Clinical and Economic Consequences of a First Major Bleeding Event in Patients Treated with Direct Factor Xa Inhibitors in Spain: A Long-Term Observational Study

Aims: Our aims were to describe the clinical characteristics, adverse clinical events, healthcare resource utilization (HCRU) and costs of patients with major bleeding during direct Factor Xa inhibitor (FXai) use. Methods: This is a retrospective cohort study that included secondary data from computerized health records of seven Spanish Autonomous Communities. Patients with a first major bleeding during treatment with a direct FXai were analyzed during a 3-year period. Results: Of 8972 patients taking a direct FXai, 470 (5.24%) had major bleeding (mean age (SD) 77.93 (9.71) years, 61.06% women). The most frequent indications for using FXais were atrial fibrillation (78.09%) and venous thromboembolism (17.66%). Among those with major bleeding, 88.94% presented with gastrointestinal bleeding, 6.81% intracranial bleeding, 2.13% trauma-related bleeding and 4.26% other major bleeding. Prothrombin complex concentrates were used in 63.19%, followed by transfusion of blood products (20.21%) and Factor VIIa (7.66%). In total, 4.26% of patients died in the hospital due to the first major bleeding. At the study end (after 3-year follow-up), 28.94% of the patients had died, 12.34% had a myocardial infarction and 9.15% an ischemic stroke. At year 3, overall bleeding cost was EUR 5,816,930.5, of which 79.74% accounted for in-hospital costs to treat the bleeding episode. Conclusions: Despite the use of replacement agents being high, major events were common, with a 29% mortality at the end of the follow up, and HCRU and costs were high, evidencing the need for new reversal treatment strategies.


Introduction
Anticoagulation is the key treatment for the management of patients with atrial fibrillation (AF), acute venous thromboembolism (VTE) and some patients soon after bioprosthetic valve implantation to prevent the development of arterial and/or venous thromboembolic complications [1][2][3][4].
This was an observational cohort study, with a retrospective design that analyzed secondary data from electronic health records in seven Spanish Autonomous Communities from the BIG-PAC ® database (initiated in 2012).Different studies have demonstrated the representativeness of this database for the Spanish population [16][17][18].
In the previous study, adults with a prescription of an oral direct FXai (i.e., apixaban, edoxaban or rivaroxaban) at a therapeutic dose, for the management of VTE, AF or nonmechanical cardiac-valve replacement, between 1 January 2013 and 31 December 2022 were analyzed [15].Additionally, to be included, patients had to have at least 12 months of data available in the database before the prescription of anticoagulant.The present study included those patients who had a first-ever major bleeding event during their treatment with a direct FXai.
Major bleeding was defined as all critical site bleeding (defined with the International Classification of Diseases (ICD)-10 codes based on the International Society on Thrombosis and Haemostasis definitions and anemia ICD-10 codes) and for other locations, if the bleeding was fatal or led to hospitalization.Trauma-related ICD-10 bleeding codes were a subset of the intracranial or other bleeding codes.Other major bleeding included genitourinary, respiratory and other bleeding not included elsewhere [15].The exclusion criteria included therapeutic non-FXai anticoagulation use or prophylactic anticoagulation use in the 60 days before the major bleeding event, the history of a major bleeding event before the start of treatment with a direct FXai, palliative care initiation any time before the major bleeding event of interest and pregnancy any time between the direct FXai treatment episode start and the major bleeding event of interest.
The index date (start of observation in these patients) was defined as the day of the first major bleed and the end of the observation period occurred when the first of any of the following events happened: 3 years after the index bleeding event, recording of pregnancy, completion of the study period, disenrollment from the database, end of data collection, death or start of palliative care.Of note, for analyses on HCRU and bleeding events, follow-up did not end when a clinical outcome of interest had occurred.No specific diagnostic or therapeutic procedures were performed for inclusion in the study, due to the retrospective observational nature of the study.The study was authorized by the Research Ethics Committee of Consorci Sanitari de Terrassa, Barcelona, Spain.The need for obtaining written informed consent was waived by the Research Ethics Committee, because the study collected secondary anonymized data.
At baseline (the day prior to index day), data from biodemographics, cardiovascular risk factors, vascular disease and other comorbidities were recorded.Clinical conditions were defined by means of the medical codes entered by practices, with the ICD, Ninth Revision (ICD-9), and Tenth Revision (ICD-10) (all converted to ICD-10 in the database).Laboratory tests closest to the index date were collected.Concomitant treatments (within 120 days prior to enrolment, 180 days in the case of anticancer drugs) were also captured.Treatments were derived from issued prescriptions coded utilizing anatomical therapeutic chemical codes (ATC medication code: "A10").Data were analyzed in the whole study population, and according to the type of major bleeding, the use of prothrombin complex concentrates, the type of FXai and the FXai indication.
The analyzed outcomes were the development of a first and subsequent clinical event (death from any cause, cardiovascular death, acute myocardial infarction, ischemic stroke, acute kidney failure or acute liver failure).Incidence and event rates of clinical events were calculated.Incidence rates were defined as the total number of incident events of interest divided by the total person time at risk and event rates, as the total number of events, including recurrent events divided by the total person time of follow-up, during the entire follow-up.Incidence and event rates were estimated in the whole study population, and according to the type of major bleeding, the use of prothrombin complex concentrates, the type of FXai and the FXai indication.Cumulative incidence curves for events (acute myocardial infarction, stroke, cardiovascular death and overall death) were calculated.In addition, risk factors for the development of these clinical events were calculated.
HCRU included the number and associated healthcare costs of all healthcare resources, including the number and length of hospital admissions, inpatient procedures and investigations (laboratory and radiology), outpatient [general practitioner (GP) and specialist] visits, prescriptions, general practice visits and referrals.Information on cost data was inferred from the eSalud database [19] and the pharmaceutical prescription costs were based on the full price of products [20].Costs of absence from work were estimated by multiplying the number of days of absence from work due to illness by the mean daily salary of a worker in Spain [21] (Supplementary Table S1).All-cause and related to bleeding total and per patient costs were calculated at 6 months, 1 year, 2 years and 3 years after the index episode of major bleeding.HCRU and costs were estimated in the whole study population, and according to the type of major bleeding, the type of FXai and the FXai indication.

Statistical Analysis
Demographic and clinical data were reported with descriptive statistics.Qualitative data were shown by their absolute (n) and relative (%) frequencies.Quantitative data were reported with the mean and standard deviation (SD).To compare continuous variables between prothrombin complex concentrates users vs. non-users, a two-sample t-test was used for variables normally distributed, and the Mann-Whitney U test was used for those non-normally distributed.The chi-square test was used for categorical variables.Wald contrast was used for the incident rates of clinical outcomes.Incidence and event rates for clinical events were described per 100 person-years with their 95% confidence intervals.Cumulative incidence curves for events (acute myocardial infarction, stroke, cardiovascular death and overall death) were calculated.The area under the curve (AUC) of these curves was calculated using the trapezoidal rule.The strength of association between potential risk factors (all baseline clinical characteristics) and the risk of clinical events (for acute myocardial infarction, stroke, cardiovascular death and all-cause death) was calculated using Cox regression models calculating the hazard ratio with the 95% confidence intervals.To investigate HCRU following major bleeding, cumulative HCRU rates were calculated, and respective costs were estimated by multiplying event numbers with costs per event.A p-value less than 0.05 was considered to be statistically significant.All statistical techniques were calculated using Stata MP Version 14.2 (StataCorp LLC., College Station, TX, USA).

Results
Out of 1.9 million patients included in the BIG-PAC ® database, 16,665 patients had received their first direct FXai prescription, of whom 13,739 had a study indication.After exclusion of 4767 patients for different causes, 8972 patients were finally analyzed.Of these, 470 (5.24%) had had a major bleed and were followed-up during a 3-year period (Figure 1).In this study, the data of patients with a major bleed are presented.
Cumulative incidence curves for events (acute myocardial infarction, stroke, cardiovascular death and overall death) were calculated.The area under the curve (AUC) of these curves was calculated using the trapezoidal rule.The strength of association between potential risk factors (all baseline clinical characteristics) and the risk of clinical events (for acute myocardial infarction, stroke, cardiovascular death and all-cause death) was calculated using Cox regression models calculating the hazard ratio with the 95% confidence intervals.To investigate HCRU following major bleeding, cumulative HCRU rates were calculated, and respective costs were estimated by multiplying event numbers with costs per event.A p-value less than 0.05 was considered to be statistically significant.All statistical techniques were calculated using Stata MP Version 14.2 (StataCorp LLC., College Station, TX, USA).

Results
Out of 1.9 million patients included in the BIG-PAC ® database, 16,665 patients had received their first direct FXai prescription, of whom 13,739 had a study indication.After exclusion of 4767 patients for different causes, 8972 patients were finally analyzed.Of these, 470 (5.24%) had had a major bleed and were followed-up during a 3-year period (Figure 1).In this study, the data of patients with a major bleed are presented.The clinical characteristics at baseline of the whole study population are reported in Table 1.The mean age (SD) was 77.93 (9.71) years old, 61.06% were females, 78.72% had arterial hypertension, 33.83% type 2 diabetes, 23.83% heart failure, 17.23% chronic kidney disease, 15.32% ischemic heart disease and 8.09% cerebrovascular disease.The most frequent indications for the use of FXai were AF (78.09%),VTE (17.66%) and non-mechanical cardiac-valve replacement (4.26%).When a major bleeding occurred, prothrombin complex concentrates were used in 63.19% of patients, followed by transfusion of blood products (20.21%) and Factor VIIa (7.66%).Surgical procedures used to correct bleeding were performed in 18.72% of the cases of major bleeding.Twenty (4.26%) patients died in hospital due to the first major bleeding event.
Table 1.Baseline characteristics on the day prior to index day 1, according to the type of major bleeding and the use of prothrombin complex concentrates.Overall, 418 (88.94%) patients presented with gastrointestinal bleeding, 32 (6.81%) intracranial bleeding, 20 (4.26%) other major bleeding and 10 (2.13%) trauma-related bleeding (included intracranial and other major bleeding).Clinical characteristics at baseline were also analyzed according to the type of major bleeding, showing that trauma-related bleeding was more common in the elderly, and prothrombin complex concentrates were more frequently used in cases of intracranial bleeding, whereas transfusion of blood products was common in cases of gastrointestinal bleeding.
In-hospital mortality during the first bleeding episode was higher in cases of traumarelated bleeding (70.0%), followed by intracranial hemorrhage (28.13%), other bleeding (10.00%) and gastrointestinal bleeding (2.15%) (Table 2).Baseline characteristics were also analyzed according to the use of prothrombin complex concentrates, the type of FXai and the FXai indication (Table 1 and Supplementary Table S2).No significant differences were found in the baseline clinical profile according to the use of prothrombin complex concentrates.S3.
Cumulative clinical outcomes during the study period are presented in Table 2.The incidence rates per 100 person-years during the index hospitalization were death from any cause 18.17 (14.68-21.66),cardiovascular death 18.17 (14.68-21.66),acute myocardial infarction 4.64 (2.74-6.54),ischemic stroke 0.93 (0.06-1.8) and acute kidney failure 0.92 (0.06-1.78) (corresponding to 4.26, 4.26, 1.06, 0.21 and 0.21%, respectively).These numbers were 17.2 (12.91-21.49),17.20 (12.91-21.49),2.94 (1.02-4.86),1.47 (0.10-2.84) and 0, respectively, among those patients treated with prothrombin complex concentrates during hospitalization.After 3 years of follow-up, incidence rates per 100 person-years were death from any cause 16.42 (13.07-19.77),cardiovascular death 9.3 (6.67-11.93),acute myocardial infarction 7.34 (4.98-9.7),ischemic stroke 5.42 (3.37-7.47),acute kidney failure 4.17 (2.36-5.98)and acute liver failure 0.97 (0.08-1.86).The incidence rates of clinical events decreased over time.There were few recurrent events, with acute myocardial infarction being the most common (with five additional cases reported).Cumulative incidence curves for events (acute myocardial infarction, stroke, cardiovascular death and overall death) were calculated (Figure 2).Cumulative clinical outcomes were evaluated according to the type of major bleeding, the use of prothrombin complex concentrates, the type of FXai and the FXai indication (Table 2 and Supplementary Table S4).Except for a higher incidence rate of acute myocardial infarction during the index hospitalization, no significant differences were found in cumulative clinical outcomes according to the use of prothrombin complex concentrates.Mortality incidence rates were higher in those patients with trauma-related bleeding, followed by intracranial hemorrhage, other major bleeding and gastrointestinal bleeding.In addition, mortality rates were higher in those patients who were not treated with prothrombin complex concentrates.The cumulative HCRU for 3 years from the index date was analyzed, including outpatient (GP and specialist) visits and hospitalizations, length of hospital stays, laboratory/radiology investigations, number of prescriptions and work absences (Tables 3 and 4).At the study end, all-cause HCRU rates of outpatient visits, laboratory/radiology investigations and hospitalization were 1437.8(1120.59-1755.01),346.02 (303.01-389.03)and 83.31 (79.94-86.68)per 100 person-years, respectively, with a decrease over time from the 6-month time point following the index date.These numbers were 898.73 (871.46-926.00),216.35 (179.12-253.58)and 63.99 (59.65-68.33)per 100 person-years, respectively, for bleedingrelated HCRU, following the same pattern of decrease (with 53% of all bleeding-related outpatient visits and 92% of hospitalizations occurring in the initial 6 months after the first major bleed).When considering all patients in the cohort, bleeding-related length of hospital stay and mean number of days (SD) due to work absences per patient were 12.20 (4.81) and 2.49 (8.65) days, respectively.The HCRU was also analyzed according to the type of major bleeding, the type of FXai and the FXai indication (Supplementary Tables S5 and S6), with hospitalization rates being higher in those patients who presented with trauma-related bleeding, followed by intracranial hemorrhage and, to a lesser extent, other bleeding and gastrointestinal bleeding.Cumulative costs for 3 years from the index date are shown in Table 5.The total overall cost reached EUR 7,843,056.4at year 3 (EUR 5,780,203.3[73.70%] in the first 6 months), of which 76.99% accounted for hospitalizations, 18.02% for outpatient care and 4.99% for indirect costs (cost of absence from work).Overall prescriptions accounted for 8.72% of the total cost.In the case of bleeding-related costs, overall bleeding cost was EUR 5,816,930.5(EUR 4,930,469.4 [84.76%] in the first 6 months) of which 79.74% accounted for hospitalizations, 14.75% for outpatient care and 5.52% for indirect costs.Overall prescriptions accounted for 7.86% of the total cost.When considering all patients, mean (SD) cost per patient was EUR 16,687.35 (7661.35)and EUR 12,376.45 (4259.28)for overall and bleeding-related costs (EUR 9868.51 for inpatient bleeding costs), respectively.Cumulative costs for 3 years from the index date were also calculated according to the type of the first major bleed, the type of FXai and the FXai indication (Supplementary Table S7).With regard to the type of bleeding, mean total cost per patient was higher in the case of gastrointestinal bleeding, followed by intracranial hemorrhage, other bleeding and trauma-related bleeding.In the case of bleeding-related cost, the order was gastrointestinal bleeding, trauma-related bleeding, other bleeding and intracranial hemorrhage.Risk factors for the development of clinical events were specifically analyzed using Cox regression models (myocardial infarction, stroke, cardiovascular mortality and allcause death).A history of type 2 diabetes increased the risk of myocardial infarction, previous major bleeding the risk of stroke, and a history of coronary artery disease the risk of all-cause death (Supplementary Table S8).

Discussion
Our data showed that in patients initiating treatment with FXais in clinical practice in Spain, mainly for AF or VTE, around 5% developed a major bleed during follow-up.Despite the high use of replacement agents, the incidence of clinical events was high and one in twenty patients died.Moreover, bleeding-related HCRU and costs were substantially high, hospitalizations being the main determinant.
In Spain, a marked increase has been observed in the use of oral FXais in the last decade (from less than 2% in 2013 to nearly 25% in 2022), but use is lower compared to other European countries, likely due to the restrictions on reimbursement of DOACs, limited to only some specific situations among patients with AF and the lack of reimbursement in the case of patients with VTE [22,23].
Our patients were old (mean age 78 years), mainly women (61%) and with many comorbidities (79% had hypertension, one third type 2 diabetes, and 24% heart failure).In the phase 3 clinical trials comparing direct FXai vs. warfarin in an AF population (ROCKET-AF, ARISTOTLE and ENGAGE AF-TIMI 48), the mean age ranged from 70 to 73 years, 35-40% of patients were females, 90-94% had arterial hypertension, 25-40% type 2 diabetes and 35-63% heart failure [24][25][26].In contrast, patients with VTE included in the phase 3 clinical trials comparing direct FXais vs. standard therapy (EINSTEIN, AMPLIFY and Hokusai-VTE) were younger, and had fewer comorbidities [27][28][29].In this context, real-life population data are mandatory in order to extend the information given by clinical trials into clinical practice [30].Of note, the baseline clinical characteristics of our study are in line with those reported by other international studies [27][28][29], confirming the generalizability of our results.
In our study, out of 8972 patients taking direct FXais, 5.2% had a major bleed.In the ROCKET-AF, ARISTOTLE and ENGAGE AF-TIMI 48 trials, 5.6%, 3.6% and 7.5% of patients taking rivaroxaban, apixaban and edoxaban, respectively, had a major bleeding event during follow-up [24][25][26].In other studies performed in a real-life population, this proportion seemed lower [27][28][29].In our case, the proportion of patients with major bleeding was similar to that found in clinical trials, but higher than that reported in real-life studies, likely due to a less restrictive definition of major bleeding in our study [31].The use of the new DOAC score could be useful to stratify patients on the basis of expected bleeding risk [32].
Of note, 70% and 28.1% of patients died after a major trauma-related and intracranial bleeds, respectively, whereas gastrointestinal bleeding was more common (89% of the total) but was associated with lower mortality (2.2%).The mortality related to other major bleeding was also low (10.0%).In the phase 3 clinical trials in an AF population, among patients with a major bleeding event, 32-57% and 14-25% of cases had a gastrointestinal and intracranial origin, respectively [24][25][26].As a result, although the majority of patients survived after major bleeding in our study, there was still a substantial proportion of patients that presented a fatal evolution, especially patients with trauma and intracranial hemorrhage, despite replacement agents being administered in most cases during the bleeding event (prothrombin complex concentrates [63%], transfusion of blood products [20%] and Factor VIIa [8%]).In addition, surgical procedures used to correct bleeding were performed in nearly 19% of the cases of major bleeding.Remarkably, mortality rates were higher in those patients who were not treated with prothrombin complex concentrates, which were more frequently used in the case of intracranial bleeding.Guidelines recommend in case of non-life-threatening major bleeding among patients treated with FXais, supportive measures (i.e., mechanical compression, endoscopic or surgical hemostasis, fluid replacement, including red transfusion, platelet replacement or consideration of adjuvant tranexamic acid) and the treatment of factors and comorbidities contributing to bleeding.In the case of life-threatening bleeding or bleeding into critical sites, these guidelines recommend the use of andexanet alfa if available, otherwise the use of replacement agents, particularly prothrombin complex concentrates [1,12,33].Different studies have shown that in cases of major bleeding related to the use of FXais, the use of andexanet alfa can reduce anti-FXa activity, with high hemostatic efficacy in most subjects [34], including a prospective, randomized trial comparing andexanet alfa to usual care (mostly PCCs), resulting in improved hemostatic efficacy and better control of hematoma expansion in FXa inhibitor-related intracerebral hemorrhage [35].More recently, the ANNEXA-I trial has shown that among patients with intracerebral hemorrhage who were receiving FXai, compared to usual care, andexanet alfa resulted in better control of hematoma expansion.However, an increased risk of thrombotic events was also observed.Despite that, the net clinical benefit clearly favored the use of andexanet alfa in this population [35].Furthermore, although there may be some differences between reversal agents, it has been reported that the use of andexanet alfa for the reversal of anticoagulation in patients with FXai-related intracranial hemorrhage may be a cost-effective approach [36,37].In our study, rates of thrombotic events were also high among those patients treated with prothrombin complex concentrates.Unfortunately, at the time we performed our study, andexanet alfa was not available in Spain.Although specific studies are warranted, it is likely that the introduction of andexanet alfa in clinical practice could be associated with a reduction in mortality rates.
Our study showed that after a major bleeding event occurred, not only mortality but also morbidity rates increased.This was particularly relevant in patients with some conditions, such as type 2 diabetes, chronic kidney disease or prior coronary artery disease.Although the incidence rates of clinical events decreased over time, these data indicate that patients with AF or VTE, taking direct FXais, who present with major bleeding are at particularly high risk for clinical events and require a comprehensive management of all comorbidities to reduce the risk of developing complications during follow-up [1][2][3].
Our study showed that HCRU, including outpatient visits, laboratory/radiology investigations and hospitalization, was high after major bleeding, with a mean bleedingrelated length of hospital stay of 12 days.Previous studies have shown that among patients taking DOACs, HCRU is high after a major bleeding event, but lower when compared to patients taking warfarin that present with a major bleed [13].In this context, although specific studies are required, the use of a specific reversal agents, such as andexanet alfa with a rapid mechanism of action, could be useful to reduce HCRU after major bleeding.
The development of major bleeding in anticoagulated patients with either AF or VTE is associated with high healthcare costs [14,38,39].In our study, the total overall cost reached EUR 7,843,056.4at year 3, and EUR 5,816,930.5 in the case of bleeding-related costs (mean cost per patient of EUR 16,687.35 and EUR 12,376.45,respectively).Since nearly 80% of bleeding-related costs were accounted for by hospitalizations, and overall prescriptions for only 8% of the total cost, better management of major bleeding, including the use of specific reversal agents, and also the use of DOACs (vs.VKAs), may reduce total costs associated with major bleeding [40].On the other hand, for comparison purposes, modern data on the clinical and economic burden of patients suffering from major bleeding without anticoagulation would be desirable and are currently lacking.
Our study has some limitations.First, since this was a cohort study, with a retrospective design, only those variables included in the electronic health record could be collected, and some data could be lacking.In addition, as this was a retrospective study, there could be some confounders that may have impacted on the results, limiting the generalizability of our data.Despite that, the high sample size could decrease this limitation.Furthermore, no control group was available, and only indirect comparisons could be performed with other studies.Lastly, our results can only be applied to patients with similar clinical characteristics and in similar healthcare systems.
In conclusion, around 5% of subjects initiating treatment with a direct FXai, mainly for AF or VTE, developed a major bleeding event during follow-up.The management of these patients is challenging.In fact, despite the extended use of replacement agents, the incidence of major clinical events, bleeding-related HCRU and costs during the study period were high, and one in twenty patients died, evidencing the need for new reversal treatment strategies.

Figure 1 .
Figure 1.Flow chart of the study.

Figure 1 .
Figure 1.Flow chart of the study.

Table 2 .
Cumulative clinical outcomes in the overall study population and according to the use of prothrombin complex concentrates.
relative (%) frequencies; AMI: acute myocardial infarction; FXai; Factor Xa inhibitor.1.Time window: cumulative events from index date (day of the first major bleeding); 2. Incidence rate defined as the total number of incident events of interest divided by the total person time at risk.Mean CHA 2 DS 2 -VASc and HAS-BLED scores in the overall group and by FXai bleeding type are shown in Supplementary Table

Table 3 .
Cumulative outpatient visits and hospitalization for 3 years from index date.

Table 4 .
Cumulative length of hospital stays, number of prescriptions and work absences for 3 years from index date.

Table 5 .
Cumulative costs for 3 years from index date.