The Bad Reputation of Digoxin in Atrial Fibrillation-Causality or Bias? Nationwide Nested Case-Control Study

Abstract

Aims: Studies have reported excess risk of mortality associated with digoxin in atrial fibrillation (AF).This study sought to investigate if these findings could be replicated and whether a potential association could be explained by bias.

Methods: Using Danish Nationwide registers, a nested-case control study from 2012 to 2022 was conducted in a cohort of patients with AF. Cases were defined as death of any cause and the exposure was treatment with digoxin compared with beta blockers/verapamil. To investigate bias, additional analyses with negative control outcomes as case definitions-in which we would not expect a plausible association (eg, nursing home admission)-were employed. Associations were reported as hazard ratios (HRs) with 95% confidence intervals (95% CI).

Results: A total of 59,748 cases were identified and matched 1:10 with controls (53% men, median age: 84 [IQR: 77-89]). Digoxin was associated with increased rates of mortality in the entire cohort (HR 1.85, 95% CI 1.78-1.92) as well as subgroups such as patients with heart failure (HR 1.84, 95% CI 1.65-2.06), diabetes (HR 1.85, 95% CI 1.6-2.14), and kidney disease (HR 1.37, 95% CI 1.04-1.8). Significant associations with all negative control outcomes were also found, most notably nursing home admissions (HR 1.79, 95% CI 1.67-1.93).

Conclusion: Digoxin use was associated with increased mortality in AF. However, negative control outcomes were also associated with digoxin use indicating that the described association between digoxin and mortality is likely not causal and being prescribed digoxin is merely a marker of more advanced disease and frailty.

Keywords: Atrial fibrillation; Digoxin; Negative control outcomes; Pharmacoepidemiology; Residual confounding.

Figure 1

Flowchart. Flow chart depicting study population inclusion and sampling of cases and controls.

Figure 2

Temporal trends in digoxin use among patients with atrial fibrillation, by history of heart failure Average use of digoxin was calculated among all patients with atrial fibrillation as marginal means with 95% CI, including competing risk of death, to give a measure of the recurrent event (prescription claim) for relative comparison across groups and calendar year. Every patient included and alive on 1 January in the year in question contributed to the analyses. History of heart failure was defined as any previous hospital visit prior to beginning of calendar year in question.

Figure 3

Hazards of all-cause death associated with concomitant rate control treatment, by subgroups of interest. Forest plot showing the association between digoxin exposure and the adjusted rates of mortality both in the entire study population as well as in predefined subgroups. HRs were calculated from the nested case-control data using conditional logistic regression associating the outcome (all-cause death) with concomitant rate-control treatment (digoxin vs beta blockers/verapamil) using software for conditional logistic regression. Exposure was defined as a claimed prescription within 180 days prior to the date of the outcome or corresponding date among controls. Adjustment variables included in the multivariable model were medical history of ischemic heart disease, heart failure, chronic obstructive pulmonary disease, diabetes mellitus, hypertension, chronic kidney disease, cancer, and thyroid disease; age; and educational level. HR = hazard ratio; CI = confidence interval.

Figure 4

Hazards of all-cause death and negative control outcomes associated with concomitant rate control treatment. Forest plot showing the association between digoxin exposure and the adjusted rates of mortality and negative control outcomes (ie, dehydration/acute kidney injury, pneumonia, septicemia, and nursing home admissions). Statistical methods as in Figure 3. HR = hazard ratio; CI = confidence interval.