Inclusion criteria

• Known hypersensitivity or specific contraindication to the use of VKA.
• History of intracranial hemorrhage (ICH), before starting oral anticoagulant therapy (OAT) or during OAT with VKA.
• Ischemic stroke with clinical and/or neuroimaging high-risk criteria for ICH (HAS-BLED score [34] of three or more, grade III-IV leukoaraiosis, multiple cortical microbleeds).
• History of serious arterial thromboembolic event while on treatment with VKA despite good INR control (time in therapeutic range, TTR >65%).
• Poor INR control while on treatment with VKA despite good therapeutic compliance (TTR <65%).
• Unable to access conventional INR monitoring.

Exclusion criteria

• Current treatment with VKA and good INR control.
• Atrial fibrillation with severe heart valve disease.
• Patients unable to guarantee collaboration: significant cognitive impairment, alcoholism or psychiatric disorders, unsupervised.
• Situations in which OAT is contraindicated: pregnancy, severe acute bleeding in the previous month, recent (ten days prior) or planned CNS surgery, severe liver or kidney disease with creatinine clearance according to Cockcroft-Gault formula (CrCl) less than 15 mL/min, severe or uncontrolled hypertension, hereditary or acquired altered hemostasis (coagulation, fibrinolysis, platelet function) with a significant risk of bleeding.

Inclusion in the study and data collection

Recruitment of patients began in July 2019 and data was collected until March 2022. The scheduling of visits, follow-up, and blood tests were the same as for patients not included in the study (usual clinical practice). They were organized chronologically as follows:

• First medical and specialized nurse visit: baseline clinical information values were collected for each patient. Information on risk of thromboembolic events (CHA₂DS₂-VASc), risk of bleeding (HAS-BLED), previous history of thrombosis or bleeding, possible contraindications, laboratory values including kidney function, CrCl and liver function, and selection and dosage of DOAC were also recorded. The selection of the drug to be prescribed was independent of inclusion in the study. Dosage was based on the technical datasheet for each drug and international guidelines [31, 35].
• Educational session: In the first month of DOAC therapy, patients received an individual or group educational session on anticoagulation to learn general concepts, specific information on DOACs, risks when taking DOACs, drug interactions, the importance of adherence, response to complications, management in case of preoperative or invasive procedures, management in case of missing a dose or double doses, and help contact phone numbers.
• Follow-up visits: A face-to-face visit took place approximately 1 month after starting DOACs for all patients. There were quarterly visits for patients with progressive renal failure, biannual for those at high thrombotic risk, annual for those with medium or low risk, and on-demand visits in case of complications, surgical or invasive procedures, or new potential pharmacological interactions. At each visit, we collected data about tolerance and adherence to the drug, weight, new diagnoses and medication, laboratory data, and thrombotic or bleeding complications since the last visit.

Statistical analyses

Descriptive analysis was conducted. A bivariate analysis was performed to assess differences in baseline characteristics between DOAC group, type of patient, age and sex. Chi-square test or Fisher exact test were used as appropriate to compare categorical variables, and non-parametric Mann-Whitney U test was used to compare continuous variables. Incidence rates and their 95% confidence intervals (exact method) were calculated for the different types of complications and mortality, and were compared between groups with Chi-square test. Finally, a Cox proportional hazards survival analysis was used to assess the relationship between baseline factors and the first occurrence of each type of complication. In all analyses, p-values less than 0.05 were considered statistically significant. Analysis was performed with SPSS 26.0 (IBM Corp.) and R 4.1.1 (R Core Team).

Baseline characteristics

Clinical characteristics of patients, by DOAC, are provided in Table 1, with statistical comparisons between edoxaban vs apixaban and edoxaban vs dabigatran. We avoided comparisons with the rivaroxaban group due to the low number of patients.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline clinical characteristics according to DOAC type.

https://doi.org/10.1371/journal.pone.0279297.t001

Mean patient age was 77.7 years (SD: 8.3) and was lower in the dabigatran cohort (74.6 years, SD 10.1, p<0.05) than the edoxaban or apixaban cohort (78.0 and 77.7 years, respectively). We did not find significant sex differences among DOAC groups (52.9% male overall). There was a higher proportion of anticoagulant-naive patients in the apixaban and dabigatran cohorts (39.7% and 63.8%, respectively) than in the edoxaban cohort (14.6%). The mean CHA₂DS₂-VASc and HAS-BLED scores were higher among those using apixaban (4.37 points, SD 1.78, and 2.45 points, SD 0.84) as reflected by the CCI (1.93 points, SD 1.65). The laboratory characteristics were similar among the different DOAC cohorts. We found significant differences in baseline hemoglobin levels, being lowest in apixaban (126.6g/L) and highest in the dabigatran cohort (139.6g/L). The same occurred for renal function 60.1 mL/min in apixaban and 62.9 mL/min in edoxaban. The CrCl in the dabigatran cohort was higher than in the edoxaban cohort and the dabigatran cohort had the highest history of previous stroke (24.1%).

We also found significant differences between edoxaban and apixaban in anti-FXa activity trough level (0.10 and 0.62 UI/Ml, P<0.01) and the trough-peak difference (1.14 and 0.67 UI/ml, P<0.01).

Between patients taking both edoxaban doses, those on the 30mg were older (81.4 vs 75.9 years), with higher CHA₂DS₂-VASc score (4.42 vs 3.69) and with worse mean CrCl (44.3 vs 74.3 mg/dL) but without statistically significant differences except for anti-FXa activity trough, peak and its difference (view Table 1).

Main outcome analyses

The median follow-up time was 13.7 months. In total there were 13 major thrombotic events, 37 major hemorrhagic events, and 62 deceased patients. The corresponding event rates were 1.28% patient-years (py) (95% confidence interval [CI]: 0.68–2.19), 3.65% py (95%CI: 2.57–5.02) and 6.11% py (95%CI: 4.68–7.83) (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Complications according to DOAC type.

https://doi.org/10.1371/journal.pone.0279297.t002

The incidence of major thrombotic complications in the edoxaban group was 0.99% py (95%CI: 0.43–1.97), in the apixaban group was 2.16% py (95%CI: 0.45–6.30), and in the dabigatran group was 3.03% py (95%CI: 0.37–10.97), without statistically significant differences. The incidence of major hemorrhagic complications was 3.49% py (95%CI: 2.32–5.05) in the edoxaban cohort, 4.31% py (95%CI: 1.58–9.38) in the apixaban cohort, and 4.55% py (95%CI 0.94–13.27) in the dabigatran cohort, without statistically significant differences. The incidence rates of clinically relevant non-major thrombosis (CRNMT) were very similar between DOACs. The CRNMB incidence rate was 10.78% py in the apixaban group, 7.98% py in the edoxaban group, and 7.58% py in the dabigatran group.

The all-cause mortality was significantly higher in the apixaban group (10.78% py, 95%CI: 6.03–17.74, p<0.05) than the edoxaban group (5.12% py, 95%CI:3.67–6.94).

The rates of thrombotic and major bleeding events were very similar among different edoxaban doses, with higher all-cause mortality in the edoxaban 30 mg cohort (8.68% py vs 2.72% py, P<0.001).

Table 3 shows the hazard ratios for thrombotic and hemorrhagic complications in the edoxaban cohort for each of the baseline and laboratory characteristics. CCI showed a statistically significant association with the first occurrence of a major thrombotic complication, major hemorrhagic complication and all bleeding complications (HR 1.61, 95%CI:1.23–2.09, p = 0.001; HR 1.27, 95%CI:1.01–1.60, p = 0.04; and HR 1.29, 95%CI:1.13–1.47, p<0.001 respectively, for each additional point of CCI). Lower hemoglobin was found to be a significant risk factor for both hemorrhagic (HR 0.97, 95%CI:0.95–0.99, p = 0.017) and all bleeding complications (HR 0.99, 95%CI:0.97–1.00, p = 0.047) and elevated anti-FXa activity trough (pre-dose) level for all bleeding complications (HR 3.73, 95%CI:1.38–10.08, p = 0.009). ALT was associated with the first occurrence of a major thrombotic complication with a HR of 1.06 (95%CI:1.01–1.11, p = 0.022). Last but not least, a history of previous bleeding was shown to be a risk factor for all bleeding complications, with a HR of 2.47 (95%CI:1.32–4.61, p = 0.005).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Clinical and laboratory characteristics related to thrombotic and hemorrhagic complications in the edoxaban cohort.

https://doi.org/10.1371/journal.pone.0279297.t003

Subgroup analyses

S1 Table in S1 File describes the characteristics of the 615-patient edoxaban population stratified by age (≤74 years, 75–84 years and ≥85 years) and by sex. When analyzed by age, the proportion of male patients decreased from 57.4% in the ≤74 year-old group to 43.9% in the ≥85 year-old group. The thrombotic risk as measured with CHA₂DS₂-VASc increased (3.03 to 4.65 points), as did the history of previous ischemic stroke (8.7% to 18.2%). The bleeding risk as measured with HAS-BLED was highest in the oldest group (1.95 in ≤74-year-olds to 2.21 points in ≥85-year-olds). Mean CrCl decreased with age (77.2 to 47.9mg/mL). The proportion of naive patients, CCI and history of previous major bleeding appeared similar across all ages.

The incidence of major thrombotic complications was similar across age groups, but the incidence of major hemorrhagic complications increased notably with age (1.69%, 3.91% and 5.13% py), with only the difference between the youngest and the oldest cohorts reaching statistical significance. The all-cause mortality rate was markedly higher in patients aged ≥85 years: 11.39% py vs 3.39% py in those aged 75–84 years (S2 Table in S1 File).

When analyzed by sex (S1 and S2 Tables in S1 File), female patients were older (79.2 y vs 76.8 y), had worse mean CHA₂DS₂-VASc score (4.40 vs 3.55) and better HAS-BLED score (2.02 vs 2.18) and CCI (1.17 vs 1.73). 12 patients with GFR <30mL/min were female vs one male. Male patients used more concomitant antiplatelet therapy (10.8% vs 2.7% in women). There were no statistically significant differences in outcomes depending on sex.

Edoxaban results

Our sample of 615 patients on edoxaban who were followed up for 15.0 months allows us to make reliable estimates of the incidence of major complications. Our population is older and with a worse CHA₂DS₂-VASc score than most previous studies (Tables 4 and 5). We observed an incidence of 4.49% py of total major complications, with a predominance of major bleeding (3.49% py). When the analysis was split by edoxaban dose, we did not observe differences in incidence of complications, but there was a higher mortality in the 30 mg group, probably due to the clinical characteristics of that group.

Focusing on the age subgroup analysis (S1 and S2 Tables in S1 File), the increased complexity of patients with increasing age was reflected in the number of major hemorrhagic complications; meanwhile, the incidence of thromboembolic events remained stable. This is important to keep in mind while treating elderly patients with DOACs. There were no differences in terms of outcomes when analyzed by sex.

Regarding demographic characteristics related to complications, we found that every increasing point of the Charlson Comorbidity Index [38] represented an increased risk of a first occurrence of a major thrombotic or hemorrhagic complication and of all bleeding complications. Risk factors related to bleeding were low hemoglobin, anti-FXa activity trough (pre-dose) level and history of previous bleeding, but not age.

In Table 6 we compare our results with other reports on edoxaban. Our findings are slightly worse than those of the pivotal ENGAGE RCT [4], which had a population with a mean age six years younger, but with a similar CHA₂DS₂-VASc score (considering that CHADS₂ score is approximately one point lower than CHA₂DS₂-VASc).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Main outcomes with edoxaban from studies in real world or RCTs, in comparison with results from MACACOD, stratified by age.

https://doi.org/10.1371/journal.pone.0279297.t006

Kohsaka et al [9] and Lee et al [6] reported slightly better results than our study but included only complications in hospitalized patients [9] or younger patients with lower CHA₂DS₂-VASc scores [6]. In contrast, the results from the ETNA registry [12], which were very good, even better than the pivotal study, could have been influenced by some of the problems mentioned above. The edoxaban results from Lee [6], Yu [29], Nielsen [30], Marston [39], and Paschke [40] are slightly worse than ours. The Korean studies [6, 29] found a higher incidence of complications with edoxaban 30 mg than edoxaban 60 mg, but this was not found in the other studies.

Table 4 lists multiple studies reporting a lower incidence of major complications [1–10, 12–14, 18–22, 41–49], and Table 5 lists those that report a higher incidence. To be conservative, we used as a threshold the incidences observed in our total sample of all four DOACs. By so doing, our results compare well with all the previous results, since those with lower incidences are generally shown to be in younger populations, with less frailty or comorbidity. Some of the studies with worse outcomes than ours are also in younger populations, although in general they have similar or higher CHA₂DS₂-VASc scores [15, 28–30, 40, 50].

Many of the studies summarized in Table 4 have remarkably good results. In their critical evaluation, some points that raise doubts about credibility should be considered. First, it is difficult to obtain similar or even lower incidences of serious complications in clinical practice [5–14, 22] than in RCTs [1–4], as already pointed out in the study by de Vries et al [51], for the simple reason that patients in real life are older and have more comorbidity. However, the data they offer on mortality (repeatedly very low) do not align with the expected mortality of 16% py in the first year [23, 24] and 8% py in the following years after AF diagnosis [25]. This 8% refers to a population with a mean age of 70 years and a CHA₂DS₂-VASc score of 2.8, that is, with a lower risk than the populations in the studies we criticize.

The data in Table 5 seem more consistent with reality and, if true, are certainly worrying due to the high incidence of serious complications. The total complications exceed 10% py in almost all the studies. The best incidences in this table come from nationwide studies from Denmark [27, 30] and a Scottish study [52] and are between 5.5 and 9.4% py, perhaps due to their better health care models. Note that the all-cause mortalities in these studies in are very close to what would be expected [23–25].

However, these high incidences of severe complications with DOACs, are lower than those observed with VKA therapy in real world [15, 28, 29, 39, 53]. In the Marston et al study [39], the composite incidence of major thrombotic and hemorrhagic events was 8.8% py with dabigatran, 6.83% py with apixaban, 7.18% py with rivaroxaban and 5.40% py with edoxaban but 11.82% py with VKA treatment.

Perhaps the key to these inconsistencies is found in the article by Bang et al [15], where a huge difference (between half and a third) was observed in the crude incidence rates of major complications when the diagnoses were limited to inpatient claims. Or in the works by Koshaka et al [9], Huybrechts et al [19], and Ujeyl et al [18], which were limited to complications that required hospitalization or deaths that occurred in hospital.

The outcomes of the MACACOD project, may indicate that the specialized management of DOACs helps minimize complications to acceptable levels, but there is undoubtedly still room for improvement. We observed a favorable trend for edoxaban, compared to dabigatran and apixaban. This trend is in line with recent nationwide publications, three of them technically well-designed with propensity-score matching comparisons in Korea [6] and Germany [39, 40], and others in Denmark [27, 30], with lower incidences of major complications with edoxaban than with dabigatran, rivaroxaban, and apixaban.

A critical appraisal of DOACs in daily practice

Marietta et al [54] warned about the difficulty of translating the results of trials into real life because, in an everyday care setting, patients have more comorbidities, are treated without a strict protocol, and are under less intensive surveillance than in an RCT. In addition, we must consider the concerning figures on non-adherence and non-persistence in the treatment with DOACs, which will inevitably worsen clinical results in the real world [55, 56].

One possible option to overcome such difficulties is the management by specialized clinics, where the clinical outcomes are expected to be as good as those presented by some reports [57] and high-quality registries [41, 42]. Our results are as good as those obtained in reliable studies such as those by Rutherford et al [43] in Norway with a population less vulnerable than ours, and even better than those reported by Nielsen et al [27, 30] in Denmark with populations very similar to ours.

This literature of DOACs suggests that the real situation is worse than is generally believed. If the figures for serious complications are like those reported by many of the studies in Table 5, then, “Houston, we have a problem…” The following inconsistency highlights this: the original CHA₂DS₂-VASc risk scale estimates the risk of thromboembolic events at 4% py for patients with a score of 3–6 [32]. A subsequent adjustment placed the risk between 4% and 6.7% py for scores of 4–5 [58]. A rate of total serious complications higher than 10% (obviously including hemorrhages) leads to an absurd situation, when what we are trying to prevent has an approximate risk of 4–6%. Our results, together with others, indicate the need for more highly-specialized clinical care when prescribing any DOAC.

Strengths and limitations

The relatively small size of our study is both a limitation and a strength. It can be considered a strength because it made possible to guarantee the registration of incidents through personalized follow-up of patients and all their clinical records, both in and out of the hospital. The single-center nature of the study also guarantees a uniform clinical protocol. The main limitation is the relatively small sample size, which mainly affected apixaban, dabigatran, and rivaroxaban, for which we could not make reliable estimates. Another limitation is, due to the nature of the prescription of DOACs in Spain, that they are used as a second-line therapy after VKAs. This translates into a low percentage of naive patients and, consequently, can bias comparisons with other studies.