Atrial fibrillation (AF) is the most common arrhythmia in the general population, with an estimated prevalence of about 2%, a figure that rises with age and comorbidities.1 Several studies have recently been conducted in Spain to understand the situation of AF in clinical practice. The VAL-FAAP study2 analyzed nearly 120 000 patients attended at primary care centers and observed that 6.1% of patients had AF. In the general population older than 40 years, the prevalence of AF in Spain was 4.4%.3 Among patients with hypertension and older than age 65 years in the Valencian Community, 10.3% had AF.4

Compared with patients without AF, patients with AF have a 2-fold risk of death and up to 5-fold risk of stroke.5 Atrial fibrillation-related stroke has a higher mortality and a higher risk of recurrences and tends to result in more sequelae.6 In most patients with AF, long-term oral anticoagulation is indicated to prevent thromboembolic complications.2 For this purpose, vitamin K antagonists (VKAs) have been widely used for decades because they can reduce the risk of stroke by about 64%.7

However, VKAs have important limitations that usually condition their use in clinical practice for nonvalvular AF (NVAF).2,8 These include narrow therapeutic window, drug and food interactions, and variable metabolism, which require regular follow-up of anticoagulation status and frequent dose titration.9,10 It is crucial that the international normalized ratio (INR) of patients receiving VKA therapy be within therapeutic range to reduce the risk of thromboembolic and hemorrhagic complications. In fact, time to onset of stroke has been shown to significantly improve in warfarin-treated patients with NVAF and CHADS2 score ≥ 2 compared with untreated patients, but only in treated patients who were within therapeutic range > 70% of the time.11

Therefore, to manage this population adequately, it is essential to identify the degree of INR control among patients with NVAF who are receiving VKA anticoagulation therapy. Several studies12,13 have investigated the degree of INR control in a specific geographic area of Spain or using only a small number of INR values. However, the PAULA study (Perspectiva Actual de la sitUación de la anticoaguLación en la práctica cínica de Atención primaria [Current perspective of anticoagulation in clinical practice in the primary care setting]) was conducted to understand the situation of anticoagulation control over a long period among patients with NVAF who are receiving VKAs in primary care clinical practice throughout Spain. This study specifically looked at anticoagulation in the clinical practice setting at primary care centers.

The PAULA study was based on an observational cross-sectional/retrospective, multicenter, national design, and its main aim was to identify anticoagulation control of patients with NVAF who were receiving anticoagulation VKA therapy at primary care centers in Spain last year. The study had the scientific backing of three Spanish primary care societies (SEMERGEN, semFYC, and SEMG).

To perform the study, a scientific committee was formed with 2 cardiologists, 1 biostatistician, and 3 general practitioners who were experts in cardiovascular disease and represented the 3 Spanish primary care societies. The general practitioners chose 9 regional coordinators (Appendix 1 of the supplementary material), who in turn selected 139 investigators (Appendix 2 of the supplementary material) from 99 health centers located throughout the different autonomous communities (except for La Rioja, for logistical reasons) according to the proportion of regional inhabitants, thus ensuring numbers of patients representative of the national territory (Table 1). The regional coordinator initially selected the investigators based on their clinical and research skills, and the choice was then approved by the scientific committee. Each investigator had to include at least the first 10 consecutive patients who met all inclusion criteria and none of the exclusion criteria, who came for routine anticoagulation follow-up, and who gave consent to participate in the study. All patients were recruited between February and June 2014.

The inclusion criteria were as follows: a) patients of either sex and 18 years of age or older; b) patients with NVAF who had been receiving VKA therapy for at least the past year at a primary health care center under routine clinical practice conditions; c) patients for whom at least 80% of INR controls were available from the past year, and d) patients who had given written informed consent to participate in the study after they had read and understood the patient information sheet. Patients were excluded if they had cognitive impairment that prevented them from correctly understanding the patient information sheet or informed consent or if they had participated in any clinical trial in the past 12 months.

The study comprised a single visit, that coincided with one of the patient's regular follow-up visits. The data were collected from the medical history and physician interview, but there was no study-specific diagnostic or therapeutic intervention. The data were entered in an electronic case report form. All electronically collected information was checked and reviewed to safeguard the quality of the registry, and the database was cleaned to avoid recording any impossible values. The case report form was designed by a specifically contracted CRO (Contract Research Organization) —Dynamic Solutions—under the supervision of the scientific committee. The CRO was responsible for ensuring the accuracy and quality of the data collected. To handle this, specific telephone follow-up and on-site monitoring were performed as needed. In addition, lists of all data with potential issues were generated and forwarded to the sites and investigators involved.

The following variables were collected: sociodemographic data, relevant medical history, comorbidities, cardiovascular events, including thromboembolic complications and major hemorrhagic events, physical examination, anthropometric data, blood work available from the previous 6 months, AF information (diagnosis date and type of AF), information on AVK therapy (drug and total weekly dose), concomitant treatments, CHADS2, CHA2DS2-VASc, and HAS-BLED scores, INR readings (including date and value), and investigator opinion on patient anticoagulation and possible risk factors associated with worse INR control. Most of the variables analyzed were collected from the patient's medical history. However, some were measured at the visit itself, in particular blood pressure, heart rate, and body mass index. To determine if patients had dietary habits that could affect INR control, patients were specifically asked if they consumed large amounts of foods rich in vitamin K (cereals, broccoli, cabbage, carrots, etc), alcohol, cranberry juice, or ginseng, if they were regular users of phytotherapy, or if they had frequent dietary transgressions. Excessive alcohol intake was considered to be more than 8 alcoholic beverages per week (1 alcoholic beverage was considered to be 10g of pure ethanol).14

To assess patients’ INR control, the time within therapeutic range in the past 12 months was calculated at a centralized site by both the direct method (percent time with INR values within therapeutic range) and by the method described by Rosendaal et al.15 Labile INR was considered to be time within therapeutic range < 60%, according to the definition proposed by Pisters et al.16

The study was approved by the Clinical Research Ethics Committee of the Hospital Universitario la Paz of Madrid. All patients were required to sign the written informed consent before inclusion, once they had read and perfectly understood the patient information sheet.

The study included 1561 patients, 37 of whom were excluded from the final analysis for different reasons. The causes for exclusion were time from first INR measurement to study start < 1 year (n = 34), history of AF < 1 year (n = 9), improperly recorded data in the electronic case report form (n = 6), and documentation of an insufficient number of INR readings (n = 5). In some cases, more than 1 cause was given for study exclusion. Finally, 1524 patients were included for assessment.

Patients’ baseline characteristics are summarized in Table 2. There was a high number of associated cardiovascular risk factors (80.2% had hypertension, 63.5% was sedentary, and 57.2% had dyslipidemia) and factors that raise the risk of both thromboembolic complications (23.9% had a history of heart failure, 13.7%, stroke/transient ischemic attack, and 9.6%, myocardial infarction) and hemorrhagic complications (12.3% had known labile INR; 8.8%, history of bleeding, anemia, or predisposition to bleeding, and 6.0%, renal failure). A total of 18.4% of patients had dietary habits potentially affecting anticoagulation control with VKAs, 7.7% were taking long-term nonsteroidal antiinflammatories, and 7.5% were receiving other antithrombotic drugs. The total mean number of tablets taken by patients per day was 7.0 ± 3.8. Table 3 lists the most relevant analytical parameters.

The mean time from AF diagnosis to study date was 6.0 ± 4.2 years. The type of AF was paroxysmal in 20.1%, persistent in 13.6%, and permanent in 64.2%; the type of AF was not documented in the remaining 2.1%. Echocardiograms were available for 81.3% of patients. A total of 75.8% of patients had a CHADS2 score ≥ 2 (mean CHADS2, 2.3 ± 1.2); 94.5% had a CHA2DS2-VASc score ≥ 2 (CHA2DS2-mean VASc, 3.9 ± 1.5), and 13.0% had HAS-BLED ≥ 3 (mean HAS-BLED, 1.6 ± 0.9) (Table 4).

Acenocoumarol was prescribed in 94.8% of the patients (n = 1445) and warfarin in 5.2% (n = 79). Anticoagulation follow-up was performed by primary care alone in 70.7%, by hematology in 8.1%, by primary care and hematology jointly in 20.9%, and by cardiology in 0.3%. The total number of INR records in the previous 12 months was 21 982, and the mean number of INR readings recorded per patient in the past year was 14.4 ± 3.8. The percent time within therapeutic range was 63.2% ± 17.9% according to the direct method and 69.0% ± 17.7% according to the Rosendaal method. A total of 56.9% of patients had adequate INR control according to the direct method and 60.6% according to the Rosendaal method (Figure). In 65.0% of patients, the physician considered that the patient had good INR control. Among patients who had poor INR control calculated by the direct method, 30.4% of INR measurements were below 2 and 25.5% were above 3. According to the Rosendaal method, these levels were 30.2% and 25.6%, respectively. In 30.5% of INR readings, the anticoagulant dose was modified. The dose was titrated manually in 46.4% and automatically/mixed (some kind of assistance or help) in the rest.

Figure.

Percent time within therapeutic range and international normalized ratio according to the direct method and the Rosendaal method. INR, international normalized ratio.

(0.08MB).

The variables associated with INR control are listed in Table 5 (bivariate analysis) and Table 6 (multivariate analysis) according to the direct method and the Rosendaal method, respectively. In the multivariate analysis, both the direct method and the Rosendaal method found that men had better control than women, that patients with dietary habits potentially affecting VKA anticoagulation were less likely to be well controlled, that patients with poor control took a higher number of tablets per day than those with good control, and that patients history of known labile INR were less likely to be controlled than those who had no known history of labile INR.

The results obtained in the PAULA study indicate that INR control in the primary care setting in Spain could clearly be improved. The study included a broad sample that was representative of the entire Spanish population with NVAF attended in primary care, with a high number of INR controls and a long follow-up period. In this study, monitoring of the work was very meticulous and rigorous controls were applied to ensure the quality of the recorded data. Additionally, the study is an example of a joint effort between primary health care physicians and cardiologists and of cooperation between scientific societies, as it is the first time in Spain that a study of these characteristics was backed by the 3 Spanish primary care societies.

The data collected in the PAULA study showed a mean patient age of 77 years; approximately half were women. A total of 80% had hypertension, 24% had heart failure, 14% had a history of stroke or transient ischemic attack, and 10% had a history of infarction. These data indicate that patients with NVAF who are receiving anticoagulation therapy in Spain are relatively older and have a significant number of associated comorbidities. However, compared with other populations with AF analyzed in Spain, particularly those attended in cardiology clinics, the proportion of patients with certain diseases, such as ischemic heart disease or renal failure, was somewhat lower in the PAULA study.2,3,12,13,20 The mean glomerular filtration rate of patients included in the PAULA study was 73.1mL/min/1.73 m2. In populations at higher risk, anticoagulation control is probably even worse.

Regarding the type of AF, several studies have shown that the most common form in patients attended in primary care is permanent AF (45.3% in VAL-FAAP, 77.5% in FIATE, and 64.2% in PAULA).2,13 Therefore, for most patients with NVAF in primary care, it seems that the therapeutic objectives should focus mainly on controlling heart rate and lowering thromboembolic risk.

Risk assessment for stroke and bleeding is fundamental in NVAF. In recent years, the use of scales has become more widespread for both thromboembolic CHA2DS2-VASc) and hemorrhagic risk stratification (HAS-BLED).21 In the VAL-FAAP study, 67.4% had a CHADS2 score ≥ 2 and 85.7% had a CHA2DS2-VASc score ≥ 2.2,8 In the ANFAGAL study12, the CHADS2 score was 2.3; CHA2DS2-VASc, 3.8 and HAS-BLED, 3.1. In the PAULA study, the figures were 2.3, 3.9, and 1.6, respectively. Therefore, anticoagulation indications appear to be on target. The difference in the HAS-BLED score between the ANFAGAL and PAULA studies may be related to the age difference of patients in the 2 studies because the inclusion criteria were somewhat different.

The guidelines21 recommend that patients with AF undergo an echocardiogram as part of the initial study. In the FIATE study,13 67.6% of patients had at least 1 echocardiogram whereas, in the PAULA study, this figure was 81.3%. This finding is highly relevant, as it indicates an improvement in the clinical management of patients with AF in primary care.

Routine INR control was conducted by primary care centers in 72% of patients in the FIATE study.13 In the PAULA study, anticoagulation control was conducted by the primary care physician in 70.7% of patients and jointly by primary care and hematology in 20.9%. Consequently, both studies found that INR control is essentially handled, as is common in Spain, by primary care. However, the prominence of primary care was even stronger in the PAULA study. This is key because although this approach has some important methodological limitations, costs may be lower when INR is controlled at the initial level of health care.22

Inadequate INR control raises the risk of both stroke and bleeding.17,23,24 Therefore, good control is essential. In clinical trials performed with the new direct-acting anticoagulants, the mean time within therapeutic range was 55% to 68.4%.25-28 However, because patients included in clinical trials have restrictive inclusion and exclusion criteria and very rigorous follow-up, the results cannot always be extrapolated to “real life.” In this context, studies conducted in the clinical practice setting provide extremely relevant information.29,30

The ANFAGAL study12 observed a mean of 13.7 readings per patient, 41.5% had less than 60% of controls within therapeutic range, and 42.7% showed a time within therapeutic range < 65%, estimated according to the Rosendaal method. In the FIATE study,13 in patients who were receiving VKA therapy and had only the last 3 INR readings, 66% were within therapeutic range at the last determination and 33% were within range at all 3 readings. Importantly, the ANFAGAL study had a small patient sample and was performed in a single geographic area (Galicia), and the FIATE study had a low number of INR readings. Consequently, the PAULA study data are of considerable interest. In that study, 43.1% of patients had a time within therapeutic range in less than 60% of controls done in the past year and 39.4% had a time within therapeutic range < 65%, with a mean of 14.4 readings per patient. In a meta-analysis of 95 articles, patients receiving anticoagulation VKA therapy remained within therapeutic range 61% of the time, and the control observed in clinical trials or specialized clinics was better than in other settings.31 Therefore, anticoagulation control in Spain is evidently poor, as has been observed in other countries. Unlike other chronic conditions, such as hypertension (where one of the reasons for poor control is a noticeably wrong perception of control), anticoagulation was seen in the PAULA study to be poorly controlled due to the inherent difficulties of maintaining the patient within an adequate therapeutic range, rather than an incorrect perception by the physician.32

Various studies have investigated the factors associated with worse anticoagulation control. The ANFAGAL study observed that poor INR control was more common in patients with high HAS-BLED scores, diabetes mellitus, renal disease, and hypertension.12 In a study performed in the United States, predictors of poor control included alcohol abuse, multidrug therapy, and multiple hospitalizations.33 In another study, the variables associated with good control were regular vitamin K intake, male sex, duration of anticoagulant treatment > 2 months, adequate family support, adequate functional and cognitive ability, and no regular alcohol intake.34 In the PAULA study, the predictors of poor control were history of labile INR, female sex, dietary habits potentially influencing VKA effects, and multidrug therapy. It is important to identify these factors to ascertain which patients may require closer INR control, including those who, at least a priori, may be more likely to potentially benefit from the use of the new direct-acting anticoagulants. For instance, female sex has been associated with worse control in several studies,34,35 probably due to differences in clinical characteristics between the male and female populations.

In clinical practice in Spain, INR control could be greatly improved, as approximately 40% of patients are not well controlled. Predictors of poor INR control were history of labile INR, female sex, dietary habits potentially influencing the effect of VKAs, and multidrug therapy.

This study was sponsored by Bayer Hispania S.L.; however, this sponsor had no influence over the course of the study or the collection or interpretation of the results, even though 2 of the authors are members of Bayer's medical department.