Clinical correlation between effective anticoagulants and risk of stroke: are we using evidence-based strategies?

Objectives: Despite evidence supporting anticoagulant use in atrial fibrillation, this modality is not fully utilized.

Methods: Retrospective chart review of 297 patients with nonvalvular atrial fibrillation between 1997 to 2000. 124 patients received warfarin and 166 did not; 91 patients suffered stroke.

Results: Age (P = 0.232) and gender (P = 0.745) were not determinant factors for starting anticoagulation prophylaxis. Whites were more likely to receive anticoagulation therapy than blacks (P = 0.043). Cardiologists were 4.5 times more likely to prescribe warfarin than neurologists and internists (P = 0.035). Neurologists (P = 0.305) and internists (P = 0.770) had similar warfarin prescription patterns and often with patients experiencing the highest rates of stroke.

Conclusion: Lack of a uniform pattern in anticoagulant administration, despite multiple guidelines, is disturbing. Continuous physician education and community awareness by local and federal medical agencies is essential and cost-effective.

Key Words: atrial fibrillation, primary prevention, stroke, thromboembolic event, warfarin

**********

Atrial fibrillation (AF) is a condition in which the heart beats rapidly and erratically. Atrial fibrillation predisposes patients to blood clots that can cause stroke by lodging in arteries of the brain. (1) The annual rate of stroke among patients with nonvalvular AF is about 5% per year. (2) If untreated, one in three people with nonvalvular AF will suffer a stroke. (3) The importance and efficacy of anticoagulation therapy in preventing a stroke is common knowledge, and has become the standard of care. (4) The cost-effectiveness of anticoagulation therapy in patients with nonvalvular AF, even in older patients who are more at risk for adverse events, is also well established. Lightowles (5) reported that anticoagulation therapy is most cost-effective in this patient group.

In a metaanalysis by Hart et al, (6) warfarin was reported to be substantially more efficacious than aspirin, as the benefit of anticoagulation therapy was not offset by the occurrence of major hemorrhage. Elaine et al (7-9) reported that anticoagulant prophylaxis is effective at the International Normalized Ratio (INR) of 2.0 to 3.0, and is associated with a lower risk of bleeding. However, despite extensive evidence and recommendations for its use, anticoagulant prophylaxis is underused in the management of AF. (7-9)

Materials and Methods

Sample

Patients with a diagnosis of AF requiring anticoagulation therapy and who were admitted between calendar years 1997 and 2000 to a community-based teaching hospital located in Southeast Georgia were identified through a retrospective chart review. However, patients who were admitted and diagnosed with hypercoagulable state, hemorrhagic stroke, carotid stenosis, peripheral vascular disease, or dilated cardiomyopathy were excluded from the study. The final sample consisted of 297 patients: 91 patients who suffered a stroke at any point during the study period, and 206 patients who did not suffer a stroke by the end of calendar year 2000. From this case-control sample, we formed two study groups: 124 patients who were prescribed warfarin, and 166 patients who were not prescribed warfarin. Without documentation of anticoagulation therapy in their records, seven patients could not be categorized into either group. We designated an INR between 2.0 and 3.0 as therapeutic in our reporting of the intensity of anticoagulation prophylaxis. (10,11)

[FIGURE 1 OMITTED]

We identified the primary and prescribing physician specialties, and categorized them in four groups: neurologists, cardiologists, general internists, and others. Of the total 62 internists, 32 were primary prescribing physicians. Eighteen cardiologists out of 28, and 9 out of 11 neurologists were prescribing physicians. The last group, designated as "others," consisted of 9 family practice physicians out of 43 prescribing, one gynecologist and one general practitioner. It seems all four groups of practicing physicians have good representation to demonstrate the practice behavior.

Nature of the variables collected

Patient demographics, clinical diagnoses, findings, and treating physicians--primary care (PMD) versus specialist or admitting physician--were extracted from admission records. History of AF, duration of the disease, treatment modalities, indicators for anticoagulation therapy, and risk factors for comorbid diseases were also obtained from hospital charts.

For each of the continuous variables, means and medians were used to assess the central tendency of the distributions. Dispersion was evaluated using standard deviation and range of values. Skewness and kurtosis distributions were assessed using their reported statistics. Normal probability plots and the Shapiro-Wilks test were used to assess the normality assumption (Figure 1). For better visualization of the distribution of age, INR, and warfarin, distribution plots that were based on kernel-smoothed density estimates according to the patient's warfarin status were used (Figure 2).

Differences in mean age and mean INR between warfarin and nonwarfarin patients, and mean age, mean INR, and mean warfarin dosage between stroke and nonstroke patients were compared using the nonparametric Wilcoxon test. A percentage of all patients and [chi square] tests of independence (of equal proportions) were utilized to compare patients prescribed and those not prescribed warfarin for each of the indicator variables, such as the patient's gender, race, age, specialty of the managing physician, comorbidities, and whether the patient had suffered a stroke or not.

[FIGURE 2 OMITTED]

Analyses

The adequacy of anticoagulation therapy by treating physicians was analyzed using an analysis of covariance. We chose warfarin and "no warfarin" prescriptions as dependent variables. Age, race, and INR values, or the percentage values from ideal INR values, were independent variables.

The risk of stroke associated with anticoagulation therapy was assessed as a preliminary analysis using 2 X 2 contingency tables and stratified Mantel-Haenszel analyses. The odds ratios of suffering from a stroke associated with not having been prescribed warfarin medication were calculated. To assess possible differences in anticoagulation treatments on the risk of stroke across physician specialties, conditional odds ratios for each stratum were calculated. A Cochran-Mantel-Haenszel [chi square] test was performed to assess the significance level of a common odds ratio, adjusted for physician specialty stratum. The appropriateness of reporting a common odds ratio was assessed by the Breslow-Day test of homogeneity of the strata-specific odds ratios. A multiple logistic regression analysis was used to identify additional predictors of stroke and to adjust the odds ratio using a full model of predictors.

The study's objective was to define correlation between effective anticoagulant prophylaxis and risk of thromboembolic stroke. Yet, our experimental design also allowed us to observe and investigate the role of other factors, such as practice patterns in different settings.

Results

Descriptive statistics for the distribution of age and INR values by warfarin status are reported in Table 1. The mean age of 72.95 (mean [+ or -] SD) for patients preseribed warfarin was not statistically different (P = 0.232) from the mean age of 74.23 for patients not prescribed warfarin. The distribution of age for the patients prescribed warfarin is skewed and highly peaked (Figure 2). The mean INR value of 2.51 for patients prescribed warfarin is significantly greater (P < 0.001) than the mean INR value of 1.19 for patients not prescribed therapeutic doses of warfarin. Yet, the proportion of patients who were prescribed warfarin does not appear to be related to gender (P = 0.745); however, we cannot reject the null hypothesis that the proportion of patients prescribed warfarin is independent of race. White patients comprise a higher proportion of those prescribed warfarin than those who were not prescribed warfarin (P = 0.043). However, patients may have none or all of the comorbidities of interest; therefore, we tested for equal proportions for each comorbidity separately. Patients with a history of stroke were proportionally more prominent in the group that was prescribed than the group who was not prescribed warfarin. All other comorbidities considered important in relation to AF and stroke were equally present in both warfarin and non-warfarin groups. Managing physicians were selected in similar proportions between those prescribed and those not prescribed warfarin (P = 0.020). Patients who were prescribed warfarin appear to have chosen cardiologists in higher proportions to manage their care. Patients not prescribed warfarin had a higher proportion of strokes than those who were prescribed warfarin.

Discussion

The degree of association and correlation coefficients for the variables of interest is shown in Table 2. Most of the pair-wise correlation coefficients are relatively small. The largest correlation coefficients exist between the following pairs: higher warfarin dosages are associated with larger values in INR (r = +0.395); higher warfarin dosages are associated with fewer strokes (r = -0.372); higher warfarin dosages are prescribed by cardiologists (r = +0.289); and greater occurrences of strokes are present in those patients under the care of neurologists (r = +0.394).

An unbalanced analysis of variance was conducted on warfarin dosages to test whether the mean dose varies significantly by the specialty of the managing physician. The results are presented in Table 3. The least square means are the diagonal elements, and the pair-wise statistical test P values are presented on the off-diagonal elements. This indicates that cardiologists prescribed, on average, 2.94 mg of warfarin, which is significantly higher than the mean prescribed warfarin dose of 1.76 mg by neurologists (P = 0.03) or the 1.80 mg by internists (P = 0.005). In fact, it was substantially higher than any other specialty type taken as a single group, whose mean does was about 1.42 mg of warfarin (P < 0.001).

Odds of Warfarin Prescription

A multiple logistic regression analysis as an alternative to a linear model was used to predict warfarin dosage (Table 4). Elderly patients were more likely to have additional comorbidities that influenced the prescription of warfarin, and there was no expectation about gender effects on the decision to prescribe warfarin across the specialty of the managing physician. We expected patients with heart disease, coronary artery disease (CAD), and congestive heart failure (CHF) to be more likely to be prescribed warfarin to reduce the risk of a stroke.

Patients with AF and history of stroke were 6.13 (P = 0.013) times more likely to be prescribed warfarin when compared with patients with CAD, CHF, or noncardiac-related concomitant diseases. Among managing physicians, cardiologists were 4.5 times more likely (P = 0.035) to prescribe warfarin than noncardiologists were. Neurologists (P = 0.308) and internists (P = 0.770) have similar patterns of prescribing warfarin when compared with cardiologists.

Standard errors for most of the predictors are slightly greater than the estimated coefficient. However, these predictors are all indicator variables, thus it is usual for the standard errors to be slightly greater than the parameter estimate. Therefore, degeneracy or collinearity in the multiple logistic regression model is marginally possible. The entropy [R.sup.2] is low at only 0.047, indicating that very little of the information available in the data explains the variations in the probability of an AF patient being prescribed warfarin.

Odds of Stroke

A Mantel-Haenszel analysis was performed to test whether a common odds ratio could be reported. Based on the result of the Breslow-Day test of homogeneity of the strata-specific odds ratio (P = 0.894), we concluded that we can report a single odds ratio, adjusted for the specialty of the managing physician and the INR value. Thereby, adequate intervention for reducing risk of stroke, utilizing conditional and unconditional odds ratios and their confidence limits with associated P values, are reported in Table 5. Conditionally, patients managed by an internist have a significantly higher risk (P = 0.012), approximately 3.05 times the odds, of suffering a stroke when not prescribed compared with those who are prescribed warfarin. Without accounting for the differences in how the care of AF patients are managed according to the physician's specialty, patients are 2.48 times more likely to suffer a stroke if not prescribed warfarin. An adjusted odds ratio of 2.91 indicates that AF patients who are not prescribed warfarin are 2.91 times more likely to suffer a stroke than AF patients who were prescribed warfarin.

A multiple logistic regression was used to adjust or to test the predictive ability relative to a stroke event. The results of the multiple logistic regressions that estimate the odds of a stroke are presented in Table 6. Entropy [R.sup.2], a measure of explained variation, is relatively good at 0.342. AF patients who are not prescribed warfarin are about 6.36 times (P = 0.001) more likely to suffer a stroke than patients who are prescribed warfarin, adjusting for the other factors included in the model. Patients with a diagnosis of CHF in addition to AF were 6.85 times (P = 0.001) less likely to suffer a stroke when compared with AF patients without CHF.

Comments

Hylek et al (7) studied 74 patients with AF who developed embolic strokes despite receiving anticoagulation therapy. They reported minimal risk of stroke in those with INR of 2.0 or higher, with a sharp increase in the rate of stroke with INR below 2.0. (7) Treatment of AF and anticoagulant prophylaxis are complex and time-consuming clinical undertakings that require long-term commitments from the patients and physicians. Review of literature and previous investigations indicates substantial benefit from anticoagulant prophylaxis, including lower rates of thromboembolic events and stroke. (12)

Our results with regard to the risk of stroke in patients with AF are consistent and in line with previously published studies. (7-9,12) Yet, despite substantial improvement in the prevention of strokes, physicians in our region are still reluctant to fully utilize anticoagulant prophylaxis. The frailty of older patients and the presence of comorbid diseases seem to play a significant role in the decision-making process. (13) However, recent evidence clearly shows improvement in outcome, and the recommendation is to administer warfarin, or a combination of warfarin and aspirin, in patients less than 65 years of age with major risk factors, as well as for all patients older than 65 years of age. (3, 14) In our study, however, age was not a major determining factor in the anticoagulation decision-making process. When adjusting for all variables, such as age, gender, and comorbid disease, cardiologists were inclined to prescribe more anticoagulation therapy and utilized higher doses, thus generating higher INR values. Yet, when compared with regional cardiologists and nationally recommended INR targets, neurologists, internists, and others had similar patterns of prescribing lower warfarin dosages, thus generating lower INR values. There were no statistically significant differences in age, gender, and comorbidity among the two groups of patients with and without anticoagulation prophylaxis. Therefore, we cannot fully explain these findings; however, referral patterns and busy practices may play a role. The lower INR target value of neurologists and internists, who provide care to the majority of patients with AF, is also surprising. We can only speculate that the level of comfort with higher doses of warfarin differs in the four groups of prescribing physicians. Perhaps this is only a conjecture; further investigations involving similar sample sizes and communities in different regions may be needed to verify our findings.

Conclusion

To our knowledge, this is the first study that critically investigated the criteria for eligibility for anticoagulation prophylaxis, including prescribing patterns among practicing physicians. However, this study was not designed to initiate anticoagulant prophylaxis by our community physicians, or to study the unwillingness of patients to stay on anticoagulation regimens and to modify their lifestyle. Although gender and race were not relative points in this study, women had good representation and African-Americans were less likely to be started on warfarin prophylaxis.

Our study has some limitations. Many patients' records revealed the presence of internists in addition to other specialists, such as cardiologists or neurologists. We tried to identify the physician of record, prescribing physician, and physician who adjusted the warfarin dosage. However, in a few instances, there were some overlaps between the primary providers in ordering warfarin dosages. Perhaps a larger sample size and another prospective study could answer these questions, and offer more insight into some of these findings.

It seems that despite extensive evidence supporting anticoagulation therapy, both physicians and patients are less inclined to more aggressive stroke prevention measures. Cost, fear of bleeding complications, drug interactions, and time seem to be among the factors influencing decision making for anticoagulation therapy. Perhaps additional educational opportunities and an aggressive public awareness campaign would be timely, and money well spent.

A people that values its privileges above its principles soon loses
both.
--Dwight D. Eisenhower

Table 1. Descriptive statistics of the study population and statistical
tests for group differences

Variables                   Warfarin (n = 124)     Non-warfarin(n = 166)

Age, years
  (mean [+ or -] SD)         72.95 [+ or -] 10.60   74.23 [+ or -] 11.92
  Median
  Range (min, max)          (29, 89)               (42, 98)
  Skewness                   -1.121                 -0.499
  Kurtosis                    2.309                 -0.452
  Non-missing observations  124                    166
INR, units
  (mean [+ or -] SD)          2.51 [+ or -] 2.03     1.19 [+ or -] 0.45
  Median
  Range (min, max)           (0.93, 15.10)          (0.76, 3.65)
  Skewness                    3.475                  3.281
  Kurtosis                   15.664                 13.006
  Non-missing observations  120                     95
Gender
  Male                       58 (46.8%)             74 (44.8%)
  Female                     66 (53.2%)             91 (55.2%)
Race
  Black                      26 (21.8%)             52 (32.9%)
  White                      93 (78.2%)            106 (67.1%)
Comorbidities
  CVA                        30 (24.2%)             22 (13.3%)
  CHF                        37 (29.8%)             45 (27.1%)
  CAD                        28 (22.6%)             30 (18.1%)
  Hypertension               68 (54.8%)             97 (58.4%)
  Diabetes                   31 (25.0%)             43 (25.9%)
Managing Physician
  Specialty
  Cardiology                 38 (30.6%)             28 (16.9%)
  Neurology                  17 (13.7%)             18 (10.8%)
  Internal Medicine          41 (33.1%)             65 (39.2%)
  Other                      28 (22.6%)             55 (33.1%)
Stroke                       24 (19.4%)             62 (37.3%)

Variables                   Statistical tests (b) P value

Age, years
  (mean [+ or -] SD)
  Median
  Range (min, max)
  Skewness                    0.232
  Kurtosis
  Non-missing observations
INR, units
  (mean [+ or -] SD)
  Median
  Range (min, max)
  Skewness                   <0.001
  Kurtosis
  Non-missing observations
Gender
  Male                        0.745
  Female
Race
  Black                       0.043
  White
Comorbidities
  CVA                         0.017
  CHF                         0.610
  CAD                         0.342
  Hypertension                0.541
  Diabetes                    0.861
Managing Physician
  Specialty
  Cardiology
  Neurology
  Internal Medicine           0.020
  Other
Stroke                       <0.001

(a) INR, International normalized ratio: CVA, cardiovascular accident;
CHF, congestive heart failure; CAD, coronary artery disease.
(b) Age and INR were tested using Wilcoxon signed-ranks sum test. All
other variables were tested using Pearson [chi square] test of
association.

Table 2. Correlation Coefficients (a, b)

                                                               Warfarin
                             Age     INR     Warfarin  Stroke  indicator
Variable                     (N)     (N)     (N)       (L)     (L)

INR (N)                       0.053
Warfarin (N)                 -0.143   0.395
Stroke (L)                    0.096  -0.200  -0.372
Warfarin indicator (L)        0.010   0.405   0.866    -0.169
White (L)                     0.195   0.071   0.138     0.039    0.113
Male (L)                     -0.110  -0.048   0.028     0.040    0.017
Cardiologist (L)             -0.220   0.063   0.289    -0.306    0.156
Neurologist (L)               0.093  -0.085  -0.095     0.394    0.051
Internal medicine (L)         0.014   0.020  -0.087     0.095   -0.055
CVA (L)                       0.114   0.000   0.088     0.107    0.144
Coronary artery disease (L)  -0.017  -0.030  -0.010    -0.155    0.056
Chronic heart failure (L)     0.000   0.181   0.120    -0.251    0.028

                             White   Male    Cardiologist  Neurologist
Variable                     (L)     (L)     (L)           (L)

INR (N)
Warfarin (N)
Stroke (L)
Warfarin indicator (L)
White (L)
Male (L)                      0.083
Cardiologist (L)              0.145   0.081
Neurologist (L)               0.032  -0.035  -0.302
Internal medicine (L)        -0.061   0.052  -0.426          -0.311
CVA (L)                      -0.047  -0.032  -0.046           0.127
Coronary artery disease (L)   0.081   0.144   0.172          -0.096
Chronic heart failure (L)    -0.138   0.020  -0.032          -0.105

                                               Coronary
                             Internal          artery
                             medicine  CVA     disease
Variable                     (L)       (L)     (L)

INR (N)
Warfarin (N)
Stroke (L)
Warfarin indicator (L)
White (L)
Male (L)
Cardiologist (L)
Neurologist (L)
Internal medicine (L)
CVA (L)                      -0.049
Coronary artery disease (L)  -0.081   0.000
Chronic heart failure (L)     0.070  -0.026     0.085

(a) INR, international normalized ratio: CVA, cardiovascular accident.
(b) Computation method, N-N pairs: Pearson: L-L pairs: square root of
the linear regression's [R.sup.2]; N-L pairs: square root of the
[R.sup.2] entropy.

Table 3. Group means of warfarin dosage by specialty of the managing
physician: analysis of covariance corrected for multiple pair-wise
comparisons, adjusted for age, gender, and race characteristics of the
patient (a)

Specialty of the
managing physician                          Internal  Other
groups               Cardiology  Neurology  medicine  specialties

Cardiology           2.938       0.0298     0.0054    0.0006
Neurology                        1.758      0.9251    0.5078
Internal medicine                           1.804     0.3001
Other specialties                                     1.423

(a) LS means are the diagonal elements while pair-wise statistical tests
P values are presented on the off-diagonal elements.

Table 4. Multiple logistic regression model of the managing physician's
specialty and patient heart comorbid conditions variables with warfarin
prescription

Independent                 Log odds ratio     Partial likelihood
variable                    (standard errors)  ratio test (p-value)

Intercept                   Not reported        Not reported
Age                         -0.010 (0.012)      0.69 (0.407)
Gender (males vs. females)   -.113 (0.262)      0.28 (0.668)
Race (white vs. black)       0.621 (0.308)      4.19 (0.041)
Cardiology                   0.792 (0.375)      4.54 (0.035)
Neurology                    0.445 (0.436)      1.04 (0.308)
Internal medicine            0.094 (0.322)      0.109 (0.770)
CVA                          0.809 (0.330)      6.13 (0.013)
CAD                          0.177 (0.322)      0.30 (0.583)
CHF                          0.309 (0.787)      1.15 (0.283)

Independent                 Entropy            Odds ratio (95%
variable                    partial [R.sup.2]  confidence interval)

Intercept                   Not calculated     Not calculated
Age
Gender (males vs. females)
Race (white vs. black)
Cardiology                  0.0204             2.883 (1.232, 7.042)
Neurology                   0.0001             1.090 (0.433, 2.760)
Internal medicine           0.0006             0.853 (0.420, 1728)
CVA                         0.0085             1.769 (0.873, 3.704)
CAD                         0.0007             0.848 (0.416, 1.734)
CHF                         0.0087             1.686 (0.889, 3.265)

(a) CVA, cardiovascular accident; CAD, coronary artery disease; CHF,
congestive heart failure.
(b) Entropy [R.sup.2] = 0.047.

Table 5. Unadjusted, conditional, and adjusted odds ratios of a stroke
by managing physician specialty

Managing physician
specialty                  Odds ratio  Warfarin  No warfarin

Cardiologist               1.38         5.3       7.1
Neurologist                3.50        58.8      83.3
Internal medicine          3.05        22.0      46.2
Other physician specialty  3.13        10.7      27.3
Unadjusted                 2.48        19.4      37.4
Adjusted (CMH) (a)         2.91

Managing physician
specialty                  95% CI, lower  95% CI, upper  P value

Cardiologist               0.18           10.48          1.000
Neurologist                0.73           16.85          0.146
Internal medicine          1.26            7.39          0.012
Other physician specialty  0.82           11.89          0.099
Unadjusted                 1.44            4.29          0.001
Adjusted (CMH) (a)         1.55            5.48          0.001

(a) We do not reject homogeneity of the stratified odds ratios, given by
the Breslow-Day test (P = 0.894).

Table 6. Multiple logistic regression model of the patient
demographics, managing physician's specialty, and patients with heart-
related comorbid conditions with stroke event (a)

                         Log odds ratio     Partial likelihood
Independent variable     (standard errors)  ratio test (P value)

Intercept                Not reported       Not reported
Age                       0.009 (0.018)      0.25 (0.618)
White                     0.412 (0.461)      0.81 (0.369)
Male                      0.429 (0.398)      1.17 (0.279)
Not prescribed warfarin   1.850 (0.466)     16.98 (0.001)
% from ideal INR in AF
  patients               -0.001 (0.005)      0.08 (0.777)
Cardiology                1.059 (0.691)      2.53 (0.112)
Neurology                 2.026 (0.630)     11.42 (0.001)
Internal Medicine         1.036 (0.472)      5.03 (0.025)
CVA                       0.701 (0.475)      2.20 (0.138)
CAD                      -0.925 (0.547)      3.05 (0.081)
CHF                      -1.927 (0.507)     17.61 (0.001)

                         Entropy            Odds ratio (95%
Independent variable     partial [R.sup.2]  confidence interval)

Intercept                Not calculated     Not calculated
Age                      0.0009             1.009 (0.974, 1.045)
White                    0.0030             1.510 (0.611, 3.730)
Male                     0.0044             1.535 (0.703, 3.350)
Not prescribed warfarin  0.0638             6.357 (2.549, 15.853)
% from ideal INR in AF
  patients               0.0003             0.999 (0.989, 1.008)
Cardiology               0.0095             0.347 (0.090, 1.343)
Neurology                0.0429             7.583 (2.204, 26.090)
Internal Medicine        0.0189             2.818 (1.117, 7.114)
CVA                      0.0083             2.016 (0.795, 5.114)
CAD                      0.0115             0.396 (0.136, 1.159)
CHF                      0.0662             0.146 (0.054, 0.393)

(a) INR, international normalized ratio: AF, atrial fibrillation; CVA,
cardiovascular accident; CAD, coronary artery disease; CHF, congestive
heart failure.
(b) Entropy [R.sup.2] = 0.342.

Accepted July 25, 2003.

Please see Vishal Bhatia's editorial on page 919 of this issue.

References

1. Fintel, DJ, Hofmann, C. The AFP guide to preventing strokes and lowering health risks in patients with atrial fibrillation [Monograph]. Lisle, Illinois: Northwestern University Medical School and Illinois Academy of Family Physicians; 2000.

2. Report of the Quality Standards Subcommittee of the American Academy of Neurology. Practice parameter, stroke prevention in patients with non-valvular atrial fibrillation. Neurology 1998;51:671-673.

3. Management of Chronic Atrial Fibrillation [Monograph]. Am Fam Physician, Fall 1993;1-23.

4. Laupacin A, Albers G, Dalen J, et al. Antithrombotic therapy in atrial fibrillation. Chest 1998;114:579S-589S.

5. Lightowlers S, McGuire A. Cost-effectiveness of anticoagulation in nonrheumatic atrial fibrillation in primary prevention of ischemic stroke. Stroke 1998;29:1827-1832.

6. Hort RG, Benarente O, McBride R, et al. Antithrombotic therapy to prevent stroke in patients with atrial fibrillation: a meta-analysis. Ann Intern Med 1999;131:492-501.

7. Hylek EA, Skates SJ, Sheehan MA, et al. An analysis of the lowest effective intensity of prophylactic anticoagulation in patients with nonrheumatic atrial fibrillation. N Engl J Med 1996;335:540-546.

8. The Boston area anticoagulation trial for atrial fibrillation investigators. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation. N Engl J Med 1990;323:1505-1511.

9. Stroke prevention in atrial fibrillation study: final results. Circulation 1991;84:527-539.

10. The European atrial fibrillation trial study group. Optimal oral anticoagulant therapy in patients with nonrheumatic atrial fibrillation and recent cerebral ischemia. N Engl J Med 1995;333:5-10.

11. Oates JA, Wood AJJ. Drug therapy. N Engl J Med 1991;324:26:1865-1874.

12. Peterson P, Boysen G, Godtfredsen J, et al. Placebo controlled, randomized trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: the Copenhagen AFASAK study. Lancet 1989;1:175-179.

13. Sudlow M, Thomson R, Thwartes B, et al. Prevalence of atrial fibrillation and eligibility for anticoagulants in the community. Lancet 1998;352:1167-1171.

14. Proceedings of the American College of Chest Physicians 5th Consensus on Antithrombotic Therapy. 1998. Chest 1998;114(5 Suppl):439S-769S.

RELATED ARTICLE: Key Points

* Despite evidence supporting the use of anticoagulation therapy in atrial fibrillation, this modality is not fully utilized.

* Physicians with different specialties have different patterns of prescribing anticoagulants.

* Costs and fear of complications appear to be among the factors influencing anticoagulant prescriptions.

* Additional educational efforts would be both timely and cost-effective.

Ali R. Rahimi, MD, Bobby Wrights, MD, Hossein Akhondi, MD, and Christian M. Richard, MSC 

From the Department of Internal Medicine, Memorial Health University Medical Center, Savannah, GA.

Reprint requests to Ali R. Rahimi, MD, FACP, AGSF, Department of Internal Medicine Education, Memorial Health University Medical Center, 4700 Waters Avenue, Savannah, GA 31404.