Immunoplatelet counting: a proposed new reference procedure

PAUL HARRISON,* ALLAN HORTO N, DONNA GRANT,* CAROL BRIGGS* A ND SAM MACH IN* *Haemostasis Research,

Department of Haematology, 98 Chenies Mews, University College London WC1E 6HX, UK, and Gulf Coast Pathology,

Cellular Analysis Division, Fort Myers, FL, USA

Received 7 October 1999; accepted for publication 11 October 1999

Summary. Given the high degree of interoperator error andpoor precision of manual platelet counting, it has recentlybeen proposed that an immunoplatelet counting methodcould become the new reference procedure. Platelets areidentied immunologically with a suitable monoclonalantibody, and the platelet count is derived from the ratio ofuorescent platelet events to collected red blood cell (RBC)events that are also counted by a reliable and calibratedstandard impedance counter (RBC ratio). In this study, wehave set up a rapid and simple method for immunoplateletcounting and simultaneously compared the RBC ratio withthe bead ratio derived from two different preparations ofcommercial calibration beads (Trucount and FlowCountbeads). Comparison of the level of imprecision of the RBCratio with either the manual count or bead ratios revealed asuperior coefcient of variation of < 5% even in samples with

a platelet count < 20 109/l. The RBC ratio correlatedextremely well with the existing manual phase referencemethod (r2 093) and especially well with three differentcommercial impedance counters and a dual-angle opticalcounter (r2 098099). However, at < 100 109/l, thecorrelation of the RBC ratio with the dual-angle opticalcount (ADVIA 120) (r2 096) was superior to allimpedance counters. This suggests that automated opticalcounting methods may be more accurate at determiningplatelet counts in thrombocytopenic samples. As the RBCratio is rapid, cheap and relatively easy to perform, wepropose that this method could replace the manual count asa new international reference method.

Keywords: immunoplatelet count, CD61, optical plateletcount, impedance platelet count, quality control.

There is an increasing clinical need for reducing the use ofprophylactic platelet transfusions. This has driven theplatelet transfusion threshold down from 20 109/l to10 109/l but without a signicant increase in spontaneousbleeding risk (Rebulla et al, 1997; Ancliff & Machin, 1998;Norfolk et al, 1998). Given the expense of platelet trans-fusions, this has resulted in a signicant decrease in bloodbank expenditure. It has also been proposed that the decisionthreshold could be reduced further to as low as 5 109/lprovided the clinicians are condent in the reliability of thecount for assessing bleeding risk (Gmur et al, 1991; Murphy,1992; Ancliff & Machin, 1998). Therefore, in order toguarantee that the correct and safe clinical decision is made,it is becoming more critical that platelet counts in severethrombocytopenia are not only precise but accurate.Although modern impedance automated cell counters arevery precise, their inability to resolve platelets from celldebris and other particulate matter often results in anoverestimate of the platelet count, particularly in severe

thrombocytopenia (Rowan, 1991; Hammerstrom, 1992;Dickerhoff & von Ruecker, 1995; Springer et al, 1998). Inorder to calibrate and assess which type of instrument canperform the most reliable platelet counts, particularly at< 20 109/l, one requires an accurate and precise referencecounting procedure (Ault, 1996).

The current international reference method for plateletcounting is performed by a manual procedure using phase-contrast microscopy (England et al, 1988). This method hasa number of signicant limitations and suffers from impreci-sion, with a typical interoperator coefcient of variation (CV)in the order of 1025%. Recently, a new immunoplateletcounting procedure has been advocated as a possiblealternative reference method and is under review by theInternational Council for Standardization in Haematology(ICSH) (Dickerhoff & Von Ruecker, 1995; Tanaka et al, 1996;Ault et al, 1997; Davis & Bigelow, 1999). The principle of themethodology involves labelling a whole-blood sample withan antiplatelet monoclonal antibody (conjugated to auorophore) and detecting the platelets by ow cytometry.The platelet count is simply calculated from the ratio ofuorescent platelet events to the number of red cells detected

British Journal of Haematology, 2000, 108, 228235

228 q 2000 Blackwell Science Ltd

Correspondence: Dr Paul Harrison, Haemostasis Research, Depart-

ment of Haematology, 98 Chenies Mews, London WC1E 6HX, UK.

229Immunoplatelet Counting

q 2000 Blackwell Science Ltd, British Journal of Haematology 108: 228235

(RBC ratio). To derive the count, one simply multiplies thisnumber by the known RBC count in the sample, asdetermined by a standard impedance analyser. The mainadvantage of the RBC ratio is that, providing the bloodsample is mixed and that coincident events (RBC/RBC andRBC/platelet) are eliminated by optimal dilution, then thecount obtained is totally independent of potential pipettingand dilution artifacts. An alternative immunoplateletcounting procedure uses a known amount of addeduorescent calibration beads to derive the platelet count(bead ratio) (Dickerhoff & Von Ruecker, 1995; Matzdorff et al,1998). However, unlike the RBC ratio, this method isdependent upon very accurate and precise pipetting.

In this study, the aim was to establish an optimized, rapidand simple immunoplatelet counting procedure that couldsimultaneously determine the RBC ratio, with the bead ratioderived from two different preparations (lyophilized andsuspension) of commercially available calibration beads(Trucount and FlowCount beads). To compare the efcacyof these three proposed immunoplatelet counting proce-dures, data were analysed using three different automatedimpedance cell counters and a recently available automatedblood cell counter using two angles of laser light scatter(ADVIA 120), and also the current manual referencemethod.

METHODS

Calibration beads. Trucount absolute count tubes (BectonDickinson, Oxford, UK) contain a lyophilized pellet of 42 mmuorescent beads and are thus very stable. FlowCount beads(Beckman Coulter, High Wycombe, Bucks, UK) are a suspen-sion of 10 mm uorescent beads and were prepared asfollows. Briey, four bottles of FlowCount were diluted withan equal volume of Isoton, allowed to stand overnight andcentrifuged at 1000 g for 10 min. The supernatant wasremoved, and the pellets were resuspended in one-third ofthe original volume and stored at 48C.

Immunoplatelet counting method. All samples were obtainedfrom standard vacutainer K3EDTA anticoagulated peripheralwhole-blood samples and kept at room temperature for 24 hor less. All quantities were aliquoted using calibrated positivedisplacement pipettes, and the outside of the plastic tip wascarefully wiped with tissue paper to remove excess blood orsolution. Mixed blood (2 ml) was pipetted into the bottom of aFalcon polystyrene tube (Becton Dickinson). Anti-CD61uorescein isothiocyanate (FITC; 2 ml; 6 mg/ml; BectonDickinson), an antiglycoprotein IIIa monoclonal antibody(clone RUU-PL 7F12) (Wong & Springer, 1995), was thenimmediately pipetted into the same tube but next to and notmixed with the blood sample. Isoton (6 ml; Beckman Coulter,High Wycombe, Bucks, UK) was then added, and the bloodwas mixed with both antibody and Isoton by pipetting gentlyup and down. The tube was then incubated for exactly 1 minand diluted to exactly 2 ml with Isoton. The relatively shortincubation time of 1 min was demonstrated not to affect thederived platelet count when compared with longer incuba-tion periods of up to 30 min (data not shown). After gentlemixing, 05 ml was removed and added to a Trucount tube

(Becton Dickinson). Isoton (04 ml) was then added, followedby 100 ml of the preprepared suspension of FlowCount(Beckman Coulter).

The nal dilution of whole blood was therefore 1:2000.Flow cytometric analysis was performed within 2 h using aCoulter XL ow cytometer (Beckman Coulter). Samples wereanalysed at medium ow rate with a forward scatterdiscrimination of 10 for either 30 s or, alternatively, until1000 platelet events (in thrombocytopenic samples) or50 000 RBC events had been collected.

Flow cytometric data analysis. Plateletred blood cell (RBC)ratios or plateletbead ratios were calculated from plottedhistograms of cell size (log forward scatter) vs. granularity(log side scatter) (Fig 1) and uorescence (log FL1 or FL3) vs.cell size (log forward scatter) (Fig 2). Figure 1 clearly showsthe platelet (A), RBC (B) and two different bead populations(C and D). FlowCount and Trucount beads are located above(C) and below the RBC cloud (D) respectively. Figure 2demonstrates that uorescent platelets (gate C) are resolvedfrom either noise/debris (gate G), RBC (gate E) and platelet/RBC coincident events (gate F).The nal platelet countsdetermined via the RBC ratio were calculated by dividingthe number of uorescent platelet events (in gate C, Fig 2) bythe number of red blood cell events (gate E, Fig 2) andmultiplying by the known RBC count determined on acalibrated Coulter STKS automated impedance cell counter(Beckman Coulter). The nal platelet counts determinedfrom the bead ratios were also determined in parallel bymultiplying the number of uorescent platelet events (gate C,Fig 2) by the ratio of total bead number to collected beadevents (Trucount beads are in gate D, Fig 2; FlowCount beads

Fig 1. Flow cytometry scattergram (log FS vs. log SS) of all detected

events (Coulter XL with a FS discriminator of 10) within a bloodsample diluted 1:2000. The platelet cloud is shown on the left-hand

side (A) and is clearly resolved from either the red blood cell cloud (B)

or the two discrete bead populations. FlowCount beads (10 mm) and

Trucount beads (4 mm) are located above (C) and below (D) the redblood cell cloud (B) respectively.

are in gate H of a plot of log FL3 vs. log FS; not shown)followed by multiplication by the dilution factor of 2000.

Manual platelet counts. Two experienced operators per-formed manual platelet counts, using a standard phasemicroscopy method (Brecher et al, 1953; England et al,1988). A 1:20 dilution of EDTA anticoagulated whole bloodwas prepared in a diluent of 1% ammonium oxalate. Thesuspension was then mixed on a mechanical mixer for 1015 min to allow for the complete lysis of all red cells. A cleandust-free Neubauer chamber was lled with the suspensionand left in a moist chamber for 20 min to allow the plateletsto settle. The preparation was then examined using the 40objective, the platelets appearing as small but refractiveparticles. The total number of platelets appearing in 80 smallsquares on the chamber were counted and were equivalentto the platelet count 109/l. Both sides of the chamber werecounted to check for reproducibility. The nal count wasthen taken as the mean value of all four counts.

Automated platelet counts. Platelet counts using an impe-dance technique were determined on the Coulter STKS, CoulterGen-S (Beckman Coulter) and SE-9500 (Sysmex UK) auto-mated cell counters. In the impedance method, a specicamount of diluted blood ows through a small aperturelocated between two sensing electrodes. The red cell andplatelet counts are performed on the same cell suspensionwith the platelets being analysed by the number of pulses ofcell size between 2 and 20 (red cells being over 36 ).Analysis of pulses for the exact platelet count varies in the

different analysers. Indeed, the same manufacturer mayanalyse impedance counts slightly differently within distinctmachine models. The Coulter STKS and GEN-S (BeckmanCoulter) apply log normal conversion of the data andextrapolate the curve to cover the normal size range up to70 . In contrast, the SE-9500 (Sysmex UK) produces a plateletsize single histogram using three thresholds or discriminators.One is xed at 12 and the other two hunt the lower (between2 and 6 ) and upper (between 12 and 30 ) ends of the plateletpopulation within these limits. These thresholds allow theanalyser efciently to distinguish platelets from debris at thelower end and red cell fragments at the upper end.

An optical platelet count was determined on the ADVIA120 (Bayer Diagnostics, Newbury, Berkshire, UK). Thisanalyser simultaneously measures laser light scatter (usinga solid-state laser diode) at low angle (238) and higherangle (5158) in the forward direction, in a similar mannerto their red cell analysis. Platelets are analysed betweenrefractive index values of 135 and 14 up to a size of 60 .This allows larger platelets to be included in the plateletcount (Stanworth et al, 1999).

RESULTS

Elimination of coincidence eventsOne of the potential problems with immunoplatelet countingusing ow cytometry is the inability of the ow cell todiscriminate platelet/RBC and/or RBC/RBC coincidenceevents (Ault, 1996; Davis & Bigelow, 1999). In order tooptimize the procedure, a series of studies was performed todetermine the inuence of sample dilution and acquisitionrate on both the RBC and bead ratio counts. For example,low sample dilutions were associated with higher numbers ofplatelet/RBC coincidence events and RBC events (Table I).Dilution curves from 1:125 to 1:8000 at three different owrates demonstrated that a 1:2000 dilution with a mediumacquisition rate (data from low and high ow rates are notshown) are the optimum conditions required to eliminatecoincidence events and thus give an accurate platelet count(Table I). This is in close agreement with the gure of 1:1800suggested by Ault (1996).

Precision studyThe precision of the RBC and bead ratios was determined byanalysis ( 10) of three samples with low (18 109/l),normal (181 109/l) and high platelet counts (751 109/l).A comparison of the 30 s analysis time with data collectionof 1000 platelet events or 50 000 RBC events was alsoundertaken. The results are summarized in Table II. The dataclearly show that the CV of the RBC ratio is superior (< 5% inthe normal sample) to either the Trucount ratio or theFlowCount ratio. With the thrombocytopenic sample, theRBC ratio CV was excellent (28%) and again superior toboth bead-derived counts (44% and 56%), but only when1000 platelet events were collected. As a result, any sampleswith a platelet count < 50 109/l were subsequently reana-lysed to collect at least 1000 platelet events. As the RBC ratioexhibited superior precision to the bead ratios, all subsequentanalyses were undertaken using this parameter.

q 2000 Blackwell Science Ltd, British Journal of Haematology 108: 228235

230 P. Harrison et al

Fig 2. Flow cytometry histogram (log FL1 vs. log FS) of all events

collected in Fig 1. The uorescent platelets (gate C) are clearly

resolved from either noise/debris (gate G), RBC (gate E), platelet/RBCcoincident events (gate F) and Trucount beads (gate D). The nal

platelet counts determined by the RBC ratio were calculated by

dividing the number of uorescent platelet events (gate C) by the

number of RBC events (gate E) and multiplying by the known RBCcount. The bead ratio counts were calculated by multiplying the

number of platelet events (gate C) by the ratio of total bead number

to collected Trucount (gate D) or FlowCount bead events (gate H; notshown) followed by multiplication by the dilution factor of 2000.

231Immunoplatelet Counting

q 2000 Blackwell Science Ltd, British Journal of Haematology 108: 228235

Comparison with manual countsAny proposed reference method should ideally be comparedwith the existing procedure. Forty-two samples with a widerange of platelet counts (7506 10 9/l) were analysed bymanual, RBC and bead-derived counts. Figure 3 comparesthe correlation of the manual method with the RBC ratio.Figure 3A shows all data (n42), and Fig 3B focuses atplatelet counts below 100 109/l (n20). The interobserverCVs between the four manual determinations varied from 2%to 48%, with an overall mean of 16%, clearly demonstratingthe problem with the existing reference method.

Comparison with impedance counters and the ADVIA 120Eighty-seven samples with a range of 4794 109/l (CoulterSTKS) were analysed by the RBC ratio, bead ratios and thethree different available impedance cell counters. Figure 4compares the correlations of RBC ratio with the impedancecounters and the ADVIA 120 over the whole range (Fig 4A)and < 100 109/l (Fig 4B). All three impedance countersand the ADVIA 120 all correlated extremely well (r2098099) with the RBC ratio over the entire range of platelet

counts. However, below 100 109/l, the correlation of theADVIA 120 optical count with the RBC ratio was clearlysuperior (r2 096). One grossly lipaemic sample, whichwas excluded from the correlation, gave very high impe-dance counts (135, 103 and 183 109/l) relative to both theimmunocount (43 109/l) and ADVIA 120 (59 109/l).

Examples of aberrant impedance countingFigure 5 illustrates a total of 41 examples with plateletcounts below < 50 10 9/l, which were found to exhibitdifferences between the RBC ratio and impedance counts.The graph illustrates the effect of two different transfusionthresholds set at either 10 or 20 109/l on clinical decision-making. Making a reasonable assumption that the immu-nocount was the truest value, then with the transfusionthreshold set at 20 109/l, seven patients would have beengiven unnecessary transfusions, as they were underesti-mated by the impedance counter. Conversely, eight patientswould not have been transfused, as they were overestimatedby the impedance counter and therefore may have exhibitedclinical bleeding problems. In contrast, if the transfusion

Table I. Inuence of blood dilution on the number of total events, RBC events, platelet events, RBC/platelet coincident events (expressed as

absolute number or %), calculated RBC ratio and platelet counts.

Dilution Events/ RBC Platelet RBC Calculated RBC/platelet RBC/platelet

factor second events events ratio platelet count coincident events (% platelets)

125 ND ND ND ND ND ND ND250 4501 120887 7329 0061 296 802 109

500 3576 93136 5785 0062 302 362 63

1000 1978 46173 2719 0059 288 136 502000 1342 29230 1642 0056 273 54 33

4000 866 14407 799 0055 268 28 35

8000 636 8272 291 0035 171 14 48

A normal sample was analysed at medium ow rate for 30 s. ND, not determined.

Table II. Comparison of coefcients of variation (CVs) obtained with three differentsamples with platelet counts of 18, 181 and 751 109/l.

Platelet count RBC ratio (%) Trucount ratio (%) FlowCount ratio (%)

30 s analysis18 83 81 81

181 21 53 36

751 50 74 73

1000 platelet events

18 28 44 56

181 40 71 71

751 56 64 76

50 000 RBC events

18 75 89 104

181 389 714 699751 653 56 44

The RBC ratio is compared with the bead ratios obtained with Trucount or

FlowCount. CVs are compared when acquisition was stopped after 30 s, or 1000platelet events or 50 000 RBC events had been collected.

threshold was set at 10 109/l, then only three patientswould have been overtransfused and four patients under-transfused. Table III illustrates specic examples in which theimpedance platelet count was either overestimated orunderestimated because of the presence of either noise orlarge platelets, resulting in counting errors compared withthe RBC ratio count.

DISCUSSION

The goal of this study was to set up an optimized plateletimmunocounting procedure and demonstrate its potential asan international reference method. Indirect platelet immu-nocounts can be derived from either the RBC or bead ratios,which require either an accurate RBC count (performed by acalibrated impedance counter; Dickerhoff & Von Ruecker,1995; Tanaka et al, 1996; Ault et al, 1997; Davis & Bigelow,1999) or the addition of a known quantity of added calibrationbeads (Matzdorff et al, 1998). Platelets are identied immuno-logically with a suitable monoclonal antibody (e.g. anti-CD61-FITC) and are thus resolved from both noise/debrisand red cell events. Providing the sample is optimally diluted(1:2000) before analysis at the medium acquisition rate onthe Coulter XL, the number of coincidence events is reducedto a level that does not inuence the nal derived plateletcount signicantly. Ault (1996) has previously recommended adilution factor of at least 1:1800, whereas Davis & Bigelow(1999) have recently demonstrated that blood should ideally

be diluted by at least 1:400 with an acquisition rate of< 4000 cells/s. The precision of the RBC ratio was superior tothat of the bead ratio derived from either a lyophilizedpreparation (Trucount) or solution of beads (FlowCount). Asthe RBC ratio is independent of potential pipetting anddilution artifacts that may inuence the bead ratio, this wasnot totally unexpected. Even within thrombocytopenicsamples, providing at least 1000 platelet events are collected,the level of precision was excellent (< 5%). Upon comparisonof the RBC ratio with the current reference procedure(manual count), both methods correlated well (Fig 3),although the manual procedure exhibited a mean CV of16% even when two experienced laboratory technologistscounted the samples in duplicate. Comparison of theproposed RBC ratio method with three different existingcommercial impedance cell counters and one optical counterdemonstrated that this method has potential as a newcalibration reference method (Fig 4A). However, comparedwith impedance counts, this method will not be interferedwith by the presence of cellular debris, which can often resultin the impedance counter overestimating the count,especially in severely thrombocytopenic samples (< 20 109/l; see Fig 5 and Table III). Conversely, in patients withmacrothrombocytopenia and ITP, most impedance counterscannot discriminate large platelets, often resulting in anunderestimate of the platelet count (see Fig 5 and Table III).Interestingly, the ADVIA 120 optical count, which not onlyeliminates noise but also counts large platelets, appears tocompare favourably with the immunocount, especially atlow counts of < 100 109/l (Fig 4B). This is in agreementwith previous studies using the CELL-DYN 4000 (Abbott),which demonstrated a superior correlation betweenimmunocounts and optical counts (Ault, 1996; Ault et al,1997).

Although the immunoplatelet count derived from the RBCratio has clear advantages over the existing referencemethod, there are a number of potential, albeit rare,problems that must be recognized. Samples containingsignicant numbers of platelet aggregates, platelet whitecell complexes, microparticles and very large platelets mayresult in gating problems. Upper and lower platelet gateswithin the log FS vs. FL1 plot can be placed to discriminatesingle platelets from aggregates, cell complexes or micro-particles (Matzdorff et al, 1998). Also, depending upon thetarget antigen of the antibody used to measure theimmunoplatelet count (e.g. CD41, CD42b or CD61), wherethere is absence of certain glycoproteins (e.g. Bernard Souliersyndrome and Glanzmann's thrombasthenia), then nouorescent platelets will be detectable. Furthermore, it isalso possible that, when patients have signicant levels ofplatelet autoantibodies or are being treated with antiplateletglycoprotein therapy (e.g. anti-GpIIb/IIIa antibodies such asReopro), these could interfere with the assay. Although theseproblems are rare, possible strategies include using a novelantibody that targets a unique epitope with no autoantibodyspecicity or using a panel of antibodies to a number ofdifferent antigens on the platelet surface.

In summary, the immunoplatelet counting proceduredescribed is a rapid, simple, reproducible and cheap method,

q 2000 Blackwell Science Ltd, British Journal of Haematology 108: 228235

232 P. Harrison et al

Fig 3. Comparison of the immunoplatelet count (derived from the

RBC ratio) with the existing manual phase reference method. The

correlations are shown for all samples (A) and for below 100 109/l(B).

233Immunoplatelet Counting

q 2000 Blackwell Science Ltd, British Journal of Haematology 108: 228235

which could be adopted by any laboratory with suitable owcytometric experience. It is suggested that this methodologycould replace the existing manual reference method andprovide important information concerning the accuracy ofplatelet counting by conventional impedance and opticalcounters in severe thrombocytopenia. Providing accuracycan be guaranteed at around the current platelet transfusionthreshold of 10 109/l, then the most appropriate clinicaldecision is more likely with regard to prophylactic plateletsupport (Kickler et al, 1998). Furthermore, improvedaccuracy of platelet counts at this level may demonstratethat the transfusion threshold could be decreased to as lowas 5 109/l without any increase in the risk of spontaneousbleeding. A further reduction in this threshold would resultin a dramatic reduction in costs by decreasing the frequencyof unnecessary platelet transfusions.

ACKNOWLEDGMENTS

The authors are grateful to Beckman Coulter, Miami, FL,USA, for providing FlowCount beads. We are also indebted toNigel Llewellyn-Smith (Becton Dickinson, Oxford, UK) andDavid Warunek (Becton Dickinson, Franklin Lakes, NJ, USA)for providing Trucount tubes and the anti-CD-61 (FITC)antibody. We would also like to thank Sue Mead (BayerDiagnostics, Newbury, UK), Andy Hay (Sysmex, Milton Keynes,UK) and Sandy Piepho (Beckman Coulter, Miami, FL, USA) forthe provision of their respective automated cell counters.

REFERENCES

Ancliff, P.J. & Machin, S.J. (1998) Trigger factors for prophylacticplatelet transfusion. Blood Reviews, 12, 234238.

Ault, K.A. (1996) Platelet counting: is there room for improvement?

Laboratory Haematology, 2, 265269.Ault, K.A., Mitchell, J., Knowles, C. & Van Hove, L. (1997)

Implementation of the immunological platelet count on a

haematology analyser, the Abbot CELL-DYN 4000. Laboratory

Haematology, 3, 125128.Brecher, G., Schneiderman, M. & Cronkite, E.P. (1953) The

reproducibility of the platelet count. American Journal of Clinical

Pathology, 23, 1521.

Davis, B. & Bigelow, N.C. (1999) Indirect immunoplatelet countingby ow cytometry as a reference method for platelet count

calibration. Laboratory Haematology, 5, 1521.

Dickerhoff, R. & Von Ruecker, A. (1995) Enumeration of platelets by

multiparameter ow cytometry using platelet specic antibodiesand uorescent platelet particles. Clinical Laboratory Haematology,

17, 163172.

England, J.M., Rowan, R.M., Bins, M., Bull, B.S., Coulter, W.H.,Groner, W., Jones, A.R., Koepke, J.A., Lewis, S.M., Shinton, N.K.,

Thom, R., Van Assendelft, O.W. & Verwilghen, R.L. (1988)

Recommended methods for the visual determination of white

cell and platelet counts. WHO LAB, 88, 1.Gmur, J., Burger, J., Schanz, U., Fehr, J. & Schnaffner, A. (1991)

Safety of stringent prophylactic platelet transfusion policy for

patients with acute leukemia. Lancet, 338, 12231226.

q 2000 Blackwell Science Ltd, British Journal of Haematology 108: 228235

234 P. Harrison et al

Fig 4. Comparison of the immunoplatelet count (derived from the RBC ratio) with three different impedance counters (impedance 1, 2 and 3) and

the ADVIA 120. (A) The correlations over for the whole range of platelet counts. (B) The correlations below 100 109/l.

Fig 5. Illustration of the effect of inaccurate impedance counts on

platelet transfusion decision-making. Aberrant samples were com-

pared by immunoplatelet counting (derived from the RBC ratio) anda conventional impedance counter. The graph illustrates two

different transfusion thresholds set at either 20 10 9/l (dottedlines) or 10 10 9/l (solid lines). Impedance counts that were eitherunderestimated or overestimated are highlighted by arrows.

Table III. Examples of impedance counting errors when comparedwith the immunocount derived from the RBC ratio.

Impedance count RBC ratio countDiagnosis (109/l) (109/l)

Lipaemia 103 43

ITP 30 68

ITP 68 104ITP 41 94

ITP 46 73

ITP 61 101

ALL 15 29ALL 22 14

ALL 75 36

AML 22 14AML 13 6

AML 75 36

AML 6 1

BS 21 33

BS 42 61

BS 6 16

NHL 33 16

ITP, idiopathic thrombocytopenic purpura; ALL, acute lympho-

blastic leukaemia; AML, acute myeloid leukaemia; BS, BernardSoulier disease; NHL, Non-Hodgkin's lymphoma.

235Immunoplatelet Counting

q 2000 Blackwell Science Ltd, British Journal of Haematology 108: 228235

Hammerstrom, H. (1992) Spurious platelet counts in acute

leukemia with DIC due to cell fragmentation. Clinical Laboratory

Haematology, 14, 239243.

Kickler, T.S., Rothe, M., Blosser, L., Schisano, T. & Van Hove, L.(1998) Improving platelet transfusion therapy using the Immu-

noPLT method on the CELL-DYN 4000. Laboratory Haematology, 4,

8087.

Matzdorff, A.C., Kuhnel, G., Kemkes-Matthes, B. & Pralle, H. (1998)Quantitative assessment of platelets, platelet microparticles and

platelet aggregates. Journal of Laboratory and Clinical Medicine, 131,

507517.Murphy, W.G. (1992) Prophylactic platelet transfusion in acute

leukemia. Lancet, 339, 120.

Norfolk, D.R., Ancliffe, P.J., Contreras, M., Hunt, B.J., Machin, S.J.,

Murphy, W.G. & Williamson, L.M. (1998) Consensus conferenceon Platelet Transfusion, Royal College of Physicians of Edinburgh,

2728 November 1997. Synopsis of background papers. British

Journal of Haematology, 101, 609617.

Rebulla, P., Finazzi, G., Marangoni, F., Avvisati, G., Gugliotta, L.,

Barbui, T., Mandelli, F. & Sirchia, G. (1997) The threshold for

prophylactic platelet transfusion in adults with acute myeloid

leukemia. New England Journal of Medicine, 337, 18701875.

Rowan, R.M. (1991) Platelet counting and assessment of plateletfunction. In: Practical Laboratory Haematology (ed. by J.A. Koepke),

p. 157. Churchill Livingstone, New York.

Springer, W., von Ruecker, A. & Dickerhoff, R. (1998) Difculties in

determining prophylactic transfusion thresholds of platelets inleukemia patients. Blood, 92, 21832184.

Stanworth, S.J., Denton, K., Monteath, J. & Patton, W.N. (1999)

Automated counting of platelets on the ADVIA 120 analyser.Clinical Laboratory Haematology, 21, 113117.

Tanaka, C., Isii, T. & Fujimoto, D. (1996) Flow cytometric platelet

enumeration utilising monoclonal antibody Cd42a. Clinical

Laboratory Haematology, 118, 265269.Wong, D.A. & Springer, T.A. (1995) CD61 (b3) cluster report. In:

Leukocyte Typing V, White Cell Differentiation Antigens (ed. by S.F.

Schlossman, L. Boumsell, W. Gilks et al), pp. 16641665. Oxford

University Press, Oxford.