Author + information
- Anthonius de Boer, MD, PhD∗ ()
- ↵∗Utrecht Institute for Pharmaceutical Sciences, P.O. Box 80082, 3508 TB Utrecht, the Netherlands
In the TREAT-AF (Retrospective Evaluation and Assessment of Therapies in AF) study (1), the effect of digoxin on overall mortality in patients with incident atrial fibrillation (AF) was studied. In this observational study, after adjustment for potential confounders with the Cox proportional hazards model and propensity score analyses, digoxin was associated with an increased risk (21% to 24%) of death. The investigators extensively discussed the potential limitations of their study, but I have 2 questions about the design choices.
First, the investigators stated that patients were placed in the digoxin group versus the reference group on the basis of use of digoxin within the first 90 days after the diagnosis of AF. Digoxin is the first choice for therapy in patients with AF complicated by heart failure and the second choice in patients whose first choices for treatment of AF, beta-blockers and calcium channel antagonists, are not effective enough. When only the first 90 days are used for exposure classification, there may be a substantial misclassification of digoxin in both the digoxin group and the reference group. The fact that the medication possession ratio of digoxin was only calculated for the digoxin group does not take into account such misclassification.
Second, the investigators stated that they adjusted for the medication possession ratio in the multivariate Cox analysis. I do not understand this: how can you adjust for a variable that is zero in all patients in the reference group and a certain number between zero and one in the digoxin group? I believe the appropriate analysis would be to stratify the digoxin group with different medication possession ratios and compare these with the reference group.
Please note: Dr. de Boer is a member of 2 steering committees of projects that are financed by GlaxoSmithKline and 2 umbrella organizations: the European Federation of Pharmaceutical Industries and Associations and the Association of the European Self-Medication Industry. These relationships are not relevant to the contents of this letter.
- American College of Cardiology Foundation