Antifibrinolytics (lysine analogues) for the prevention of bleeding in people with haematological disorders

Summary of findings for the main comparison. Antifibrinolytics (lysine analogues) compared to placebo to prevent bleeding in people with haematological disorders

Antifibrinolytics (lysine analogues) compared to placebo to prevent bleeding in people with haematological disorders
Patient or population: people with haematological disorders Settings: hospital Intervention: antifibrinolytics (lysine analogues) Comparison: placebo
Outcomes	Illustrative comparative risks^a (95% CI)		Relative effect (95% CI)	No of participants (studies)^b	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Placebo	Antifibrinolytics (lysine analogues)
Number of participants with any bleeding	See comment	See comment	Not estimable	56 (1 study)	⊕⊝⊝⊝ very low^1,2	We are uncertain whether antifibrinolytics reduce bleeding. Only Shpilberg 1995 reported the number of bleeding episodes per participant, results were reported separately for induction and consolidation chemotherapy and no difference was seen.
Number of participants with thromboembolism	See comment	See comment	Not estimable	68 (2 studies)	⊕⊝⊝⊝ very low^1,2	We are uncertain whether antifibrinolytics increase the risk of thromboembolism. No patients within the Avvisati 1989 or Shpilberg 1995 studies had an episode of thromboembolism. Gallardo 1983 only reported no deaths due to thrombosis.
Mortality (all‐cause) ‐ not reported	Not estimable		Not estimable	0 (0)	See comment	None of the studies reported all‐cause mortality. Shpilberg 1995 reported no deaths due to bleeding and Gallardo 1983 reported no deaths due to thrombosis.
Number of platelet transfusions per participant ‐ not reported	Not estimable		Not estimable	0 (0)	See comment	None of the studies reported the number of platelet transfusions per participant
Adverse events of transfusions ‐ not reported	Not estimable		Not estimable	0 (0)	See comment	None of the studies reported adverse events due to blood transfusions
Adverse events of antifibrinolytic agents	See comment	See comment	Not estimable	74 (2 studies)	⊕⊝⊝⊝ very low^1,2	We are uncertain whether antifibrinolytics increase the risk of adverse events. Shpilberg 1995 reported no side effects were observed and Gallardo 1983 stated side effects were minimal but no further information was provided
Quality of life ‐ not reported	Not estimable		Not estimable	0 (0)	See comment	None of the studies reported quality of life.
^aNo meta‐analyses were performed within this review and therefore no comparative risks could be calculated. ^bOne study (Fricke 1991) was included within the review but no data were extracted from this study. This was because Fricke 1991 had significant flaws within the study design (see Risk of bias in included studies and Characteristics of included studies) and no viable data could be extracted from the study report. CI: confidence interval
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹We downgraded the quality of the evidence by 1 for risk of bias. A full assessment of the quality of the evidence was limited by a lack of reporting. However, selective outcome reporting was present in Gallardo 1983. ²We downgraded the quality of evidence by 2 for imprecision. Within all three studies there were only 86 participants. This is significantly below the optimal information size (OIS) (Pogue 1997).

Background

Description of the condition

People with haematological disorders are frequently at risk of severe or life‐threatening bleeding as a result of thrombocytopenia (reduced platelet count). This is commonly a result of the underlying pathology, a side effect of treatment with chemotherapeutic agents, or both. People are administered therapeutic platelet transfusions to treat bleeding and prophylactic platelet transfusions to prevent bleeding once the platelet count falls below a certain threshold (10 x 10⁹/L, or higher if the risk of haemorrhage is raised, e.g. sepsis or the presence of another bleeding diathesis) (BCSH 2003). A large, recent randomised controlled trial (RCT) of platelet transfusions involving 1272 participants with haematological or solid tumours showed that the baseline level of clinically significant bleeding does not appear to depend on the platelet count once it is above 5 x 10⁹/L. The risk of bleeding was 17% per day at platelet counts above 5 x 10⁹/L, compared with 25% per day at counts below this level (Slichter 2010). This confirms earlier work from a large, retrospective observational study of 2942 adults with thrombocytopenia that showed that the platelet count did not affect the risk of severe or life‐threatening bleeding (World Health Organization (WHO) scale 3 or 4) (Friedmann 2002).

Platelet transfusions are not without risks. Adverse events may range from mild reactions, such as fever (one in five transfusions) (Heddle 2009), to more serious and even life‐threatening events such as bacterial sepsis from transfusion‐transmitted infection (one in 10,000 transfusions) (Heddle 2009), or transfusion‐related acute lung injury (TRALI) (Popovsky 1985). Patients may also become refractory to platelet transfusions; the incidence increases with the number of platelet transfusions administered (Slichter 2005). Once refractory, the ability to treat bleeding with platelet transfusions becomes more difficult, requiring expensive, specially matched platelets that can be difficult to source. Furthermore, the financial cost of platelet transfusions is considerable. Around 302,000 adult doses of platelets are issued in the UK each year (Bolton‐Maggs 2012) at an annual cost of approximately GBP 68.5 million (Llewelyn 2009) and up to two‐thirds (67%) of these are given to people with haematological malignancies (Cameron 2007; Greeno 2007; Pendry 2011).

Clearly interventions that can safely prevent bleeding in people with thrombocytopenia, whilst minimising exposure to allogeneic platelets and reducing financial costs, would be welcomed. One possible adjunct, or even an alternative, to prophylactic platelet transfusions is the use of antifibrinolytics, specifically lysine analogues, such as tranexamic acid (TXA) and epsilon aminocaproic acid (EACA).

This systematic review has been designed to establish the safety and efficacy of these agents specifically in people with haematological disorders who are at risk of thrombocytopenia and bleeding, either due to the disorder itself, its treatment, or both.

Description of the intervention

There have been several Cochrane reviews examining the efficacy of antifibrinolytics in preventing bleeding in other patient groups (Gurusamy 2011; Henry 2011; Martin‐Hirsch 2010; Novikova 2011; Roos 2008; Tzortzopoulou 2008). The largest of these involved over 25,000 participants, and assessed the use of TXA, EACA, and another type of antifibrinolytic, aprotinin, with respect to the minimisation of perioperative allogeneic blood transfusions (Henry 2011). However, no systematic review has addressed the use of antifibrinolytics to prevent bleeding in people with haematological disorders.

TXA and EACA are effective in surgical patients (Henry 2011; Ker 2012). They have been used widely in both elective and emergency surgery and have been shown to reduce both blood loss and the need for blood transfusions. In the largest Cochrane review, 65 trials compared TXA with control and comprised a total of 4842 participants of whom 2528 were randomised to TXA. TXA versus control showed a relative reduction in the need for allogeneic blood transfusion of 39% (Henry 2011). A significant effect was also observed for EACA; in 16 trials comparing EACA with control (with a total of 1035 participants, of whom 530 were randomised to EACA), there was a relative reduction in the need for allogeneic blood transfusion of 19% (Henry 2011). In the literature there appears to be a paucity of direct comparisons between TXA and EACA but they appear comparable in terms of safety and efficacy. A recent study comparing TXA and EACA in 234 children undergoing cardiac surgery found no significant differences between the two in terms of transfusion requirement, rates of revision for re‐bleeding, postoperative complications (such as seizures, renal failure and thrombosis) and in‐house mortality (Martin 2011).

TXA is effective in trauma patients. In a recent large RCT (CRASH‐2), TXA has been shown to significantly reduce the risk of death due to bleeding in trauma patients with significant haemorrhage (Shakur 2010). TXA has also been found to be highly cost‐effective: it is relatively inexpensive and its use in preventing bleeding may obviate the need for additional transfusion of blood products and longer stays in hospital. In a recent cost‐effectiveness analysis of the CRASH‐2 trial, Guerriero 2011 reported that the administration of TXA within three hours of injury to bleeding trauma patients has been estimated to save 755 life‐years (LYs) per 1000 trauma patients in the UK, and the incremental cost of giving TXA versus not giving TXA was estimated at USD 48,002 in the UK, equivalent of a cost of around USD 64 per life‐year saved.

TXA and EACA are commonly used to treat bleeding in patients with haematological disorders (Lozano 2013). They are also used to prevent bleeding in patients who are refractory to platelet transfusions (Lozano 2013). It therefore seems possible that lysine analogues may also be cost‐effective in preventing bleeding in patients with haematological disorders with severe thrombocytopenia who are not refractory to platelet transfusions.

TXA and EACA are the only antifibrinolytics in common use. Aprotinin, a naturally occurring serine protease inhibitor, was once commonly used as a blood‐sparing agent, particularly in cardiac surgery. However, it is now used rarely due to concerns of an increased risk of cardiovascular complications and death (Henry 2011). This was because the BART (Blood Conservation Using Antifibrinolytics in a Randomized Trial) multi‐centre blinded RCT was terminated early when a higher rate of death was seen in patients receiving aprotinin (Fergusson 2008). This study was designed to determine whether aprotinin was superior to either TXA or EACA in decreasing massive postoperative bleeding in patients undergoing high‐risk cardiac surgery. A modest and non‐significant reduction in the risk of massive bleeding was observed in the aprotinin arm compared to TXA or EACA, but the rate of death from any cause at 30 days was 6.0% in the aprotinin group, compared with 3.9% (relative risk 1.55; 95% CI 0.99 to 2.42) and 4.0% (relative risk 1.52; 95% CI 0.98 to 2.36) in the TXA and EACA groups, respectively. The authors concluded the negative mortality trend associated with aprotinin, as compared with the lysine analogues, precluded its use in high‐risk cardiac surgery (Fergusson 2008).

Although TXA and EACA have been shown to be effective in other patient groups, there is a concern that these drugs may increase the rate of thromboembolism (Henry 2011). This is particularly important in haematology patients, as patients with an underlying malignancy already have a higher rate of thromboembolic disease than the general population. In a retrospective cohort study of thromboembolism in hospitalised neutropenic cancer patients, 4% (593/14,600) of acute leukaemia patients developed venous thromboembolism and 1.9% (279/14,600) of acute leukaemia patients developed arterial thromboembolism (Khorana 2006).

Furthermore, TXA and EACA may increase the risk of disseminated intravascular coagulation (DIC). In a subsequent exploratory analysis of the CRASH‐2 trial, late treatment with TXA (> three hours) seemed to increase the risk of death in trauma patients due to bleeding (Roberts 2011). The mechanism underlying this could not be readily explained, but the authors noted that one possibility related to the evolution of DIC, a condition in which lysine analogues could be contraindicated (Prentice 1980; Roberts 2011; Sawamura 2009). This highlights a serious need for caution in the use of these agents in patients with haematological malignancies as they are at increased risk of DIC (Franchini 2010). Overt cases of DIC are diagnosed in approximately 15% of people with acute leukaemia and bleeding manifestations tend to prevail over thrombosis (Franchini 2010).

Despite these important concerns, it should be noted that in the recent large Cochrane review of over 25,000 participants (Henry 2011), the use of TXA or EACA was not associated with an increased risk of mortality, myocardial infarction, deep vein thrombosis, stroke, incidence of renal dysfunction or length of hospital stay, although the data were sparse. In addition, there have been small RCTs assessing the efficacy of TXA versus placebo in haematology patients that did not report an increase in thromboembolic complications, although sample sizes were small (Avvisati 1989; Shpilberg 1995).

How the intervention might work

TXA and EACA are synthetic analogues of the amino acid lysine that act by blocking the lysine binding sites on plasminogen. This inhibits the formation of plasmin and therefore prevents fibrinolysis, leading to improved haemostasis (Okamoto 1997). In vitro TXA is approximately 10 times more potent than aminocaproic acid and binds much more strongly to the sites on the plasminogen molecule (Faught 1998). It is plausible that if these lysine analogues are effective and safe, the bleeding risk in people with haematological disorders could be reduced, and the requirement for prophylactic platelet transfusions could be minimised.

Why it is important to do this review

Clearly, it is essential to reduce the risk of bleeding in patients with haematological disorders and thrombocytopenia as effectively and as safely as possible. Since the CRASH‐2 trial (Roberts 2011) and two large systematic reviews (Henry 2011; Ker 2012) have shown antifibrinolytics to be effective in other patient groups there has been renewed interest in using this drug to prevent bleeding in patients with haematological disorders.

The following key questions need to be addressed.

What is the efficacy of lysine analogues in preventing bleeding people with haematological disorders who are thrombocytopenic?
Can the number of prophylactic platelet transfusions be minimised?
Does the use of lysine analogues lead to a significant increase in the incidence of thromboembolism?

If lysine analogues are shown to be effective whilst demonstrating an acceptable safety profile, there would be a strong case for their routine use in patients with haematological disorders at significant risk of severe thrombocytopenia. A systematic review is therefore required before any proposed introduction of these agents in patients with haematological disorders.

Objectives

To determine the efficacy and safety of antifibrinolytics (lysine analogues) in preventing bleeding in people with haematological disorders.

Methods

Criteria for considering studies for this review

Types of studies

We only included RCTs in this review, irrespective of language or publication status.

Types of participants

People of any age, with a haematological disorder (malignant or non‐malignant) who were severely thrombocytopenic due to bone marrow failure (secondary to the disease or to its treatment) and required platelet transfusions. We excluded people with immune thrombocytopenic purpura (ITP) because they are not usually treated with platelet transfusions.

Types of interventions

We only reviewed antifibrinolytic agents that act by competitively inhibiting the conversion of plasminogen to plasmin (lysine analogues), i.e. tranexamic acid (TXA) and epsilon aminocaproic acid (EACA). Aprotinin is a serine protease and has a different mechanism of action. We included the following comparisons:

TXA versus placebo;
EACA versus placebo;
TXA versus EACA.

We included any dose of the medication, administered either orally or intravenously.

Types of outcome measures

Primary outcomes

Number, site and severity of bleeding (i.e. any bleeding, clinically significant bleeding, life‐threatening bleeding)
Thromboembolism (venous and arterial)

Secondary outcomes

Mortality (all causes)
Mortality (secondary to bleeding)
Mortality (secondary to thromboembolism)
Laboratory assessment of fibrinolysis
Number of platelet transfusions or platelet components
Number of red cell transfusions or red cell components
Adverse events of antifibrinolytic agents
Adverse events of transfusions (e.g. transfusion reactions, antibody development)
Disseminated intravascular coagulation (DIC)
Quality of life (QoL)

We listed both primary outcomes in the 'Summary of findings' table, as well as the number of red cell and platelet transfusions.

Search methods for identification of studies

We formulated search strategies in collaboration with the Cochrane Haematological Malignancies Group.

Electronic searches

The Systematic Review Initiative Information Specialist (CD) updated search strategies in collaboration with the Cochrane Haematological Malignancies Review Group based on those used in the previous version of this review (Wardrop 2013). We searched for relevant RCTs in the following electronic databases:

Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2016, Issue 3 to 07 March 2016) (Appendix 1);
MEDLINE (from 1948 to 07 March 2016) (Appendix 2);
Embase (from 1974 to 07 March 2016) (Appendix 3);
CINAHL (from 1937 to 07 March 2016) (Appendix 4);
PubMed (e‐publications as of 07 March 2016 only) (Appendix 5);
LILACS (from 1982 to 07 March 2016) (Appendix 6);
KoreaMed (from 1982 to 07 March 2016) (Appendix 6);
PakMediNet (from 2001 to 07 March 2016) (Appendix 6);
IndMed (from 1985 to 07 March 2016) (Appendix 6);
Transfusion Evidence Library (www.transfusionevidencelibrary.com) (from 1950 to 07 March 2016) (Appendix 7);
Web of Science, Conference Proceedings Citation Index (from 1990 to 07 March 2016) (Appendix 8).

We updated searches from the original search in January 2013 (Wardrop 2013). We did not re‐screen the original search strategies and instead placed date restrictions from the date of the final search in the preceding review (10 January 2013) to 07 March 2016 for four databases (CENTRAL, MEDLINE, Embase and CINAHL). The other databases had no date restrictions.

We combined searches in MEDLINE, Embase, and CINAHL with adaptations of the Cochrane RCT search filters, as detailed in the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2011).

Databases of ongoing trials

We also searched ongoing trial databases (all years) on 07 March 2016:

ClinicalTrials.gov (Appendix 9);
WHO International Clinical Trials Registry Platform (ICTRP) (Appendix 9);
ISRCTN Register (Appendix 10);
EU Clinical Trials Register (EUDRACT) (Appendix 10);
UMIN‐CTR Japanese Clinical Trials Registry and the Hong Kong Clinical Trials Registry (Appendix 10).

Searching other resources

We augmented database searching with the following.

Handsearching of reference lists
- We checked references of all identified trials, relevant review articles and current treatment guidelines for further literature.
- We limited these searches to the 'first generation' reference lists.
Personal contacts
- We contacted authors of relevant studies, study groups and experts worldwide who are known to be active in the field for unpublished material or further information on ongoing studies.

Data collection and analysis

Selection of studies

The selection of studies was updated from the selection of studies performed for the original version of this review (Wardrop 2013).

Two review authors (MD, LE) screened independently all electronically‐derived citations and abstracts of papers identified by the review search strategy for relevance. Studies clearly irrelevant were excluded at this stage. The full texts of all potentially‐relevant trials were formally assessed independently for eligibility by two review authors (MD, LE) against the criteria outlined above. All disagreements were resolved by discussion without the need for a third review author (SS). Further information was sought from study authors if the article contained insufficient data to make a decision about eligibility. A study eligibility form was designed for this review to help in the assessment of relevance. The reasons why potentially‐relevant studies failed to meet the eligibility criteria were recorded.

Data extraction and management

We updated the data extraction from the data extraction performed for the previous version of this review (Wardrop 2013). This included data extraction for the new ongoing studies included since the previous review.

Two review authors (LE, MD) conducted the data extraction according to the guidelines proposed by The Cochrane Collaboration (Higgins 2011a). Potential disagreements between the review authors were resolved by consensus. The review authors were not blinded to names of authors, institutions, journals, or the outcomes of the trials. The data extraction forms were piloted in the previous version of this review (Wardrop 2013). We were not blinded to the names of authors, institutions, journals or the outcomes of the trials. We used a standardised data extraction form to record the following items.

General information: review author's name, date of data extraction, study ID, first author of study, author's contact address (if available), citation of paper, objectives of the trial.
Trial details: trial design, location, setting, sample size, power calculation, inclusion and exclusion criteria, reasons for exclusion, comparability of groups, length of follow‐up, stratification, stopping rules described, results, conclusion and funding.
'Risk of bias' assessment: sequence generation, allocation concealment, blinding (participants, personnel, outcome assessors), incomplete outcome data, selective outcome reporting and other sources of bias.
Characteristics of participants: age, gender, ethnicity, total number recruited, total number randomised, total number analysed, types of haematological disease, lost to follow‐up numbers, drop outs (percentage in each arm) with reasons, protocol violations, current treatment, previous treatments.
Interventions: experimental and control interventions, type of antifibrinolytic given, timing of intervention, compliance to interventions, additional interventions given especially in relation to platelet and red cell transfusions, any differences between interventions.
Outcomes measured: number, site and severity of bleeding episodes; thromboembolism (venous and arterial); mortality (all causes); mortality due to haemorrhage; mortality due to thromboembolism; laboratory assessment of fibrinolysis; number of platelet transfusions; number of red cell transfusions; adverse effects of antifibrinolytic agents; adverse effects of transfusions (e.g. transfusion reactions, development of platelet antibodies); DIC.

We retrieved the data from both full‐text and abstract reports of studies. Where these sources did not provide sufficient information, we contacted authors and study groups for additional details.

Assessment of risk of bias in included studies

The risk of bias assessment was not updated from the previous version of this review because no new completed studies were identified (Wardrop 2013). We assessed all completed studies for possible risk of bias as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c). The assessment included information about the design, conduct and analysis of the trial. We evaluated whether the studies are at a low risk, high risk or unclear risk of bias. To assess risk of bias, the following questions were included in the 'Risk of bias' table for each included study:

Was the allocation sequence adequately generated?
Was allocation adequately concealed?
Was the knowledge of the allocated intervention adequately prevented during the study?
Were incomplete outcome data adequately addressed?
Are reports of the study free of selective outcome reporting?
Was the study apparently free of other problems that could put it at risk of bias?

Measures of treatment effect

We performed this according to the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). For dichotomous outcomes we recorded the numbers of outcomes in treatment and control groups.

For continuous outcomes, we recorded the mean and standard deviations.

Unit of analysis issues

We did not pre‐specify in the protocol how we would deal with any unit of analysis issues. No unit of analysis issues arose in this review.

Dealing with missing data

We performed this according to the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). We contacted one author by email in order to obtain information that was missing or unclear in the published report. This author responded to our email request but was unable to provide any further information (Gallardo 1983). We were unable to contact the other authors. We recorded the number of patients lost to follow‐up for each trial.

Assessment of heterogeneity

We did not perform a formal assessment of heterogeneity because it was not possible to perform meta‐analyses due to the nature of the data reported by the included studies. (Deeks 2011).

Assessment of reporting biases

We did not perform a formal assessment of reporting biases because there were not enough data to support such an assessment and no meta‐analyses of outcome data were performed (Sterne 2011).

Data synthesis

We performed a narrative synthesis of the findings from the included studies, structured around the type of antifibrinolytic. No statistical analyses were performed because the studies reported outcomes in different ways and these results could not be integrated.

We used the GRADE profiler to create a 'Summary of findings' table as suggested in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2011). We did not pre‐specify in the protocol the outcomes to be reported in the 'Summary of findings' table (Wardrop 2012).

Subgroup analysis and investigation of heterogeneity

We pre‐specified three subgroup analyses.

Age (children/adults).
Underlying haematological diagnoses.
Type of treatment (e.g. chemotherapy, autologous and allogeneic transplantation, immunosuppression).

However, we did not perform any subgroup analyses due to a lack of outcome data. The three studies included in the outcome analysis only included adults who had acute leukaemia who were receiving chemotherapy.

Sensitivity analysis

We did not perform a formal sensitivity analysis because we performed no meta‐analyses.

Results

Description of studies

See Characteristics of included studies; Characteristics of excluded studies and Table 1.

Table 1. Study characteristics

Study

Type of study

No. of participants

No. of participants receiving antifibrinolytic

No. of participants platelet refractory/ allo‐immunisation

Diagnosis of patients

Treatment of underlying disease

Antifibrinolytic dose

Antifibrinolytic frequency

Antifibrinolytic route

Treatment started

Treatment stopped

Platelets given

Tranexamic acid studies

RCT

APML

Chemo

2 g

8‐hourly

IV in 500 mL 5% glucose

1st day of antileukaemic Rx

After 6 days

Prophylactic

Fricke 1991

RCT Cross‐over

Only 3 completed study

At least 3

7 AA

1 MDS

20 mg/kg

8‐hourly

Oral

After 4‐day trial period to assess drug tolerance

Successive 4/52 courses or until WHO grade 2 bleeding

Therapeutic

RCT

AML

38 induction

18 consolidation

Chemo

1 g

6‐hourly

Oral

Platelets < 20 or rapidly falling and < 50

Platelet count > 20 for 2 consecutive counts

Therapeutic

EACA studies

See: Summary of findings for the main comparison Antifibrinolytics (lysine analogues) compared to placebo to prevent bleeding in people with haematological disorders

RCT

15 AML

4 ALL

(1 patient not evaluable)

Chemo

Loading dose 100 mg/kg

12 to 24 g/day thereafter

Platelet count < 20 x 10⁹/L

Platelet count ≥ 20 x 10⁹/L

Platelet count < 20 x 10⁹/L

AA: aplastic anaemia
ALL: acute lymphoblastic leukaemia
AML: acute myeloid leukaemia
APML: acute promyelocytic leukaemia
EACA: epsilon aminocaproic acid
IV: intravenous
MDS: myelodysplastic syndrome
NR: not reported
RCT: randomised controlled trial
TXA: tranexamic acid
WHO: World Health Organization

Results of the search

See PRISMA diagram Figure 1. The original search (conducted January 2013) identified 953 records through database searching with an additional 12 records identified through other sources (principally the handsearching of reference lists of included studies). After duplicates were removed, we screened 470 records in abstract form for eligibility and excluded 436 records. Of the remaining records, we retrieved and assessed 34 full‐text articles for eligibility and excluded 29 due to either: not being an RCT (N = 12), wrong participant group (N = 8), because the article was a review (N = 6) or ineligible intervention (N = 3). The original search identified four included studies (Avvisati 1989; Fricke 1991; Gallardo 1983; Shpilberg 1995).

Figure 1

Study flow diagram.

This updated search (conducted 07 March 2016) identified an additional 2219 potentially relevant records. There were 322 records after duplicates were removed and 317 records were able to be excluded on the basis of the abstract by two review authors (LE, MD). Six full‐text articles were retrieved for relevance and assessed by two review authors (LE, MD). This updated search identified three studies (A‐TREAT 2015; PROBLEMA 2014; TREATT 2015) that are either recruiting participants or are not yet recruiting participants and are due to be completed in 2017 and 2020. We excluded three studies.

No studies were identified that compared tranexamic acid (TXA) with epsilon aminocaproic acid (EACA).

In total, six studies were assessed and deemed eligible for inclusion (Avvisati 1989; Fricke 1991; Gallardo 1983; Shpilberg 1995; PROBLEMA 2014; TREATT 2015).

Included studies

There were seven trials (within eight articles) eligible for inclusion (A‐TREAT 2015; Avvisati 1989; Fricke 1991; Gallardo 1983; PROBLEMA 2014; Shpilberg 1995; TREATT 2015). Two of these trials are still recruiting participants (PROBLEMA 2014; TREATT 2015), and one trial has not yet started recruiting participants (A‐TREAT 2015).

Completed studies

The four completed trials were published between 1983 and 1995. Two were conducted in the USA, one in Israel, and a further one in Italy and The Netherlands. The trials randomised a total of 95 participants (8 to 56). See Characteristics of included studies for full details of each study and Table 1 for a comparison between studies.

Three studies evaluated the effect of TXA therapy in the reduction of bleeding during treatment of acute myeloid leukaemia (AML) (Avvisati 1989; Fricke 1991; Shpilberg 1995). One cross‐over TXA study (eight patients) was excluded from the outcome analysis because the data were uninterpretable due to major methodological flaws in the study design (see Table 1 for details of the study design) (Fricke 1991). We included data from this study in the 'Risk of bias' assessment.
One study evaluated the effect of EACA for bleeding control during induction chemotherapy for acute leukaemia (Gallardo 1983).

In the remainder of this review, these sub‐categories will be reported in separate sections.

Tranexamic acid (TXA) versus placebo

Three trials evaluated this comparison (Avvisati 1989; Fricke 1991; Shpilberg 1995) (Table 1).

Participants

In total, 76 participants were randomised to receive TXA or placebo (Table 1). The population characteristics varied between the studies. In Avvisati 1989, the 12 participants randomised were all diagnosed with acute promyelocytic leukaemia (APML) and were all undergoing induction chemotherapy. In Fricke 1991, seven participants had aplastic anaemia (AA) and one participant had myelodysplastic syndrome (MDS); all were out‐patients but no other treatments were reported. In Shpilberg 1995, all 56 participants had acute myeloid leukaemia (AML), however, only one of the participants randomised was diagnosed with APML (consolidation group). Thirty‐eight of the participants randomised were undergoing induction chemotherapy and 18 were undergoing consolidation chemotherapy. None of the participants were children.

Intervention

All three trials compared TXA versus placebo (Table 1). In Avvisati 1989, TXA or placebo began at the same time as the antileukaemic therapy (day 1) and lasted for 14 days. In Fricke 1991, all patients served as their own control and, after a four‐day trial period to test drug tolerance (followed by a one‐week interval without the drug), each patient began a course of either TXA or placebo that lasted for four weeks or until a platelet transfusion was required to control bleeding. In Shpilberg 1995, TXA or placebo was given when the platelet count was less than 20 x 10⁹/L or in a falling trend and less than 50 x 10⁹/L.

Co‐interventions

In Avvisati 1989, platelet transfusions (6 to 8 units/m²) were given routinely during the first seven days and additionally for overt haemorrhage, and packed red cells were administered to maintain the haemoglobin concentration above 90 g/L. In Fricke 1991, each patient's personal physician was permitted to determine the need for platelet transfusion based on "some form of bleeding, such as severe petechiae, blood blisters, and gum or nose bleeding". No red cell transfusion policy was stated. In Shpilberg 1995, platelet transfusions (4 units/m²) were given irrespective of the count, but only when clinically significant bleeding occurred and packed red cells were given to maintain the haemoglobin concentration above 90 g/L.

Outcomes

Efficacy endpoints in Avvisati 1989 were severity of bleeding, thromboembolism, laboratory assessment of fibrinolysis, packed red cell and platelet concentrate transfusion requirement. In Fricke 1991, the endpoints were number of bleeding episodes, severity of bleeding episodes, site of bleeding episodes, red cell and platelet transfusion requirements, and drug side effects. Shpilberg 1995 reported the number of bleeding events and severity of bleeding (using a scoring system), red cell and platelet concentrate transfusion requirement, thromboembolism and adverse events of drug, duration of hospitalisation, duration of significant thrombocytopenia (< 20 x 10⁹/L) and days with fever.

Epsilon aminocaproic acid (EACA) versus placebo

There was only one trial evaluating this comparison (Gallardo 1983). It was an abstract published in 1983 detailing a randomised two‐arm study with adults undergoing remission induction for acute leukaemia.

Participants

In total 19 adults undergoing remission induction for acute leukaemia were randomised to receive EACA or not; 15 with AML and four with acute lymphoblastic leukaemia (ALL). One patient was not evaluable for unstated reasons, leaving nine patients in each study arm.

Intervention

All patients received platelet transfusions (multiple, single donor or human leukocyte antigen (HLA)‐matched) in the event of thrombocytopenia (< 20 x 10⁹/L); this count defined the "days at risk of bleeding". One arm received EACA (100 mg/kg loading dose and 12 to 24 g/day thereafter in divided doses) alongside platelet transfusions whilst the other arm did not.

Co‐interventions

All patients received platelet transfusions (multiple, single donor or HLA‐matched) in the event of thrombocytopenia (< 20 x 10⁹/L); this count defined the "days at risk of bleeding".

This study reported no other co‐interventions.

Outcomes

Outcomes reported included bleeding; either as capillary bleeding (CB; skin, mucous membranes, conjunctivae, nose, occult blood in gastrointestinal (GI) or genitourinary (GU) tract), or major bleeding (MB; nose bleeding requiring posterior packing, gross GI or GU bleeding, and central nervous system (CNS) bleeding), monitoring of antifibrinolytic therapy using I¹²⁵ fibrinogen plasma clot lysis, platelet transfusion requirement, adverse events of drug and thromboembolism.

Tranexamic acid versus epsilon aminocaproic acid

No RCTs that evaluated this comparison were identified.

Ongoing Studies

There are three ongoing clinical trials (A‐TREAT 2015; PROBLEMA 2014; TREATT 2015). Please see Characteristics of ongoing studies for further details.

Excluded studies

In this update, we excluded a further three trials.(Antun 2013; NCT01980355; Rathi 2015) See Characteristics of excluded studies for further details.

Fourteen studies were not randomised controlled trials.

(Antun 2013; Bartholomew 1989; Ben‐Bassat 1990; Cattan 1963; Chakrabarti 1998; Dean 1997; Fossa 1978; Gardner 1980; Garewal 1985; Kalmadi 2006; Rathi 2015; Sanz 2010; Schwartz 1986; Wassenaar 2008)

Nine studies examined different patient groups.

(Amar 2003; Byams 2007; Celebi 2006; McConnell 2011; McConnell 2012; Mevio 1983; Movafegh 2011; NCT01980355; Yang 2001)

Six studies were reviews.

(Bates 2011; Breen 2012; Brown 2002; Levy 2005; Marti‐Carvajal 2011; Rickles 2007)

Three studies examined a different intervention.

(Bedirhan 2001; Jeserschek 2003; Katzel 1998)

Risk of bias in included studies

See Figure 2 and Characteristics of included studies for further details.

Figure 2

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

All completed studies (Avvisati 1989; Fricke 1991; Gallardo 1983; Shpilberg 1995) had some threats to validity. The majority of these threats were due to a lack of detail provided on the specific criteria and we therefore judged them as 'unclear' using the Cochrane grading system. However Fricke 1991 had significant flaws in study design which we considered 'high risk', including attrition bias, reporting bias and other sources of bias (see Figure 2 and Characteristics of included studies).

Allocation

None of the studies reported the method of sequence generation or allocation concealment and we deemed all studies reported in this review as having an unclear risk of bias.

Blinding

We deemed all studies to have an unclear risk of bias (Avvisati 1989; Fricke 1991; Gallardo 1983; Shpilberg 1995).

In Avvisati 1989, the risk of performance bias and detection bias was unclear as the article states that attending physicians were blinded to the treatment groups and that bleeding assessments were examined by the same investigator, but it is not stated whether the investigator was one of the attending physicians. In Gallardo 1983, the risk of performance and detection bias was unclear as the abstract did not state whether the investigators and or patients were blinded and does not state who carried out the assessments of bleeding. In Shpilberg 1995, the threat of performance and detection bias was unclear as although the study states that it was double‐blinded, no further details were given as to who was blinded. In Fricke 1991, although it was stated that study was double‐blinded, the article states that drug levels were obtained during 38 of the 49 courses. It does not state that the investigators or patients were blinded to this information. It also states that the study defined overall success of TXA in a patient as either five failures of placebo and none of drug or seven failures of placebo and one of drug and defined overall failure of TXA as two failed courses of drug. Sequential courses continued until overall success or failure of TXA could be determined. However, it did not state how this assessment of success or failure was performed without un‐blinding study personnel.

Incomplete outcome data

There were no missing outcome data in Avvisati 1989, so we deemed the article to have a low risk of attrition bias. We deemed Fricke 1991 to have a high risk of bias ‐ the article states that "three [of eight] patients completed the randomised portion of the study ... five of the eight patients did not complete enough courses to determine the efficacy of the drug". The two remaining studies (Gallardo 1983; Shpilberg 1995) were deemed to have an unclear risk of bias as there were insufficient data to assess incomplete outcome data.

Selective reporting

We deemed two studies to have an unclear risk of bias (Avvisati 1989; Shpilberg 1995) and two studies to have a high risk of bias (Fricke 1991; Gallardo 1983).

There were insufficient data to assess the risk of selective reporting (reporting bias) in Avvisati 1989 and in Shpilberg 1995 and we deemed these to have an unclear risk of bias. There was a high risk of reporting bias in two studies (Fricke 1991; Gallardo 1983).

In Fricke 1991, one patient died of intracranial haemorrhage four days after starting the first randomised course. Data from this course were not reported. There were two courses of TXA or placebo interrupted in two separate patients; one due to an upper respiratory tract infection and the other in which the patient developed an oesophageal haematoma after starting antibiotic treatment for an infection. Data from these courses were not included in the analysis as the investigators felt that the infection/antibiotic treatment may have compromised haemostasis. Furthermore, in Fricke 1991, the article states that severity of bleeding (as well as number and site) were recorded by the assessor (the patient), but this outcome is not reported in the article. Finally, five of the eight patients were reported as not completing enough sessions to determine the effectiveness of TXA.

In Gallardo 1983, there are data for thromboembolism and death ("no patient died of thrombosis"), but no data given on the number episodes of thromboembolism or number of deaths. There were also no data reported on the monitoring of antifibrinolytic therapy using I¹²⁵ fibrinogen plasma clot lysis. This may have been because the article was an abstract and there was limited space available.

Other potential sources of bias

We deemed two studies to have a low risk of bias (Avvisati 1989; Shpilberg 1995), one an unclear risk of bias (Gallardo 1983) and one a high risk of bias (Fricke 1991).

Avvisati 1989 and Shpilberg 1995 seemed to be free of other sources of bias and we deemed them to be at low risk of bias. In Gallardo 1983, the “at risk of bleeding days” were much higher in the EACA group – 158 versus 80 due to more severe thrombocytopenia and more cycles of chemotherapy for refractory disease. There may be bias in the randomisation procedure but the method of randomisation is not stated and we deemed the study overall to have an unclear risk of bias.

There were other sources of potential bias in Fricke 1991. The overall success of TXA was defined as either five failures of placebo and none of the drug or seven failures of placebo and one of the drug and overall failure of TXA was defined as two failed courses of drug. Failure of a course was defined as a patient receiving a platelet transfusion for bleeding during a four‐week study period. Patients received a variable number of courses of drug/placebo. The three patients who completed the study received between three (two TXA, one placebo) and nine courses (five TXA, four placebo) of treatment. The five patients who did not complete the study received between zero and 20 courses (10 TXA, 10 placebo) of treatment. Of the three patients who completed the study, two did not have any successful courses of treatment. The third patient had three of five successful courses with TXA and one of four successful courses with placebo, however this was classified by the study as a failure of TXA (two failed courses with TXA). Interim analysis of the data was therefore performed after each course of treatment for each patient, with completion of the study being biased against TXA (only two failures of TXA are required, whereas five failures of placebo and none of TXA for study to be classified as completed).

In Fricke 1991, failure of a course of treatment would be classified in the same way whether the patient was on study drug for one day before bleeding or 27 days before bleeding that required treatment with a platelet transfusion. More bleeding episodes seen in the TXA arm may have been due to more days on study drug before bleeding requiring a platelet transfusion. Number of days on study drug before bleeding was not reported for individual courses.

No protocol deviations were commented upon in any of the studies. However, in Fricke 1991, one patient began receiving HLA‐matched platelet transfusions two months after enrolment and was kept in the study because these platelet transfusions failed to control bleeding. Definition of failure of a course of treatment for this patient was the need for additional platelet transfusions. It is unclear whether this represented a protocol violation but two other patients were withdrawn from the study after they started to receive HLA‐matched platelet transfusions.

Effects of interventions

(See Table 2 and Table 3)

Table 2. Results of studies (primary outcomes of review)

Study	Type of study	Number of participants	Type of participants	Number, site and severity of bleeding		Thromboembolism (venous and arterial)
Tranexamic acid studies
Avvisati 1989	RCT	12	APML	Cumulative haemorrhagic scores TXA 3 C 42 (P = 0.0045)		No thromboembolic events observed
Shpilberg 1995	RCT	56	AML 38 induction 18 consolidation	Mean number of bleeding events per participant: Induction TXA 6.2 SD 2.9 C 4.5 SD 3.6 Consolidation TXA 1.1 SD 1.4 C 2.6 SD 2.2 (P < 0.05)	Cumulative haemorrhagic scores: Induction TXA 8.3 SD 4.8 C 5.6 SD 4.8 Consolidation TXA 1.3 SD 1.8 C 5.1 SD 3.6 (P < 0.05)	No thromboembolic events observed
EACA Studies
Gallardo 1983	RCT	19	15 AML 4 ALL	Capillary bleeding^a EACA 31% of days at risk^b Placebo 50% of days at risk^b	Major bleeding^c EACA 15% Placebo 19%	No deaths due to thromboembolic disease

^aCapillary bleeding defined as bleeding in skin, mucous membranes, conjunctivae, nose and guaiac in gastrointestinal or genitourinary tract.

^bDays at risk defined as days when platelet count fewer than 20 x 10⁹/ L.
^cMajor bleeding defined as nose bleeding requiring posterior packing, gross gastrointestinal or genitourinary tract bleeding and central nervous system bleeding.

ALL: acute lymphoblastic leukaemia
AML: acute myeloid leukaemia
APML: acute promyelocytic leukaemia
C: control
EACA: epsilon aminocaproic acid
RCT: randomised controlled trial
SD: standard deviation
TXA: tranexamic acid

Table 3. Results of studies (secondary outcomes of review)

Study

Type of study

Number of participants

Type of participants

Mortality (all causes)

Mortality (secondary to bleeding)

Mortality (secondary to thromboembolism)

Laboratory assessment of fibrinolysis

Number of platelet transfusions

Number of red cell components

Adverse events of antifibrinolytic agents

Adverse events of transfusions

(e.g. transfusion reactions, antibody development)

DIC

QoL

Tranexamic acid studies

RCT

APML

No difference in the coagulation and fibrinolysis indices between the 2 groups apart from FDPs^a. FDPs decreased in TXA arm but increased in the placebo arm

(P < 0.01)

Platelet Tx

TXA = 45 Tx

C = 246 Tx

(P = 0.045)

Reduction in overall RBC components required in TXA group:

(units)

TXA = 28

C = 56

(P = 0.016)

RCT

AML

38 induction

18 consolidation

No fatal bleeding in either group

Induction (units)

TXA 22.1 SD 13.2

C 23.1 SD 11.7

Consolidation

(units)

TXA 3.7 SD 4.1

C 9.3 SD 3.3

(P < 0.05)

No reduction RBC transfusion requirements

Induction

(units)

TXA 7.5 SD 4.7

C 7.3 SD 3.3

Consolidation

(units)

TXA 4.1 SD 2.8

C 4.1 SD 3.4

No side effects were observed

EACA studies

RCT

15 AML

4 ALL

No participant died of thrombosis

Monitored with the I¹²⁵ fibrinogen plasma clot lysis assay but no further data described.

EACA

1 every 13.3 days at risk^b

Placebo

1 every 10.5 days at risk^b

Side effects were stated as minimal

^aBlood coagulation tests were prothrombin time, activated partial thromboplastin time, fibrinogen, FDP, antithrombin III activity, thrombin‐antithrombin III complexes, protein C activity and α₂‐antiplasmin. These were carried out daily for the first 10 days.
^bDays at risk defined as days when platelet count fewer than 20 x 10⁹/L.

ALL: acute lymphoblastic leukaemia
AML: acute myeloid leukaemia
APML: acute promyelocytic leukaemia
C: control
DIC: intravascular coagulation
EACA: epsilon aminocaproic acid
FDPs: fibrin degradation products
GI: gastrointestinal
GU: genitourinary
QoL: quality of life
NR: not reported
RBC: red blood cell
RCT: randomised controlled trial
SD: standard deviation
TXA: tranexamic acid
Tx: transfusion

Tranexamic acid (TXA) versus placebo

Three studies evaluated this comparison (Avvisati 1989; Fricke 1991; Shpilberg 1995).

We did not extract any data from the Fricke 1991 study due to major methodological problems in the study design. In addition to the high risk of bias in terms of attrition bias, reporting bias and other bias (see text section above, Figure 2 and the Risk of bias in included studies table), there was a variable number of study cycles depending on the results of previous cycles of treatment. All these factors meant that it was impossible to fully understand the data in this trial and we took the decision to not include this trial in the assessment of 'effects of interventions'.

Number, site and severity of bleeding (i.e. any bleeding, clinically significant bleeding, life‐threatening bleeding)

The two remaining studies (68 participants) both reported bleeding (Table 2), but bleeding was reported in different ways.

Shpilberg 1995 reported the mean number of bleeding events per participant. During induction chemotherapy, there was no difference in the number of bleeding events per participant between TXA and placebo (38 participants, mean difference (MD) 1.70 bleeding events, 95% confidence interval (CI) ‐0.37 to 3.77) (Analysis 1.1). During consolidation chemotherapy, there was no difference in the number of bleeding events per participant between TXA and placebo (18 participants; MD ‐1.50 bleeding events (95% CI ‐3.25 to 0.25) (Analysis 1.1). This differs from the study authors who reported a statistically significant difference in the number of bleeding events per participant (P < 0.05). We are unable to explain the reason for this difference because we do not know what statistical analyses were performed by the study authors.

Avvisati 1989 (12 participants) did not report bleeding per participant and instead reported bleeding as a cumulative score in the first observation period (days two to seven), second observation period (days eight to 14) and as an overall cumulative score (Table 2). The study stated that there was a reduction in the overall cumulative haemorrhagic scores in the TXA study arm (P = 0.0045).

Thromboembolism (venous and arterial)

Both studies (68 participants) reported thromboembolism but did not distinguish between arterial or venous events (Table 2). Shpilberg 1995 (56 participants) reported that no thromboembolic events occurred in either group throughout the study. Avvisati 1989 (12 participants) reported that there were no thromboembolic complications and there was no evidence of enhanced thrombin generation (as assessed by thrombin‐antithrombin‐III complexes). TXA was only given for the first six days out of 14 days of observation because of what the authors described as the "known increase of cerebral thromboembolic disease with prolonged therapy".

Mortality (all causes)

Neither study reported all‐cause mortality.

Mortality (secondary to bleeding)

Only Shpilberg 1995 (56 participants) reported mortality (secondary to bleeding) and stated that there was no fatal bleeding in either group (Table 3).

Mortality (secondary to thromboembolism)

Neither study reported mortality secondary to thromboembolism.

Laboratory assessment of fibrinolysis

Only Avvisati 1989 (12 participants) reported laboratory assessment of fibrinolysis. There were no statistically significant differences in the coagulation and fibrinolysis indices between the two groups apart from the results for fibrin degradation products. The study stated that median fibrin degradation products decreased in the TXA arm but increased in the placebo arm during the first week of observation (P < 0.01) (Table 4).

Table 4. Laboratory assessment of fibrinolysis ‐ Avvisati 1989

Coagulation factors	Timing	Treatment groups Median (range)
Coagulation factors	Timing	Tranexamic acid	Placebo
Fibrinogen (mg/dL)	Baseline	80 (55 to 395)	70 (50 to 190)
	Day 3	55 (20 to 125)	62 (45 to 150)
	Day 5	53 (35 to 80)	78 (36 to 150)
	Day 7	60 (50 to 65)	85 (17 to 160)
Thrombin‐antithrombin complex (ng/mL)	Baseline	32 (4 to 58)	22 (15 to 26)
	Day 3	50 (20 to 100)	22 (11 to 52)
	Day 5	41 (5 to 51)	17 (5 to 34)
	Day 7	10 (5 to 31)	21 (3 to 37)
α₂‐antiplasmin (%)	Baseline	39 (29 to 80)	50 (29 to 58)
	Day 3	27 (17 to 40)	35 (27 to 47)
	Day 5	32 (27 to 34)	36 (29 to 51)
	Day 7	33 (19 to 53)	42 (32 to 59)
Fibrin/fibrinogen degradation products (μg/dL)	Baseline	40 (10 to 80)	80 (40 to 160)
	Day 3	NR	NR
	Day 5	NR	NR
	Day 7	NR	NR

NR: not reported

Number of platelet transfusions or number of platelet components

Neither study reported the number of platelet transfusions per participant.

Both studies reported the number of platelet components (Table 3) but only Shpilberg 1995 (56 participants) reported the number of platelet components per participant. During induction chemotherapy, there was no difference in the number of platelet components per participant in the TXA arm versus the placebo arm (38 participants, MD ‐1.00 platelet units, 95% CI ‐9.11 to 7.11) (Analysis 1.2). During consolidation chemotherapy there was a decrease in the number of platelet components per participant (18 participants; MD ‐5.60 platelet units, 95% CI ‐9.02 to ‐2.18) (Analysis 1.2).

Avvisati 1989 (12 participants) did not report the number of platelet transfusions per participant and instead reported the total number of platelet transfusions used in each study arm (Table 3). The study stated that there was a reduction in the number of platelet transfusions in the TXA study arm (P = 0.045).

Number of red cell transfusions or number of red cell components

Neither study reported the number of red cell transfusions per participant.

Both studies reported the number of red cell components (Table 3), but only Shpilberg 1995 (56 participants) reported the number of red cell components per participant. During induction and consolidation chemotherapy, there was no difference in the number of red cell components per participant in the TXA arm versus the placebo arm (induction chemotherapy; 38 participants, MD 0.20 red cell components, 95% CI ‐2.48 to 2.88: consolidation chemotherapy; 18 participants, MD 0.00 red cell components, 95% CI ‐2.93 to 2.93) (Analysis 1.3).

Avvisati 1989 (12 participants) did not report the number of red cell components per participant and instead reported the total number of red cell components used in each study arm (Table 3). The study stated that there was a reduction in the number of red cell components used in the TXA study arm (P = 0.016).

Adverse events of antifibrinolytic agents

Avvisati 1989 did not report adverse events of antifibrinolytic agents. Shpilberg 1995 (56 participants) reported that no side effects were observed (Table 3).

Adverse events of transfusions (e.g. transfusion reactions, antibody development)

Neither study reported the adverse events of transfusions.

Disseminated intravascular coagulation (DIC)

Neither study reported DIC.

Quality of life

Neither study reported quality of life.

Epsilon aminocaproic acid (EACA) versus placebo

There was only one study evaluating this comparison (Gallardo 1983). It was a randomised two‐arm study published in 1983 involving participants undergoing remission induction for acute leukaemia.

Number, site and severity of bleeding (i.e. any bleeding, clinically significant bleeding, life‐threatening bleeding)

Gallardo 1983 did not report bleeding per participant instead it reported bleeding as the proportion of days at risk of bleeding (defined as where the platelet count was < 20 x 10⁹/L) (Table 2). This was 158 days for the group on EACA compared to only 80 for the group on no EACA, but the abstract noted that the patients on EACA had more severe thrombocytopenia and more cycles of chemotherapy for refractory disease. Capillary bleeding (i.e. bleeding in skin, mucous membranes, conjunctivae, nose and occult blood in gastrointestinal (GI) or genitourinary (GU) tract) was present in 31% of days at risk with patients on EACA compared to 50% of patients not receiving EACA (P value not reported). There was no difference in major bleeding (defined as nose bleeding requiring posterior packing, gross GI or GU bleeding and bleeds within the central nervous system (CNS)) between the two groups (15% versus 19%) (P value not reported).

Thromboembolism (venous and arterial)

There were no reports of thromboembolism although the study stated that no patient died of thrombosis (18 participants).

Mortality (all causes)

The study did not report all‐cause mortality.

Mortality (secondary to bleeding)

The study did not report mortality secondary to bleeding.

Mortality (secondary to thromboembolism)

The study stated that no patient died of thrombosis (18 participants).

Laboratory assessment of fibrinolysis

The study reported that antifibrinolytic therapy was monitored with the I¹²⁵fibrinogen plasma clot lysis assay although no further data were described regarding this outcome.

Number of platelet transfusions or platelet components

The study did not report the number of platelet transfusions or platelet components per participant.

The study reported that platelet transfusions per days at risk were decreased in the patients on EACA, one every 13.3 days at risk versus one every 10.5 days at risk. However, the authors noted that these were not statistically significant (P value not reported). The abstract detailed a projection that the results would achieve statistical significance at a P value of < 0.05 with 25 patients in each group. However, no subsequent study has since been published. It is therefore important to note that there were insufficient patients within this study to show statistical significance for any clinically meaningful true difference (Table 3).

Number of red cell transfusions or platelet components

The study did not report the number of red cell transfusions per participant.

Adverse events of antifibrinolytic agents

No specific adverse events were described, although the study stated that side effects were minimal (18 participants).

Adverse events of transfusions (e.g. transfusion reactions, antibody development)

The study did not report adverse events of transfusions.

Disseminated intravascular coagulation (DIC)

The study did not report DIC.

Quality of life

The study did not report quality of life.

Tranexamic acid versus epsilon aminocaproic acid

We did not identify any RCTs that evaluated this comparison.

Discussion

The overall aim of this review was to determine the efficacy and safety of antifibrinolytics (lysine analogues) in the prevention of bleeding in people with haematological disorders.

Specifically, we aimed to address the following questions:

i) Do lysine analogues help to prevent bleeding in people with haematological disorders who are thrombocytopenic?

ii) Can the number of prophylactic platelet transfusions be minimised?

iii) Do lysine analogues increase the incidence of thromboembolism?

Our primary outcomes were bleeding and the occurrence of thromboembolism. Our secondary outcomes were mortality, laboratory assessment of fibrinolysis, number of platelet transfusions, number of red cell transfusions, adverse events of antifibrinolytic agents, adverse events of transfusions (e.g. transfusion reactions, antibody development), disseminated intravascular coagulation (DIC) and quality of life.

Summary of main results

Seven trials met our inclusion criteria and three of these trials are on‐going. Of the four completed trials, one had to be excluded from the assessment of Effects of interventions due to major methodological flaws in its design and a high risk of bias across several criteria. Of the remaining three randomised controlled trials (RCTs), a total of 86 patients were investigated. There were two studies comparing tranexamic acid (TXA) and placebo (Avvisati 1989; Shpilberg 1995) and one trial comparing epsilon aminocaproic acid (EACA) with placebo (Gallardo 1983). There were no studies comparing TXA with EACA.

i) Do lysine analogues help to prevent bleeding in people with haematological disorders who are thrombocytopenic?

We are uncertain whether antifibrinolytics reduce the risk of bleeding. Only one trial reported the number of bleeding episodes per participant and no difference was seen (Shpilberg 1995). The two other trials suggested bleeding was reduced in the antifibrinolytic study arm (Avvisati 1989; Gallardo 1983), but this was statistically significant in only one of these studies.

ii) Can the number of prophylactic platelet transfusions be minimised?

No trial reported the number of prophylactic platelet transfusions per participant. Only one trial reported the number of platelet components per participant and there was a reduction in platelet component usage during consolidation chemotherapy but not during induction chemotherapy (Shpilberg 1995). One of the other two trials also found that the total number of platelet components administered to participants in the antifibrinolytic arm was lower than in the placebo arm (Avvisati 1989).

iii) Do lysine analogues increase the incidence of thromboembolism?

We are uncertain whether antifibrinolytics increase the risk of developing a thromboembolism as the quality of the evidence has been assessed as very low. Two trials reported the presence or absence of thromboembolism and no events occurred in either of these trials (Avvisati 1989; Shpilberg 1995). It was reported in Gallardo 1983 that no patient died of thrombosis.

Other results

Two of the trials (Avvisati 1989; Shpilberg 1995) reported red blood cell transfusion requirements. Only one trial reported the number of red cell components per participant and there was no difference in red cell component usage between the TXA and placebo arms (Shpilberg 1995).

None of the trials reported on overall (all‐cause) mortality. However, one of the trials reported on mortality due to bleeding (Shpilberg 1995) and only one reported on mortality due to thromboembolism (Gallardo 1983); none occurred in either category.

Two trials (Gallardo 1983; Shpilberg 1995) reported on the side effects of antifibrinolytics. Gallardo 1983 stated that the side effects of EACA "were minimal" but did not provide any further detail on this within the abstract. Shpilberg 1995 also reported on side effects of TXA and stated that none were observed.

None of the trials reported on our other outcomes of interest: adverse events of transfusion, presence or development of DIC, or quality of life.

Overall completeness and applicability of evidence

This review provides the most up‐to‐date assessment of the effectiveness and safety of antifibrinolytics in people with haematological disorders who are thrombocytopenic and would usually be treated with platelet transfusions. This review identified three ongoing RCTs that are due to be completed by 2020.

There is very low quality evidence that adults with acute leukaemia who are receiving chemotherapy may have a decreased risk of bleeding with antifibrinolytics and may decrease their use of platelet components.

The results of this review should take into consideration the impact of the following factors.

The completed studies included in this review were all conducted at least twenty years ago (1983 to 1995), during which time chemotherapy protocols, predicted overall survival rates, and supportive care, including transfusion, have changed substantially.
Only 86 participants were included in the three studies that provided data for this review's outcomes. Two studies (68 participants) compared TXA with placebo and one study (18 participants) compared EACA with placebo.
The recording of bleeding is subjective, and between centres there is variability in the assessment, grading, investigation and recording of bleeding. The same bleeding scale may even be interpreted and applied differently, particularly with respect to red cell transfusion.
We could not perform any meta‐analyses because of the various ways in which the outcomes had been reported.
We were unable to obtain any additional data from study authors to be used quantitatively in a meta‐analysis. Due to the age of all three studies only one author could be located. This author no longer had data available because the trial was conducted over 30 years ago (Gallardo 1983).
There were several review outcomes that were not reported by any of the studies. These included: adverse events of transfusion, the presence or development of DIC and quality of life. This highlights the paucity of data in this area.
All of the evidence in this review was in adults with acute leukaemia receiving chemotherapy, there is currently no evidence to support the use of antifibrinolytics in adults with other haematological disorders, people requiring haematopoietic stem cell transplants, or children.
All studies were suggestive of a role of antifibrinolytics in reducing platelet usage, but it is important to note that in Shpilberg 1995 the effect was only seen in the consolidation group and not in the induction group, and in Gallardo 1983, although the number of platelet transfusions required per days at risk was less for the EACA arm, the results were not statistically significant. The study authors projected that their results would "achieve statistical significance at a P value of < 0.05 with 25 patient in each group". However, despite an extensive literature search, and direct contact with the study authors no further data for Gallardo 1983 were available.
Two studies (68 patients) reported the presence or absence of thromboembolic events and no events occurred in either study (Avvisati 1989; Shpilberg 1995). It was reported in Gallardo 1983 that no patient died of thrombosis, but it is unclear whether any non‐fatal thromboembolic events occurred at all. Although there is no evidence within these three studies to suggest that there is an increased risk of thromboembolism with antifibrinolytics there are insufficient data to conclude that this risk does not exist

Quality of the evidence

We assessed the quality of the evidence using the GRADE approach and this was either very low or the outcome was not reported by any of the studies (summary of findings Table for the main comparison). We were unable to gain additional information via direct author contact and therefore could not improve the quality of the data.

Only three outcomes could be assessed using the GRADE approach.

Number of participants with any bleeding
Number of participants with thromboembolism
Adverse events of antifibrinolytic agents

All were assessed as very low grade quality evidence due to serious risk of bias and very serious imprecision.

One study (Fricke 1991) had significant methodological problems with its design (seeRisk of bias in included studies and Figure 2). The overall success of TXA was defined as either five failures of placebo and none of the drug, or seven failures of placebo and one of the drug. Overall failure of TXA was defined as two failed courses of the drug. Failure of a course was defined as a patient receiving a platelet transfusion for bleeding during a four‐week study period. Patients received a variable number of courses of drug/placebo. The three patients who completed the study received between three (two TXA, one placebo) and nine courses (five TXA, four placebo) of treatment. The five patients who did not complete the study received between zero and 20 courses (10 TXA, 10 placebo) of treatment. Of the three patients who completed the study, two did not have any successful courses of treatment. The third patient had three of five successful courses with TXA and one of four successful courses with placebo, however this was classified by the study as a failure of TXA (two failed courses with TXA). Interim analysis of the data was therefore performed after each course of treatment for each patient, with completion of the study being biased against TXA (only two failures of TXA are required, whereas five failures of placebo and none of TXA for study to be classified as completed).

All of the other three included studies (Avvisati 1989; Gallardo 1983; Shpilberg 1995) had some threats to validity and in most cases this was graded as 'unclear' due to lack of detail in the study to determine the level of risk. One of these studies (Gallardo 1983) was at high risk of selective reporting. The data were presented in an abstract form and the problem was likely to have been due to limited space. However, important data were omitted, most notably number of episodes of thrombosis since the comment "no patient died of thrombosis" is suggestive of the presence of thrombosis in the study. This is clearly of particular importance when considering the safety of antifibrinolytics when thrombosis is a noted side effect of these agents.

Avvisati 1989 appeared to be the study most free of bias with low risk when considering attrition bias and other bias. Other than those mentioned, the other parameters for bias were listed as unclear due to a lack of data to determine risk as high or low.

One negative aspect that all the studies had in common was small sample sizes which reduced their statistical power. This means that even if a clinically meaningful true difference was present it may not be detected due to the small number of patients within each study. This could not be overcome via the use of meta‐analysis because the data had been reported in different ways. In one study in particular (Gallardo 1983), the small sample size was insufficient to permit statistical significance for at least one outcome (number of platelet transfusions in each arm)

Potential biases in the review process

There were no clear biases identified in the review process. We conducted a comprehensive search of data sources (including multiple databases and clinical trial registries) to ensure that we would capture all relevant trials. We made no restrictions for the language in which the paper was originally published. We carefully assessed the relevance of each paper identified and performed all screening and data extractions in duplicate.

The systematic methods of searching, data extraction and result analysis were followed with reference to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a).

One important consideration was whether our decision not to include Fricke 1991 in the narrative review of the included studies may represent some risk of publication bias. Fricke 1991 was the only study that showed a lack efficacy of the antifibrinolytic (TXA). However, given the methodological flaws and high levels of risk across several criteria as mentioned in Assessment of risk of bias in included studies, it was felt that it should be excluded from the narrative review.

Two of the authors of this review are primary investigators of an ongoing study included in the systematic review (TREATT 2015). This study has not been completed yet and 'Risk of bias' assessment and analysis of the completed studies was performed prior to the initiation of TREATT 2015 in the previous version of this review (Wardrop 2013).

Agreements and disagreements with other studies or reviews

The fact that there were so few RCTs that were suitable for inclusion in our narrative review demonstrates the lack of efficacy and safety data for the use of antifibrinolytics in haematology patients who are thrombocytopenic.

Comparison to other systematic reviews

To our knowledge, there are no other systematic reviews examining this topic.

Comparison to non‐randomised trials

Several small non‐randomised studies have used TXA or EACA in haematology patients (Bartholomew 1989; Ben‐Bassat 1990; Chakrabarti 1998; Dean 1997; Gardner 1980; Garewal 1985; Kalmadi 2006; Schwartz 1986; Wassenaar 2008). However, virtually all of these studies did not have a comparator arm, making it difficult to draw any valid conclusions on the effectiveness and safety of antifibrinolytics. One larger study of TXA in patients with acute promyelocytic leukaemia (APML) used an historical control (Sanz 2010); there was no difference in deaths due to bleeding between those patients who received TXA and those who did not. Death due to bleeding is a rare event and this study may not have had sufficient power to detect a difference. This study (Sanz 2010) did show a statistically significant increase in the number of patients who developed thromboembolic complications. However, because Sanz 2010 used an historical control there may have been other confounding factors including changes to the chemotherapy regimen used that could have also affected the number of thromboembolic events.

Figure 1

Study flow diagram.

Figure 2

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Analysis 1.1

Comparison 1 Tranexamic acid versus placebo, Outcome 1 Number of bleeding events per participant.

Analysis 1.2

Comparison 1 Tranexamic acid versus placebo, Outcome 2 Number of platelet components per participant.

Analysis 1.3

Comparison 1 Tranexamic acid versus placebo, Outcome 3 Number of red cell components per participant.

Summary of findings for the main comparison. Antifibrinolytics (lysine analogues) compared to placebo to prevent bleeding in people with haematological disorders

Antifibrinolytics (lysine analogues) compared to placebo to prevent bleeding in people with haematological disorders
Patient or population: people with haematological disorders Settings: hospital Intervention: antifibrinolytics (lysine analogues) Comparison: placebo
Outcomes	Illustrative comparative risks^a (95% CI)		Relative effect (95% CI)	No of participants (studies)^b	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Placebo	Antifibrinolytics (lysine analogues)
Number of participants with any bleeding	See comment	See comment	Not estimable	56 (1 study)	⊕⊝⊝⊝ very low^1,2	We are uncertain whether antifibrinolytics reduce bleeding. Only Shpilberg 1995 reported the number of bleeding episodes per participant, results were reported separately for induction and consolidation chemotherapy and no difference was seen.
Number of participants with thromboembolism	See comment	See comment	Not estimable	68 (2 studies)	⊕⊝⊝⊝ very low^1,2	We are uncertain whether antifibrinolytics increase the risk of thromboembolism. No patients within the Avvisati 1989 or Shpilberg 1995 studies had an episode of thromboembolism. Gallardo 1983 only reported no deaths due to thrombosis.
Mortality (all‐cause) ‐ not reported	Not estimable		Not estimable	0 (0)	See comment	None of the studies reported all‐cause mortality. Shpilberg 1995 reported no deaths due to bleeding and Gallardo 1983 reported no deaths due to thrombosis.
Number of platelet transfusions per participant ‐ not reported	Not estimable		Not estimable	0 (0)	See comment	None of the studies reported the number of platelet transfusions per participant
Adverse events of transfusions ‐ not reported	Not estimable		Not estimable	0 (0)	See comment	None of the studies reported adverse events due to blood transfusions
Adverse events of antifibrinolytic agents	See comment	See comment	Not estimable	74 (2 studies)	⊕⊝⊝⊝ very low^1,2	We are uncertain whether antifibrinolytics increase the risk of adverse events. Shpilberg 1995 reported no side effects were observed and Gallardo 1983 stated side effects were minimal but no further information was provided
Quality of life ‐ not reported	Not estimable		Not estimable	0 (0)	See comment	None of the studies reported quality of life.
^aNo meta‐analyses were performed within this review and therefore no comparative risks could be calculated. ^bOne study (Fricke 1991) was included within the review but no data were extracted from this study. This was because Fricke 1991 had significant flaws within the study design (see Risk of bias in included studies and Characteristics of included studies) and no viable data could be extracted from the study report. CI: confidence interval
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹We downgraded the quality of the evidence by 1 for risk of bias. A full assessment of the quality of the evidence was limited by a lack of reporting. However, selective outcome reporting was present in Gallardo 1983. ²We downgraded the quality of evidence by 2 for imprecision. Within all three studies there were only 86 participants. This is significantly below the optimal information size (OIS) (Pogue 1997).

Summary of findings for the main comparison. Antifibrinolytics (lysine analogues) compared to placebo to prevent bleeding in people with haematological disorders

Table 1. Study characteristics

Study

Type of study

No. of participants

No. of participants receiving antifibrinolytic

No. of participants platelet refractory/ allo‐immunisation

Diagnosis of patients

Treatment of underlying disease

Antifibrinolytic dose

Antifibrinolytic frequency

Antifibrinolytic route

Treatment started

Treatment stopped

Platelets given

Tranexamic acid studies

RCT

APML

Chemo

2 g

8‐hourly

IV in 500 mL 5% glucose

1st day of antileukaemic Rx

After 6 days

Prophylactic

Fricke 1991

RCT Cross‐over

Only 3 completed study

At least 3

7 AA

1 MDS

20 mg/kg

8‐hourly

Oral

After 4‐day trial period to assess drug tolerance

Successive 4/52 courses or until WHO grade 2 bleeding

Therapeutic

RCT

AML

38 induction

18 consolidation

Chemo

1 g

6‐hourly

Oral

Platelets < 20 or rapidly falling and < 50

Platelet count > 20 for 2 consecutive counts

Therapeutic

EACA studies

RCT

15 AML

4 ALL

(1 patient not evaluable)

Chemo

Loading dose 100 mg/kg

12 to 24 g/day thereafter

Platelet count < 20 x 10⁹/L

Platelet count ≥ 20 x 10⁹/L

Platelet count < 20 x 10⁹/L

Table 1. Study characteristics

Table 2. Results of studies (primary outcomes of review)

Study	Type of study	Number of participants	Type of participants	Number, site and severity of bleeding		Thromboembolism (venous and arterial)
Tranexamic acid studies
Avvisati 1989	RCT	12	APML	Cumulative haemorrhagic scores TXA 3 C 42 (P = 0.0045)		No thromboembolic events observed
Shpilberg 1995	RCT	56	AML 38 induction 18 consolidation	Mean number of bleeding events per participant: Induction TXA 6.2 SD 2.9 C 4.5 SD 3.6 Consolidation TXA 1.1 SD 1.4 C 2.6 SD 2.2 (P < 0.05)	Cumulative haemorrhagic scores: Induction TXA 8.3 SD 4.8 C 5.6 SD 4.8 Consolidation TXA 1.3 SD 1.8 C 5.1 SD 3.6 (P < 0.05)	No thromboembolic events observed
EACA Studies
Gallardo 1983	RCT	19	15 AML 4 ALL	Capillary bleeding^a EACA 31% of days at risk^b Placebo 50% of days at risk^b	Major bleeding^c EACA 15% Placebo 19%	No deaths due to thromboembolic disease
^aCapillary bleeding defined as bleeding in skin, mucous membranes, conjunctivae, nose and guaiac in gastrointestinal or genitourinary tract. ^bDays at risk defined as days when platelet count fewer than 20 x 10⁹/ L. ^cMajor bleeding defined as nose bleeding requiring posterior packing, gross gastrointestinal or genitourinary tract bleeding and central nervous system bleeding. ALL: acute lymphoblastic leukaemia AML: acute myeloid leukaemia APML: acute promyelocytic leukaemia C: control EACA: epsilon aminocaproic acid RCT: randomised controlled trial SD: standard deviation TXA: tranexamic acid

Table 2. Results of studies (primary outcomes of review)

Table 3. Results of studies (secondary outcomes of review)

Study

Type of study

Number of participants

Type of participants

Mortality (all causes)

Mortality (secondary to bleeding)

Mortality (secondary to thromboembolism)

Laboratory assessment of fibrinolysis

Number of platelet transfusions

Number of red cell components

Adverse events of antifibrinolytic agents

Adverse events of transfusions

(e.g. transfusion reactions, antibody development)

DIC

QoL

Tranexamic acid studies

RCT

APML

No difference in the coagulation and fibrinolysis indices between the 2 groups apart from FDPs^a. FDPs decreased in TXA arm but increased in the placebo arm

(P < 0.01)

Platelet Tx

TXA = 45 Tx

C = 246 Tx

(P = 0.045)

Reduction in overall RBC components required in TXA group:

(units)

TXA = 28

C = 56

(P = 0.016)

RCT

AML

38 induction

18 consolidation

No fatal bleeding in either group

Induction (units)

TXA 22.1 SD 13.2

C 23.1 SD 11.7

Consolidation

(units)

TXA 3.7 SD 4.1

C 9.3 SD 3.3

(P < 0.05)

No reduction RBC transfusion requirements

Induction

(units)

TXA 7.5 SD 4.7

C 7.3 SD 3.3

Consolidation

(units)

TXA 4.1 SD 2.8

C 4.1 SD 3.4

No side effects were observed

EACA studies

RCT

15 AML

4 ALL

No participant died of thrombosis

Monitored with the I¹²⁵ fibrinogen plasma clot lysis assay but no further data described.

EACA

1 every 13.3 days at risk^b

Placebo

1 every 10.5 days at risk^b

Side effects were stated as minimal

Table 3. Results of studies (secondary outcomes of review)

Table 4. Laboratory assessment of fibrinolysis ‐ Avvisati 1989

Coagulation factors	Timing	Treatment groups Median (range)
Coagulation factors	Timing	Tranexamic acid	Placebo
Fibrinogen (mg/dL)	Baseline	80 (55 to 395)	70 (50 to 190)
	Day 3	55 (20 to 125)	62 (45 to 150)
	Day 5	53 (35 to 80)	78 (36 to 150)
	Day 7	60 (50 to 65)	85 (17 to 160)
Thrombin‐antithrombin complex (ng/mL)	Baseline	32 (4 to 58)	22 (15 to 26)
	Day 3	50 (20 to 100)	22 (11 to 52)
	Day 5	41 (5 to 51)	17 (5 to 34)
	Day 7	10 (5 to 31)	21 (3 to 37)
α₂‐antiplasmin (%)	Baseline	39 (29 to 80)	50 (29 to 58)
	Day 3	27 (17 to 40)	35 (27 to 47)
	Day 5	32 (27 to 34)	36 (29 to 51)
	Day 7	33 (19 to 53)	42 (32 to 59)
Fibrin/fibrinogen degradation products (μg/dL)	Baseline	40 (10 to 80)	80 (40 to 160)
	Day 3	NR	NR
	Day 5	NR	NR
	Day 7	NR	NR
NR: not reported

Table 4. Laboratory assessment of fibrinolysis ‐ Avvisati 1989

Comparison 1. Tranexamic acid versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Number of bleeding events per participant Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

1.1 Induction chemotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 Consolidation chemotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2 Number of platelet components per participant Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

2.1 Induction chemotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.2 Consolidation chemotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
3 Number of red cell components per participant Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

3.1 Induction chemotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]
3.2 Consolidation chemotherapy	1		Mean Difference (IV, Fixed, 95% CI)	0.0 [0.0, 0.0]

Comparison 1. Tranexamic acid versus placebo