Oral Anticoagulation for Nonvalvular Atrial Fibrillation: Are Scientific Recommendations Effective and Safe in Daily Clinical Practice?

INTRODUCTION

The American College of Cardiology/American Heart Association/European Society of Cardiology¹ and the Sociedad Española de Cardiología^2,3 guidelines are clear in recommending oral anticoagulation in patients with nonvalvular atrial fibrillation (NVAF), high risk of embolic complications, and no contraindications for this treatment. This treatment is, nevertheless, frequently underused.^4-12 This may be because clinicians believe the treatment to be less effective and safe in clinical practice than it has been shown to be in clinical trials, in which patients are highly selected and controls are optimal. There have been few observational studies of the effectiveness of oral anticoagulation treatment in non-selected NVAF patients in clinical practice^13-16 and although the majority of those studies^13-15 indicate that the treatment is equally effective in the "real world" as it is in clinical trials, at least one study concluded that the treatment might be less effective in this situation than in clinical trials.¹⁶

We performed a prospective study to assess the effectiveness and safety of anticoagulation treatment used in accordance with the guidelines from scientific societies to reduce the rate of thromboembolic complications in a non-selected population of NVAF patients. We have previously described the usefulness of a treatment protocol in an outpatient clinic to increase the prescription of anticoagulants in this population¹⁷ and we have previously published the results of follow-up in a subgroup of patients aged >=75 years.¹⁸ In the present article, we describe the final results from the full series.

METHODS

Inclusion Criteria

The study included all consecutive patients with permanent nonvalvular atrial fibrillation who attended 2 outpatient cardiology clinics between February 1, 2000 and February 1, 2002. Patients were referred from primary care, the emergency department, or the cardiology and internal medicine departments. The possibility of pharmacological or electrical cardioversion was evaluated for all patients and those in which a reversion to sinusal rhythm was achieved were excluded from the study. Patients with atrial flutter were also excluded.

Study Protocol

The protocols used in this study and those used to guide thromboembolic prophylaxis¹⁷ were based on a consensus reached between the study investigators after analyzing the scientific evidence available at the time and after a review of guidelines for antithrombotic treatment published by the Sociedad Española de Cardiología² before the project was designed. While the study was in progress, the American College of Cardiology/American Heart Association/European Society of Cardiology guidelines for the treatment of atrial fibrillation were published,¹ as were the guidelines from the Sociedad Española de Cardiología for cardiac arrhythmias.³ After a thorough review of both documents, it was decided not to modify the already existing study protocol as it was felt to be consistent with the basic principles of the 3 guidelines.

All patients were given a complete clinical work-up, including clinical history, physical examination, blood tests (hematocrit, glucose, urea, creatinine, sodium, potassium, hepatic enzymes, thyroid hormones, and coagulation study), electrocardiogram, and chest x-ray. When necessary, an echocardiogram was performed in patients suspected of having structural heart disease.

The following were considered to be cardioembolic risk factors: older age (≥75 years), hypertension, diabetes mellitus, previous cardioembolic event, history of ischemic heart disease, heart failure, left atrial dilatation (anteroposterior diameter ≥50 mm) and left ventricular dysfunction (ejection fraction ≤0.45).

Contraindications for anticoagulation were recent severe bleeding, poorly controlled severe hypertension, gastrointestinal disease with a high risk of bleeding, likelihood of poor treatment compliance, severe anemia of unknown origin, high probability of frequent injuries, and the patient's refusal to take the treatment. Lack of treatment compliance was considered likely if there was persistent lack of compliance in the past, illiteracy, severe visual or cognitive deficits with a lack of adequate family or social support. All risk factors and absolute contraindications for anticoagulation were recorded prospectively.

Protocol for Thromboembolic Prophylaxis

In patients with no cardioembolic risk factors or in whom anticoagulation was absolutely contraindicated, treatment with aspirin or other platelet aggregation inhibitors was prescribed, or no treatment, depending on clinician judgment. Patients without an absolute contraindication for anticoagulation treatment and with 2 or more risk factors were offered anticoagulation treatment. Sufficient time was allowed for an explanation of the advantages of treatment in order to avoid patient refusals due to incomplete or inadequate information. In patients with no absolute contraindication for anticoagulant treatment but only one risk factor, the decision as to whether to prescribe treatment was left to the attending cardiologist. The treatment given to each patient was recorded and 2 study groups were defined: anticoagulated patients on the one hand, and the rest of the series on the other. In the latter group, no distinction was made between patients who received platelet aggregation inhibitors and those who did not.

Anticoagulant Treatment and Controls

Patients assigned to anticoagulation treatment were sent to the hospital hematology service. The great majority of those patients received acenocumarol; the remainder received warfarin. Blood extractions to determine the INR (international normalized ratio) were performed in the hospital or in the patient's health care center and the samples were sent to the hematology service, where the results were assessed by expert hematologists who were unaware that the study was being performed. The target INR was generally between 2 and 3, closer to 2 in patients whose only risk factor was age, and closer to 3 in patients who had had a previous embolic event or who had several risk factors. Patients either collected their INR results and treatment plan in the hematology service, or it was sent by fax to the corresponding health care center on the same day.

Follow-Up

Patients were followed with yearly check-ups and the appearance of major events was recorded including stroke, transient ischemic attack (TIA), peripheral embolism, severe hemorrhage, or death through any cause. Diagnosis of stroke or TIA was based on an acute neurological deficit of over or under 24 hours, respectively, which could not be explained by other causes such as hemorrhage, injury or infection, and which was confirmed by a neurologist. Peripheral embolism was defined as the presence of clinical status compatible with the condition and an embolism identified by vascular echography, intraoperative exam, or anatomopathological findings. Confirmation of peripheral embolism by a vascular surgeon was required in all cases. Hemorrhages were considered severe when they required transfusion or hospital admission. On completion of the study, data on patients who had been lost to follow-up were actively sought by mail or telephone interview or by interviewing their primary care clinician.

Statistical Analysis

All baseline and follow-up data were entered in an SPSS database (version 8.0, SPSS Inc. Chicago, Illinois, USA). Quantitative data are presented as means and standard deviations (SD) or median and interquartile ranges for parametric and non-parametric data, respectively. Non-categorical data are presented as percentages. Comparisons between sub-groups were performed using Student's t test for parametric qualitative variables and the Mann-Whitney U test for non-parametric variables. Non-categorical data were compared using the χ² test, and Fisher's exact test was used when needed. The probabilities of an embolic event, mortality, severe bleeding and event-free survival for each treatment group were estimated for the median survival time using the Kaplan-Meyer actuarial method and the log-rank test for sub-group comparisons. We also calculated the probabilities of these events after adjusting for age, sex, and previous embolisms using a Cox analysis model. We also present the unadjusted per 100 patient-year rates for each of the events analyzed. Independent predictors of embolic events, survival or severe bleeding were identified using a stepwise Cox proportional hazard model in which covariables were age, sex, hypertension, diabetes, a previous embolic event, ischemic heart disease, heart failure, left atrial dilatation, left ventricular dysfunction, number of cardioembolic risk factors, and anticoagulant treatment. These variables were selected on the basis of the results from a literature review, and every effort was made not to exclude any known variable with possible prognostic significance. Results are presented as hazard ratios (HR) and corresponding 95% confidence intervals (95% CI). P values of <.05 were considered significant. The analyses were performed using the SPSS statistical package.

RESULTS

General Characteristics and Symptoms

A total of 624 patients were included from February 1, 2000 to February 1, 2002. Mean age was 73 (SD, 8) years and 43% were male. Of the sample, 50% were aged 70-79, 25% were aged 60-69, 20% were around 80 years old, and 5% were under 60; 73% of the patients were asymptomatic, 23% had dyspnea, 2% had palpitations, and 2% had chest angina.

Cardioembolic Risk Factors and Contraindications for Anticoagulation

Nonvalvular atrial fibrillation was most frequently associated with arterial hypertension (60%) followed by the absence of structural heart disease (23%), ischemic heart disease (8%), cardiomyopathy (7%), and other heart disease (2%). The most frequent risk factors were hypertension, older age and diabetes (Table 1). A total of 48 patients (8%) did not show any risk factors, 154 patients (25%) had one risk factor, 241 patients (38%) had 2 factors, 118 patients (19%) had 3 factors, and 63 patients (10%) had 4 or more risk factors. Therefore, 576 patients (92%) had one or more cardioembolic risk factors. Of these, 115 (20%) had at least one absolute contraindication for anticoagulation (table 2), the most frequent being a high risk of non-compliance with treatment.

Percentage of Anticoagulation

Of the 461 patients with one or more risk factors and no contraindication for anticoagulation, 425 patients actually received anticoagulant treatment (92% in this group, 68% of the overall series). A total of 199 patients did not receive anticoagulation: 115 because of contraindications, 48 because of an absence of cardioembolic risk factors, and 36 as a result of a decision by the attending cardiologist. The majority of the latter group had only one cardioembolic risk factor (32/36; 89%).

General Characteristics of the Treatment Group

Anticoagulated patients more frequently had hypertension (75% vs 53%; P<.001), diabetes (28% vs 15%; P=.001), previous embolism (19% vs 5%; P<.001) and ischemic heart disease (11% vs 4%; P=.003), as well as a higher number of cardioembolic risk factors (2.2±1.1 vs 1.6±1.2; P<.001) and a lower frequency of NAVF without structural heart disease or hypertension (15% vs 40%; P<.001), even though the mean age was lower (72±7 vs 75±9 years; P<.001). There were no statistically significant differences between groups in terms of proportion of males, heart failure, left atrial dilatation, or left ventricular dysfunction. Overall, 93% of nonanticoagulated patients received platelet aggregation inhibitors (92% received aspirin).

Events During Follow-Up

After a median follow-up of 21 months (interquartile range, 10-33 months), data were available for 593 patients (95%). Of these, 407 were anticoagulated (729 patient-years) and 186 were not (293 patient-years). In spite of a more unfavorable cardioembolic profile, the probability of an embolic event (TIA, stroke, or peripheral embolism) and death, estimated for the median time of follow-up, was lower in anticoagulated patients, with no significant differences in the probability of severe bleeding (Table 3). The results remained unchanged after adjusting for differences in age, sex, and previous embolic events using the Cox proportional hazard model (Table 3). The majority of events in nonanticoagulated patients were TIA or non-fatal stroke, although the rate of fatal stroke was also higher in this group (Table 4). On admission, the INR in the 6 anticoagulated patients with embolic events was between 1.7 and 2.7. The majority of non-fatal severe hemorrhages were of gastrointestinal origin in both groups (9/14; 64%), but there were also 2 non-fatal intracraneal hemorrhages in anticoagulated patients. There were 2 fatal episodes of bleeding, with 1 of gastrointestinal bleeding in an anticoagulated patient and 1 cerebral hemorrhage in a nonanticoagulated patient. This last patient had not received platelet antiaggregation treatment, but had had a prior episode of intracranial bleeding. On admission, the mean INR of the 10 anticoagulated patients with severe bleeding was 4.4±2.1 (range, 2.7-8.8). The majority of deaths from other causes in both groups were due to cardiovascular disease (8/26; 31%) or neoplasms (6/26; 23%). Survival time free of embolism or severe bleeding was significantly better in anticoagulated patients (Figure).

Figure. Probability of survival free of embolism or severe bleeding, by study group.

Independent Predictors of Embolic Events

The multivariate analysis showed that anticoagulant treatment (HR=0.07; 95% CI 0.03-0.19; P<.0001) and previous embolism (HR=5.63; 95% CI, 2.19-14.50; P=.0003) were independent predictors of embolic events. These 2 factors, and age, also independently predicted mortality (HR=0.42; 95% CI, 0.20-0.92; P=.03 for anticoagulation; HR=2.59; 95% CI, 1.10-6.10; P=.03 for previous embolism, and HR=1.10; 95% CO, 1.04-1.15; P=.0006 for age). Being male (HR=6.26; 95% CI, 1.70-23.05; P=.006) and age (odds ratio=1.09; 95% CI, 1.01-1.18; P=.04) were independent predictors of severe bleeding.

DISCUSSION

The effectiveness of oral anticoagulation for NAFV in daily clinical practice has been analyzed in previous studies. Abdelhafiz et al¹⁵ followed 402 patients for 19 months in an anticoagulation clinic and found similar rates of stroke, severe bleeding and anticoagulation control as those observed in clinical trials. They did not include nonanticoagulated patients, however. Aronow et al¹³ reported on the 3 year follow-up of 350 patients with a mean age of 83 years, 40% of whom were anticoagulated. They found a significant reduction in the incidence of cardioembolic stroke (34% vs 64%; P<.0001), and no increase in the rate of severe bleeding. Go et al¹⁴ described a reduction of 43% (51% after adjusting for confounding factors) in the risk of stroke and peripheral embolism (1.16 vs 2.03 per 100 patient-years) in a large series (>11 000 patients) of selected patients who were followed using several data bases in an integrated health-care system in northern California. In this case, the mean follow-up was 2.2 years and 55% of the patients were anticoagulated. On the other hand, in a population based study with a similar design carried out in Denmark and which included >5 000 patients, 27% of whom were anticoagulated, Frost et al¹⁶ reported that the effectiveness of oral anticoagulation in clinical practice could be less than that observed in clinical trials. When compared with these studies, our study is the largest, single-center, observational cohort in patients with NAVF which has analyzed major events related with the use of anticoagulants. As far as we are aware, there have been no similar studies in Spain and the present study supports the notion that oral anticoagulation is safe and effective in non-selected patients with NAVF in daily clinical practice. Although the anticoagulated patients had a worse cardioembolic profile, the rate of embolisms and mortality were significantly lower than in nonanticoagulated patients, and there was no significant increase in the rate of hemorrhage. These results remained unchanged after adjusting for age, sex, and previous embolism.

The extreme results obtained with anticoagulation in the present study are surprising. However, in a meta-analysis of 6 randomized trials based on "effectively administered treatment,"¹⁹ the relative risk reduction in embolic events was over 80%, which is not very different to the results observed here.

In the present study, we observed a crude rate of embolic events of 7.16 per 100 patient-years in nonanticoagulated patients (Table 4). The placebo group in the meta-analysis by the Atrial Fibrillation Investigators²⁰ had a rate of embolic events (stroke, TIA or embolism) of 6.3 per 100 patient-years. That was in a selected population comprising patients from 5 clinical trials, with a lower frequency of cardioembolic risk factors than in the present series (20% of patients >75 years of age, 46% with hypertension, 15% with diabetes and 6% with prior cerebral embolism). Incidence was 64% after 3 years in Aronow et al's study,¹³ which might be explained by the fact that patients in that study were older, and that there was a high rate of previous stroke (44%). The low incidence observed by Go et al¹⁴ (only 2.03 cases per 100 patient-years in nonanticoagulated patients) is surprising, but could be explained by a low rate of cardioembolic risk factors (48.6% of hypertensive patients, 15.1% of patients with diabetes, and 4.2% with prior cerebral embolism). There were also methodological differences in Go et al's study, including the fact that they did not include TIA among the embolic events, data collection was performed via databases and mortality registers, but not through a follow-up of patients in the clinic. The population included in our study therefore presents an intermediate level of embolic risk compared with the other studies mentioned.

We considered age >=75 years to be a cardioembolic risk factor. However, age is a continuous variable in terms of risk, such that risk increases with age²⁰; in the recommendations from scientific societies, there are discrepancies regarding the cut-point at which age should be considered an embolic risk factor.^1-3 After analyzing these recommendations, it was agreed that 75 would be used as the cut-point for treating age as a risk factor.

A significant proportion of patients (n=199, 32%) did not receive anticoagulants, the majority (n=115) because anticoagulation was contraindicated, principally because the patient was considered unlikely to adhere to treatment (47%). This may be explained by the fact that the study had a high proportion of very elderly patients living alone, or who had visual or cognitive deficits which would make compliance with treatment difficult. Many of the patients also lived in relatively deprived areas. Nevertheless, before rejecting outright the possibility of anticoagulant treatment, efforts were made to provide support mechanisms through relatives or social and health-care services, though in some cases the final decision was that anticoagulant treatment was not appropriate. Lack of treatment compliance is a predictor of excessive anticoagulation in ambulatory patients with heart disease, and increases the risk of bleeding.²¹ However, the 20% of patients with contraindications in the present study compares favorably with the 34% of patients considered unsuitable for anticoagulation in the European Atrial Fibrillation Trial.²² A further 48 patients did not have cardioembolic risk factors and 36 patients were not anticoagulated on the decision of the attending clinician, the majority because they only had one risk factor.

Although guidelines from the Sociedad Española de Cardiología^1,3 recommend anticoagulation in this sub-group, recommendations from other scientific bodies, such as the American College of Cardiology/American Heart Association/European Society of Cardiology² guidelines, recognize that anticoagulation may not always be appropriate in intermediate risk patients, i.e. in patients with only one of the following risk factors: age 65-75 years, diabetes, ischemic heart disease, and hypertension. Patients with only one risk factor have the lowest level of embolic risk of all of those with at least some risk factors² and therefore obtain the least absolute and relative benefit from treatment. Clinician judgment therefore plays an important role in this group, and the clinician needs to take minor contraindications and patient preferences into consideration, aspects which would not contraindicate anticoagulation if the risk was greater.

This study had some limitations. In the first place, it is impossible to avoid some subjectivity when assessing contraindications, especially when deciding about likely adherence to treatment. On the other hand, a small percentage of patients with an indication for anticoagulation and no absolute contraindication for the treatment did not, in the end, receive anticoagulation. Although the majority only had one cardioembolic risk factor, making the need for anticoagulation arguable, there may have been patients in this group who would have benefited from treatment indicating a need to improve prescription of anticoagulation treatment in this group. Thirdly, INR data were not available for the whole of the study sample. It would have been useful to have this information when judging the rate of events observed. Finally, this was an observational cohort study, and is therefore subject to certain biases; nevertheless, the data obtained are consistent with evidence from clinical trials and other observational studies.^13,14

Three new approaches to thomboembolic prophylaxis in NVAF patients have shown promising results recently: the thrombin inhibitor ximelagatran,²³ or combined therapy with antiplatelets and anticoagulants,²⁴ and treatment with a combination of aspirin and clopidogrel.²⁵ Until these new treatments have been thoroughly evaluated and become firmly established, the best way to reduce embolic events in patients with NVAF will continue to be through the judicious use of oral anticoagulation following guidelines from scientific societies.

CONCLUSIONS

In spite of its limitations, we believe that this study helps to confirm that oral anticoagulation, when administered following the recommendations from scientific societies and with the appropriate controls by the hematology service, is effective and safe for the prevention of thromboembolic complications in patients with NVAF in daily clinical practice.

ACKNOWLEDGEMENTS

We would like to acknowledge the patient work of nurses Mercedes Ortiz Funes and Inmaculada Muñoz Álamo, without whom this study would not have been possible.

This study was awarded a research grant from Fundación Mapfre Medicina.

---

Correspondence: Dr. M. Ruiz Ortiz.
Huerta del Recuero, 1, portal 7, 3.^o, 2. 14011 Córdoba. España.
E-mail: [e-mail]

Received November 15, 2005.
Accepted for publication March 15, 2006.