Journal of the American College of Cardiology
The prevalence of atrial fibrillation in incident stroke cases and matched population controls in Rochester, MinnesotaChanges over three decades
Author + information
- Teresa S.M Tsang, MD, FACC*,* (tsang.teresa{at}mayo.edu),
- George W Petty, MD†,
- Marion E Barnes, MSc*,
- W.Michael O’Fallon, PhD‡,
- Kent R Bailey, PhD‡,
- David O Wiebers, MD†,§,
- JoRean D Sicks, MS‡,
- Teresa J.H Christianson, BSc‡,
- James B Seward, MD, FACC* and
- Bernard J Gersh, MB, ChB, DPhil, FACC*
- ↵*Reprint requests and correspondence:
Dr. Teresa S. M. Tsang, Mayo Clinic, 200 First Street SW, Rochester, Minnesota 55905, USA.
Abstract
Objectives We sought evidence of a change in the prevalence of atrial fibrillation (AF) over a 30-year period among residents of Rochester, Minnesota.
Background Atrial fibrillation is increasingly encountered in clinical practice, but there is limited data on secular trends of AF over time.
Methods Within a longitudinal case-control study of ischemic stroke, the prevalence of AF and of selected comorbid conditions among incident stroke cases and age- and gender-matched controls between 1960 and 1989 was determined.
Results The mean age ± standard deviation for the 1,871 stroke cases (45% men) and matched controls was 75 ± 11 years. For cases, age-adjusted estimates of AF prevalence for 1960 to 1969, 1970 to 1979, and 1980 to 1989 were 11%, 13%, and 16%, respectively, for men, and 13%, 16%, and 20% for women. For controls, the rates were 5%, 8%, and 12%, respectively, for men, and 4%, 6%, and 8% for women. Increasing AF prevalence was associated with increasing age (doubling of odds per decade of age in both cases and controls) and calendar time adjusted for age and gender (cases: odds ratio [OR] per 5 years 1.13, 95% confidence interval [CI], 1.05 to 1.22; controls: OR per 5 years 1.24, 95% CI 1.12 to 1.37). The rates of increase with calendar time were significant for cases (p = 0.001) and controls (p < 0.001) and comparable between the genders.
Conclusions The prevalence of AF increased significantly in ischemic stroke patients and their controls from 1960 to 1989 in Rochester, Minnesota, independent of age and gender. The rate of increase did not differ significantly between men and women.
Atrial fibrillation (AF) is a growing epidemic (1). Approximately 2.3 million individuals in the U.S. currently carry this diagnosis (2), and it is projected that 5.6 million Americans will have AF by 2050. Aside from being an independent predictor of mortality (3), AF is also a cause of substantial morbidity, including stroke (4–7), and has a major impact on health care costs (5).
Atrial fibrillation is strongly associated with age (5), and this, in combination with a burgeoning older population, provides the basis for the developing epidemic. The Framingham investigators identified a significant increase in the prevalence of AF in men, but not in women, over the period 1968 to 1989, even after adjusting for age and comorbid conditions (8). To our knowledge, there have been no other published studies on trends of AF over time. We conducted an analysis of data collected in a longitudinal, population-based study of ischemic stroke in the community of Rochester, Minnesota, to determine whether the prevalence of AF changed among patients with incident ischemic stroke and among age- and gender-matched control subjects over a 30-year period (1960 to 1989).
Methods
The study received review and approval of ethics by the Institutional Review Board of the Mayo Foundation. The community of Rochester, Minnesota, is well suited to population-based studies because of a number of unique features. Geographically, the community is relatively isolated from other urban centers, and medical care is delivered by only a few health care providers. These include the Mayo Clinic and its two affiliated hospitals, which provide primary through tertiary care services in all specialties and subspecialties, the Olmsted Medical Center and its affiliated hospital, and a small number of private practitioners. The majority of care is provided by the Mayo Clinic, which, for more than 80 years, has maintained a unified medical record containing details of all inpatient and outpatient encounters, laboratory reports, and patient-related correspondence for each registered patient. Within each medical record, diagnoses made during office visits, clinic consultations, emergency room visits, hospital admissions, nursing home care, and autopsy examinations, as well as surgical procedures, are listed on a master sheet and subsequently coded. Coded diagnoses are then transferred to a central diagnostic index. This diagnostic index allows all patients with a diagnosis of interest to be readily identified. The resources of the Rochester Epidemiology Project (9)also include an established medical linkage system, which allows access to medical information on patients seen in practices in the surrounding communities. Thus, all patients in the community with a particular disease or diagnosis can be readily identified (10).
With the use of the resources of the Rochester Epidemiology Project and those of the Mayo Stroke Center (11), a case-control study of the incidence of ischemic stroke in the Rochester community was previously conducted (11,12). The definitions of ischemic stroke and methods for selecting control subjects have been reported (11,13). Briefly, a pool of potential age- (within ±2 years) and gender-matched control subjects from the general Rochester population were identified using the medical records of the Rochester Epidemiology Project (9,11). Potential control subjects included persons without any documented history of ischemic stroke and with a health care contact within ±1 year of the diagnosis of the relevant stroke case, as demonstrated by a recorded blood pressure measurement. The date of blood pressure assessment became the reference date for the control subject. Control subjects also had to have been residents of Rochester for at least one year before the reference date. The first control subject was defined as the person from the pool of potential controls whose registration number was the closest to that of the case, thus ensuring more comparable duration of medical records. If the first eligible control subject failed any validation checks, another potential control subject with the next closest registration number would be selected.
Within the framework of the case-control study of incident ischemic stroke in Rochester, Minnesota (11,12), we determined the prevalence of a history of AF (hereafter referred to as “prevalence of AF”) and selected comorbid conditions as of the date of stroke diagnosis for case patients and reference date for control subjects. Clinical data for each study subject were abstracted from the entire medical record by a trained nurse abstractor supervised by a physician. References to possible AF in the clinical record had to be substantiated by an electrocardiogram (ECG) documenting at least one episode of the arrhythmia in order for a diagnosis of AF to be coded. For purposes of this analysis, no distinction was made between lone, intermittent/paroxysmal, chronic/persistent, or valvular/nonvalvular AF. Atrial flutter alone, without any intervening episodes of AF, was not included in this definition. Cases in which AF was first detected at the time of stroke diagnosis were not included in prevalence calculations, as it would not be possible to determine whether the arrhythmia preceded or followed the onset of stroke. For comparability, AF initially detected on the reference date of control subjects was not used in prevalence calculations.
Standard definitions of comorbid conditions (see July 2 issue online for Appendixat www.cardiosource.com/jacc.html) were applied over the entire period, as previously reported (12). Comorbid conditions were coded as present or absent and the date of first mention recorded, allowing prevalence calculations for each of the three decades (1960 to 1969, 1970 to 1979, 1980 to 1989) to be performed. However, the precise time order of these conditions was unavailable, precluding analysis of temporal relationships between comorbid conditions and AF. The use of ECG and echocardiography within ±30 days of stroke/reference date was available from the original case-control study and provided a crude estimate of the influence of rates of utilization of these diagnostic modalities on ascertainment of AF and comorbid conditions for the study period.
Statistical analyses
Logistic regression models (14)were used to assess the effects of age, gender, and calendar time on the odds of a history of AF and of each comorbid condition separately among case patients and control subjects. Age (both linear and quadratic), gender, and their interactions were also considered. Then, calendar year, both as a continuous and as a discrete (by decade) variable, was assessed, adjusting for best-fit age and gender models. Interactions of calendar year with age and gender were considered. Only those significant interactions with calendar year that affected the qualitative interpretation were reported. Once the best logistic model was determined for each condition, the proportions of case patients and control subjects with each condition (adjusted for the demographic variables in the model) were estimated by use of the logit function (14). In order to summarize the age-adjusted results, prevalence estimates derived from the multivariate models are reported for age 75 years, because this was the mean age for the cohort (median age 76 years). Model-based prevalence estimates of AF for other ages were also provided in a table for men and women with and without ischemic stroke. For all model building, two-sided p values were used and a variable was considered statistically significant only if p < 0.05.
Results
Prevalence of AF: effects of age, gender, and calendar time
A total of 3,742 residents of Rochester, mean age 75 ± 11 years (median age 76 years, range 45 to 105 years), were included in the study. The age and gender distribution, stratified by ischemic stroke status, is shown in Table 1. Of 1,871 Rochester residents (838 men, 1,033 women) who had a confirmed ischemic stroke between 1960 and 1989, 327 (105 men, 222 women), or 17%, had a history of AF before the date of stroke diagnosis (Table 2). Among the 1,871 age- and gender-matched control subjects, 176 (9.4%) had a history of AF (72 men, 104 women) prior to the reference date. The observed prevalence of AF among case patients and control subjects, stratified by age, gender, and calendar decade is outlined in Figure 1.
Prevalence of atrial fibrillation in 1,871 patients with ischemic stroke and their age- and gender-matched controls, stratified by age, gender, and calendar decade.
Age and Gender Distribution for Ischemic Stroke Cases and Control Subjects
Observed Prevalence (%) of Atrial Fibrillation Among Rochester Residents, Stratified by Gender, Age, Presence or Absence of Ischemic Stroke, and Decade
In the stroke patients, there was a significant increase in the prevalence of AF with increasing age (odds ratio [OR] per 10 years of age 1.81, 95% confidence interval [CI] 1.58 to 2.07; p < 0.001), but there was no significant difference between the genders. After adjustment for age and gender, there was a significant calendar-time effect: the OR per five years was 1.13 (95% CI 1.05 to 1.22, p = 0.001). This change in the prevalence of AF over calendar time did not vary according to gender (p = 0.702) or age (p = 0.349) (Fig. 2). In this multivariate model, the age-adjusted estimates of prevalence for the three decades (1960 to 1969, 1970 to 1979, 1980 to 1989) were 11%, 13%, and 16%, respectively, for men and 13%, 16%, and 20% percent for women.
Trends for age-adjusted prevalence of atrial fibrillation in 1,871 patients with ischemic stroke and their age- and gender-matched controls, stratified by gender. *Significant (p = 0.001) time trend among case patients; insignificant (p = 0.702) difference between men and women. †Significant (p < 0.001) time trend among controls; insignificant (p = 0.098) difference between men and women.
In the control subjects, there was also a significant increase in the prevalence of AF with advancing age (OR per 10 years 2.22, 95% CI 1.84 to 2.68), and male gender was associated with a greater likelihood of AF (OR 1.50, 95% CI 1.06 to 2.11). After adjustment for age and gender, there was also a significant calendar-time effect: the OR per five years was 1.24 (95% CI 1.12 to 1.37, p < 0.001). This change in the prevalence of AF over calendar time also did not vary according to gender (p = 0.098) or age (p = 0.617) (Fig. 2). In this multivariate model, the age-adjusted estimates of prevalence for the three decades (1960 to 1969, 1970 to 1979, 1980 to 1989) were 5%, 8%, and 12%, respectively, for men and 4%, 6%, and 8% for women.
Over the 30-year period, age-adjusted prevalence of AF increased significantly (p < 0.001) for the total cohort of 3,742 individuals (Fig. 2). The rates of increase were significant for both case patients (p = 0.001) and control subjects (p < 0.001) but did not differ between the two groups (p = 0.154) and did not differ between the two genders within groups (p = 0.702 for cases, p = 0.098 for controls). Although the main analyses of the overall trends of prevalence of AF were based on age-adjusted prevalence values for age 75 years (the mean age for both cases and controls), the modeled prevalence estimates of AF for men and women of various ages with and without ischemic stroke also showed trends of progressive increase over the three decades (Table 3).
Modeled Prevalence (%) of Atrial Fibrillation Among Rochester Residents, Stratified by Gender, Age, Presence or Absence of Ischemic Stroke, and Decade
Time trends of comorbid conditions
From 1960 to 1989, ⇓significant increases in prevalent hypertension, coronary artery disease, and valvular heart disease were evident in both cases and controls. Prevalent myocardial infarction was found to increase significantly in cases and older controls, whereas prevalent congestive heart failure, diabetes mellitus, and history of cardiac surgery increased significantly only in the cases (Table 4, Fig. 3).
Prevalence of atrial fibrillation (AF) and comorbid conditions in 1,871 patients with ischemic stroke and their age- and gender-matched controls, stratified by calendar decade. CAD = coronary artery disease; CHF = congestive heart failure; CS = cardiac surgery; DM = diabetes mellitus; HTN = hypertension; MI = myocardial infarction; VHD = valvular heart disease.
Change in Prevalence (Odds Ratio) for Five-Year Changes in Calendar Time for Various Comorbid Conditions
Use of ECG and echocardiography
For both ischemic stroke cases and their controls, there was an increase in the use of ECG within the ±30-day window of the stroke/reference date from the first decade studied to the second, but no further increase was evident thereafter (Table 5). The overall rates of utilization of ECG were higher in the cases, but the trends over time appeared similar for both cases and controls. The utilization of echocardiography increased substantially over the three decades. It was 0%, 5%, and 53% for the stroke cases, and 0%, 2%, and 18% for the controls, for 1960 to 1969, 1970 to 1979, and 1980 to 1989, respectively.
Use of Electrocardiography Within 30 Days of Stroke in Case Patients or Reference Date in Control Subjects for Three Decades
Discussion
Prevalence of AF and factors contributing to development of an epidemic
This study evaluated the prevalence rates of AF in patients with incident stroke and in age- and gender-matched controls over a 30-year period. Although the observed and the modeled prevalence rates should not be assumed to be identical to the actual prevalence of AF in the general population of Rochester, Minnesota, the secular trends of AF prevalence observed for these unique groups are informative of the epidemiology of this arrhythmia over the time period examined. Consistent with other studies (15–19), age is a potent risk factor for the development of AF: each decade of age was associated with a doubling of the odds for AF (controls: OR 2.22, 95% CI 1.84 to 2.86; cases: OR 1.81, 95% CI 1.58 to 2.07). This fact, coupled with the steadily increasing numbers of persons reaching very old age, most certainly contributes to the growing epidemic of AF.
Second, a “calendar-time effect” on the prevalence of AF, independent of age, was evident. The age-adjusted prevalence estimates were those of the predicted prevalence from the multivariate models for age 75 years, which was the mean age for the cohort. Over the 30-year period, the age-adjusted prevalence of AF increased significantly (p < 0.001) for the total cohort of 3,742 individuals (Fig. 2). The rates of increase were significant for both case patients (p = 0.001) and control subjects (p < 0.001) but did not differ between the two groups (p = 0.154). In contrast to the findings from the Framingham cohort, this effect did not differ significantly by gender. Within case and control groups, the rates of increase over time were significant for both men and women but did not differ between the two genders (p = 0.702 for cases, p = 0.098 for controls).
Third, the precise contribution of ascertainment bias to the observed changes in prevalence of AF in the study population could not be readily determined; however, these changes were unlikely to be largely accounted for by the utilization pattern of ECG. Relative to 1960 to 1969, the frequency with which ECG was performed within ±30 days of the stroke/reference date was somewhat higher in 1970 to 1979: cases had an increase of 12%, and controls an increase of 9% (Table 5). Yet the increase in observed AF prevalence was 6% in cases and 3% in controls (Table 2). Further, there was no increase in the utilization of ECG for either the cases or the controls from the second to the third decade (Table 5). Yet the observed increase in prevalence of AF for both cases and controls was 5% (Table 2). While acknowledging the limitation of assessing utilization of ECG within ±30 days of the stroke/reference date, to attribute the change in prevalence to the degree of increased utilization of ECG suggested by these data requires an implausibly high rate of AF among the additional subjects having an ECG in later time periods.
The rates of utilization of ECG were higher in cases than in controls for all three decades, although the trends over time were similar. The higher utilization rate in the cases probably reflected the greater cardiovascular disease burden in these subjects, and therefore more frequent indication for ECG assessment. For symptomatic AF, both cases and controls should have had similar likelihood of receiving ECG assessment, and detection rate should be comparable. A systematic under-detection of silent AF in the controls remains possible, because these individuals were less disease burdened and, therefore, possibly less likely to undergo ECG assessment as part of evaluation for other conditions they might have had.
Changes over time in the prevalence of comorbid conditions, however, may play an important role in the genesis and propagation of this epidemic (Fig. 3). The risk of AF development is a function not only of advancing age but also of the total disease burden and the presence among those who survive to older ages of specific comorbid conditions that are themselves risk factors for AF. However, the precise effect of changes in the prevalence of comorbid conditions on the secular trends of AF could not be readily determined in this study.
The role of comorbid conditions in the epidemic of AF
Hypertension, coronary artery disease, diabetes mellitus, and heart failure are among the age-related comorbid conditions that likely play a central role in the propagation of the AF epidemic. In addition to increasingly greater numbers of persons reaching the age at which the risk of AF escalates, a significant proportion of the elderly population have survived these conditions (20–31), which in themselves are risk factors for the development of AF. Diastolic dysfunction, which occurs commonly in patients with these conditions, was recently identified as a robust predictor of nonvalvular AF and may serve as the common pathophysiologic link between these conditions and AF development (32). There was also evidence of an increase in prevalence of valvular heart disease over time, both for cases and for controls. This could in part be related to an increase in detection rate as a result of greater sophistication and utilization of echocardiography, although an actual increase in the number of older individuals who have developed degenerative valve disease could also have contributed (33).
Variations in the prevalence of AF in the U.S.
In this study, the age-adjusted prevalence rates of AF among Rochester men in the control group were 5% for the period 1960 to 1969 and 12% for the period 1980 to 1989. The age-adjusted prevalence rates were 3.2% for 1968 to 1970 and 9.1% for 1987 to 1989 for men in the Framingham study (8). In the Cardiovascular Health Study, the prevalence of AF was 6.2% for men, all of whom were 65 years of age or older (15). The differences in the prevalence rates could be related to the differences in study criteria, methods of ascertainment, and techniques of age adjustment for estimation of the prevalence rates, although true variations in the geographic distribution of AF cannot be excluded.
Study limitations
Rochester is predominantly a white community, and the extent to which the findings of this study may be generalized to the population at large is unknown. We acknowledge that the prevalence of AF in neither the stroke cases nor the controls is necessarily representative of the prevalence rate of AF in the general Rochester community. This notwithstanding, the trends of AF identified in this study remain informative with respect to the changes in prevalence of AF over the 30-year period in men and women who had an ischemic stroke and in age- and gender-matched controls. The precise contribution of ascertainment bias on the overall change in AF prevalence could not be readily determined. Also, AF first detected on stroke/reference date was not counted as part of the “prevalence of history of AF,” and, therefore, the true prevalence of AF in both stroke cases and control subjects could have been underestimated.
Although we evaluated concurrent trends in prevalence for a number of important comorbid conditions, we did not assess other putative risk factors such as alcohol consumption, sleep apnea, thyroid disorders, infection, or pulmonary disease, which could also have affected the trends in AF. The reasons for the observed trends in AF prevalence are undoubtedly multifactorial, and the impact of each contributing factor could not be readily addressed in this study.
Clinical implications
Analyzing multiple concurrent trends is a highly complex process, and cause-and-effect interpretations are not possible. Nevertheless, simple descriptive analyses of secular trends, such as those reported here, are vital benchmarks for our health care efforts. The data from this study contributes to the understanding of the epidemiology and magnitude of AF as a public health problem. The projected 2.5-fold increase in the number of patients with AF over the next 50 years, based on cross-sectional census data (2), may potentially underestimate the magnitude of the problem as it evolves over time.
Conclusions
The prevalence of AF increased significantly over the three decades from 1960 to 1989 in Rochester residents who had ischemic stroke and in their age- and gender-matched controls. The magnitude of increase did not differ significantly between men and women. A concomitant increase in the prevalence of certain comorbid conditions might have contributed in part to the identified trends in AF. The observations are alarming with potentially far-reaching consequences not only in terms of the growing AF burden, but also its impact on stroke, the quality of life in many elderly, and the attendant socioeconomic implications. Aggressive AF risk reduction and management are pivotal to the containment of this epidemic. Studies are needed to advance our understanding of the forces shaping this epidemic, to quantify the impact of AF on resource utilization and health outcomes, and to determine effective primary prevention strategies for the control and mitigation of the effects of this arrhythmia.
APPENDIX
Myocardial infarctionwas based on three criteria: clinical history indicative of acute ischemia; presence of serial ECG changes indicative of myocardial damage; and diagnostic increase in serum creatinine phosphokinase (introduced in 1963), lactic dehydrogenase (introduced in 1963), and serum glutamic-oxaloacetic transaminase (introduced in 1955).
Coronary heart diseasewas defined by presence of angina or myocardial infarction.
Hypertensionwas considered present if the subject was given antihypertensive treatment and/or if a systolic blood pressure ≥160 mm Hg or a diastolic blood pressure ≥95 mm Hg on two or more measurements occurred before the stroke or reference date. An increased pressure occurring in the context of an acute illness or injury was considered an exception.
Congestive heart failurewas defined by the presence of at least four of the following 11 criteria: dyspnea on ordinary exertion, paroxysmal nocturnal dyspnea, acute pulmonary edema, distended neck veins, bilateral ankle edema (not due to a condition other than cardiac failure), hepatomegaly (not due to liver disease), rales in the absence of pulmonary disease, presence of third heart sound or summation gallop, radiographic evidence of pulmonary congestion (pleural fluid, pulmonary venous congestion, prominent pulmonary veins) with or without cardiomegaly, circulation time >24 s (arm to tongue), pulmonary edema, visceral congestion, and cardiomegaly at autopsy.
Valvular heart diseasewas defined by clinical diagnosis with or without ancillary evidence on chest radiography and echocardiography.
Diabetes mellituswas considered present if the National Diabetes Data Group criteria were met before the date of stroke or reference date.
Atrial fibrillationwas considered present only if there was ECG confirmation.
Footnotes
☆ This work was supported by the American Heart Association Scientist Development Grant.
- Abbreviations
- AF
- atrial fibrillation
- CI
- confidence interval
- ECG
- electrocardiogram/electrocardiographic
- OR
- odds ratio
- Received July 24, 2002.
- Revision received November 24, 2002.
- Accepted December 18, 2002.
- American College of Cardiology Foundation
