Keywords
venous thromboembolism - anticoagulants - risk assessment - hemorrhage
Background
Venous thromboembolism (VTE), a disease entity including deep vein thrombosis (DVT)
and pulmonary embolism (PE), is the third most common cardiovascular disease, following
coronary heart disease and ischemic stroke.[1]
[2] VTE is associated with significant morbidity due to acute symptoms and long-term
complications of DVT and PE, such as the postthrombotic or post-PE syndrome, and contributes
to a major global disease burden.[3]
[4]
[5] The incidence of VTE, PE (±DVT), and DVT alone is approximately 1 to 2, 0.6, and
0.9 per 1,000 inhabitants per year, respectively.[1]
[6]
[7] VTE affects all age groups; however, the incidence is increasing with age; for example,
the yearly incidence in people older than 55 years is 5 to 6 events per 1,000 persons.
The mainstay of VTE treatment is anticoagulation for at least 3 months. In patients
with a transient or reversible risk factor, treatment can be stopped after 3 months.
Extended anticoagulation therapy is suggested in patients at high risk of VTE recurrence
(e.g., patients with a persistent risk factor) and those in whom the index episode
occurred in the absence of any identifiable risk factor, the latter referred to as
unprovoked VTE.[7]
[8] The main complication of anticoagulation is bleeding, and major bleeding is the
complication that limits extended/long-term oral anticoagulation to prevent VTE recurrence.[9] In this narrative review, we aimed at providing an overview of bleeding risk and
discuss bleeding risk assessment tools, clinical factors, and biomarkers for prediction
of bleeding events in the VTE population.
Balancing Risk of VTE Recurrence versus Risk of Bleeding for Decision-Making
Balancing Risk of VTE Recurrence versus Risk of Bleeding for Decision-Making
The treatment of choice for the treatment of VTE and prevention of recurrence in patients
at high risk of VTE recurrence is anticoagulation. To determine the recurrence risk,
VTE is usually categorized into unprovoked and provoked. Provoked VTE events are further
divided into those by a transient or persistent risk factor. Recently, there has been
an effort in adopting the terminology from “unprovoked VTE” to VTE in the absence
of identifiable risk factors.[7] In patients without risk factors, risk of recurrence is approximately 10% at 1 year,
25 to 30% at 5 years, and 30 to 40% at 10 years after stopping oral anticoagulation.[9]
[10]
[11]
[12] In patients who had presented with a major transient risk factor, VTE recurrence
risk is only 1% after 1 year.[13] Accordingly, international guidelines suggest extended anticoagulation in patients
without identifiable risk factors (unprovoked VTE), with persistent risk factors,
or a minor transient or reversible risk factor.[7]
[8]
However, anticoagulation therapy comes at the expense of increasing the risk of bleeding,
the most feared complication of all currently available anticoagulants. The management
of bleeding may require reversal agents or other interventions and hospitalization,
which again may boost risk of VTE recurrence. Overall, occurrence of bleeding may
be associated with mortality and increased health care costs.[14]
In clinical practice, the decision to initiate and continue anticoagulation (e.g.,
for the prevention of VTE recurrence) is based on evaluation and balancing both risk
of recurrent VTE and risk of bleeding on anticoagulation. Furthermore, patient preference,
which is not further precisely defined yet, is also highlighted as an important factor
in the decision of extended/long-term treatment.[8]
To categorize the bleeding events, the International Society on Thrombosis and Haemostasis
(ISTH) has established criteria for the definitions of “major bleeding (MB),”[15] “clinically relevant non-major bleeding (CRNMB),”[16] and “non-clinically consequential minor bleeding” to assess the severity of bleedings
in nonsurgical studies. Major bleeding is defined as bleeding that occurs in a critical
organ such as intracranial, intraspinal, intraocular, retroperitoneal, intra-articular
or pericardial, or intramuscular with compartment syndrome or bleeding that lead to
a fall in hemoglobin of 2 g per deciliter or more, or leading to a transfusion of
two or more units of whole blood or red blood cells. Fatal bleedings are also categorized
as major bleedings.[15] Hemorrhage that does not fit the criteria of major bleeding but requires medical
intervention, leads to hospitalization, or requires face-to-face evaluation of a health
care professional is defined as CRNMB.[16] Other bleedings are referred to as nonclinically consequential minor bleedings.
Predicting major bleeding is crucial for clinical decision-making. However, CRNMB
is also important as patient-centric outcomes resulting in low quality of life may
lead to discontinuation of anticoagulation treatment.[15]
[16] CRNMB has become an important primary or secondary safety endpoint in clinical studies
of VTE as it reflects time-consuming management, extended medical care, and increased
treatment costs.[16]
[17] However, CRNMB may not be appropriate as a surrogate parameter for major bleeding.[18]
Patients on anticoagulation for VTE treatment are approximately at a 2% risk to develop
major bleeding during the first 3 months according to a meta-analysis published in
2003.[19] The introduction of direct oral anticoagulants (DOACs) in clinical practice has
impacted the discussion of bleeding risk. When compared with vitamin K antagonists
(VKAs), DOAC showed a better safety profile in randomized-controlled phase III trials,
in which consistent definitions of bleeding outcomes were reported across the different
studies, with an absolute major bleeding risk of 1.1% during the first 3 to 12 months
of treatment period.[20]
[21] However, in RIETE, a prospective registry study, 19% of patients had at least one
criterion (e.g., renal insufficiency, high risk of bleeding, pregnancy), which would
have excluded them from the randomized clinical trials of DOAC. Such patients had
substantially higher rates of VTE recurrences, major bleedings, and deaths.[22] Consequently, there is a need to further elaborate the bleeding risk in real-world
patients on DOAC. Of note, a systematic review even found major bleeding events in
patients receiving placebo (incidence: 0.42/100 patient-years), which should be taken
into account as the baseline risk of major bleeding without anticoagulation when assessing
risk of bleeding.[23]
Case-Fatality Rates of Major Bleeding and VTE Recurrence
Case-Fatality Rates of Major Bleeding and VTE Recurrence
Bleeding events most frequently occur during the first 3 to 6 months after the initial
event, while the risk of VTE recurrence increases after discontinuation of anticoagulation.[24]
[25] Interestingly, the majority of fatal events related to both, bleeding and recurrence,
are observed within the first month after starting therapy.[26] For clinical decision-making, case-fatality rates of major bleeding events and VTE
recurrence are usually taken into consideration. The case-fatality rate is a measure
of disease severity representing the proportion of patients who die from a specific
condition, over a certain period.
In a meta-analysis, risk of VTE recurrence was higher than major bleeding, while case-fatality
rate for both major bleeding and VTE recurrence was 11.3% during the first 3 months
of anticoagulation.[27] Accordingly, the initiation of anticoagulation and long-term treatment during the
first 3 months after the VTE event is crucial. However, case-fatality rate for recurrence
drops after the initial 3 months, while case-fatality remains stable for major bleeding.[19] The RIETE registry, a worldwide “all-comers” registry, found a case-fatality rate
of 2% for recurrent VTE and 18% for major bleeding after the first 3 months of anticoagulation
in a large cohort of 42,000 patients with a first VTE event.[26] This highlights the clinical relevance of major bleeding in VTE patients treated
in clinical practice. Therefore, extended (i.e., no scheduled stop date) anticoagulation
treatment is still a weak recommendation in the current guidelines.[8] Note, data on case-fatality rates after major bleeding in patients on anticoagulants
have been mostly reported in patients receiving VKA (mostly warfarin). Direct comparisons
of DOAC and VKA showed a better safety profile for DOAC which was also reflected by
significantly lower case-fatality rates of bleeding in the phase III trials (10.4
vs. 6.1%).[28]
To exemplify the importance of individual assessment, we provide the following example:
A patient suffering from first unprovoked VTE/PE faces a recurrence risk of approximately
25% and a case-fatality rate of 4% within 5 years, according to a recent meta-analysis.[9] Thus, his/her 5-year risk of death due to a recurrent VTE event would be approximately
1%. Given a 10% risk to die of major bleeding, his/her yearly major bleeding risk
estimation should be below 2% to show a mortality benefit for anticoagulation treatment.
Therefore, tools to predict major bleeding for the decision-making on extended/long-term
treatment are essential in clinical practice.
Prediction of Major Bleeding—Risk Factors, Risk Assessment Models, and Biomarkers
Prediction of Major Bleeding—Risk Factors, Risk Assessment Models, and Biomarkers
While assessment of bleeding risk in patients on anticoagulation is considered a vital
element of VTE management, its implementation is challenging. Common risk factors
for bleeding are older age, anemia, history of bleeding, abnormal renal function,
history of stroke, hypertension, antiplatelet agents, cancer, abnormal liver function,
alcohol abuse, female sex, diabetes, labile INR, poor anticoagulant control, thrombocytopenia,
increased fall risk, and nonsteroidal anti-inflammatory drugs (NSAIDs; [Fig. 1]).
Fig. 1 Bleeding risk assessment—balancing bleeding risk and recurrence risk. Numbers reflect
the risks within the first year of a patient suffering from first unprovoked VTE.
Risk factors on the right are modified according to the ACCP guidelines.[8]
Current guidelines do not recommend a specific approach to predict major bleeding.
The guidelines of the American College of Chest Physicians (ACCP/CHEST) suggest a
list of 18 risk factors to indicate high risk of bleeding,[8] while the 2019 European Society of Cardiology (ESC) guidelines lists 8 risk factors
and 5 prediction models (OBRI,[29] Kuijer et al,[30] RIETE,[31] HAS-BLED,[32] and VTE-BLEED[33]) to assess bleeding risk.[7] Furthermore, the ESC guidelines implicate to reassess the bleeding risk in high-risk
patients every 3 or 6 months. In total, 16 clinical prediction scores for major bleeding
are available.[34] Seven were developed specifically in VTE cohorts,[30]
[31]
[33]
[35]
[36]
[37] while 9 were developed in atrial fibrillation (AF)[32]
[38]
[39]
[40]
[41]
[42] or mixed cohorts.[29]
[43]
[44] The available tools differ in a variety of features. Most of the scores and models
are designed to predict major bleeding as defined by the ISTH. However, some were
developed to predict clinically relevant bleeding (major bleeding and CRNMB) or bleeding
events with different definitions. As some scores were developed in patients with
different indications for anticoagulation (e.g., AF), they may lack validity due to
different patient demographics and underlying disease. Furthermore, the vast majority
of scores had a derivation cohort of patients solely treated with VKA. Only three
were derived in a population including patients who were treated with DOAC for VTE.[33]
[35]
[45]
Assessment of major bleeding during anticoagulation varies throughout the different
stages of anticoagulation treatment and is, therefore, discussed separately.
Initial Bleeding Risk Assessment
Initial Bleeding Risk Assessment
In 1960, anticoagulation treatment has been shown to be effective for treating acute
VTE and prevent recurrent events for the first time.[46] Given the high case-fatality rate of acute VTE, all guidelines recommend anticoagulation
treatment for at least 3 months.[7]
[8] Only in patients presenting with subsegmental PE, or isolated distal DVT and without
severe symptoms or risk factors for extension, anticoagulation treatment might be
withheld. However, bleeding risk assessment in the acute phase is less relevant for
the decision to start anticoagulation but might be important to evaluate the use of
systemic thrombolysis, choose the appropriate anticoagulant drug,[47] and help identify patients at high risk for major bleeding during anticoagulant
treatment and, thereby, improve patient care. The optimal risk model for the initial
risk assessment should specifically characterize patients at very high risk of bleeding
in which withholding anticoagulation, despite the risk of thrombus extension and recurrent
VTE, might be acceptable. However, high risk of bleeding is often accompanied by high
risk of VTE recurrence. Therefore, an ideal bleeding risk assessment model would adjust
for the VTE recurrence risk and identify those at higher risk for major bleeding.
In the following, we discuss two selected tools for the initial bleeding risk assessment
in patients with VTE, which have been extensively studied and validated ([Table 1]).
Table 1
Clinical scores to predict major bleeding in patients with venous thromboembolism
|
HAS-BLED[32]
|
RIETE[31]
|
VTE-BLEED[33]
|
|
Derived in AF population
|
Derived in VTE population
|
Derived in VTE population
|
Risk factors
|
Age ≥ 60 y
|
|
|
1.5 points
|
Age > 65 y
|
1 point
|
|
|
Age > 75 y
|
|
1 point
|
|
History of bleeding
|
1 point
|
|
1.5 points
|
Recent bleeding
|
|
2 points
|
|
Active cancer
|
|
1 point
|
2 points
|
Abnormal renal function
|
1 point
|
1.5 points
|
1.5 points
|
Abnormal liver function
|
1 point
|
|
|
History of stroke
|
1 point
|
|
|
Anemia
|
|
1.5 points
|
1.5 points
|
Hypertension
|
1 point
|
|
1 point
|
Labile INR
|
1 point
|
|
|
Antiplatelets/NSAID
|
1 point
|
|
|
Alcohol abuse
|
1 point
|
|
|
Clinically overt PE
|
|
1 point
|
|
Risk stratification[a]
|
Low risk
|
0 points
|
0 points
|
0–2 points
|
Intermediate risk
|
1–2 points
|
1–4 points
|
|
High risk
|
3–9 points
|
4.5–8 points
|
2–9 points
|
Pros
|
• Best validated score for major bleeding
• Established in clinical practice
|
• Derived in “real-world” patients
• Consistent moderate predictive value throughout validation studies
|
• Extensively validated and studied in the VTE population
|
Cons
|
• Variable “labile INR” is rarely useful in extended VTE treatment
• Implementation of the variable “cancer” might be beneficial in VTE population
|
• Intended to predict bleeding in first 90 days
• Not sufficiently validated in DOAC patients
|
• Poor positive predictive value
|
Abbreviations: AF, atrial fibrillation; DOAC, direct oral anticoagulant; INR, international
normalized ratio; NSAID, nonsteroidal anti-inflammatory drugs; PE, pulmonary embolism;
VTE, venous thromboembolism.
a Percentage of patients with the respective bleeding outcome in each derivation cohort
stratified by risk: HAS-BLED, 0.59% in the low-risk, 1.7% in the intermediate-risk,
and 19.6% in the high-risk group; RIETE, 0.3% in the low-risk, 2.6% in the intermediate-risk,
and 7.3% in the high-risk group; VTE-BLEED, 2.8% in the low-risk and 12.6% in the
high-risk group.
HAS-BLED Score
In clinical practice, the HAS-BLED score is broadly used to assess bleeding risk in
patients on anticoagulation treatment. It was originally developed in 3,987 AF patients,
who were followed up for 1 year,[32] and has been validated in several AF cohorts, and is decently balanced in terms
of sensitivity and specificity.[48]
[49] Few studies also evaluated the HAS-BLED score in VTE patients. One study compared
8 bleeding risk scores and showed that HAS-BLED best predicted clinically relevant
bleeding (major bleeding and CRNMB) during the first 3 months.[50] However, in this study, none of the scores were better than chance in predicting
major bleeding alone. Two other studies showed that patients with a HAS-BLED score
≥3 were at increased risk for major bleeding, but major bleeding rates in the high-risk
group widely differed between 2.4 and 9.6% within the first 6 months of VTE treatment.[51]
[52] On the contrary, the HAS-BLED score performed poorly in a study of elderly patients
(≥80 years) with a c-statistic of 0.55.[53] Furthermore, the score's cutoff point for high risk of major bleeding is debatable.[54] A score of ≥4 indicated a clear delineation for those at high risk for major bleeding
with higher positive predictive values but very poor sensitivity.[51]
[52]
RIETE Bleeding Risk Score
The RIETE bleeding score consists of six variables (age >75 years, recent bleeding,
cancer, creatinine levels >1.2 mg/dL, anemia, and PE) and was derived and internally
validated in a registry with more than 19,000 VTE patients.[31] A variety of studies already validated the score and found decent discriminative
ability between the 3 bleeding risk categories, but poor to moderate predictive value.[50]
[53]
[55]
[56]
Risk Assessment for Extended Anticoagulation Treatment
Risk Assessment for Extended Anticoagulation Treatment
After the initial treatment phase, extended/long-term anticoagulation is recommended
in every patient who suffers from unprovoked VTE or VTE that was associated with a
persistent major thrombotic risk factor except in those at high risk for bleeding.
Therefore, a bleeding risk assessment for the decision and reevaluation of extended/long-term
anticoagulation treatment is of utmost importance for further patient management.
However, predictive values of current assessment tools vary substantially throughout
validation studies and independent validation studies mostly report c-statistics between
0.5 and 0.6 with most models not predicting better than chance in at least one study.[45]
[51]
[57]
[58]
[59]
[60]
[61] Importantly, most scores were evaluated in studies focusing on the initial 3 or
6 months of the treatment period. Only a few studies evaluated the performance of
clinical prediction scores for bleeding after the first months.[35]
[50]
[62] We are, therefore, discussing the only adequately validated score for the extended
anticoagulation period.
VTE-BLEED
VTE-BLEED, which consists of six clinical variables, is the only rule that was designed
and validated to predict bleeding events during “stable” (treatment period of 30 days
after the initial event) anticoagulation.[33] It was tested and validated in both patients with DOAC and VKA of two phase III
DOAC trials, one cohort, and one registry and showed good to moderate discrimination
between risk groups.[33]
[57]
[62]
[63] However, in the validation cohort, the positive predictive value of only 1.5% for
the high-risk group of the VTE-BLEED limits the decision to withhold anticoagulation
solely based on the score. Due to the high negative predictive value, the strength
of this score could lie in identifying patients with very low risk of major bleeding
and, therefore, safe administration of extended anticoagulation.
In conclusion, no score showed sufficient discriminative ability throughout the reported
studies. Thus, regardless of the predictive value of any bleeding score, withholding
anticoagulants is unacceptable in patients with acute VTE, including those with high
bleeding risk. Similarly, no currently available bleeding risk score for extended
anticoagulation can be recommended without doubts. Therefore, identification of further
risk factors and biomarkers for risk of bleeding is needed to improve bleeding assessment.
Nevertheless, bleeding scores can be used to identify patients at high risk and modify
and reduce possible risk factors. We suggest elaborating all risk factors listed by
the ACCP guidelines and modify or provide adequate treatment of blood pressure; improve
INR monitoring and control; stop the long-term use of NSAID or—if appropriate—of platelet
inhibitors; and check diabetes, anemia, renal, and liver function. Furthermore, patient
education and self-monitoring have been shown beneficial in patients treated with
VKA, and could lead to a decrease in bleeding risk.[64]
[65] Of note, a study of AF patients suggests that continuous bleeding risk assessment
better predicts bleeding than baseline assessment only.[66] Given that this may also be the case in VTE patients, continuous assessments might
help improve prevention of major bleeding.
Additionally, patient preferences as part of the individual decision-making should
be taken into account. A recent study investigated the patients' attitude toward secondary
prevention in VTE patients without an identifiable risk factor. Patients reported
the willingness to endure four major bleeds to prevent one recurrent event which highlights
the considerable fear of VTE in this population.[67] A similar ratio of major bleeding versus stroke was reported by a study conducted
in AF patients.[68]
Biomarkers for Risk Prediction
Biomarkers for Risk Prediction
In current bleeding prediction models for VTE patients, simple cutoff values for creatinine
clearance, creatinine, hemoglobin, and platelet count are used. However, continuous
markers overcome dichotomous variables and may better reflect susceptibility to bleeding.
Numerous biomarkers such as N-terminal pro-brain natriuretic peptide (NT-proBNP),
high-sensitivity cardiac troponin T (hs-cTnT), markers of renal function, hemoglobin,
low platelets, inflammatory markers (e.g., interleukin-6 and C-reactive protein),
growth differentiation factor-15 (GDF-15), vitamin E, D-dimer, von Willebrand factor,
and genetic polymorphisms have been associated with increased bleeding risk in cardiovascular
disease patients.[42]
[69]
[70]
[71]
[72]
[73]
[74] Only one bleeding risk score, which has been recently developed in AF patients,
has implemented continuous biomarkers to predict major bleeding. This new model, termed
“ABC-bleeding risk score,” includes three biomarkers associated with bleeding risk
(e.g., GDF-15, hs-cTnT, and NT-proBNP) and yielded higher c-indices than the HAS-BLED
score.[42]
To our knowledge, no predictive biomarker-based model has been developed for VTE patients.[75] However, the biomarkers of the ABC bleeding risk model including GDF-15 might be
a promising approach for the evaluation in VTE patients.
Conclusion
Medical care of patients suffering from VTE includes balancing two opposing aims,
namely, preventing VTE recurrence while minimizing risk of bleeding. As high risk
of bleeding might contradict anticoagulation treatment, clinical decision-making is
primarily based on estimating bleeding risk. Unfortunately, we are still far from
an optimal bleeding risk assessment tool. Although extensive research in the field
identified a variety of risk factors, current risk assessment models lack predictive
value and have not been tested in prospective interventional management studies. In
patients with acute VTE, anticoagulation is, without a doubt, the treatment of choice.
Current bleeding risk scores do not provide enough evidence to withhold anticoagulation.
However, they may provide information on critical patients. Such patients should be
followed up thoroughly to manage modifiable risk factors for bleeding and to assess
changes in bleeding risk over time. After the initial treatment phase, bleeding risk
scores (e.g., VTE-BLEED) may help in identifying patients with very low risk of bleeding
to safely administer extended anticoagulation, if indicated. Notably, traditional
risk factors seem to be simultaneously linked to bleeding and thromboembolic risk.
Therefore, repeated individual assessment of both risks, VTE recurrence and bleeding,
is key for extended secondary thromboprophylaxis.