Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients
Sam Schulman MD, HHS-General Hospital, 237 Barton Street East, Hamilton, ON, L8L 2X2, Canada.
Summary. A variety of definitions of major bleeding have been used in published clinical studies, and this diversity adds to the difficulty in comparing data between trials and in performing meta-analyses. In the first step towards unified definitions of bleeding complications, the definition of major bleeding in non-surgical patients was discussed at the Control of Anticoagulation Subcommittee of the International Society on Thrombosis and Haemostasis. Arising from that discussion, a definition was developed that should be applicable to studies with all agents that interfere with hemostasis, including anticoagulants, platelet function inhibitors and fibrinolytic drugs. The definition and the text that follows have been reviewed and approved by the cochairs of the subcommittee and the revised version is published here. The intention is to also seek approval of this definition from the regulatory authorities.
Hemorrhage is by far the most frequent complication of anticoagulant therapy, only remotely followed by other complications such as allergic reactions, thrombocytopenia, skin necrosis, purple toes (cholesterol microembolization) or alopecia. It is important to be able to make valid comparisons of the rates of bleeding between studies, particularly when performing systematic reviews. However, this task has been hampered by heterogeneity in the definition of bleeding in different studies . Bleeding events have been classified variously as ‘fatal’, ‘major’, ‘life-threatening’, ‘excessive’, ‘clinically significant’, ‘minor’, ‘overt’, ‘trivial’ or ‘nuisance’. Although it is of some interest to compare the frequency of all bleeding events, the emphasis is generally on the more clinically important outcome of ‘major bleeding’, which constitutes a primary endpoint for the evaluation of safety in most trials on antihemostatic agents.
The perspective of regulatory authorities
The European Agency for the Evaluation of Medicinal Products (EMEA) has issued two statements that relate to classification of bleeding with antihemostatic medications: (A) ‘Points to consider on clinical investigation of medicinal products for prophylaxis of intra- and postoperative venous thromboembolic risk’; and (B) ‘Notes for guidance on clinical investigation of medicinal products for the treatment of venous thromboembolic disease’. Their objective is to provide advice for the development of medicinal products; document A contains a ‘recommendation’ for the definition of major bleeding, whereas document B provides some criteria for major bleeding that ‘can be used’. In addition to statement A providing a ‘recommendation’, whereas statement B provides only a ‘suggestion’, there is the difference that intraspinal and intraocular hemorrhages are listed as critical sites that qualify as major bleeding in statement B but not in statement A. However, it seems even more appropriate to include intraspinal hemorrhage in a document that addresses postoperative use of antihemostatic agents in view of the frequent use of neuraxial anesthesia in this setting.
The Food and Drugs Administration (FDA) of the United States has not issued guidelines for definition of major bleeding and accepts different definitions. However, the FDA encourages the use of definitions that have been applied previously.
In the chapter on ‘Hemorrhagic complications of anticoagulant treatment’ in the latest update of the American College of Chest Physicians (ACCP) Consensus Conference on Antithrombotic Therapy, the variation in the rates of major hemorrhage between trials is evident (Table 1) . Although this variability is explained partly by different patient populations, intensities of treatment, lack of validation of claimed international normalized ratio (INR), timing of anticoagulant therapy in relationship to cardiac surgery and combinations of antihemostatic agents, inconsistency in the definition of major bleeding adds to these differences.
Table 1. Range of rates of major hemorrhage on anticoagulant therapy (vitamin K antagonists or heparins) in percentage per year in randomized clinical trials according to indication 
|Prosthetic heart valves||1.0–19.2|
|Acute ischemic coronary syndrome||0–6.8|
|Ischemic heart disease, long-term||0.6–14.5|
|Venous thromboembolism, initial treatment||0–7.0|
|Venous thromboembolism, long-term||0–16.7*|
The objective of this ISTH SSC statement is to provide a uniform definition of major bleeding in the evaluation of antihemostatic agents, including inhibitors of coagulation, inhibitors of platelet function and fibrinolytic agents, for acute treatment of thromboembolism or for long-term prophylaxis against thromboembolism. Long-term prophylaxis may be primary, such as in patients with prosthetic heart valves and atrial fibrillation, or secondary, such as after systemic arterial embolism, myocardial infarction or venous thromboembolism. This definition is not intended to include postoperative bleeding in association with prophylaxis against venous thromboembolism, because that requires a definition of major bleeding at the surgical site. A working group is currently validating surgical site hemorrhage and specific guidelines on this topic are anticipated. Furthermore, we have not attempted to define non-major bleeding in this statement. Another working group will evaluate clinically significant non-major bleeding in four ongoing trials with the long-acting pentasaccharide, idraparinux. That program will include about 12 000 patients, many of whom will be treated with vitamin K antagonists, and approximately 1200 of bleeds of this type are anticipated.
The following definition is based on a discussion held at the meeting of the subcommittee of Control of Anticoagulation at the 50th Scientific and Standardization Committee Meeting of the ISTH in Venice, 18 June 2004. Presenters at this session included R. Beyth, C. Francis, C. Kearon, L. Linkins, M. Prins and S. Schulman (C. Kearon and S. Schulman were cochairs). Attendees also contributed comments, and a general principle reached received wide acceptance. A draft definition was then sent to the cochairmen of the subcommittee, who were requested to review and comment on the text within 1 month. Responses were obtained from six cochairpersons and the text has been revised by taking this input into account.
It is intended that these guidelines will be presented to the European Agency for the Evaluation of Medicinal Products (EMEA) and the United States Food and Drugs Administration (FDA), and it is hoped that their widespread use in future clinical trials will facilitate comparisons between trials and the performance of systematic reviews of the literature.
Definition of major bleeding
As general principles, a definition of major bleeding needs to be based on objective criteria, and major bleeds are those that result in death, are life-threatening, cause chronic sequelae or consume major health-care resources. With this in mind, the Control of Anticoagulation Subcommittee recommends the following criteria for major bleeding in non-surgical patients:
- 1Fatal bleeding, and/or
- 2Symptomatic bleeding in a critical area or organ, such as intracranial, intraspinal, intraocular, retroperitoneal, intra-articular or pericardial, or intramuscular with compartment syndrome, and/or
- 3Bleeding causing a fall in hemoglobin level of 20 g L−1 (1.24 mmol L−1) or more, or leading to transfusion of two or more units of whole blood or red cells.
The following comments should be made in relationship to the above definition.
Fatal bleeding has not always been counted among major bleeds in previous trials [5,6]; instead, these bleeds have sometimes been included among reported deaths without identifying that bleeding was the cause of death. Similarly, intracranial bleeding has not always been counted among major bleeds; instead, these bleeds have sometimes been included as strokes without identifying that bleeding had occurred. The practice of including such bleeds as part of composite outcomes without also identifying that bleeding is the cause of these outcomes should be avoided.
All major bleeds should be described in a trial report. If this is not feasible, as a minimum requirement authors should specify the number of fatal bleeds and intracranial bleeds that are included among the total number of major bleeds.
In order for bleeding in a critical area or organ to be classified as a major bleed it must be associated with a symptomatic clinical presentation. For example, it is not uncommon that patients with venous thromboembolism have computed tomography scanning to detect a malignancy. If, in the absence of symptoms to suggest bleeding, a small retroperitoneal bleed was found on such an examination it would not be classified as a major bleed.
In addition, in patients with symptoms and bleeding [e.g. present on a computerized tomography (CT) scan], it is recognized that bleeding may not be the cause of the symptoms (e.g. symptoms may be due to known cancer). If symptoms are thought to be unrelated to bleeding, patients with symptoms and bleeding in a critical area or organ need not be considered to have had a major bleed. Consequently, it is acknowledged that subjective assessments by the clinical centers and by the central adjudicators may influence the assessment of bleeding severity.
The guidelines of EMEA have also included ‘bleeding causing permanent treatment cessation’ among the criteria for major bleeding. However, this decision is extremely subjective and is influenced by a variety of factors other than the severity of bleeding. Therefore, this criterion has been excluded.
Some classification schemes have included ‘bleeding resulting in admission to hospital’ or ‘bleeding resulting in serious medical complications (e.g. myocardial infarction or ischaemic stroke)’ among the criteria for major bleeding. These criteria have not been included in the definition of major bleeding described above for the following reasons: (i) admission to hospital is influenced strongly by other factors such as support in the community, hospital bed availability and presence of comorbidity; and (ii) bleeding is extremely unlikely to be the primary cause of serious medical consequences without also satisfying one of the included criteria for major bleeding, and inclusion of this criterion has the potential to falsely classify a minor bleed as a major bleed because of the coincidental occurrence of conditions such as myocardial infarction or ischemic stroke.
‘Bleeding resulting in surgical intervention’ has not been included as a separate criterion as, if no other criteria for major bleeding have been satisfied, performance of surgical interventions will often be minor and not justify categorization as a major bleed.
As noted previously, even with the use of a standard definition of major bleeding, there is a subjective component to the assessment of the presence of bleeding and its severity, both at the clinical center and central adjudication levels. Furthermore, factors other than antithrombotic therapy and the method of assessment of bleeding also influence the frequency of bleeding in clinical trials, such as differences in the patient population (e.g. age, presence of comorbidity). For these reasons, the most valid comparison of frequency of bleeding with different antithrombotic therapies is obtained from double-blind randomized comparisons (i.e. within the same study).