Oral anticoagulants are the treatment of choice for the prevention of strokes and mortality in patients with nonvalvular atrial fibrillation (NVAF) who are at significant thromboembolic risk according to the CHA2DS2-VASc scale.1 Although vitamin K antagonists (VKA) lower the risk of stroke by up to 64% compared with placebo,2 they require careful regular monitoring with frequent dosage adjustments due to a narrow therapeutic margin. There is solid evidence that patients with poor anticoagulation control throughout treatment (time in therapeutic range [TTR] <65% according to the Rosendaal method) are at higher risk of stroke, bleeding complications, and death3–6 than patients with adequate control.

In Spain, the prevalence of NVAF is around 4.4% in persons aged ≥ 40 years and rises gradually with age, reaching 17.7% in patients older than 80 years.7 According to the 2014 European PREFER registry on thromboembolic events, around 80% of patients with NVAF in Spain are receiving VKA therapy,8 although in recent years this percentage has dropped due to the use of new direct-acting anticoagulants.

Several publications have reviewed anticoagulant control in patients with NVAF and have identified a high number of patients with poor control, with a prevalence of 39.4% to 57.2% (TTR <65% according to the Rosendaal method),9,10 leading to a higher risk of stroke, bleeding complications, and all-cause mortality.11 In terms of economics, the CONOCES12 and CODICE13 studies provided data on the high cost of NVAF complications (stroke, hemorrhagic stroke, acute cardioembolic cerebral infarction, and major bleeding). Although several Spanish studies report data on the real-world situation of NVAF patients treated with VKA and the cost of their complications, no information is available on the potential socioeconomic impact of poor anticoagulant control in patients treated with VKA in the Spanish National Health Service (NHS).

The aim of this study was to evaluate the burden and cost of clinical events for the Spanish NHS that could arise due to poor anticoagulation control (determined by TTR according to the Rosendaal method) in NVAF patients treated with VKA.

The hypothetical cohort for this study included 594 855 patients with a mean age of 75.4±4.3 years (48.9% women) at high cardioembolic and bleeding risk: 74.4% with CHADS2 ≥ 2 (91.8% if CHA2DS2-VASc scale ≥ 2 is used) and 58.4% with HAS-BLED ≥ 2. The mean prevalence of poor anticoagulant control was 48.3% (n=287 089), between 39.4% (n=234 373) and 57.2% (n=340 257).

Compared with adequate control, poor anticoagulant control was associated with an increase of 4557 events per year (2321 ischemic strokes and 2236 major bleeds) and 14 463 all-cause deaths (table 3).

The annual cost of events due to poor anticoagulation control, considering only direct health costs, was €29 578 306 (€116 /patient-year), mainly due to the cost associated with ischemic stroke (€20 979 365/year), which accounted for 70.9% of event costs.

When considering the societal perspective (including direct nonhealth costs and indirect costs), the impact of poor anticoagulant control was €75 737 451 (€293/patient-year), with ischemic stroke accounting for an even larger portion of the cost (€67 138 510/year; 88.6%) (table 4 and figure 3). The impact on mortality for the entire population analyzed was 170 502.05 PYLL (0.287 PYLL/patient-year).

Figure 3.

Cost distribution, according to the perspective used. A, National Health Service perspective. B, societal perspective.

(0.07MB).

The results of this study, the first in Spain to analyze the costs associated with oral anticoagulation quality, reveal the clinical and economic impact of poor control among patients receiving VKA therapy, a common clinical situation in Spain9,10,17–24 and a risk factor for stroke in patients with NVAF.8 Another relevant finding of the analysis is the estimated mortality among these patients, expressed as PYLL.

The analysis started with the prevalence figures according to age bracket from the OFRECE,7 CARDIOTENS,29 PREV-ICTUS,30 and AFABE31 studies, but the prevalence data from Gómez-Doblas et al.7 were finally used because the study population was the most relevant. The percentage of VKA use was taken from the Val-FAAP study,15 which described a population receiving oral anticoagulants in a similar sample with similar characteristics to the cohort analyzed (population with CHADS2 ≥ 2, 74.4%). Although this study did not specify the type of anticoagulant therapy, it is assumed to be VKA in all patients because the study was conducted between 2009 and 2010, when direct-acting oral anticoagulants were rarely used. However, although the European PREFER registry has reported on VKA usage in Spain (80%),8 the present study was finally based on the levels from Barrios et al.,15 because the patients in this study were at high cardioembolic risk (CHADS2 ≥ 2), similar to the cohort analyzed.

This analysis included studies evaluating the quality of anticoagulant therapy based on TTR according to the Rosendaal method using 65% as a cutoff point, as the method is considered most orthodox, is widely used in our field,21 and is recommended by health authorities.32 The PAULA9 and ESPARTA10 studies provided the extreme values for the analysis.

Despite the magnitude of the levels analyzed, the estimated impact of poor anticoagulant control could be lower than the real-world impact. According to the AFABE study31 results, 23.5% of patients with known AF and moderate-to-high cardioembolic risk (CHA2DS2-VASc ≥ 2) did not receive anticoagulant therapy and, therefore, a large patient population would not be included in the analysis. However, no cost was assigned to mortality, as no data were found on the amounts. In addition, the direct nonhealth and indirect costs of systemic embolism and major bleeding have not been identified and, therefore, these costs were not included in the analysis.

The limitations of this analysis include aspects related to calculating the target population (NVAF prevalence) and event rates. First, prevalences were applied to each age bracket (10-year intervals) from the OFRECE study7 to the population prevalences taken from the Spanish Statistics Institute (5-year intervals), which could underestimate the prevalences in the older groups of each bracket. Second, in the real-world study conducted in Spain by Rivera et al.,11 event rates were limited to ischemic stroke, major bleeding, and all-cause mortality, unlike the study used in the sensitivity analysis (post hoc analysis of the SPORTIF III and V clinical trials performed in 23 countries in Europe, Asia and Oceania, United States, and Canada),3 which also included ischemic stroke and systemic embolism rates. Third, questions could be raised regarding the suitability of the White et al.3 study for the sensitivity analysis, as it was an international multicenter study with a different definition of poor anticoagulant control from ours (poor control, TTR <60%; good control, TTR> 75%). However, the differences in rates used in the 2 studies were very similar between the groups (difference between adequate and poor control in ischemic stroke, 0.89% vs 0.82%; major bleeding, 0.93% vs 2.27%; mortality, 5.48% vs 2.51% in the base case and sensitivity analysis, respectively). Hence, it can be stated that the event and mortality rates used are valid in our model. Nevertheless, although the prevalence of NVAF has been applied by age bracket, the same exercise could not have been undertaken with the prevalence of poor control and the event rates because the information was not segregated by age. Furthermore, it would have been interesting to evaluate the clinical and economic impact of poor anticoagulant control according to possible differences between various regions of Spain, as mentioned in the PAULA study.33 Last, the analysis was limited to a 1-year time horizon, considering the 1-year unit costs of ischemic stroke and hemorrhage for these events (CONOCES study),12 which may not be representative of the entire natural course of the disease.

Despite the above-mentioned limitations, the study results underscore the high socioeconomic impact of poor control of VKA therapy in Spain, essentially based on acenocoumarol (although no differences were observed in the quality of anticoagulation control or events compared with warfarin).34 This poor control may have various causes: the presence of comorbidities (eg, kidney failure, type 2 diabetes mellitus, hypertension), factors such as labile international normalized ratio (INR), high bleeding risk, female sex, dietary habits, multidrug therapy, and prolonged use of antiplatelet agents.10,17,21,35 Obviously, nonadherence is a key factor (as with all long-term therapy), as observed in the REACT-AF study, which observed adherence> 80% in only 34.9% of patients.36 The last possible cause is therapeutic inertia, reported in studies such as ESPARTA, in which therapy was not switched to a direct-acting oral anticoagulant in 65.6% of NVAF patients at high risk of stroke (CHA2DS2-VASc ≥ 2) and treated with VKA who did not maintain INR within therapeutic range.10

Irrespective of these considerations, poor anticoagulant control in Spain, in addition to affecting NVAF patients by compromising their clinical prognosis and producing higher mortality and disability, makes the NHS less efficient. The results of this study reveal the clinical and economic impact of poor anticoagulant control in NVAF patients treated with VKA, highlighting the potential importance of a change in clinical practice to lower the disease burden.

This study shows that higher rates of thrombotic, bleeding events, and mortality among patients with poor anticoagulant control treated with VKA represent a cost of more than €75 million per year (from the societal perspective) and a high impact of mortality, expressed in PYLL (up to 170 502 per year). Because this is a major issue for patients and the NHS, measures should be implemented to reduce the number of patients receiving VKA therapy who have poor anticoagulant control, thus minimizing the socioeconomic impact in our setting.

BMS and Pfizer provided unconditional funds for this project.

V. Barrios, S. Cinza-Sanjurjo, O. Gavín, and I. Egocheaga declare no conflicts of interest. R. Burgos-Pol and M.A. Casado are employees of PORIB. C. Polanco and J. Suárez are employees of BMS. J. Soto is an employee of Pfizer.