A total of 896,422 mirabegron episodes and 7,040,463 antimuscarinic episodes were identified prior to applying the inclusion and exclusion criteria (Fig. 1). Of these, 178,150 mirabegron and 691,548 antimuscarinic episodes were eligible for PS matching. A total of 152,026 mirabegron and 152,026 antimuscarinic episodes were subsequently matched one-to-one; the reasons for exclusion of the episodes are shown in the ESM. Furthermore, no episodes of oxybutynin use were identified in Denmark.
The median duration of current exposure for the matched treatment episodes ranged from 67 days (Humana database) to 165 days (Danish data) for mirabegron, and from 45 days (CPRD-linked) to 114 days (Danish data) for antimuscarinics. Median length of episode follow-up for the matched treatment episodes ranged from 173 days (ORD) to 564 days (Danish data) for mirabegron, and from 183 days (ORD) to 401 days (Danish data) for antimuscarinics. The median number of prescriptions per episode of current exposure ranged from one (Swedish data and Humana) to two (Danish data, CPRD-linked, CPRD-unlinked and ORD) for mirabegron, and also from one (Danish data, Swedish data, CPRD-linked, CPRD-unlinked and Humana) to two (ORD) for antimuscarinics.
Within each data source, the majority of the 268,144 total patients contributed only one matched episode, ranging from 79.6% of patients from the Danish register to 92.7% of patients from the Swedish register. Of the 31,231 total patients who contributed two or more matched episodes, approximately 60% contributed both mirabegron and antimuscarinic episodes (from 44.8% of patients from the Swedish register to 69.5% of patients from the CPRD-linked database).
Strong balance was achieved for each data source across a range of measured baseline covariates (Table 1). Owing to sex and age group category matching, identical data were generated for these parameters in the mirabegron and antimuscarinic cohorts. Overall, 72.6% (220,798) and 63.1% (192,006) of the episodes were from patients who were ≥ 65 years old and female, respectively. Although differences in CV disease risk factors were observed across data sources, the matched mirabegron and antimuscarinic episodes within each data source were well balanced with respect to these risk factors.
CV and Mortality Outcomes
The results below are from pooled analyses unless otherwise stated. As the findings were usually similar with the fixed-effects and the random-effects models, the results shown are primarily from the fixed-effects model. When substantial heterogeneity was observed across data sources, results from both models are included.
The incidence rates per 1000 person-years during current mirabegron exposure were 21.99 (95% CI 20.97–23.05) for MACE, 5.12 (95% CI 4.63–5.64) for AMI, 12.78 (95% CI 12.00–13.59) for stroke, 5.67 (95% CI 5.17–6.22) for CV mortality and 34.95 (95% CI 33.66–36.28) for all-cause mortality (Table 2). During current antimuscarinic exposure, the incidence rates were 22.71 (95% CI 21.59–23.86) for MACE, 5.32 (95% CI 4.80–5.89) for AMI, 12.05 (95% CI 11.25–12.90) for stroke, 7.09 (95% CI 6.48–7.74) for CV mortality and 41.10 (95% CI 39.60–42.65) for all-cause mortality.
The incidence rates of MACE, AMI and stroke were similar among current users of mirabegron relative to current users of antimuscarinics, with HRs close to 1 and the 95% CIs encompassing 1 (MACE: HR 0.94, 95% CI 0.88–1.01; AMI: HR 0.94, 95% CI 0.81–1.08; stroke: HR 1.00, 95% CI 0.91–1.10). The incidence rates of CV mortality and all-cause mortality were no higher with mirabegron compared with antimuscarinics, with HRs of 0.83 (95% CI 0.73–0.95) and 0.80 (95% CI 0.76–0.84), respectively. The HRs for each outcome in the individual study populations (non-pooled data) were generally similar to those in the combined population (Fig. 2 and ESM).
In the analysis according to age category, the HRs for MACE, AMI and stroke ranged from 0.74 to 1.04 (< 65 years) and from 0.94 to 0.98 (≥ 65 years), and the corresponding 95% CIs all encompassed 1 (Table 3). Among those aged < 65 years, the HR for all-cause mortality was 0.76 (95% CI 0.62–0.93). In contrast to the overall results, the 95% CI was relatively large and encompassed 1 for the HR for CV mortality in this age category (HR 0.92, 95% CI 0.53–1.59). Among those aged ≥ 65 years, the HRs for CV mortality and all-cause mortality displayed the same associations as the overall results.
For the analysis by prior OAB medication use, results showed that the HRs among non-naïve OAB medication users ranged from 0.97 to 1.09 and the 95% CIs encompassed 1 (Table 4). In naïve users, the HRs for MACE, AMI and stroke ranged from 0.89 to 0.96; the 95% CIs for AMI and stroke encompassed 1, whereas the upper limit of the CI for the MACE estimate was slightly less than 1 (HR 0.90, 95% CI 0.83–0.97). The HR for all-cause mortality for naïve users was 0.73 (95% CI 0.69–0.78). Substantial heterogeneity was observed across the data sources in terms of CV mortality for naïve users (I2 = 62%). The HR for CV mortality was 0.79 (95% CI 0.68–0.92) using the fixed-effects model and 0.75 (95% CI 0.56–1.02) using the random-effects model.
Approximately 70% of the mirabegron and antimuscarinic episodes were considered at high risk for CV events (Table 5). When the data were restricted to these episodes, similar findings to the overall results were observed.
For the analysis by sex, the HRs among female patients were similar to the overall results (ESM). The results among male patients were also generally similar to the overall results. However, substantial heterogeneity among male patients was observed across the data sources for both MACE (I2 = 59%) and CV mortality (I2 = 53%). For male patients, the HRs for the fixed-effects model were 0.95 (95% CI 0.86–1.04) for MACE and 0.91 (95% CI 0.76–1.08) for CV mortality. With the random-effects model, the HRs were 0.92 (95% CI 0.76–1.12) for MACE and 0.89 (95% CI 0.65–1.21) for CV mortality.
For the analysis by prior history of AMI or stroke, similar results were observed to the overall findings, with the exception of CV mortality (ESM). For patients with a prior history of AMI or stroke, the 95% CIs for CV mortality encompassed 1 (HR 0.84, 95% CI 0.69–1.03) and for patients without a prior history, the upper confidence limit was equal to 1 (HR 0.84, 95% CI 0.71–1.00). In addition, substantial heterogeneity was observed for CV mortality in patients who had a prior history of AMI or stroke (I2 = 59%) and the HR for the random-effects model was 0.80 (95% CI 0.57–1.14). Substantial heterogeneity was also observed for MACE in patients without a prior history of AMI or stroke (I2 = 53%). The HR for MACE was 0.97 (95% CI 0.88–1.06) using the fixed-effects model and 0.96 (95% CI 0.82–1.12) using the random-effects model.
The residual confounding analysis indicated that the presence of bias was < 9% for smoking, alcohol/substance abuse and obesity in each of the data sources (ESM). When the data were censored for all CV outcomes upon occurrence of the first event, the results were nearly identical to those generated in the main analysis (ESM).