Scott Norris, Haematology Department, Level 4, John Radcliffe Hospital, Oxford Radcliffe NHS Trust, Headington, Oxford, OX3 9DU, UK. E-mail: Scott.Norris@orh.nhs.uk
Summary. Research is required to determine the optimal approach for prophylactic platelet transfusions in patients with haematological malignant disorders. It has been suggested that thresholds for prophylactic platelet transfusions of platelet counts below 10 × 109/l should be investigated, as these may be equivalent in clinical effectiveness and associated with lower costs and fewer complications. An important concern in such investigation is the accurate estimation of platelet counts below 10 × 109/l. This study aimed to further examine the potential reduction in platelet usage that could be made if a lowered platelet transfusion threshold of 5 × 109/l was used in conjunction with an immunoplatelet counting method. Clinical and laboratory data from 130 haematology patients were used. Standard platelet counting was performed using Bayer H3 and ABX Argos analysers. Immunoplatelet counting was performed by flow cytometry using anti-CD61. The potential for reducing platelet transfusions included consideration of clinical criteria that influence prophylactic platelet transfusion use. The results indicated that the use of an immunoplatelet count with a 5 × 109/l platelet transfusion threshold would potentially reduce the number of transfusions by 10·4% in comparison with a 10 × 109/l threshold and standard automated platelet counting with the ABX Argos analyser, and increase the number of transfusions by 5·4% in comparison with the same threshold using the Bayer H3 analyser. The immunoplatelet count may aid the clinical decision to transfuse platelets, but would not necessarily lead to a reduced use of platelet transfusions.
The use of prophylactic platelet transfusions for patients without bleeding is based on a threshold principle, whereby platelets are given to a patient when their platelet count drops to a preset transfusion trigger level. Until the 1990s that level was defined as a platelet count of 20 × 109/l. The validity of the 20 × 109/l level was subsequently brought into question and lower trigger levels were suggested (Gmür et al, 1991; Beutler, 1993).
The 1990s saw the publication of studies (Heckman et al, 1997; Rebulla et al, 1997; Wandt et al, 1998) that supported the work of Gmür et al (1991) in providing further data that a trigger level of 10 × 109/l is safe. Use of a threshold of 10 × 109/l rather than 20 × 109/l resulted in a considerable decrease in the use of platelet transfusions. Gmür et al (1991) advocated a lower threshold of 5 × 109/l in some patients but, because of difficulties associated with accurate platelet counting at this low level, the use of the 5 × 109/l threshold has not been further studied.
In 1997, a Consensus Conference on Platelet Transfusion, Edinburgh (Norfolk et al, 1998), agreed that a threshold of 10 × 109/l is as safe as higher levels for most patients without additional risk factors. A survey of the use of platelet transfusions in centres participating in the Medical Research Council leukaemia trials, carried out earlier in 1997, found that the majority of centres had already switched to a threshold of 10 × 109/l (Murphy et al, 1998). The Consensus Conference also recommended that, for uncomplicated patients, evidence on the safety of even lower thresholds should be sought, but accepted that problems in the accurate counting of platelets below 10 × 109/l may make this difficult.
The problems associated with counting platelets at low levels are due to the inability of current auto-analysers to discriminate platelets from non-platelet particles such as microcytic erythrocytes, schistocytes, debris, chemical precipitation, air bubbles, electronic noise, giant platelets and platelet clumps (Dickerhoff & Von Ruecker, 1995). While manual platelet counting has been used in some studies, this is too labour intensive and is impractical for routine use.
The aim of this study was to examine whether the use of the immunoplatelet count would actually result in a reduction of platelet transfusions, should a transfusion threshold of 5 × 109/l be used instead of the current level of 10 × 109/l. Gmür et al (1991) indicated that considerable savings in the use of platelet transfusions could be made by reducing the threshold for platelet transfusions from 20 to 5 × 109/l, but the potential savings from reducing the threshold from 10 to 5 × 109/l may be somewhat different from those found with the larger reduction from 20 to 5 × 109/l. This study, unlike some other groups who have suggested potential savings could be made in platelet transfusions using the immunoplatelet count, included consideration of clinical criteria that influence the use of prophylactic platelet transfusions.
Materials and methods
All samples were collected from tripotassium ethylenediamine tetra-acetate (K3EDTA) peripheral whole blood samples, taken as routine samples on the haematology ward of the John Radcliffe Hospital, Oxford. The samples were kept at room temperature and analysed within 6 h of collection. Only samples from patients with a platelet count of < 50 × 109/l (using the Bayer H3 analyser) were used in the study.
The study was a retrospective study of patients' platelet counts and transfusions and in no way affected the decisions of the clinicians involved in the transfusion of patients, who routinely use H3 platelet counts to monitor patients' platelet counts.
Automated platelet counts. Automated platelet counts were performed on H3 (Bayer, Newbury, UK) and ABX Argos (ABX, Shefford, UK) analysers. The H3 provided a platelet count that was derived from an optical technology while the ABX provided a count based on impedance counting technology. The H3 is used routinely in the laboratory as the main analyser and, as such, the actual platelet transfusion decisions are based on H3 data.
Immunoplatelet count. The method used was that as published by Harrison et al (2000). Mixed whole blood (2 µl) was placed in a cytometer tube followed by 2 µl of anti-CD61 (Clone RUU-PL 7F12; Becton Dickinson, Oxford, UK). FACS Flow (6 µl) (Becton Dickinson) was then added, and the reagents were mixed gently with the pipette and left to incubate for 1 min. After incubation, 2 ml of FACS Flow was added to the tube to give a final dilution of 1/1000. Before analysis on the flow cytometer (FACSCaliber; Becton Dickinson), the cell suspension was diluted 1/2 in FACS Flow to give a final dilution of 1/2000. This high dilution negates the need to correct the final calculations for coincident events (Harrison et al, 2001).
The immunoplatelet count was calculated by counting the number of platelet events and red cell events using bitmap analysis of the dot plots generated from the flow cytometer (see Fig 1). The actual platelet count was calculated using the peripheral RBC count from either the H3 or ABX, depending upon which machine was being compared, using the following formula:
Indicators for platelet transfusion. The patients used in the study were monitored daily for clinical criteria that would be used to indicate the need for either a therapeutic or prophylactic platelet transfusion. These indicators included fever, bleeding, sepsis, the use of heparin or the need for a minor procedure to be performed. Whether or not the patient was actually transfused was also recorded.
Data analysis. The data from the immunoplatelet count was compared with that of the H3 and ABX count with the student's paired t-test, and correlation and regression analysis to look for significant differences in the data.
The results were compared to determine differences in the decision to transfuse using a threshold for platelet transfusion of a platelet count of 5 × 109/l and 10 × 109/l using immunoplatelet counting, and 10 × 109/l using automated counting with H3 and ABX analysers. This was done using the results of platelet counts and clinical data. The clinical data were applied using the criteria in the Gmür protocol that was suggested to be safe for the use of a 5 × 109/l threshold (Gmür et al, 1991):
• transfuse all patients with a platelet counts of < 5 × 109/l;
• transfuse patients with a platelet count of 6–10 × 109/l, if fever or minor haemorrhage is present;
• transfuse patients with a platelet count of 11–20 × 109/l, if heparin is being administered or coagulation disorders present, or if bone marrow biopsy or lumbar puncture is to be performed;
• transfuse patients with a platelet count of > 20 × 109/l, if minor surgery is to be performed, or to control major bleeding.
Threshold data comparisons. The platelet counts were plotted as graphs to compare the difference in platelet counts in relation to the transfusion thresholds.
The data were grouped by the number of platelet counts that fell within the platelet count thresholds corresponding to the protocol: 0–5, 6–10, 11–20 and > 20. It was then determined whether the patient would have been transfused or not, depending on the protocol used. The two protocols (1 and 2), which do not take into account the patients clinical indications, were used for comparison of the protocols with clinical indications (3, 4 and 5). Protocols 1 and 2 are included to show that potential savings based solely on lowering the transfusion threshold may not be realistic when compared with protocols that take into account the patient's clinical indications. The first five protocols are shown in the bars of the graphs of Figs 2Band 3B. Protocols 6 and 7 relate to comparison of thresholds of a platelet count of 10 × 109/l using the immunoplatelet count, with a threshold of 10 × 109/l using the H3 and ABX analysers. These data are not shown in the figures:
Actual transfused patients (using Oxford Radcliffe platelet transfusion protocol). This was applied only in relation to the H3 count and not the ABX count, as the H3 count was used to make decisions on the use of platelet transfusions.
Costs. The potential savings or increase in cost was calculated, based on the known costs of platelet concentrates.
Comparison of immunoplatelet count and H3 count (see Fig 4)
One hundred and thirty samples were analysed with the immunoplatelet count giving significantly (P = < 0·01) lower results than the H3 across the range of values (1–53 × 109/l immunoplatelet count). The results showed a good correlation (r2 = 0·89) and a constant bias of 4·85. The bias may indicate that the H3 is unable to resolve platelets below a value of approximately 5 × 109/l.
Comparison of immunoplatelet count and ABX count (see Fig 5)
One hundred and sixteen samples were analysed with the immunoplatelet count giving significantly (P = < 0·01) higher results than the ABX across the range of values (1–47 × 109/l immunoplatelet count). The results showed a good correlation (r2 = 0·83) and a small constant bias of 0·87. The bias may indicate that the ABX has the potential to resolve platelets to a lower level than the H3.
Figure 2B illustrates the overestimation of platelet counts by the H3 analyser in the 0–5 × 109/l range, as shown by only four counts in this range compared with 25 using the immunocount. The bias to higher counts using the H3 is shown by 21 more counts falling in the 11–20 × 109/l and > 20 × 109/l ranges than with the immunocount. The ABX data (shown in Fig 3B) shows, in contrast, nine more counts than the immunocount in the 0–5 × 109/l range, indicating an underestimation of platelet counts.
Even using the immunoplatelet count, 18/25 (72%) of the patients with platelet counts in the 6–10 × 109/l range (counted by the H3 analyser) and 16/22 (73%) of patients with platelet counts in this range (counted by the ABX analyser) had clinical indicators for platelet transfusion which would have triggered a platelet transfusion.
Comparison of 5 × 109/l threshold, immuno against 10 × 109/l threshold, H3 (seeFig 2)
Data without clinical indications taken into account. Reducing the platelet threshold from 10 × 109/l using the H3 to 5 × 109/l using the immunoplatelet count would have potentially saved four transfusions overall (3%). This is derived from eight counts that were below 10 × 109/l on the H3 count but above 5 × 109/l on the immunoplatelet count, and four counts that were above 10 × 109/l on the H3 count but below 5 × 109/l on the immunoplatelet count (see shaded areas on Fig 2A).
Data with clinical indications taken into account. Our local transfusion protocol is identical to the Gmür protocol at platelet counts above 10 × 109/l, and it would be expected to give similar results. Using the Gmür et al (1991) protocol indicators, the number of transfusions using a 5 × 109/l threshold using the immunoplatelet count would have been 59 compared with 52 using a 10 × 109/l threshold and platelet count from the H3 analyser. This is an increase of seven transfusions (5·4%) through using the lower threshold (see Fig 2B). The actual number of transfusions performed was 39; this is 13 less than would be expected using the Gmür et al (1991) protocol with zero non-compliance. The 13 non-transfused patients were not transfused according to the clinical judgement of the medical staff, but this only represented a 10% non-compliance with the decision to transfuse, using a strict application of the Gmr protocol.
Comparison of 5 × 109/l threshold, immuno against 10 × 109/l threshold, ABX (see Fig 3)
Data without clinical indications taken into account. Reducing the platelet threshold from 10 ×
109/l using the ABX to 5 × 109/l using the immunoplatelet count would have potentially saved 29 transfusions overall (25%). Again this is derived from 29 counts that were below 10 × 109/l on the H3 count but above 5 × 109/l on the immunoplatelet count, and zero counts that were above 10 × 109/l on the ABX count but below 5 × 109/l on the immunoplatelet count (see shaded areas on Fig 3A).
Data with clinical indications taken into account. Using the Gmür et al (1991) protocol indicators, the number of transfusions using a 5 × 109/l threshold would have been 52 compared with 64 using a 10 threshold and ABX count. This is a decrease of 12 (10·4%) transfusions through using the lower threshold (see Fig 3B). If the ABX count had been used routinely, the number of platelet transfusions would have been 8% higher than was the case using the H3 count.
Comparison of 10 × 109/l threshold, immuno against 10 × 109/l threshold, H3
Data without clinical indications taken into account. Comparing a platelet threshold of 10 × 109/l for both the immunoplatelet count and the H3 platelet count, the use of the immunocount would have potentially increased the number of transfusions by 21 transfusions overall (16%). This is derived from 29 counts that were below 10 × 109/l on the H3 compared with 29 counts below 10 × 109/l with the immunoplatelet count (50–29 = 21).
Data with clinical indications taken into account. Using the Gmür et al (1991) protocol indicators, the number of transfusions using a 10 × 109/l threshold and the immunocount would have been 66 compared with 52 using the H3 count. This would amount to an increase of 14 transfusions (10·8%).
Comparison of 10 × 109/l threshold, immuno against 10 × 109/l threshold, ABX
Data without clinical indications taken into account. Comparing a platelet threshold of 10 × 109/l for both the immunoplatelet count and the ABX platelet count, the use of the immunocount would have potentially decreased the number of transfusions by seven transfusions overall (6%). This is derived from 49 counts that were below 10 × 109/l on the ABX compared with 42 counts below 10 × 109/l with the immunoplatelet count (49–42 = 7).
Data with clinical indications taken into account. Using the Gmür et al (1991) protocol indicators, the number of transfusions using a 10 × 109/l threshold and the immunocount would have been 58 compared with 64 using the ABX count. This would amount to a decrease of six transfusions (5·2%).
The cost of platelet transfusions for haematology patients for the year 2000 at the Oxford Radcliffe Hospital was around £150 000. Changing from the H3 10 × 109/l threshold to the immunoplatelet count 5 × 109/l threshold would have increased costs by 5·4% (total £158 100).
If the ABX had been used as the routine platelet count analyser with a threshold of 10 × 109/l, the cost of transfusions would have been 8% higher (total £162 000) than using the same threshold with the H3 analyser. The use of a threshold of 5 × 109/l and immunoplatelet counting would have reduced the cost of platelet transfusions by 10·4% to £145 152.
The ready availability of platelet concentrates has undoubtedly made a major contribution to modern clinical practice, in particular, in allowing the development of intensive treatment regimens for haematological and other malignancies. Considerable advances have been made in platelet transfusion therapy in the last 30 years, but some areas continue to provoke debate especially the use of prophylactic platelet transfusions. Research is still required to determine the optimal approach for prophylactic platelet transfusions in patients with haematological malignant disorders, both in terms of their clinical benefit and their cost-effectiveness.
It has been suggested that thresholds for prophylactic platelet transfusions of platelet counts lower than 10 × 109/l should be investigated (Norfolk et al, 1998), as these may be equivalent in their clinical effectiveness and associated with lower costs and fewer complications.
An important concern in this investigation is the accurate estimation of platelet counts below 10 × 109/l. The traditional reference method for platelet counting has been the manual microscopic chamber count, which is labourious and imprecise for estimating low platelet counts. Standard haematology analysers using single-channel impedance technology are similarly unsuitable for estimating low platelet counts. Recent studies have shown that platelet counting of very low platelet counts is more accurate with new methods such as two-dimensional optical analysis and immunological counting using a fluorescein-labelled CD61 monoclonal antibody (Dickerhoff & Von Ruecker, 1995; Harrison et al, 2000; Kunz et al, 2000). A recent multinational task force established that a flow cytometric method using the labelling of platelets with two monoclonal antibodies met the criteria for a reference platelet count (Harrison et al, 2001; ICSH Expert Panel on Cytometry and International Society of Laboratory Hematology Task Force on Platelet Counting, 2001).
The aim of this study was to examine whether the use of the immunoplatelet count would result in a reduction of platelet transfusions, should a transfusion threshold of 5 × 109/l be used instead of the current level of 10 × 109/l. The potential for reducing platelet transfusions included consideration of clinical criteria that influence the use of prophylactic platelet transfusions. The results obtained gave conflicting answers, with the H3 comparison indicating that the number of transfusions would actually increase, while had the ABX been used for routine platelet counting then the number of transfusions would have decreased. In both cases, however, the most noticeable finding was that the potential reduction in patients' transfusion were not as obvious as had previously been suggested, when taking into account the patient platelet count alone. The data from this study suggest that the clinical criteria for prophylactic platelet transfusion play an important role in determining the number of transfusions and may outweigh any savings that may be made by using a lower transfusion threshold.
In this study, comparing the use of a 5 × 109/l platelet count threshold using the immunoplatelet count with the H3 analyser and a 10 × 109/l threshold, the overall number of platelet transfusions would be increased with a greater potential risk of platelet transfusion complications, while having a substantial effect on costs. This study examined a broad picture with over 130 patients, but did not following specific patients from day to day. Some patients may actually see a reduction in the number of transfusions they receive, providing they were stable and did not have any clinical indications for transfusion.
Some patients were found to be under-transfused using the H3 platelet count; this is indicated by the fact that four patients had counts above 10 × 109/l using the H3 count but below 5 × 109/l with the immunoplatelet count. All the four patients were transfused as they all had clinical indications for transfusion, but had they been stable they may not have been transfused. The appropriate transfusion of patients who may be currently being under-transfused should be considered a benefit of the immunoplatelet count, even if the lower platelet transfusion threshold is not used. The reduction in the platelet threshold should be considered separately from the use of the immunoplatelet count, even though the more accurate platelet counting by the immunoplatelet count is necessary to implement the lower threshold of 5 × 109/l for prophylactic platelet transfusions.
The ABX results were in favour of using the lower threshold, as there would have been an overall reduction in the number of transfusions and costs. If the ABX had been used as the routine analyser, it would be easy to argue in favour of changing to the immunoplatelet count with the lower threshold, as long as there was no significant increase in bleeding, as determined by future trials. The cost to the laboratory of implementing an immunoplatelet count, and the labour and time necessitated, would be met by the potential savings in the cost of transfusions. As with the H3 counter, patients would benefit, but the justification for any laboratory using an ABX for routine platelet counting would be far simpler.
The use of the immunoplatelet count without changing the transfusion threshold from 10 × 109/l would have given results similar to using the lower 5 × 109/l threshold, in that savings in transfusions would have been seen if changing from the ABX and an increase in transfusions if changing from the H3. The major difference between the two thresholds would be that more patients would be transfused regardless of clinical criteria for prophylatic platelet transfusion and, therefore, any potential savings in the number of transfusions would be less with the higher 10 × 109/l threshold than the 5 × 109/l threshold.
The conclusions that can be drawn from examining both sets of data together are less straightforward. It is reasonable to assume that the immunoplatelet count is the correct platelet count, as it is the international reference method. Accepting this, the H3 was overestimating the platelet count at low platelet counts, while the ABX was underestimating the counts. The ABX, being an impedance analyser, would have been expected to have overestimated the platelet count because of non-platelet particle interference (Hammerstrøm, 1992). The fact that it is underestimating the count may say more about the calibration and differences between auto-analysers. Both the H3 and ABX are monitored by the National External Quality Assurance Scheme (NEQAS) scheme, with both machines comparing favourably with other machines in the user groups. Alternatively, the results may indicate that the calibration of the analysers at low platelet counts may not be accurate, either locally or nationally. The precision of the analysers may also be an issue at these low levels, but the clear distinction of one analyser overestimating and the other underestimating tends to imply that the precision even at these levels is acceptable (the immunoplatelet count had a coefficient of variance of < 3·0% across the platelet range). Had the analysers been imprecise, the clear biases in the results would probably have been less obvious. The NEQAS scheme does not usually provide materials with platelet counts at very low counts, which are at the level where transfusion decisions are likely to be made. The study highlights the possibility that a low platelet count (below 10 × 109/l) would be useful as a NEQAS sample and may indicate the need for the use of the immunoplatelet count by NEQAS as a measure of the samples they are distributing. Whether the differences are due to calibration or machine differences, the results show that the decision to transfuse may be biased depending upon the type of analyser used in the laboratory. This implies that nationally there could be a large variation in the platelet transfusion usage from hospital to hospital.
The most obvious conclusion from the study is that the benefits of the introduction of the new immunoplatelet count will vary depending on the hospital and the type of auto-analyser they are currently using. Whether savings can or cannot be made, the argument for the immunoplatelet count to be used for low-level platelet counting is still strong. An accurate platelet count allows clinicians to make a more informed judgement about patients' risk of bleeding. If this means an increased cost to the hospital for some laboratories, it is surely justifiable, as the patients may be at a reduced risk of bleeding.
It will obviously be impractical and costly for smaller laboratories to use the immunoplatelet count to monitor their haematology patients, and we would not suggest that they do so. The immunoplatelet count, if used in other larger multicentre studies, may, however, prove useful in indicating to smaller laboratories how their analysers compare with both other analysers and other hospitals around the country, and encourage them to calibrate them better for low platelet counts.
The study shows that there can be a large difference between analyser platelet counts at the low platelet count levels used to make clinical decisions. The immunoplatelet count may prove useful to manufacturers of full blood count analysers in the standardization of platelet counting technology.
In conclusion, the study shows the need for further investigation of the use of platelet transfusions in relation to platelet counting methods. Multicentre laboratory studies of the immunoplatelet count will be necessary to compare other auto-analysers with larger sample numbers than is possible with a single centre study, such as this. Once the immunoplatelet count has been further evaluated, clinical trials of lower thresholds for prophylactic platelet transfusions could be undertaken.
This work was part of the studies for the MSc Haematology (Transfusion Science) at the University of Westminster. We thank Dr Paul Harrison (Haemophilia Centre, Oxford) for his critical reading of the manuscript.