (0.6/1,000 person-years) occurred in the metformin group and 35 cases (1.0/1,000 person-years) occurred in the sulfonylurea group (adjusted HR = 0.71, 95% CI: 0.50, 1.01) (Table 2). Similar findings were observed for DVT (adjusted HR = 0.64, 95% CI: 0.48, 0.87). Sensitivity analyses (i.e., "intention-to-treat" analysis, further adjustment for HbA1c level or hematocrit, restricting to patients with incident type 2 diabetes, restricting to patients without a history of cancer, varying the prescription refill to a 90-day supply, and adjustment for all potential covariates in the Cox proportional hazards model) did not change the results materially (Table 3).
Cumulative incidence of venous thromboembolism among adults with type 2 diabetes according to metformin or sulfonylurea use, United Kingdom, 2000–2019. A) All venous thromboembolism; B) pulmonary embolism; C) deep vein thrombosis.
All-cause Mortality and Incident CVD
The risk of all-cause mortality was lower in the metformin initiators than in the sulfonylurea initiators. As Web Table 3 shows, 5,164 deaths (11.9/1,000 person-years) occurred in the metformin group and 1,412 deaths (37.7/1,000 person-years) occurred in the sulfonylurea group over the entire follow-up period (adjusted rate difference = –21.0 (95% CI: −22.8, −19.2); adjusted HR = 0.35 (95% CI: 0.34, 0.37)). Similar findings were observed when the risk of incident CVD in the metformin group was compared with that in the sulfonylurea group (adjusted rate difference = –6.5 (95% CI: −7.9, −5.1); adjusted HR = 0.64 (95% CI: 0.59, 0.68)).
Am J Epidemiol. 2022;191(5):856-866. © 2022 Oxford University Press