Effect of the quality of anticoagulation on the risk of stroke, thrombotic events, hemorrhagic events, and death in patients with nonvalvular atrial fibrillation on acenocoumarol in Real-World Data

Introduction

Over the past fifty years, vitamin K antagonists (VKAs) have been the mainstay of anticoagulation therapy for the primary and secondary prevention of venous and arterial thromboembolic events.¹ They have proven to be highly effective and are therefore used today by millions of people around the world with nonvalvular atrial fibrillation (NVAF), venous thromboembolic disease, mechanical valve prosthesis and other conditions.^2,3 In Spain, the type of VKA used is acenocoumarol which inhibits the reduction of vitamin K by vitamin K reductase. This prevents carboxylation of certain glutamic acid residues near the N-terminals of clotting factors II, VII, IX and X, the vitamin K-dependent clotting factors. Its mechanism of action is to interfere with the vitamin K cycle, decreasing levels of the biologically active forms of coagulation factors, thus hindering the formation and growth of thrombi.¹ Given its complex pharmacokinetics and pharmacodynamics, there is a wide individual variability in the therapeutic effect of anticoagulation with VKAs. Among other factors, genetic characteristics, medication interactions, environmental conditions, vitamin K intake, hypermetabolic states, diseases, comorbidities, and age can interfere with the coagulation process.^4–6 Therefore, interactions with other medications, the anticoagulant effect and any reduction in thromboembolic events are factors to be taken into account in clinical practice together with the risk of bleeding associated with this therapy.^1,7

Monitoring and strict control of anticoagulation therapy reduces the risk of thromboembolic or hemorrhagic events. The international normalized ratio (INR) is the standardized measure for expressing the results of prothrombin time tests and has served to validate the effectiveness of oral anticoagulation in numerous studies.⁸ The target therapeutic range is established based on the reason why a patient is anticoagulated, and is typically between 2.00 and 3.00, except in patients with mechanical valve prosthesis, in whom it is 2.50 to 3.50.³ The INR value at a given time determines the effectiveness of anticoagulant therapy with VKAs and is related to the risk of bleeding or thrombosis.^2,8 We can also evaluate the quality of anticoagulation by monitoring the time in therapeutic range (TTR).^8–11 TTR, expressed as the percentage of time that INR values remain within the therapeutic range, is commonly used as a measure of the quality of anticoagulation control over a certain time interval due to its association with the occurrence of bleeding and thromboembolic events.^8,12

The results in the literature concerning the impact of good anticoagulation control differ by study design. In a systematic review, TTR was found to be negatively correlated with rates of bleeding and thromboembolic events. This effect was significant in observational studies, but not in randomized controlled trials.⁸ For this reason, it is crucial that the correlation between good control of anticoagulation and its benefits in terms of thrombotic and hemorrhagic events avoided be evaluated in clinical practice.

The objective of this study was to measure the association of the quality of anticoagulation control using VKAs with the occurrence of stroke, thrombotic or hemorrhagic events, or death during follow-up in patients diagnosed with NVAF.

Methods

Results

We recruited 949 patients who had been started on anticoagulation therapy with acenocoumarol. Among these patients, 59 had not been on VKA therapy for a minimum of 6 months and 34 had missing data regarding control of anticoagulation in the TAOnet registry; hence, they were excluded from the study. Therefore, the final sample studied comprised 856 patients.

Table 1 summarizes the characteristics of these patients disaggregated by anticoagulation control according to TTR. Only 286 (33%) patients had TTR values ≥60% indicating good control. The mean age was similar in both groups by level of control. The analysis of crude differences showed that the group with TTR ≥60% had significantly fewer thrombotic events, strokes and deaths, while the differences in hemorrhagic events were not statistically significant. The distributions of socioeconomic variables, such as income level and occupational status, also did not differ significantly with TTR level. Overall, 41% of patients were switched to DOACs. Notably, switching was much less common in the good control group (TTR ≥60%) than in the poor control group (TTR <60%) (11% vs 57%). The average TTR in the entire sample was 52.6%, rising to 68.8% in the good control group and falling to 44.5% in the poor control group.

Figures 1-3 show the Kaplan-Meier curves for each subsample by level of control (TTR< or TTR≥60%) for the four events analyzed (stroke, thrombotic and hemorrhagic events, and death). The differences between the subsamples were statistically significant for all four events according to the log-rank tests.

Figure 1. Probability of stroke (a) and a thrombotic event (b) over the follow-up as represented by Kaplan-Meier curves.

TTR: time in therapeutic range. X-axis: time in days. Y-axis: probability. These figures were drawn by ourselves from the results.

Figure 2. Probability of hemorrhagic event over the follow-up represented by Kaplan-Meier curves.

TTR: time in therapeutic range. X-axis: time in days. Y-axis: probability. This figure was drawn by ourselves from the results.

Figure 3. Probability of death over the follow-up represented by Kaplan-Meier curves.

TTR: time in therapeutic range. X-axis: time in days. Y-axis: probability. This figure was drawn by ourselves from the results.

The multivariate survival analyses using Cox regression reported in Table 2 confirmed that, when adjusting for age, sex, income level, occupational status and CCI, the risk of each of the four events was much higher in the subsample with TTR <60% than in the good control group. Specifically, the HRs were 1.94 (CIs: 1.13-3.30) for stroke, 1.60 (CIs: 1.10-2.33) for thrombotic events, 1.61 (CIs: 1.08-2.42) for hemorrhagic events and 2.97 (CIs: 1.86-4.75) for death. Occupational status was not a significant factor in the risk of any event, while a low-income level was significantly associated with a higher risk of hemorrhage but not with the other events.

In the crude analysis, the percentages of women and higher comorbidity categories were significantly lower in the good control group (Table 1). These results were confirmed in the multivariate analysis analyzing the probability of good control with logistic regression in that the odds ratios (ORs) for women (OR:0.69; CIs: 0.50-0.93) and the higher comorbidity categories, namely, CCI of 3-4 (OR: 0.61; CIs: 0.39-0.97) and CCI >4 (OR: 0.50; CIs: 0.30-0.81), indicated significantly poorer control in these groups (Table 3).

Discussion

The findings of this study confirm the benefits of achieving good anticoagulation control in clinical practice, the risks of the three cardiovascular events and death being significantly higher in the group with poor control (TTR <60%). Further, the HR (1.92) for the risk of stroke was almost twice as high in the poor control group, reflecting a greater preventive effect of anticoagulation when a patient’s INR is within the therapeutic range. Notably, in the survival analysis, the highest risk was for death (HR of 2.96), despite this being adjusted for ICC, underscoring the high excess mortality in the poor control group. These results are consistent with those published in the literature in which good control has been associated with HRs greater than 2 for the risk of events.^21–23 On the other hand, a systematic review suggested that a higher mean TTR on VKA therapy is associated with more effective prevention although the strength of the association decreases when adjusted for patient clinical characteristics.²⁴ Control of anticoagulation in patients with NVAF is a key factor for achieving effectiveness by preventing strokes and safety by reducing hemorrhages. Achieving both these objectives increases the survival and quality of life of patients.²⁵ Therefore, TTR is a key indicator to monitor to maximize the effectiveness of VKA therapy.²⁶ Along the same lines, a systematic review indicated that 1 major hemorrhage and 1 thromboembolic event could be avoided per 100 patient-years if TTR were increased by 7% and 12%, respectively.⁸ Conversely, poor control of anticoagulation with VKA is associated with higher rates of cardiovascular events, and in turn, substantial increases in healthcare costs.²⁷

Hardly any associations were found between socioeconomic variables and events in our results. Previously, it has been noted that some occupations may increase the risk of injury or trauma, which may influence the need to adjust anticoagulant doses to prevent serious bleeding.²⁶ Moreover, the nature of patients’ work activity can affect their consistency in taking medications and, therefore, make it necessary to adapt the anticoagulation regimen to ensure treatment adherence.²⁶ On the other hand, socioeconomic status may influence access to medical care, including follow-up visits and laboratory tests necessary to adjust acenocoumarol doses appropriately.^28,29 Additionally, economic factors may affect patients’ ability to acquire and take medications.²⁸ Nonetheless, as in a Spanish multicenter study,³⁰ the risk of events in our results was not significantly associated with patients’ socioeconomic level or occupational status. Only when analyzing the risk of hemorrhage did a lower level of income appear as a significant risk factor. This lack of significant differences in events is consistent with the lack of significant differences in the probability of having an adequate TTR. It is plausible that social factors have little influence in the context of a Beveridge-type health system like ours with a single public system that ensures universal access to health care.³¹ Further, it may indicate that equity in the provision of anticoagulation therapy underlies the lack of differences in outcomes measured in terms of risk of events.

Good anticoagulation control was only associated with male sex and lower levels of comorbidity. The lack of significance of age is surprising. On the contrary, and as in the literature, the results of the Cox models indicated that in addition to good control, the risk of events depends on age and comorbidity which is consistent with the growing risk of cardiovascular events with age.^16,32 A high level of comorbidity implies the use of other medications that may interact with the action of anticoagulation, making it necessary to adjust the doses of acenocoumarol to avoid hemorrhagic or thromboembolic complications,^23,26 while conditions such as kidney or liver failure affect the metabolism of oral anticoagulants, meaning that doses need to be adjusted to maintain therapeutic levels.²⁴ There are mixed results in the literature concerning adherence to acenocoumarol and good control, significant associations not always being found between adherence and therapeutic control.^33,34 Genetic differences seem to influence the metabolism of vitamin K and cytochrome P450.³⁵ Genetic variants in the warfarin-metabolizing enzyme, cytochrome P-450 2C9 (CYP2C9), and a key drug target of warfarin, vitamin K epoxide reductase (VKORC1), contribute to differences in patient responses to various doses of warfarin, though the role of these variants during initial anticoagulation is unclear.³⁶ All this indicates that factors such as drug-diet interactions, genetic variability in drug metabolism, and frequent dose adjustments, which were not measured in this study, may hinder anticoagulation control, in turn, increasing the risk of events.³⁷

Unlike the results of a meta-analysis of anticoagulated NVAF patients in which the mean TTR was 61%,³⁸ our mean TTR was just 52.6% and two-thirds (67%) of our patients had a TTR <60%. In other studies conducted in Spain, the rate of good anticoagulation control has been similar (between 53% and 59%).^30,39,40 Although the prevalence of poor control is higher in clinical registries that include patients from routine practice,⁸ even in relevant clinical trials such as RE-LY⁴¹ and ROCKET AF,⁴² between 30-40% of patients with NVAF who receive VKA have been outside the range, with INRs <2 or >3, more than 60-65% of the time. Similar to our results, two studies in Korean and Thai populations found good control rates of around 30%.^10,43

As well as noncompliance with treatment or changes in dosage, potential causes of an inadequate INR include interruptions due to procedures requiring discontinuation of acenocoumarol, recent administration of vitamin K, diarrhea, or vomiting, the emergence of other diseases, changes in other medications, and changes in diet and lifestyle, such as increased dietary intake of vitamin K, increased tobacco consumption, and/or increased levels of physical activity.^43,44 As in another study,³⁹ we found that female sex and greater comorbidity were significantly associated with poor control. Other authors have considered female sex to be a possible modifier of TTR, and many reasons have been suggested for poorer control of oral anticoagulation in women, such as pharmacokinetics, hormonal effects, family and social factors, and less access to health services.^19,45 Gender differences in VKA pharmacokinetics have been partially attributed to average body size and liver fat, as well as intrinsic differences in VKA metabolism. Furthermore, the metabolism carried out by cytochrome P450 can be altered by the effect of sex hormones, leading to an increase in the response to a VKA dose in women.

The results of our study indicate that a high proportion of patients with NVAF currently receiving VKA in our province are poorly anticoagulated. This implies that despite being “protected”, they continue to have a high risk of stroke, embolism, hemorrhagic complications and death. We believe it is necessary to develop and implement strategies to tackle this problem. In the clinical practice context in which the study was conducted, the consequence of poor control was that more than half of the patients (56.8%) were switched to DOACs. Given this high figure, the question arises as to whether patients with more comorbidities should have been prescribed a DOAC instead of acenocoumarol at the outset when NVAF was diagnosed.

One limitation of this study is its observational design associated with the data having been obtained from clinical practice. Nevertheless, our findings noting poor control of anticoagulation in real world data underline the need to closely monitor adherence as measured by time within the appropriate INR range. Another limitation to note is that the diagnosis of NVAF has been identified based on the ICD codes, without having validated the criteria for its diagnosis. On the other hand, the strengths of this study are that it included patients from routine clinical practice and that the results were adjusted for comorbidity, occupational status, and socioeconomic level.

As a positive conclusion of the study, we point out the great benefit observed in terms of events and deaths avoided when TTR reaches figures ≥60%. On the contrary, the low rate of good control found represents a key opportunity for improvement. This is important because achieving effective control of anticoagulation means decreasing the risk of cardiovascular events and therefore both increasing quality-adjusted life expectancy for patients and reducing costs for the health system.

Authors’ contributions

MFP, GP, CP and JM conceived, designed the research and wrote the first version of introduction and discussion. OI obtained the data, and interpreted the data, helping to write results and discussion. MFP and JM designed the methods, performed the analyses, interpreted the data and wrote the methods and results. All authors revised the manuscript for important intellectual content and approved the final manuscript.