Thromboembolism Prophylaxis in Chronic Atrial Fibrillation
Practice Patterns in Community and Tertiary-Care Hospitals
Background and Purpose By 1992, several prospective trials established the efficacy of anticoagulation (AC) and to some extent antiplatelet (AP) agents in the prevention of stroke in the setting of atrial fibrillation (AF). The objective of this study was to determine whether practice patterns in AF stroke prophylaxis reflect the findings of clinical trials and whether stroke prophylaxis in AF differs between community hospitals and tertiary teaching hospitals.
Methods Retrospectively, 1250 hospital charts were reviewed. After patients who had undergone recent surgery, received treatment for malignancy, or were not in chronic AF on discharge were eliminated, 651 remaining records were analyzed for the presence of 26 clinical factors influencing the selection of thromboembolism prophylaxis. Descriptive statistics and logistic regression were used to analyze the association between clinical and demographic factors and the decision to treat with AC, AP, or no specific antiembolic therapy.
Results Of the 651 patients in AF, 273 (42%) received no emboli prophylaxis while 219 (34%) were treated with AC (warfarin), 146 (22%) were treated with AP, and 13 (2%) received both agents. Patients discharged in AF from community hospitals were significantly less likely to be treated with either AC or AP agents than patients discharged from tertiary centers. A strong bias against thromboembolism prophylaxis with either AC or AP agents in AF existed with age over 45 years. Multivariate logistic regression indicated that the decision to treat was associated only with the presence of prosthetic valve, history of prior stroke, mitral disease, and absence of a recent gastrointestinal bleed or occult blood in stool. Even after adjustment for these factors, a significant bias against treatment with either AC or AP agents with advancing age and discharge from community hospitals remained.
Conclusions Thromboembolism prophylaxis with either AC or AP agents is underutilized in the setting of AF. Furthermore, factors known to increase the risk of embolization in AF such as age, hypertension, diabetes, and heart disease were not associated with decisions to treat with either AP or AC agents. This study suggests that the use of clinical guidelines suggested by trials of thromboembolism prophylaxis in AF could reduce the incidence of stroke.
Atrial fibrillation has long been known to be a major risk factor for stroke. The prevalence of AF in the United States for those aged 55 years or younger is less than 0.5% but rises to 6% for individuals older than 65 years.1 With advancing age, the percentage of total strokes attributable to the presence of AF also increases. Of patients aged 50 to 59 years with AF, 6.7% of all strokes are attributable to AF, while in those older than 80 years, the figure rises to 36%.2 The prevention of stroke in AF has been studied extensively. Long-term AC is the established standard of care in the setting of AF and prosthetic valves or mitral valve disease. Nonvalvular AF is also a major age-related risk factor for stroke. By 1992, five prospective trials established that AC reduced the risk of stroke in nonvalvular AF by 42% to 86% compared with placebo.3 4 5 6 7 While AC is clearly more effective than placebo in stroke prophylaxis, the role of AP agents in nonvalvular AF remains less clear. The Copenhagen AFASAK study demonstrated a benefit in stroke reduction with 75 mg of aspirin daily, which failed to achieve statistical significance compared with placebo.3 In contrast, in 1994 the second SPAF study (SPAF-II) reported a 44% decrease in stroke in those treated with 325 mg of aspirin daily over placebo (95% confidence interval, 7% to 66%; P=.02) but suggested that aspirin may be ineffective in those older than 75 years.8 Whether in selected patient groups AP therapy is equivalent to or better than AC therapy is currently being studied in several large trials (AFASAK-II, SPS, and PATAF). Recently, SPAF-III results demonstrated a superior benefit of adjusted-dose warfarin to low, fixed-dose warfarin plus aspirin in high-risk AF patients.9
A multivariate analysis of the SPAF-I data identified clinical and echocardiographic factors associated with a high risk for thromboembolic events. The presence of either congestive heart failure within the previous 3 months, a history of hypertension, or a previous arterial embolic event increased the risk of stroke from 2.5% to 7.2% yearly. Any two or three of these risk factors were associated with a yearly stroke risk in excess of 17%.10 Echocardiographic evidence of left atrial enlargement or left ventricular dysfunction was also found to increase the risk of thromboembolic stroke.11 Conversely, the risk of ICH in AC prophylaxis of AF has been related to uncontrolled hypertension and AC beyond therapeutic ranges.12 13 14 Similarly, major bleeding episodes can be expected to occur more commonly in a host of clinical conditions ranging from frequent falls to bleeding diatheses. However, in the patients randomized to the five trials of AC in nonvalvular AF reported by 1992, the benefit of AC prophylaxis emerged. Collectively, the published studies strongly support the use of AC and perhaps AP agents in patients with nonvalvular AF older than 60 years. In 1992 and again in 1995, the American College of Chest Physicians recommended that all patients except those younger than 60 years who have no associated cardiovascular disease (“lone AF”) receive AC therapy.15 16
Identification of patient factors associated with either high risk for thromboembolic events in AF or, if anticoagulated, high risk for ICH or other major hemorrhage may influence the clinical decision between AC or AP prophylaxis. The objective of this study was to determine whether current clinical practice reflects the risk stratification suggested by clinical trials. We further sought to compare practice patterns in two tertiary-care hospitals with large teaching services with those of two community hospitals within a single geographic region.
Subjects and Methods
Medical information systems at four regional hospitals were queried for all patients discharged during a 12-month period extending from January 1994 through February 1995 with an International Classification of Diseases, 9th Revision, code of 427.31 (AF). Approximately 1250 charts were then reviewed. Patients with transient or paroxysmal AF, a recent major surgical procedure, or undergoing treatment for active malignancy were excluded. The remaining 651 charts were from patients with chronic AF and were reviewed for clinical information relevant to choice of thromboembolism prophylaxis.
Two community hospitals and two tertiary referral hospitals within a single geographic region were surveyed. The community hospitals were acute-care facilities with medical and surgical services but no supporting teaching services. The tertiary-care hospitals were university affiliated and had large medical and surgical teaching units. The two tertiary teaching hospitals were located in an urban environment. The two community hospitals were located 40 miles from the tertiary hospitals and served rural populations. To prevent confounding of comparisons between the two hospital groups, the participating hospitals were selected so that few, if any, of the attending physicians of one hospital group had privileges within the other group. Patients were admitted to either hospital group on the basis of the attending physician's affiliation. All four hospitals had advanced medical information systems, allowing accurate identification of AF patients.
A data collection form was developed, piloted, and modified for the extraction of pertinent historical and diagnostic information. Before data collection began, a set of records from each center was reviewed independently to ensure uniformity of data reporting. Along with age and sex, factors suggesting a benefit as well as factors suggesting risk for AC or AP therapy were searched for by review of the entire medical record (Table⇓). For patients with multiple admissions, only the most recent was analyzed.
The diagnosis of coexisting medical conditions, such as hypertension and diabetes, was made if this could be identified in the admission history or was actively treated on discharge. Congestive heart failure/left ventricular dysfunction was identified if a patient received treatment for congestive heart failure or if an echocardiogram demonstrated an ejection fraction of less than 30%. A previous GI bleeding event was established by history. GI bleeding requiring invasive investigation or therapy within 3 months of the admission was considered a major bleed. The presence of occult blood was recorded if found in the stool during admission or otherwise noted in the record. The determination of antithrombotic therapy was made by reviewing medication orders in effect at the time of discharge and specific discharge medications listed in the discharge orders or summary. Descriptive statistics and logistic regression were used to analyze the association between the clinical data and selection of AP, AC, or no specific antiembolic therapy.
Of the 651 charts reviewed, 42% (273) received no specific antithrombotic therapy (AC or AP). Of those whose physicians selected antiembolic therapy, 34% (219) were treated with AC (warfarin), 22% (146) with AP (aspirin or ticlopidine), and 2% (13) were treated with both AC and AP agents. Of the 651 AF patients reviewed, 609 had no history of previous major bleed (including intracranial), terminal illness, thrombocytopenia, bleeding disorder, aneurysm, or arteriovenous malformation. In this selected group without major risk factors for AC or AP therapy, 40% received no thromboembolism prophylaxis. Additionally, 129 patients were identified with at least one clinical factor supporting thromboembolism prophylaxis and none associated with increased risk for AC or AP use. Of these 129 patients, 36% received no thromboembolism prophylaxis, 39% were treated with AC, 23% with AP, and 2% with both agents.
Patients were progressively less likely to be treated with AC with advancing age (Fig 1⇓). This bias against AC with advancing age was seen in both community and tertiary referral hospitals. The mean age between the two hospital groups was similar. In community hospitals, 57% of AF patients did not receive either AC or AP thromboembolism prophylaxis on discharge compared with 36% in tertiary hospitals (Fig 2⇓). Patients discharged from tertiary-care hospitals in AF were twice as likely to be anticoagulated than those discharged from community hospitals (42% versus 21%).
Multiple logistic regression was performed to explore the relationship between choice of antithrombotic therapy and clinical factors. Using a two-outcome-state model, comparing any treatment (AC, AP, or both) versus no treatment, we found that five parameters were significantly associated with the decision to treat with antithrombotic therapy: (1) presence of occult blood, (2) recent evidence of GI bleed, (3) presence of prosthetic valve, (4) history of prior stroke, or (5) mitral disease (P<.025). A history of a prior arterial embolus or ICH just missed statistical significance in predicting whether patients received AC/AP or no therapy in this model.
In this regression model, patient age and hospital group correlated best with the decision to treat or not treat (P<.0001). Both discharge from a community hospital and advancing age were associated with the decision not to treat with either AC or AP agents. These two variables alone correctly predicted the physician's choice to treat or not with an accuracy of 66%. All other clinical risk factors did not improve the ability of the model to predict the physician practice patterns.
Further analysis was done with a three-outcome-state model to determine whether any factor was associated with the decision to treat with AC (with or without AP) versus AP versus no prophylaxis. In this model no clinical variable, age, or hospital group could predict the choice of AC versus AP versus no emboli prophylaxis. Factors predicting the choice of AC were identical with the factors predicting the choice of AP agents. The three-outcome-state regression model thus implies that none of the clinical factors examined correlated with the choice between AC or AP as an agent for thromboembolism prophylaxis in AF.
Our study demonstrated that in patients with AF with no identified risk factors for AC or AP therapy, 40% remained untreated. Furthermore, the results of the regression analysis above indicate that the clinical decision centered on whether to treat at all; the choice between AC or AP agents could not be significantly associated with any clinical factor. As established in the Framingham data and supported by nonvalvular AF studies, the risk of stroke in AF is small in those younger than 60 years but increases with age.2 Our study found that patients aged 45 to 54 years were twice as likely to receive AC as those aged 75 to 84 years. Advancing age was associated with a decrease in the number of patients treated with either AC or AP prophylaxis in both community and tertiary teaching hospital groups, thus reflecting a general practice bias against any therapy with age not supported by clinical trials. Furthermore, the percentage of patients with AF who remained untreated in community hospitals was significantly higher than in tertiary hospitals.
A meta-analysis of the five prospective clinical trials of stroke prophylaxis in nonvalvular AF published in 1994 demonstrated that in the selected group of patients randomized in these trials, treatment with AC reduced stroke rates by 68% and overall mortality by 33%. The overall stroke rate was 4.5% for control subjects (n=1236) and 1.4% yearly for the warfarin-treated group (n=1225). In this pooled analysis, aspirin reduced the risk of stroke by 3.6% but had no significant effect on death. The pooled rate of ICH was 0.1% yearly for placebo and 0.3% yearly for the AC group.17
A multivariate analysis of these pooled data identified the following predictors of stroke (and their associated relative risk): (1) prior stroke/transient ischemic attack (2.5); (2) history of hypertension (1.6); (3) age (1.4 based on decades of age); (4) history of diabetes (1.7); (5) congestive heart disease (1.4); and (6) history of myocardial infarction (1.2). For patients older than 75 years in SPAF-II, the rate of ICH was 1.8% yearly (95% confidence interval, 0.8 to 3.7), suggesting an increased risk for this age group. However, in data pooled from the AFASAK, BAATAF, CAFA, and SPINAF trials, the ICH rate of patients older than 80 years was 0.3% (95% confidence interval, 0.04 to 2.1), suggesting that ICH may not be increased in patients older than 75 years. Data from the pooled analysis was insufficient to allow a conclusion regarding the effect of age on the risk of ICH, but the risk of ICH remained less than that of stroke for all ages.17
In contrast to the findings of the pooled analysis of AF trials, in which advancing age was associated with an increased risk for stroke with no clear increase in ICH, the regression model of our practice pattern data disclosed that advancing age was most significantly associated with the decision not to treat with antithrombotic therapy. Discharge from a community hospital was also associated with a decision not to treat with either AC or AP drugs. When age and hospital group factors were removed, the choice of therapy was statistically associated with the presence of occult blood, history of GI bleed, prior stroke, mitral disease, and the existence of a prosthetic valve. In an additional analysis in which patients with prosthetic valves, known mitral valvular disease, and prior stroke were excluded, no clinical factor could be associated with the decision to treat, and only the presence of occult blood or history of GI bleed was associated with the decision not to treat. This implies that when making therapeutic decisions, clinicians do not consider risk factors increasing the likelihood of stroke in nonvalvular AF but exclude patients with advancing age, occult blood in stool, or remote GI bleed from either AC or AP prophylaxis. Our study also indicated that more than 50% of patients younger than 65 years received AC prophylaxis, which is at variance with the findings of the clinical trials, which found little benefit of AC for patients younger than 65 years who had no risk factors for thromboembolism.
Our findings imply that physicians in both hospital groups studied have not fully embraced the results of clinical trials indicating the benefit of AC or AP prophylaxis in AF. Information on risks and benefits associated with the decision to treat with AC or AP agents suggested by several clinical trials is not currently reflected in clinical practice. This disparity between clinical practice and the results of clinical trials is not unique to stroke prophylaxis in AF. In a study of prophylaxis of venous thromboembolism, only 19% of patients treated at nonteaching hospitals and 44% of patients at teaching hospitals received adequate thromboembolic prophylaxis.18 Our results, however, differ from those of Gottlieb and Salem-Schatz,19 who reported AC practice in AF for a large staff model health maintenance organization. In this study of 238 patients with AF, 85% were offered and 79% received AC therapy. The number of patients treated with AP prophylaxis was not reported. The reasons for a substantially higher treatment rate in this study than in ours may reflect regional differences in attitudes toward AC between this urban environment of greater Boston and ours in western New York. Alternatively, the higher rates of AC prophylaxis reported may reflect care patterns in a large staff model health maintenance organization with previous experience in a trial of AC in AF compared with the diverse practice settings of western New York physicians.
The reasons physicians elected not to treat with AC were not explored in this study but probably relate to a perceived excessive risk of AC in the practice environment compared with that of clinical trials.20 Physicians have demonstrated concerns about long-term AC, particularly in the elderly, despite results of the clinical trials.21 The results of SPAF-II suggest that the risk of ICH increases with high international normalized ratios. This potential risk of hemorrhage may bias physicians against the use of AC in patients in whom compliance and therapeutic monitoring are problematic. This perceived excess risk, however, did not translate to an increased use of AP agents. AC prophylaxis in AF has been established as clinically effective in reducing both stroke and overall mortality in high-risk patient groups. Furthermore, stroke prophylaxis with AC agents is cost-effective and does not adversely affect quality of life.22 23 The selection and monitoring of thromboembolism prophylaxis in AF should be determined by clinical practice guidelines that stratify the risk and benefits of prophylaxis as determined by clinical trials. Such an assessment is currently not apparent in the broad clinical practice studied here but could be expected to reduce the number of cardioembolic strokes and improve outcomes.
Selected Abbreviations and Acronyms
|AFASAK||=||Atrial Fibrillation, Aspirin, Anticoagulation|
|BAATAF||=||Boston Area Anticoagulation Trial for Atrial Fibrillation|
|CAFA||=||Canadian Atrial Fibrillation Anticoagulation Study|
|PATAF||=||Prevention of Arterial Thromboembolism in Nonvalvular Atrial Fibrillation Study|
|SPAF||=||Stroke Prevention in Atrial Fibrillation|
|SPINAF||=||Stroke Prevention in Nonrheumatic Atrial Fibrillation|
|SPS||=||Stockholm Prospective Study|
This study was supported in part by DuPont Pharma and the Western New York Stroke Consortium.
- Received July 16, 1996.
- Revision received September 16, 1996.
- Accepted September 24, 1996.
- Copyright © 1997 by American Heart Association
The SPAF Investigators. The Stroke Prevention in Atrial Fibrillation study: final results. Circulation.. 1991;84:527-539.
Ezekowitz MD, Bridgers SL, James KE, Carliner NH, Colling CL, Gornick CC, Krause-Steinrauf H, Kurtzke JF, Nazarian SM, Radford MJ, Rickles FR, Shabetai R, Deykin D. Warfarin in the prevention of stroke associated with nonrheumatic atrial fibrillation. N Engl J Med. 1992;327:1406-1412.
SPAF Investigators. Predictors of thromboembolism in atrial fibrillation, I: clinical features of patients at risk. Ann Intern Med.. 1992;116:1-5.
SPAF Investigators. Predictors of thromboembolism in atrial fibrillation, II: echocardiographic features of patients at risk. Ann Intern Med.. 1992;116:6-12.
Kase CS, Robinson RK, Stein RW, DeWitt LD, Hier DB, Harp DL. Anticoagulant-related intracerebral hemorrhage. Neurology. 1985;35:943-948.
Albers GW. Intensity of anticoagulant treatment and risk of intracerebral hematoma. Stroke. 1990; 21:1758. Letter.
Laupacis A, Albers G, Dunn M, Feinberg W. Antithrombotic therapy in atrial fibrillation. Chest. 1992;102(suppl):426S-433S.
Laupacis A, Albers G, Dalen J, Dunn M, Feinberg W, Jacobson A. Antithrombotic therapy in atrial fibrillation. Chest. 1995;108(suppl):352S-359S.
Anderson FA, Wheeler HB, Goldberg RJ, Hosmer DW, Forcier A, Patwardhan NA. Changing clinical practice: prospective study of the impact of continuing medical education and quality assurance programs on use of prophylaxis for venous thromboembolism. Arch Intern Med.. 1994;154:669-677.