Methods

We analyzed a public-use copy of the Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) data obtained from the National Heart, Lung, and Blood Institute. The design and primary results of AFFIRM have been previously published. Briefly, AFFIRM was a multicenter randomized clinical trial for rate versus rhythm control treatment strategies for AF conducted in 213 centers in the United States and Canada. Patients with recurrent AF without contraindication to anticoagulant therapy (as determined by their physicians) and with high risk for stroke were recruited. Because age was considered a risk factor for stroke in AF, those ≥65 years of age could be enrolled regardless of other risk factors. However, to be eligible for enrollment, those <65 years were required to have ≥1 other risk factor for stroke, which included previous stroke or transient ischemic attacks, hypertension, heart failure, diabetes mellitus, increased left atrial enlargement, and left ventricular systolic dysfunction. AFFIRM participants had a mean age of 65 years (range 49 to 80) and 76% of patients (3,091 of 4,060) were ≥65 years of age.

The present analysis was restricted to 2,248 patients (55% of 4,060) who were 70 to 80 years of age. We chose a cutoff of 70 years because of the high prevalence of AF in this age group. Of the 2,248 patients, 1,901 (85%) were receiving warfarin, with a goal International Normalized Ratio from 2.0 to 3.0. Patients were followed up for up to 6 years (mean follow-up time 3.4 years) with interval follow-up visits every 4 months. All outcomes were blindly adjudicated by the AFFIRM events committee. The primary outcome for the present analysis was all-cause mortality. Secondary outcomes included all-cause hospitalization, ischemic stroke, and major bleeding defined as bleeding requiring transfusion and/or surgery and/or permanent cessation of warfarin.

Considering the significant imbalances in baseline characteristics between the 2 groups ( Table 1 ), we used propensity scores to assemble a matched cohort. Propensity scores for warfarin use were estimated for each of the 2,248 patients using a nonparsimonious multivariable logistic regression model. We were able to match 227 of the 347 patients not receiving warfarin to 616 patients receiving warfarin using a greedy algorithm to match those on warfarin to sets of 1, 2, or 3 patients not receiving warfarin with similar propensity scores. The matched cohort of 843 patients was well balanced between warfarin recipients and nonrecipients on the 45 baseline characteristics used in the propensity score model. Absolute standardized differences were estimated to evaluate the imbalance before matching and balance after matching and are presented in a Love plot ( Figure 1 ) . Absolute standardized differences directly quantify biases in means (or proportions) of covariates across groups and are expressed as percent pooled SDs. An absolute standardized difference of 0% indicates no residual bias and differences <10% are considered inconsequential.

Absolute standardized differences of 45 baseline characteristics between patients receiving and not receiving warfarin before and after propensity score matching. ^* Symptoms reported during atrial fibrillation in the previous 6 months; ^** hospitalization for qualifying episodes of atrial fibrillation. ACE = angiotensin-converting enzyme; HTN = hypertension; NYHA = New York Heart Association; PND = paroxysmal nocturnal dyspnea.

For descriptive analyses, we used Pearson chi-square and Wilcoxon rank-sum tests for prematch comparisons and paired-sample t tests for postmatch comparisons of baseline characteristics of patients with and without warfarin use, as appropriate. We used Kaplan–Meier plots and Cox regression analyses to determine associations between warfarin use and outcomes during follow-up. We conducted formal sensitivity analyses to quantify the degree of hidden bias that would need to be present to invalidate our conclusions based on a significant association between use of warfarin and all-cause mortality in matched patients. Subgroup analyses were conducted to determine the homogeneity of association between use of warfarin and all-cause mortality. To assess the generalizability of the findings of the present study based on trial-eligible AFFIRM participants 70 to 80 years old with AF to community-dwelling patients with AF in that age group, we compared baseline characteristics and outcomes of participants included in our study to patients with AF 70 to 80 years in the Cardiovascular Health Study (CHS). All statistical tests were 2-tailed with a p value <0.05 considered statistically significant and all data analyses were performed using SPSS 18 for Windows (SPSS, Inc., Chicago, Illinois).

Results

Patients (n = 843) had a mean ± SD age of 76 ± 3 years, 45% were women, and 7% were nonwhite. Before matching patients receiving warfarin were more likely to have heart failure and valvular heart disease, have higher CHADS2 scores, but similar CHA2DS2VASc scores. These and other baseline imbalances were balanced after matching ( Table 1 and Figure 1 ).

All-cause mortality occurred in 18% and 33% of matched patients on and not on warfarin, respectively, during 6 years of follow-up (hazard ratio [HR] when use of warfarin was compared to its nonuse 0.58, 95% confidence interval [CI] 0.43 to 0.77, p <0.001; Table 2 and Figure 2 ) . A hidden covariate that is a near-perfect predictor of mortality would need to increase the odds of warfarin use by 48% to explain away this association. The association of warfarin use with mortality in various subgroups of patients is displayed in Figure 3 . Associations of warfarin use with various cause-specific mortalities are presented in Tables 3 and 4 .

Kaplan–Meier plots for (A) all-cause mortality and (B) all-cause hospitalization by warfarin use.

Association of warfarin use with all-cause mortality in subgroups of propensity-matched patients with atrial fibrillation 70 years of age. Abbreviation as in Figure 1 .

Table 3

Associations of warfarin use with other outcomes in 843 propensity-matched patients with atrial fibrillation ≥70 years of age

	Events (%)		Absolute Risk Difference ⁎	HR (95% CI)	p Value
	No Warfarin	Warfarin
	(n = 227)	(n = 616)
Cardiovascular mortality	27 (12%)	56 (9%)	−3%	0.80 (0.51–1.27)	0.346
Cardiac causes	20 (9%)	43 (7%)	−2%	0.84 (0.49–1.43)	0.516
Arrhythmic	12 (5%)	24 (4%)	−1%	0.76 (0.38–1.52)	0.432
Nonarrhythmic	8 (4%)	19 (3%)	−1%	0.96 (0.42–2.20)	0.931
Vascular causes	7 (3%)	13 (2%)	−1%	0.70 (0.28–1.75)	0.442
Noncardiovascular mortality	40 (18%)	48 (8%)	−10%	0.45 (0.30–0.70)	<0.001
Cancer	14 (6%)	24 (4%)	−2%	0.61 (0.31–1.19)	0.609
Pulmonary	8 (4%)	10 (2%)	−2%	0.43 (0.18–1.07)	0.068
Others	17 (8%)	15 (2%)	−6%	0.33 (0.17–0.67)	0.002
All-cause hospitalization	152 (67%)	394 (64%)	−3%	0.93 (0.77–1.12)	0.423
Cardiovascular causes	103 (44%)	252 (41%)	−3%	0.90 (0.72–1.14)	0.386
Noncardiovascular causes	104 (46%)	258 (42%)	−4%	0.90 (0.71–1.13)	0.355
Ischemic stroke	17 (8%)	26 (4%)	−4%	0.57 (0.31–1.04)	0.068
Major bleeding †	22 (10%)	44 (7%)	−3%	0.73 (0.44–1.22)	0.229

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Association of Frontal QRS-T Angle–Age Risk Score on Admission Electrocardiogram With Mortality in Patients Admitted With an Acute Coronary Syndrome
2. Vitamin D Deficiency and Supplementation and Relation to Cardiovascular Health
3. Analysis of Echocardiograms in a Large Heterogeneous Cohort of Patients With Friedreich Ataxia
4. Comparison of Patients With Pulmonary Arterial Hypertension With Versus Without Right-Sided Mechanical Alternans

Stay updated, free articles. Join our Telegram channel

Join