The frequency of bleeding complications in patients with haematological malignancy following the introduction of a stringent prophylactic platelet transfusion policy


Colin R. Callow, Department of Haematology, Christie NHS Trust, Wilmslow Road, Manchester, M20 4BX, UK. E-mail:


Summary. Indications for platelet transfusion remain controversial and are frequently based on arbitrary numerical criteria. In October 2000, we introduced a stringent prophylactic-platelet transfusion policy < 10 × 109/l for stable patients and < 20 × 109/l in the presence of major bleeding or additional risk factors. A trigger of < 50 × 109/l was introduced for patients undergoing invasive procedures. A prospective analysis was performed measuring the frequency of minor and major bleeding events, morbidity, mortality and duration of pancytopenia. Blood product usage was assessed and health care savings measured. A total of 98 patients were evaluated on 2147 patient study days and 271 bleeding episodes were recorded. Major bleeding occurred on 1·39% (30/2147) of the study days when platelet counts were < 10 × 109/l and 2·3% (50/2147) of the study days when platelet counts were 10–20 × 109/l. In patients with platelets > 20 × 109/l, there were 117 major bleeding episodes observed on 5·4% of the study days. In patients with no identified additional risk factors present, major haemorrhages were recorded in 0·51% (11/2147) of the study days in patients with platelet counts ≥ 10 × 109/l. There was a 36% reduction in platelet units transfused compared with retrospective data when an arbitrary transfusion trigger of 20 × 109/l was in place (P = < 0·02). Of note, a 16% reduction in red cell transfusions was recorded. These data confirm that the introduction of a transfusion trigger of < 10 × 109/l in the absence of fresh bleeding and sepsis (> 38°C) is safe and has a significant impact on overall hospital transfusion costs.

The pioneering use of platelet transfusions during the 1960s enabled the introduction of more aggressive chemotherapy regimens, which were associated with improved clinical efficacy.

The relationship between haemorrhage and patient platelet count was defined by a landmark study (Gaydos et al, 1962). Subsequently, it was considered that the risk of haemorrhagic complications was unacceptable when the platelet count was reduced to below 20 × 109/l. This was ata time when there was still widespread administration ofaspirin to patients with acute leukaemia.

During the 1980s, however, increasing evidence emerged to suggest that such high thresholds for prophylactic platelet transfusion were, in fact, harmful to patients, and in 1987 the National Institutes of Health (NIH) formally questioned the validity of the 20 × 109/l trigger (NIH, 1987; Beutler, 1993).

In the last 10 years, significantly lower thresholds have been accepted for prophylactic platelet transfusions. Studies by Gmur et al (1991), Gil Fernandez et al (1996) and Rebulla et al (1997) have all indicated that the classical trigger of 20 × 109/l could be safely reduced in certain, well-defined patient groups with possible economic benefits and reduced transfusion requirements for patients. Other studies have investigated the effects of changing platelet dose in order to reduce the costs of platelet transfusions; Ackerman et al (2000) suggested that efforts to use lower-dose single-donor platelets result in an increase in patient transfusion requirements with a corresponding overall increase in overall hospital transfusion costs.

Despite these studies, indications for the transfusion of platelets have remained controversial, and in many cases continue to be based on arbitrary and poorly justified numerical criteria (Patten, 1992). This practice has the potential for unnecessary exposure to the infectious and immunological risks of transfusion.

In December 1997, there was cautious agreement in the Royal College of Physicians Consensus Conference statement on platelet transfusion that a platelet threshold of 10 × 109/l is as safe as higher levels in managing patients without additional risk factors (Contreras, 1998). However, the UK demand for platelet transfusions has remained relatively static over the last 5 years and this suggests that many centres have not changed their platelet transfusion practice.

Given the widespread, considerable variations in clinical practice, there remains a need to develop prospective studies that evaluate the impact of changing platelet transfusion policies for all haematological patients. We therefore decided to evaluate our prophylactic platelet transfusion practice by defining new protocols and policies that altered the platelet transfusion threshold from 20 × 109/l to 10 × 109/l.

In total, 98 patients receiving intensive chemotherapy for haematological malignancies were evaluated over a 3 month period. We performed a non-randomised prospective analysis of the frequency of bleeding events and measured the effects on blood product usage. We compared the use of blood products during this period with data from the preceding 9 month period.

Patients and methods

Patients.  All patients with haematological malignancy admitted to the unit for administration of cytotoxic therapy were recruited to the study. Ninety-eight patients underwent in-patient treatment for haematological malignancy during the 3 month study period. Three patients presented with acute myeloid leukaemia (AML) French–American–British (FAB) Type M3, known to be associated with disseminated intravascular coagulopathy and major bleeding complications.

The median age of patients was 52 years (range 16–78); 36 were females and 62 were males, and their main characteristics are summarized in Table I. The majority of patients had acute leukaemia and all patients received intensive chemotherapy regimes that resulted in prolonged pancytopenia.

Table I.  Patient characteristics.
No of patients:98
Median Age:52 (16·6–78·2)
Acute myeloid leukaemia34
Acute lymphoblastic leukaemia11
Chronic myeloid leukaemia 8
Chronic lymphocytic leukaemia 1
Hodgkin's disease 3
Non-Hodgkin's lymphoma 6
Myelodysplastic syndrome 4
Solid tumour 1

Duration of cytopenia.  The median duration of neutropenia [white blood cell (WBC) count < 0·5 × 109/l] and thrombocytopenia (< 20 × 109/l) in patients during the study period was 6 days (ranges 1–38 and 1–44 respectively). (Table II)

Table II.  Duration of pancytopenia.
 Number of
WBC < 0·1 × 109/l273(1–25)
WBC < 0·5 × 109/l636(1–38)
Plts < 10 × 109/l641(1–18)
Plts < 20 × 109/l846(1–44)

Transplants.  The number and type of transplants undertaken were documented and compared to the preceding 9 month period. Twenty-seven patients underwent bone marrow/peripheral blood stem cell transplants (BMT/PBSCT) during the study period, 16 of which were autologous PBSCT and nine were allogeneic (six BMT, three PBSCT). Three of the BMT transplants were from matched unrelated donors (see Table III).

Table III.  Transplant data comparison.
November 2000–January 2001January 2000–October 2000
Transplant typen = 27Transplant typen = 73
Auto PBSCT16Auto PBSCT45
Auto BMT 0Auto BMT 1
Allo PBSCT 3 (all sibling)Allo PBSCT11 (10 sibling, 1 MUD)
Allo BMT 6 (3 sibling, 3 MUD)Allo BMT12 (4 sibling, 8 MUD)

Platelet transfusion protocol.  A protocol taking account of daily platelet count, evidence of bleeding and the presence of additional risk factors was devised and used to determine platelet transfusion requirements (Table IV). In the absence of sepsis, bleeding complications and additional risk factors, a platelet count threshold of < 10 × 109/l was introduced. Higher transfusion thresholds were reserved for patients with additional risk factors and/or fresh major bleeding or patients undergoing surgical procedures such as central venous catheter (CVC) insertion, lumbar puncture and endoscopy. A summary of the indications for platelet transfusion is shown in Table IV.

Table IV.  Platelet transfusion protocol.
Platelet count (× 109/l)Prophylactic platelet transfusion indicated
< 10In every case
10–20In the presence of:
 Pyrexia > 38°C
 Coagulation disorder
 Major bleeding
 Bone marrow biopsy
> 20–50In the presence and until the control of:
 Major bleeding
 Soft tissue bleeding requiring RBC's
 Fresh retinal haemorrhage with impaired vision
 Disseminated intravascular coagulation
 Haemorrhagic cystitis
> 50Prior to surgical intervention (CVC insertion, endoscopy, lumbar puncture, etc.)

Patients were closely observed for evidence of bleeding and the presence or absence of associated risk factors for increased platelet consumption. Bleeding complications were defined in two main categories based on the World Health Organization (WHO) classification (Table V). Major and minor bleeding manifestations of the gastrointestinal, urinary, pulmonary and central nervous system were included. Menstrual loss was excluded from the analysis.

Table V.  Bleeding definitions.
Minor bleedingMucocutaneous haemorrhages or haematomas not requiring RBC transfusion,
retinal haemorrhage without impaired vision
Major bleedingMelaena, haematemesis, petechiae, haematuria, haemoptysis, epistaxis, intercranial haemorrhage, retinal haemorrhage with impaired vision.
Soft tissue bleeding requiring red cell support.

Sampling.  Anticoagulated blood samples were obtained at 06.00 hours each morning for full blood count analysis. Automated full blood counts were performed using automated analysis (Advia 120; Bayer Diagnostics, Leverkusen, Germany). Post-platelet-transfusion full blood count analysis was performed prior to invasive investigations or procedures and in cases where the presence of specific human leucocyte antigen (HLA) antibodies were suspected. In patients who were HLA antibody positive, the average platelet count increment was measured using the following formula:


in which Tx, transfusions; PLT, platelet.

The use of aspirin and other non-steroidal anti-inflammatory drugs was strictly avoided during the study period. The use of prophylactic antimicrobial, antiviral and antifungal therapies was permitted on study. Paracetamol was used as an antipyretic. Packed red cell transfusions were permitted when haemoglobin levels were reduced to less than 9 g/dl.

Blood products.  All blood products were leucocyte depleted to less than 5 × 106 residual white blood cells. Standard red blood cell (RBC) concentrates in saline–adenine–glucose–mannitol (SAG-M) preservative solution were used. Platelet concentrates were either buffy-coat-derived pooled platelets (four) or single donor apheresis platelets with a mean platelet content of 2·4 × 1011 platelets per adult dose. Platelets were normally issued after midday and were stored on a flat bed agitator until transfused later that day.

Data recording and statistical analysis.  Blood product transfusions were documented daily at the patients bedside along with a record of each bleeding assessment. This record was checked against the transfusion record in the hospital blood-bank database (Technidata, Grenoble, France). Blood product usage was analysed on a weekly and monthly basis, and compared with the previous 9 month period when the 20 × 109/l trigger was in place. RBC usage was also recorded during the study period and a comparison made with the proceeding 9 month period. We also calculated the number of platelet and red cell units transfused per patient admission per month. A comparison was again made between the study period and the previous 9 months.

Statistical analysis was performed using statistical package for social sciences (SPSS) Version 9·0. The Mann–Whitney U-test was used to compare differences in platelet usage between the study period and the preceding 9 months when the 20 × 109/l trigger was in place.


Incidence and causes of bleeding

A total of 98 patients who were being treated for a haematological malignancy were prospectively evaluated over a 3 month period. This represented 2147 patient study days. All bleeding episodes were recorded and included in our data analysis. Overall, 271 bleeding episodes were recorded (Table VI), of which 217 were recorded in patients with platelet counts ≥ 10 × 109/l. There were 198 major bleeding episodes as defined by the bleeding classification in Table V and this accounted for 9·2% of the total study days recorded.

Table VI.  Bleeding complications.
Bleeding typePlatelet count (× 109/l)
< 10> 10 < 20> 20N/ANumber of patients
  1. N/A, platelet count not recorded.

Soft tissue bruising2028204 72
Petechiae2427124 67
Retinal 1 8 30 12
Retinal with impaired vision 0 1 20  3
Haematemesis 110 50 16
Epistaxis1415110 40
Haemoptysis 1 7 70 15
Melaena 0 0 40  4

Major bleeding occurred in 13 patients on 1·39% (30/2147) of the study days when the platelet count was < 10 × 109/l. There were 27 patients who had a major bleeding episode on 2·32% (50/2147) of the study days when the platelet count was 10–20 × 109/l.

There were 117 major bleeding episodes in 28 patients observed on 5·45% of the study days in patients whose platelet counts were > 20 × 109/l. Of these, 18 were associated with disseminated intravascular coagulation (DIC) and 54 associated with haemorrhagic cystitis (HC).

Four patients died during the study period; one died within 24 h of admission from disease sequelae and bleeding complications associated with DIC (AML FAB type-M3). One patient died from central nervous system (CNS) aspergillus infection after a second mini-matched unrelated donor (MUD) BMT for T-cell lymphoma. Two patients died from end-stage disease, one AML with progressive CNS disease, and one patient died following acute lymphoblastic leukaemia (ALL) relapse complicated by bony metastases and a renal space occupying lesion. Coagulation abnormalities were present in 18 of the patients studied, one of which was diagnosed with disseminated intravascular coagulopathy.

Risk factors and bleeding

In addition, other risk factors such as pyrexia > 38°C, sepsis and antifungal treatments were present in 23 patients who required platelet transfusions. One patient had 32 major bleeding episodes (16% of the total number of major bleeding episodes) which were associated with the development of post BMT microangiopathy that required daily plasma exchange on a cell separator (COBE Spectra; Gambro BCT, Lakewood, CO, USA). Twenty-four patients developed haematuria during the study, four of whom were diagnosed with HC, accounting for 40% of the total haematuria episodes recorded (54/134).

Overall, major bleeding occurred on 0·51% of the study days (11/2147) in the absence of any additional risk factors in patients with platelet counts ≥ 10 × 109/l.

Blood product usage

There were 339 platelet transfusion episodes administered during the study period, accounting for 410 adult transfusion doses; 171 of these transfusion episodes occurred when the platelet count was < 10 × 109/l. In total, 167 platelet transfusion episodes (49%) took place at platelet counts of ≥ 10 × 109/l. Eighty-one of these transfusion episodes 48·5% (81/167) were in the presence of major bleeding, 6% (10/167) were associated with DIC and 11·3% (19/167) associated with HC. There were 82 platelet transfusion episodes given at platelet counts > 20 × 109/l, of which 15 episodes were prior to invasive procedures, and 10 of these were associated with CVC insertion.

Overall, 39 platelet transfusion episodes occurred in breach of protocol i.e. in the absence of fresh bleeding or additional risk factors when the platelet count was ≥ 10 × 109/l. This accounted for 11·5% of all platelet transfusions given.

HLA antibody samples were taken weekly during the study period and were detected in 20 patients. This group of patients received 98 platelet transfusions which accounted for 24% of all transfusions given.

We calculated the average platelet recovery for each patient to determine whether the presence of HLA antibodies adversely affected transfusion outcomes. Of the 20 HLA antibody patients, only 15 received a platelet transfusion. The mean change in platelet count in each patient was 21 × 109/l. There were 48 HLA antibody-negative patients who received a platelet transfusion and the mean change in platelet count in this group was 16 × 109/l. The difference between the groups was not significant (P = 0·77; Mann–Whitney U-test).


The number of platelet units transfused was reduced by 36% in comparison with retrospective data from the previous 9 months when an arbitrary transfusion trigger of 20 × 109/l was in place. To ensure a fair comparison was being made, the cost of platelet transfusion support was related to patient activity as measured by the number of bed days occupied per month (Table VII). During the 3 months of the study when stringent prophylactic platelet transfusion policies were in place, the median platelet transfusion cost per bed day per month was £ 49·60, which was significantly different to a median platelet cost of £ 81·32 for the preceding 9 month period when the traditional policies were in place (Mann–Whitney U-test P < 0·02). Of note, there was also a 16% reduction in the number of RBC transfusions given, despite no change in the transfusion trigger point. This reduction in blood product usage resulted in a saving of £ 61 953 over the 3 month study period with projected savings of £ 250 000 over the next financial year.

Table VII.  Blood product usage.
 Platelet unitsPlatelet cost (£)Bed daysPlt units/
bed day
Plt cost/
bed day (£)
February 200021432 3724670·569·32
March 200026339 7845370·574·09
April 200030546 1374930·693·58
May 200030646 2895600·582·66
June 200026740 3895390·574·93
July 200029344 3225450·581·32
August 200026840 5405380·575·35
September 200027942 2045190·581·32
October 200029744 9275460·582·28
New policy
November 200016524 9605030·349·62
December 200018928 5905450·352·46
January 200117726 7755740·346·65
February–October 200027942 2045380·581·32
November 2000–January 200117726 7755450·349·62


Despite several reports indicating the safety of reducing transfusion triggers for prophylactic platelet transfusions, optimal transfusion strategies for supporting patients undergoing treatment for haematological malignancy have still yet to be defined. To date, few studies have set out to analyse the criteria for prophylactic platelet transfusion methodically and to evaluate the pharmoeconomic impact of such policies.

Gmur et al (1991) reported results on the safety of introducing a stringent prophylactic platelet transfusion policy in leukaemia patients during induction chemotherapy. In this study, a threshold of 5 × 109/l was shown to be safe for this group of patients in the absence of fresh bleeding or fever (> 38°C). Gil Fernandez et al(1996) reported the clinical results from 190 patients undergoing bone marrow transplantation, where there was a reduction in the transfusion trigger from 20 to < 10 × 109/l. In this study, no significant differences were reported in the relative risk of death from haemorrhage or bleeding between the classical threshold group and the stringent threshold group.

Despite these studies, a survey by the American Association of Blood Banks (AABB) of practices at 630 hospitals involved in the care of onco-haematology patients continued to demonstrate widely varying practice (Pisciotto et al, 1995). Sixty percent of hospitals continued to set thresholds for prophylactic transfusion at 20 × 109/l and 20% higher than 20 × 109/l.

In 1997, results from a multicentre randomised clinical trial by Rebulla et al (1997) further supported the safety of decreasing the classical platelet transfusion threshold. This study of 255 patients undergoing induction chemotherapy for acute myeloid leukaemia demonstrated a 21% reduction in platelet use, following a reduction in the platelet transfusion threshold to 10 × 109/l.

In our study, a threshold of < 10 × 109/l was adopted for all patients undergoing intensive chemotherapy who were not actively bleeding or septic with acceptable relative risk levels of bleeding. Other studies have shown that a reduction in the traditional threshold is safe without increasing the bleeding risk but only in certain well-defined patient groups and settings.

Our results concur with previously published data that a transfusion trigger of < 10 × 109/l in the absence of fresh bleeding and sepsis (> 38°C) is safe. However, the presence of additional risk factors significantly increases the bleeding risk.

There were 44 major haemorrhagic episodes associated with these additional risk factors, 14 of these complications occurred at platelet counts of 10–20 × 109/l. Four patients presented with DIC. Platelet counts were generally maintained above 50 × 109/l in the presence of major bleeding and while clotting times remained abnormal (prothrombin time > 17secs and activated partial thromboplastin time > 34 s), to minimize the risk of major bleeding complications. Despite this, there were 18 major bleeding episodes associated with these patients.

We also found a high incidence of haematuria during our study period, primarily in four patients with platelet counts above 10 × 109/l. This was largely attributable to HC and accounted for 40·2% of all the episodes of haematuria recorded during the study (54/134).

Overall, 36% of major haemorrhagic episodes (72/117) were associated with DIC and HC alone, and all of these complications occurred at platelet counts of ≥ 10 × 109/l. We have since instituted a higher trigger point of > 20 × 109/l in patients with additional risk factors, to reduce the incidence of major bleeding complications.

We demonstrated a significant reduction in platelet usage in comparison with a previous 9 month period. If this reduction was sustained over the next financial year, this would represent health-care savings of £ 181 322 with the potential for reinvestment in patient care (Table VIII). We also established a reduction in red cell usage of 15%. This is difficult to explain but this is probably due to the heightened awareness of appropriate blood product usage as a consequence of the audit. If these reductions in red cell usage were sustained, projected savings would exceed £ 250 000 over the next financial year.

Table VIII.  Blood product costs.
Blood productPrior to changes
(annual cost) (£)
Post changes
(projected annual cost) (£)
savings (£)
  1. CMV, cytomegalovirus.

RBC cost292 380244 86047 520
Platelet cost502 619321 297181 322
CMV-testing supplement (RBC)798668129
CMV-testing supplement (Plt)387324751397
HLA-testing supplement (Plt)48 38230 92817 454
Total848 054·01600 230·47247 823·54

In summary, our findings indicate that our threshold of 10 × 109/l for prophylactic platelet transfusions is safe in all patients being treated for haematological malignancy who are not actively bleeding or septic. Where additional risk factors are present, the threshold should be > 20 × 109/l and in the presence of major bleeding complications platelet counts should be maintained > 50 × 109/l. In addition to achieving substantial health care savings, a more restrictive transfusion policy will also reduce the risk of transfusion-associated immunological complications with the potential to improve patient outcome.


Dr Raj Chopra is supported by Cancer Research UK.