Digoxin initiation in advanced heart failure with reduced ejection fraction and atrial fibrillation: a nationwide target trial emulation study

Background

The safety and efficacy of digoxin in patients with heart failure with reduced ejection fraction (HFrEF) and atrial fibrillation (AF) remain highly controversial, with limited randomized evidence. Observational data have frequently associated digoxin with a higher mortality risk. However, these analyses were often biased or incorrectly adjusted.

Aim

We aimed to evaluate the effectiveness of digoxin initiation in a HFrEF and AF population by applying a target trial emulation (TTE) framework to a nationwide registry on the primary outcome of time to first cardiovascular death or HF hospitalization (CVD/HHF), while also assessing all-cause death/HHF (ACD/HHF), as well as ACD, CVD, and HHF. We then compared these results against a classic, biased, prevalent-user analysis.

Methods

Using the Italian Network of Heart Failure (IN-HF) registry, we applied strict clinical equipoise criteria, including only patients with HFrEF, AF, severe symptoms (New York Heart Association class III/IV), and who were not already on digoxin treatment (new/incident user design). Missing data were handled via multiple imputation by chained equation. Super Learner-derived overlap weights were calculated to achieve a pseudo-population under strict equipoise (all absolute Standardized Mean Differences <0.1). Population-average risk ratios (RR) and absolute risk reductions (ARR) were estimated using doubly robust G-computation.

Results

Among 1,480 eligible TTE patients (median age 74 years, 78% males, median EF 30%), 127 (8.6%) initiated digoxin. Over a median follow-up of 1.1 years, incident digoxin use was associated with a significant 33% relative reduction in the primary composite outcome of CVD/HHF (RR 0.67, 95% CI 0.50-0.91; ARR 13%) and ACD/HHF (RR 0.69, 95% CI 0.53-0.90; ARR 12%). This was mostly driven by a reduction in HHF (RR 0.58, 95% CI 0.39-0.85; ARR 12%), with neutral effect on ACD (RR 0.95, 95% CI 0.70-1.30) and CVD (RR 0.95, 95% CI 0.64-1.42) (Figure 1). When subgroups were assessed, only baseline heart rate showed a significant interaction on the primary outcome of CVD/HHF (p-interaction 0.03). Restricted cubic spline analysis showed the greatest protective effect of digoxin in patients with a baseline heart rate between 75 and 110 bpm (Figure 2).

Conversely, when an analysis considering also prevalent digoxin users was performed (n=2,198, median age 72 years, 75% males, median EF 30%), replicating previously done observational study designs, no association was found with HHF, CVD/HHF, ACD/HHF and CVD, while a spurious 23% increased risk of ACD was found (RR 1.23, 95% CI 1.01-1.51).

Conclusions

Digoxin initiation was associated with a lower risk of CVD/HHF and ACD/HHF, mainly driven by HHF, in severely symptomatic patients with HFrEF and AF. Patients with baseline heart rates between 75 and 110 bpm had the greatest benefit from digoxin initiation.