Care Patterns and Outcomes in Atrial Fibrillation Patients With and Without Diabetes: ORBIT-AF RegistryFree Access

Data source and study population

Our study cohort was formed from the ORBIT-AF registry, which is an observational, prospective, quality improvement program. The rationale and design of ORBIT-AF have been previously described in detail (14). In brief, the registry enrolled adults older than 18 years with electrocardiographic evidence of AF. Patients were excluded if they were diagnosed as having AF secondary to an easily reversible condition or if they had a life expectancy <6 months. Patients were enrolled from heterogeneous practices across the United States, including internal medicine, neurology, cardiology, and electrophysiology clinics. Data were collected by various health care providers, including internists, primary care physicians, neurologists, cardiologists, and electrophysiologists. The patient’s medical record was entered into an online case report form, and patients were to be followed every 6 months for 2 to 3 years. All participants gave informed consent, and all study sites had approval from an institutional review board for inclusion in the ORBIT-AF registry.

We identified 10,137 patients with AF from a total of 176 U.S. clinic practices enrolled between June 29, 2010, and August 9, 2011. We excluded those patients without information on follow-up (n = 388), leaving a final study sample of 9,749 patients.

Diabetes mellitus was defined by previous medical history or new clinical diagnosis during the enrollment visit.

Outcomes

Patients were followed at 6-month intervals for up to 3 years. Outcomes assessed at follow-up were all-cause mortality, CV death, non-CV death, hospitalization (all-cause, CV-related, bleeding-related, and non-CV and nonbleeding-related), stroke or non-central nervous system systemic embolism or transient ischemic attack (TIA), incident new-onset HF, AF progression, bleeding events, AF symptoms, health status, and use of targeted therapies. AF progression was defined as either: 1) progression from paroxysmal AF at baseline to either persistent or permanent AF reported at any subsequent follow-up visit; or 2) progression from persistent AF at baseline to permanent AF reported at any subsequent follow-up visit (15,16). Paroxysmal AF was defined as recurrent AF episodes that terminate spontaneously within 7 days; persistent AF as recurrent AF that is sustained for more than 7 days; and permanent AF as AF in which the presence of the AF is accepted. The assessment of the use of targeted therapies included the management strategy (rate control, rhythm control), medication use (warfarin, non-vitamin K oral anticoagulants [OACs], aspirin, clopidogrel, beta-blockers, calcium-channel blockers, digoxin, amiodarone, sotalol, dofetilide, propafenone, flecainide), and procedures (catheter ablation for AF or atrial flutter, atrioventricular node ablation cardioversion).

Statistical analysis

We compared patient (demographic and clinical) and hospital characteristics, as well as proportional use of different AF management strategies between patients with and without diabetes. Categorical variables are presented as count and proportion, and differences were tested using the Pearson chi-square test. Continuous variables are presented as median (interquartile range [IQR]), and differences between groups were tested using the Wilcoxon rank sum test.

The event rate for each outcome was estimated both overall and by diabetes status, with rates presented as the number of events per 100 patient-years of follow-up. We used Cox proportional hazards survival models to test the association between diabetes and each clinical outcome. For AF progression, we estimated the frequency and percentage of AF progression, and a logistic regression model was used to estimate the odds ratio (OR) for the association between diabetes and AF progression. All models used a robust variance estimate to account for correlation within sites. Covariates for the multivariable modeling were chosen based on their clinical relevance or known association with diabetes and outcome(s). A common set of variables was used in models for adjustment and is listed in the Online Appendix. Adjusted associations for outcomes are displayed as hazard ratio (HR) or OR with corresponding 95% confidence interval (CI). We assessed the influence of age, race (white vs. nonwhite), use of anticoagulants, renal function (as defined by estimated glomerular filtration rate), type of AF (paroxysmal vs. nonparoxysmal), history of ablation procedure, and pulse pressure on the relationship between diabetes and outcomes, by testing the interaction of each of these factors with diabetes. In models, all continuous variables were tested for linearity with each outcome, and any nonlinear associations are accounted for using linear splines. Missing covariate data in the regression analysis were handled by multiple imputation using Markov Chain Monte Carlo and regression methods. Final estimates and associated standard errors reflect the combined analysis over 5 imputed datasets.

Analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, North Carolina). All p values were based on 2-sided tests and were considered statistically significant at p < 0.05.

Symptoms, functional status, and quality of life

In the ORBIT-AF cohort, patients with diabetes had greater functional impairment as reflected by higher European Heart Rhythm Association scores (Table 1). Across the AF symptom checklist, patients with diabetes less often reported palpitations (29.2% vs. 33.6%; p < 0.001) or syncope (3.3% vs. 4.9%; p < 0.001), but they reported significantly more dyspnea upon exertion (29.7% vs. 26.6%; p = 0.002) or at rest (11.1% vs. 9.8%; p = 0.06), exercise intolerance (11.9% vs. 9.2%; p < 0.001), and fatigue (27.9% vs. 25.6%; p = 0.02). There were no differences between the diabetes and nondiabetes groups in terms of lightheadedness or dizziness and chest tightness or discomfort.

Regarding health-related quality of life (QOL) (Table 1), patients with diabetes had a slightly lower overall median Atrial Fibrillation Effect on Quality-of-Life scores compared to those without diabetes (80.6; IQR: 62.5 to 92.6 vs. 82.4; IQR: 67.6 to 93.6; p = 0.03). This difference was mainly driven by the daily activities domain score, as the symptoms and treatment concern domain scores were similar. These differences persisted at 12- and 24-month follow-up.

AF management and outcomes

Patients with diabetes had persistent and permanent AF or a combination of persistent or permanent AF (47.9% vs. 43.6%) more frequently than did those without diabetes (p < 0.001) (Table 1), but they were less likely to have undergone cardioversion or ablation (Table 2). The use of rate-control strategies was higher among people with diabetes compared to those without diabetes contrary to use of rhythm-control therapies, which was lower (Table 2). Among rate-control agents, the use of beta-blockers (67.8% vs. 62.8%; p < 0.001), calcium-channel blockers (31.8% vs. 29.8%; p = 0.05), and digoxin (29.0% vs. 21.3%; p < 0.001) was more prevalent among people with diabetes. For rhythm-control therapies, flecainide was less commonly used among people with diabetes (1.0% vs. 3.7%; p < 0.001). The overall use of anticoagulants (warfarin or dabigatran) was significantly greater among people with diabetes. The use of warfarin only was higher among those with diabetes compared to those without diabetes (74.3% vs. 70.4%; p < 0.001), but use of dabigatran was similar in the 2 groups. Aspirin was more commonly used among people with diabetes (46.5% vs. 43.4%; p = 0.006).

The incidence rates for all outcomes during follow-up are given in Table 3. Over the 2-year period, mortality rates were higher among people with diabetes compared to those without diabetes. The rates of CV death, non-CV death, and sudden cardiac death (SCD) also were higher among patients with diabetes than in those without diabetes (Table 3, Central Illustration, Figure 1, and Online Figures 1 to 12).

There was a significant interaction between diabetes and age for all-cause death (p = 0.025) and CV mortality (p = 0.002) outcomes. After multivariable adjustment, diabetes was associated with a higher risk of mortality, both among those age <70 years (adjusted hazard ratio [aHR]: 1.63; 95% CI: 1.04 to 2.56) and among those age ≥70 years (aHR: 1.25; 95% CI: 1.09 to 1.44) (Table 3). Diabetes was also related to a higher risk of CV death, more so among those <70 years than in those ≥70 years, as well as an increased risk of non-CV death or SCD. Diabetes was similarly associated with a higher risk of all-cause (aHR: 1.15; 95% CI: 1.09 to 1.22), CV (aHR: 1.13; 95% CI: 1.05 to 1.22) and non-CV and nonbleeding-related hospitalizations (aHR: 1.19; 95% CI: 1.10 to 1.30). However, diabetes was not associated with a higher risk of thromboembolic events (including stroke, TIA, and non-central nervous system embolism) or hospitalization related to bleeding (Table 3). Also, there was no association between diabetes and incident new-onset HF (aHR: 1.08; 95% CI: 0.80 to 1.47) or AF progression (adjusted OR: 0.96; 95% CI: 0.85 to 1.08). OAC use modified the association of diabetes and all-cause hospitalization (p for interaction = 0.018). Diabetes was more strongly related to all-cause hospitalization (aHR: 1.21; 95% CI: 1.12 to 1.29) among those using an OAC than among those not using OAC (aHR: 0.98; 95% CI: 0.85 to 1.13). Among patients with diabetes, there were no significant differences in outcomes of AF between patients using both OAC and antiplatelet therapy and those taking an OAC alone (Online Table 1).

There was a significant interaction of diabetes and pulse pressure (≤60 mm Hg vs. >60 mm Hg) for the following outcomes (Online Table 2): new-onset HF (p < 0.001), CV hospitalization (p = 0.015), bleeding hospitalization (p = 0.029), and non-CV or nonbleeding hospitalization (p = 0.002).

Study limitations

The strengths of our study include a large nationwide cohort, a standardized methodology for data collection, and the examination of AF subtypes, regular follow-up, detailed information on comorbid illness, and several clinically relevant outcomes. However, there were limitations to our study. First, reliance on clinical practice data for diabetes ascertainment might have missed patients without a pre-hospitalization history of diabetes and who were not screened during their hospitalization for diabetes, especially as up to one-fourth of all persons with diabetes in the United States are undiagnosed (33). Diabetes might have been underreported by using clinically diagnosed and documented diabetes rather than using strict laboratory criteria. Likewise, we did not ascertain cases of diabetes that developed during study follow-up. The consequence of underestimation of diabetes frequency would be to bias the influence of diabetes on the outcomes of AF toward the null.

Second, the main endpoint for this analysis was all-cause mortality; the causes of death are not known because we did not have access to this information (especially the CV causes). However, differences in overall and CV-related hospitalization rates pointed to a potentially high contribution of CV diseases to deaths. Third, we did not have information on the degree of glycemic control (as represented by HbA_1c) or duration of diabetes, which might influence outcomes. The levels of HbA_1c are associated with stroke risk among those with AF and improve the predictive accuracy for stroke in diabetic patients with AF (24). The duration of diabetes might also be important for assessing the risk of ischemic stroke among patients who have diabetes and AF (10,11). We did not have information on the medications used for glycemic control (insulin or noninsulin-based therapies), which may influence AF outcomes, especially given that some medication classes (the thiazolidinediones) may be related to better AF outcomes (34).

Fourth, the ORBIT-AF registry relied on voluntary participation from sites and patients. Thus, the results might not be generalizable to all U.S. patients, particularly those who are managed in different clinical settings compared with that of the ORBIT-AF participants. Fifth, the follow-up duration in our study (2 years) might not reveal the longer-terms effects of diabetes. Furthermore, given the average age of our participants (>70 years) and that younger people with diabetes might be at greater risk for developing AF (35), there may have been a selection bias related to the enrollment process. However, it is important to bear in mind that age is a significant risk factor for AF. Finally, we may have had a limited power to detect a difference in some of the outcomes, and, despite appropriate adjustment for confounders in our models, residual confounding may have persisted, potentially affecting our findings.