Baseline characteristics

A total of 14,201 patients [median age: 71.0 (IQR 64.0–77.0) years; 46.2% female] who were prescribed with beta-blockers and/or digoxin were included in the study. Supplementary Fig. 1 shows how the study cohort was assembled from the GLORIA–AF database. Table 1 shows the baseline characteristics of the study cohort and compared the characteristics of the different groups of beta-blockers and/or digoxin prescription.

Table 1 Baseline characteristics and the comparison among different groups of beta-blockers and/or digoxin users

History of congestive heart failure (53.6% vs 22.4%), Heart rate (median: 84 vs 75) and use of diuretics (67.2% vs 40.0%) were higher in the group with combination therapy, compared to the group with beta-blockers prescription alone. After PS matching, all baseline covariates were well-balanced between beta-blockers and combination therapy with a standardised mean difference less than 0.1 as shown in the Love plot (Supplementary Fig. 2). In the PS weighted cohort, a similar balance was also achieved among the baseline covariates (Supplementary Fig. 3).

Incidence of MACE and all-cause mortality

After a median follow-up period of 3.0 (IQR 2.4–3.1) years, 864 MACEs and 988 all-cause deaths were recorded. Among 14,201 patients, 864 MACE events occurred during 38,491 person-years of follow-up which corresponded to an IR of 22.4 (95%CI 21.0–24.0) per 1000 person-years, while the IR of all-cause death was 25.4 (95%CI 23.8–27.0) per 1000 person-years.

In the original cohort, 723 (5.8%), 35 (6.6%) and 106 (8.7%) cases of MACE developed among the group with beta-blocker prescription alone, digoxin prescription alone and combination therapy, respectively. The incidence of MACE was significantly higher (p < 0.001) in the combination therapy group, compared to the beta-blockers only group (Table 2). All-cause mortality was 812 (6.5%), 53 (10.0%), 123 (10.1%) in the beta-blocker group, digoxin group, and combination therapy group, respectively. The incidence of all-cause mortality was significantly higher in both the digoxin group and the combination therapy group, compared to the beta-blockers only group (Table 2).

Table 2 Incidence rate and incidence rate ratio among different treatment groupsNNTharm with digoxin and beta-blockers

In the original cohort, NNTharm to get one additional MACE in the digoxin alone group and the combination therapy group was 125 (95%CI: reached infinity and became not significant) and 35 (95% CI 21.4–71.5), respectively. NNTharm to get one additional all-cause death was 29 (95% CI 15.8–87.2) in the digoxin alone group and 28 (95% CI 18.3–50.8) in the combination therapy group.

In the PS matched cohort, NNTharm to get one additional MACE in the digoxin alone group and the combination therapy group was 266 (95% CI: reached infinity and became not significant) and 45 (95% CI 22.3–878.8), respectively. NNTharm to get one additional all-cause death was 255 (95% CI not significant) in the digoxin alone group and 38 (95% CI 20.4–261.5) in the combination therapy group.

The risk of MACE with digoxin and beta-blockers

After multivariate adjustment with Cox regression, the risk of MACE (HR: 1.35, 95% CI 1.09–1.68) was significantly higher in the combination therapy group, compared to the beta-blockers only group. However, the risks were not significantly different between the beta-blocker only group and the digoxin only group in the original cohort (Fig. 1).

Fig. 1

Forest plot illustrating the multivariate Cox regression for predicting MACE (top) and all-cause mortality (bottom) in the whole cohort. BMI Body Mass Index, HTN Hypertension, DM Diabetes Mellitus, CHF Congestive Heart Failure, CAD Coronary Artery Disease, AKF Abnormal Kidney Function, COPD Chronic Obstructive Pulmonary Disease, OAC Oral Anti-Coagulant, ACE Angiotensin-converting-enzyme, AF Atrial fibrillation

The elevated risk in the combination therapy remained significant in both the PS matched cohort (HR 1.43, 95% CI 1.06–1.92) and the PS weighted cohort (HR 1.57, 95% CI 1.17–2.11) (Table 3). The survival distribution significantly differed among different drug groups as shown by the Kaplan–Meier survival curves and the log-rank test (p < 0.001) with the combination therapy group having the lowest MACE-free survival rate (Fig. 2).

Table 3 Multivariate Cox regression for predicting risk of MACE and all-cause mortality in the propensity score matched and propensity score weighted cohortFig. 2

Kaplan-Meier Survival Curves for illustrating the MACE-free survival distribution (top) and the survival distribution (bottom) among different groups of drug users

The risk of all-cause death with digoxin and beta-blockers

In the original cohort, the risk of all-cause death (HR 1.28, 95% CI 1.04–1.57) was significantly higher in the combination therapy group, compared to beta-blockers only group (Fig. 1). The observed increase in the risk of all-cause death remained significant in the PS matched cohort (HR 1.42, 95% CI 1.08–1.86) (Table 3). The risk of all-cause death in the combination therapy group (HR 1.33, 95% CI 1.01–1.75) were still significantly higher, compared to the beta-blockers only group in the PS weighted cohort. However, the risks were not significantly different between the beta-blockers only group and the digoxin only group after multivariate adjustment.

Kaplan–Meier survival curves and log-rank test demonstrated the significant differences in the survival rates among different groups of drug prescription (p < 0.001). The survival rate was the lowest in the combination therapy group (Fig. 2).

Post-hoc interaction analysis

To further investigate if the increased risk of death associated with digoxin and/or beta-blockers was influenced by HF, abnormal kidney function, and angiotensin-converting enzyme inhibitors (ACEIs) or angiotensin II receptor blockers (ARBs), a post-hoc interaction analysis was performed to evaluate if the risk was modified by different sub-groups of patients. Supplementary Fig. 4 summarises the analysis and the result indicated that the risk was not different between patients with or without HF, patients with or without abnormal kidney function, and patients with or without ACEIs or ARBs (p > 0.05).