SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES
  8. Appendix

BACKGROUND: Hematology analyzers are designed to count whole blood samples, but are also used by blood centers to perform quality control on blood components. In platelet (PLT) concentrates, the number of PLTs is approximately fivefold higher and red blood cells are absent, causing variable PLT counting results. It was our aim to compare currently used hematology analyzers for counting PLTs in PLT concentrates using fixed human PLTs.

STUDY DESIGN AND METHODS: PLT samples were fixed, diluted into seven concentration levels (plus one blank), aliquoted, and shipped to 68 centers. Evaluable data were obtained for 89 hematology analyzers. All samples were counted six times, and results were reported to the coordinating center. The overall group mean was calculated, and the percentage deviation from this mean was calculated for each analyzer.

RESULTS: At PLT levels relevant for blood centers, 750 × 109 to 2000 × 109 per L, analyzers gave results that were between 35 percent lower and 16 percent higher than the overall group mean. Within a group of analyzers, results were comparable with coefficient of variations usually below 10 percent, indicating that the observed differences were caused by instrument characteristics. A smaller study with fresh, unfixed PLT samples showed that analyzers behaved similarly for fixed and fresh PLTs.

CONCLUSION: With a wide array of currently used hematology analyzers, a marked difference was determined for the PLT counts of fixed human-based identical samples provided to 68 laboratories by a centralized facility. A gold standard method is needed to allow for more valid interlaboratory comparisons between hematology analyzers.

ABBREVIATION:
PPP

platelet-poor plasma

Hematology analyzers are designed to count patient blood samples. The methods to measure blood cells, as well as the calculations that are made by the analyzer, are aimed at the analysis of whole blood. These analyzers have analytical ranges sufficient to count cells in the majority of normal and patient whole blood samples.

Blood centers also use these hematology analyzers to perform quality control tests on the blood components that they make. However, the composition of the components is unlike that of whole blood. For example, platelet (PLT) concentrates have a fivefold higher PLT concentration compared to whole blood, whereas red blood cells (RBCs) are (almost) absent.

A previous study of the BEST Collaborative has shown that, when counting PLT concentrates, large differences in outcomes were seen when different hematology analyzers were used.1 In that study, comparisons between analyzers were done using either fixed porcine PLTs or goat RBCs to simulate human PLTs. However, the source of the materials affected the PLT counting outcomes. Samples with identical PLT concentrations could yield variable results on the same analyzer, depending on the source material. This made it difficult to find a “true” value. In that study, the discrepancy between two brands of hematology analyzers reduced from 230 × 109 to 110 × 109 per L at the level of 1500 × 109 per L when a common calibrant of porcine PLTs was used. At that level, the coefficient of variation (CV) ranged from 2 to 12 percent without common calibration but 2 to 7 percent with common calibration.1 Similar discrepancies between analyzers have been seen in other studies.2-4

The earlier BEST study was done with fixed nonhuman PLTs in an international study and with unfixed human PLTs in a North American study.1 With the availability of a fixation protocol for human PLTs developed by the United Kingdom National External Quality Assessment Scheme for General Haematology (UK NEQAS(H)),5 a multicenter comparative study was designed, using fixed human PLTs to study the effect of the type of hematology analyzer used on the observed PLT count. The aim of the current study was to address the extent of variation in PLT counts, in a wide range of currently used hematology analyzers with human PLTs. In the absence of a “gold standard” PLT counting method, accuracy was calculated by determining the deviation from the overall group mean,6 rather than using a common calibrant.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES
  8. Appendix

Sample preparation, shipping, and counting

PLT concentrates were derived from four pooled buffy coats and 1 unit of plasma and provided by the National Blood Service (Birmingham, UK). The PLTs were fixed using a proprietary fixation protocol by UK NEQAS(H) (Watford, UK)5 based on glutaraldehyde (final concentration, 0.75 µL/mL PLT concentrate) and formaldehyde (final concentration, 0.54 µL/mL). The PLT samples were either centrifuged to obtain a higher concentration or diluted with (unfixed) PLT-poor plasma (PPP) for a lower concentration, to intended levels of 150 × 109, 250 × 109, 500 × 109, 750 × 109, 1000 × 109, 1500 × 109, and 2000 × 109 PLTs per L (designated Samples A through G, see Table 1). Samples A through E thus were resuspended in unfixed PPP, whereas Samples F and G were suspended in fixative-containing supernatant. A sample of PPP, prepared by hard-spin centrifugation, was included to check background and was assumed to contain 0 × 109 PLTs per L. All samples were aliquoted in portions of 2 mL and packed in insulating packaging. Shipping to non-UK countries took place by DHL and to UK centers by Royal Mail. Each center received one set of samples for each analyzer to be tested. The samples were coded so the centers were unaware of the PLT counts to be expected.

Table 1. Intended and actual PLT concentrations in 109 PLTs per mL in fixed PLT samples, used for comparison of hematology analyzers*
SampleIntended valueActual value
  • * 

    Actual value shown as mean ± SD (n = 89 analyzers).

A15081 ± 22
B250152 ± 37
C500337 ± 76
D700561 ± 113
E1000805 ± 133
F15001522 ± 136
G20002000 ± 249

A protocol, describing in detail the procedure to be followed upon arrival of the samples, was sent to the centers a few days before the intended study date. This protocol stipulated that the centers were to hold the samples at room temperature until the designated day of analysis. If the samples arrived later, the samples were to be analyzed upon arrival; a window of analysis of 2 days before and after the intended analysis date was indicated in the protocol. Each sample had to be counted six times per the center's standard procedures. In addition, to investigate the effect of dilution on the PLT counts, the samples containing 1000 × 109, 1500 × 109, and 2000 × 109 PLTs per L (i.e., Samples E, F, and G) had to be diluted 1:5 with Isoton (Beckman Coulter, Miami, FL) or a similar solution, and the diluted samples were counted six times. The use of PPP for dilution of the samples was specifically excluded.4

An Excel file was sent to each of the centers for entering the raw data, the date of analysis, and the type of analyzer used. The files were returned by email to the coordinating center for analysis.

Stability testing

To determine the effect of shipping on the PLT preparations and to assess the stability of the PLT concentrations, additional samples were counted by the center that performed the aliquoting and shipment. A hematology analyzer (Sysmex K4500, TOA, Tokyo, Japan) was used for PLT counting. On the day of preparation (Day −6; Day 0 was the day of intended analysis) all Samples A through G were counted. One set of samples was sent to the center on Day −6 and received on Day −5, whereas another set of samples was kept in the laboratory. On Day −5, PLT counts were done on these self-shipped and laboratory-stored samples. Counting was repeated on the day of intended analysis (Day 0), on the final day of the window of analysis (D +2), and on D +6.

In addition, two centers that received two sample sets but had only one analyzer in the study performed PLT counts on the intended day of analysis using one sample set six times and used the extra set for additional measurements on Day 2 and Day 3 in triplicate. These two centers both used the same analyzers (CellDyn 1700, Abbott Diagnostics, Santa Clara, CA).

Data analysis

All data were combined in a spreadsheet for analysis. Data points of measurements outside the intended study period (which was between Day −2 to Day +2) were excluded. If the analyzer had more than one mode of operation, the impedance mode was used for the primary analysis. The overall group mean of a particular PLT level was calculated by averaging replicate measurements of all analyzers included in the primary data set. Then, for each of the individual data points, the percentage deviation from the group mean was calculated using the formula

  • image

These percentage deviations from the group mean were used to calculate a mean and standard deviation (SD).

The CV of an analyzer was calculated from the mean and SD of the sixfold measurements. CVs of an analyzer group were obtained by calculating mean and SD from all sixfold measurements of a particular PLT level.

The dilution effect was calculated by dividing the result of the diluted sample with the mean value of the undiluted measurement and multiplied by 100 to give the percentage correspondence from the undiluted results. This calculation was done for each individual machine and then averaged per machine group.

Secondary analysis consisted of comparing impedance versus optical measuring methods; the percentage of correspondence was calculated as given for determination of the dilution effect.

Comparative study using freshly prepared PLTs

To determine if the fixation procedure had influence on the outcome of the international comparative study with fixed PLTs as described above, a smaller, national study was performed with fresh unfixed PLTs. Freshly prepared buffy coat–derived PLT concentrates in plasma were concentrated by centrifugation or diluted to contain similar levels as used in the international study and were aimed to be 0 × 109, 81 × 109, 153 × 109, 336 × 109, 556 × 109, 798 × 109, 1498 × 109, and 1998 × 109 per L. On the same day, the samples were aliquoted and transported by car to the five sites in the Netherlands. Upon arrival, the samples were immediately counted in triplicate. The results were reported back to the coordinating blood center for further analysis. The percentage correspondence was calculated as given for determination of the dilution effect.

Statistical analysis

Statistical functions of a spreadsheet program (Excel, Microsoft Corp., Redmond, WA) were used. A p value of less than 0.05 was used to indicate a significant difference.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES
  8. Appendix

In total, 68 centers (see Appendix, available as supporting information in the online version of this article) participated in this comparative study, encompassing 94 hematology analyzers, two flow cytometric methods, and one manual method. The results of 5 analyzers had to be excluded because the samples had arrived late and consequently were counted outside the intended window of analysis. One center never received its samples. A summary of the analyzers used in the study is given in Table 2.

Table 2. Hematology analyzers used in the study, comparing the counting of PLTs in PLT concentrates
ManufacturerNumberInstrument groupType
Abbott Diagnostics (Santa Clara, CA)1CellDyn 16CellDyn 16
15CellDyn 1700CellDyn 1700
2CellDyn 3200CellDyn 3200
6CellDyn 3700CellDyn 3700
3CellDyn 4000CellDyn 4000
Bayer (Leverkusen, Germany)11Advia 120/2120Advia 120
  Advia 2120
Beckman Coulter (Miami, FL)7Coulter ActAcT8
  AcT10
  ActDiff
7Coulter LH750LH750
2Coulter MD16MD16
2Coulter OnyxOnyx
Horiba ABX (Montpellier, France)3Cobas MicrosMicros 60
1Pentra 60Pentra 60
2Pentra 120Pentra 120
Mallinckrodt Baker (Phillipsburg, NJ)3Baker 9110/91209110+
  9120
Medonic (Stockholm, Sweden)2Medonic 620Medonic 620
Sysmex (Kobe, Japan)2Sysmex K1000K1000
7Sysmex K4500K4500
4Sysmex KX21KX21
1Sysmex SE9000SE9000
3Sysmex XE2100XE2100
5Sysmex XT1800/2000XT1800
  XT2000
Total89  

In the absence of a gold standard PLT counting method, the results of the 89 analyzers were averaged to give an overall group mean (Table 1). This calculated overall group mean was used as reference value to assess the variability in PLT counts on the span in PLT counts with the various hematology analyzers. Figure 1 shows the results of specific groups of analyzers relative to that of the overall group mean. There was considerable disagreement between the analyzers, ranging from an underestimation of 40 percent to an overestimation of 47 percent in the low PLT ranges (<150 × 109/L) in comparison with the overall mean. With higher PLT concentrations the divergence from the overall group mean became smaller. However, even in the range of PLT concentrations found in PLT concentrates, from approximately 750 × 109 to 2000 × 109 PLTs per L, some analyzers failed to give results that were within 10 percent of the mean group outcome. Nonlinearity at very high PLT concentrations is visible as an underestimation relative to the overall group mean. Other analyzers do not report the counting results when the concentration is above a preset limit.

image

Figure 1. PLT counts of various hematology analyzers. The deviation from the overall group mean was calculated and plotted. (A) ◆, Advia 120/2120; ▪, Baker 9110/9120; ▵, Cobas Micros; □, Medonic 620; ○, Pentra 120; ●, Pentra 60. (B) ◆, CellDyn 16; ▪, CellDyn 1700; ▵, CellDyn 3200; □, CellDyn 3700; ○, CellDyn 4000. (C) ◆, Coulter AcT; ▪, Coulter LH750; ▵, Coulter MD16; □, Coulter Onyx. (D) ◆, Sysmex K1000; ▪, Sysmex K4500; ▵, Sysmex KX21; □, Sysmex SE9000; ○, Sysmex XE2100; ●, Sysmex XT1800/2000.

Download figure to PowerPoint

The CellDyn 16 and CellDyn 1700 analyzers gave results that were approximately 20 and 35 percent lower than the overall mean values over the whole range of PLT concentrations. To investigate the possibility that shipping had had an effect on the outcomes, an analysis was made for one center that had done measurements both with the CellDyn 1700 and with an Advia 120, and these were compared with the overall results as mentioned above. In this center, for example in Sample E, the PLT count was 609 × 109 ± 12 × 109 per L on the CellDyn versus 901 × 109 ± 14 × 109 per L on the Advia. These results were not significantly different from results obtained from these analyzers in other centers: 579 × 109 ± 90 × 109 per L for CellDyn (n = 14 analyzers) and 892 × 109 ± 28 × 109 per L for Advia (n = 10; p > 0.05 by unpaired t test). For the other PLT levels similar results were obtained (not shown), indicating that shipping itself was probably not the cause of the lower results found with the CellDyn analyzers.

Most analyzers had reasonably low CVs to count PLTs, as indicated in Fig. 2. At PLT levels greater than 337 × 109 PLTs per L, intercenter CVs lower than 10 percent and even 5 percent were often observed. One outlying group was the CellDyn 4000, where one analyzer in one center gave an overestimation between 26 and 50 percent (depending on the level tested) from the overall group mean, whereas two analyzers at another center gave results that were between 34 percent lower and 9 percent higher than the group mean. However, the CV of each individual analyzer was 5 percent or less at any of the levels tested, emphasizing the importance of standardization between centers.

image

Figure 2. CVs of hematology analyzers used for counting PLTs in PLT concentrates. The CV in percentage per group of analyzers is shown. (A) bsl00039, Advia 120/2120; bsl00008, Baker 9110/9120; □, Cobas Micros; bsl00021, Medonic 620; ▪, Pentra 120; bsl00052, Pentra 60. (B) bsl00039, CellDyn 16; bsl00008, CellDyn 1700; □, CellDyn 3200; bsl00021, CellDyn 3700; ▪, CellDyn 4000. (C) bsl00039, Coulter AcT; bsl00008, Coulter LH750; □, Coulter MD16; bsl00021, Coulter Onyx. (D) bsl00039, Sysmex K1000; bsl00008, Sysmex K4500; □, Sysmex KX21; bsl00021, Sysmex SE9000; ▪, Sysmex XE2100; bsl00052, Sysmex XT1800/2000.

Download figure to PowerPoint

The flow cytometric methods gave no coherent results; one method gave high background levels and thus an overestimation of the PLT counts, whereas the other method gave an overestimation at high levels and an underestimation at low levels. Further studies have been initiated to optimize a flow cytometric PLT counting method. With the manual (microscopic) method the samples were counted only once; thus, these results were excluded from the analysis.

Sample stability

Sample stability was determined by the coordinating center; both stability during shipping and stability over time were investigated. There was no effect of shipping. The self-shipped samples showed a PLT count that was on average 4 percent lower compared to the samples that were kept in house, but this difference was not significant (paired t test). The stability over time is shown in Fig. 3, where the PLT values of the intended day of analysis are set at 100 percent. Especially in the low PLT range, a considerable drop in PLTs was observed between the day of sample preparation (Day −6) and the intended day of analysis (Day 0). Until Day 2, the maximum allowable analysis date, the decrease in PLT count was on average 18 ± 2 percent for values up to 805 × 109 PLTs per L; the higher samples with 1522 × 109 and 2000 × 109 per L showed a smaller decrease of 7 and 2 percent, respectively.

image

Figure 3. Sample stability of fixed human PLTs, used for a comparative PLT counting study. Results are shown as PLT counts on (bsl00039) Day −6, (bsl00052) Day −5, and (▪) Day +2 and (bsl00021) Day +6, as percentages of the counting results on (□) Day 0.

Download figure to PowerPoint

The two centers that performed PLT counts on Days 0, +1, and +2 with CellDyn 1700 analyzers showed a mean 16 ± 2 and 26 ± 2 percent lower PLT value on Days +1 and +2, respectively, when compared to the Day 0 value (irrespective of the PLT level).

One consideration is that these results may have been confounded with the specific analyzers, and so also results of the comparative study were analyzed per day of analysis. Compared with the values found on Day 0 (n = 68 analyzers), Day 1 gave values that were on average 3 percent higher (n = 17, not significant). For individual Levels A through G on Day +1, the differences relative to the Day 0 results were +8, +4, −1, −2, −2, +5 (p < 0.001), and +7 percent (p < 0.001), respectively. On Day +2 (n = 7 analyzers), the mean PLT results were 1 percent higher than the Day 0 value (not significant); for individual levels the differences were +11 percent (p < 0.05) for Level A and +3, +1, −2, −4, −1, and −1 percent for levels B through G (not significant).

Dilution

The three highest PLT samples were diluted 1 to 5 in Isoton or a similar diluent and the results were compared with values obtained from undiluted samples. The diluted samples showed slightly lower results, on average 98, 95, and 98 percent for Levels E, F, and G, respectively, of the values obtained when counting undiluted samples (Table 3).

Table 3. Effect of dilution on PLT counting results, shown as percentages of the diluted result relative to the undiluted for each group of analyzers
Instrument groupNumberSample
EFG
  • * 

    d.l. = PLT concentration exceeded upper detection limit for undiluted sample.

Act79194108
Advia 120/2120119899101
Baker 9110/9102310997101
CellDyn 40003999799
CellDyn 16170d.l.*d.l.
CellDyn 170014110d.l.d.l.
CellDyn 32002103100d.l.
CellDyn 3700610198102
Cobas Micros3103104102
Coulter MD16299d.l.d.l.
Coulter Onyx28787104
Medonic 620210510199
Sysmex K10002868490
Sysmex K45007949696
Sysmex KX2149392d.l.
Sysmex LH7505838988
Sysmex XE210031049999
Sysmex XT1800/20005949390
Total82989598

Analysis mode

For both the XE2100 analyzer and the CellDyn 4000 analyzer, results were obtained using the impedance and optical mode to enumerate PLTs. For the XE2100 (n = 2), with increasing PLT concentrations, the discrepancy between the impedance and optical mode became smaller. For Samples A through G, the optical values were 37, 26, 19, 11, 5, −1, and 1 percent higher compared to the impedance values (paired t test, p < 0.001 except Samples F and G, not significant). The CellDyn 4000 showed an opposite effect, with optical results 8, 3, 7, 15, 19, 32, and 33 percent lower compared with the impedance mode (all p < 0.001). A comparison with the overall group mean showed similar results.

Comparative analysis of “fresh” PLTs

To investigate whether the fixation procedure and subsequent shipment had had an effect on the results, a national comparative study was performed with freshly prepared unfixed PLTs at approximately the same levels as in the international study. Results of analyzers that had participated in both studies were compared (n = 11 analyzers). The PLT concentrations were on average 4 percent lower in the fresh samples versus the fixed samples (unpaired t test, not significant). As shown in Fig. 4, the majority of the analyzers behaved similarly whether fixed or fresh PLTs were used for comparison of analyzers, with a deviation often within 10 percent from the overall group mean, both when fixed and when fresh PLTs were used. The fixation procedure thus was not likely to have an effect on the outcome of the comparative studies.

image

Figure 4. Comparison of fixed or freshly prepared PLT samples. A subset of 11 analyzers were compared and the deviation from the overall group mean is shown.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES
  8. Appendix

This multicenter study again demonstrated that there are large differences between hematology analyzers when counting PLTs in PLT concentrates. Using fixed human PLTs, results could be up to 50 percent off the overall group mean when samples with low PLT concentrations were counted. At PLT concentrations relevant for blood centers (>750 × 109 PLTs/L), the various analyzers gave outcomes with an underestimation of 35 percent up to an overestimation of 16 percent. A combination of underlying causes can explain the results observed in this comparative study, and these are mostly related to the different composition of PLT concentrates relative to whole blood and the different ways that the software packages process the data output.

An important factor is the absence of RBCs in PLT concentrates. Hematology analyzers use the bell-shaped curve of the PLT and RBC size distribution to discriminate the cell populations. A threshold—the “discriminator”—is set to indicate which events should be counted as PLTs and which as RBCs. Some analyzers use a preset, fixed threshold; others use a “floating” discriminator, whose value depends on the specific size distribution of the PLTs and RBCs of a patient. This “floating” discriminator should result in accurate outcomes when whole blood is counted; however, when RBCs are absent, the analyzer uses default “fixed” discriminators, potentially giving inaccurate results.2 In addition, the absence of RBCs may result in an overestimation of the PLT count. Normally, when counting whole blood, a correction factor is needed to compensate for PLT-RBC coincidence. In PLT concentrates, this erroneously leads to an overcorrection of the actual number of PLTs.7 Other additional correction factors or software algorithms of the analyzer, intended to give accurate outcomes when counting whole blood, may in fact introduce inaccuracy when counting PLT concentrates.

An additional source of variance was the very high concentration of PLTs. Some analyzer groups showed nonlinearity at high PLT concentrations, usually at or above 1500 × 109 per L. This is caused by PLT-PLT coincidence, where multiple PLTs are counted as one single impedance event, resulting in an underestimation of the actual number of PLTs. Indeed, a number of analyzers do not report PLT numbers above a certain value.

Consequently, when comparing diluted versus undiluted samples, it is expected that (when multiplied with the dilution factor) the diluted samples would yield higher values than the undiluted ones. However, though not for all analyzers, the dilution experiment in this study showed lower values for the diluted samples relative to the undiluted ones for a number of analyzers. We have no clear explanation for this finding, but it may be related to the dilution in Isoton,4 which may lead to PLT swelling. For three analyzers, results were provided on mean PLT volume (MPV), which increased from 11.0 ± 0.4 fL in the undiluted samples to 13.2 ± 0.4 fL in the diluted ones (p < 0.0001, paired t test). When analyzers use a fixed discriminator (i.e., all events with an MPV greater than a preset value), this increase in MPV potentially results in rejection of a swollen PLT as a PLT event. Thus, we hypothesize that PLT swelling caused by suspension in Isoton explains the underestimation found in diluted samples. More study is needed, but considering the good linearity of most current analyzers, dilution is in general not needed. Despite the absolute deviation from the overall group mean, most individual analyzers gave precise results, with CVs below an arbitrary limit of less than 5 percent, especially at higher PLTs concentrations.

Two analyzers in this study used an alternate mode of measurement in addition to impedance measurement. When used in the optical mode, the CellDyn 4000 gave lower results with increasing PLT concentration when compared with the impedance mode. This is probably caused by the effect of different anticoagulants on PLT properties. Patient samples are almost always in ethylenediaminetetraacetate, whereas in blood collections alternative solutions are used, such as citrate-phosphate-dextrose or acid-citrate-dextrose. PLTs anticoagulated in these solutions have different properties and exhibit a broader PLT scattering pattern.7 Owing to default cutoff limits in the analyzer,7 part of these PLT populations fall outside the limits and are rejected, resulting in lower PLT counting results.

Hervig and colleagues2 extended these investigations to explain in depth the differences between impedance and optical modes of hematology analyzers. In that study, a CD61-based counting assay was used as immunoplatelet reference method. That study showed an overestimation of approximately 20 percent for the CD4000 impedance measurement, while optical measurement was approximately 20 percent lower compared to the immunoplatelet reference. The XE2100 gave a mean 5 percent underestimation for the impedance mode versus an overestimation of about 5 percent for the optical mode. The current study shows the same trend. A likely explanation for the higher PLT count in the impedance mode is the inclusion of small, CD61-negative cell debris in the impedance count that are not included in the immunoplatelet count.2 For the underestimation found for the optical mode, there is no definite answer, although this again appears to be related to cell debris.2,8 When counting patient samples, no differences are observed between the two modes of measurement.3,8

As indicated previously, hematology analyzers use the presence of RBCs to accurately count PLTs. The presence of RBCs is also an integral part of the currently defined gold standard PLT counting method9 and thus could not be used in the current study. Also, though manual counting has been considered a gold standard, the accuracy and precision may show poor results.10 Therefore, in the absence of a functional gold standard PLT counting method, the results of all 89 analyzers were averaged to give an overall group outcome. Although by doing so the overall group outcome is affected by the analyzers themselves, the sample size is large enough to allow this approach.

In the previous BEST study, goat RBCs or porcine PLTs were used to compare analyzers.1 One of the complicating factors was that the outcome of the PLT counting depended on the source of the PLTs: at similar PLT concentrations the analyzer results could differ depending on the source material. Therefore, in the current study, fixed human PLTs were used to overcome this source of variation. A comparison with freshly isolated, unfixed PLTs in a subset of the analyzers indicated that the fixation itself had only a small effect on the PLT counting results. When considering PLT stability, the results of both the coordinating center and the two study centers showed a decrease in PLT concentration by approximately 10 percent per day of storage for Samples A through E; Samples F and G showed stable PLT concentrations for the duration of the study. A likely cause for the decreasing PLT concentrations in part of the samples was the method of dilution of the samples during preparation. Samples F and G were prepared with the supernatant that was obtained during the fixation procedure, whereas Samples A through E were made with PPP from an unrelated donation and did not contain fixatives. In any case, when comparing the overall group means from day to day, the mean values differed only by a few percent. When interpreting the outcomes of this study it should be remembered that PLT stability in this study was not absolute, but the impact on the interpretation of the results may be moderate.

In this study, the CellDyn 16 and CellDyn 1700 analyzers showed significantly lower PLT concentrations when compared to the overall group mean. The data show that sample deterioration due to shipping was not likely to be the cause of the observations. Also, the use of fixed PLTs rather than fresh ones was not a likely explanation for the result, as shown previously. The CellDyn 1700 has shown to give excellent comparability with other analyzers when PLTs were counted in blood samples.11 So a specific combination of factors of this instrument apparently leads to an underestimation of the number of PLTs when PLT concentrates are counted. These results emphasize the importance of performing additional validations when counting samples other than whole blood. Accuracy of a method can be determined by comparing the results of a particular analyzer with those obtained with a gold standard counting method. By definition, this status is currently assigned to a method based on staining PLTs with fluorescent conjugated CD61 antibodies and subsequent detection usually with a flow cytometer.10,12,13 Because the sample is continuously diluted with sheath fluid, the exact sample volume that is counted is not known, and therefore RBCs are counted simultaneously. By counting RBCs on a hematology analyzer and dividing the flow cytometer count by the hematology analyzer count, the actual PLT concentration can be calculated. However, because PLT concentrates contain negligible number of RBCs, counting beads must be added. The development and application of such a method is challenging, as the results of two users in this study demonstrated. The BEST Collaborative is developing a bead-based, flow cytometric PLT counting method for this purpose. In addition, guidelines are being developed for validation of hematology analyzers that are used to count PLT concentrates.

In summary, this international collaborative study on PLT counting shows large variation between hematology analyzers. When comparing results from different blood centers, the effect of the analyzer on the results must be taken into account. Adequate validation of an analyzer before use, and additional adjustments to come to comparable counting results (either by comparison with a gold standard method or a gold standard PLT sample) should be undertaken to improve accuracy and allow fair comparison of results between different centers.

ACKNOWLEDGMENTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES
  8. Appendix

We thank the staff of UK NEQAS(H) for performing the fixation procedure at no cost and for arranging shipment of the samples. Neil Beckman (National Blood Service, UK) is acknowledged for supplying PLT concentrates at no cost. We express our thanks to all 68 laboratories for performing the analyses so well and for sharing their enthusiasm.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES
  8. Appendix

Appendix

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES
  8. Appendix

List of centers that participated in the BEST comparative PLT counting study.

Australia: Red Cross Blood Service, Sydney.

Brazil: Hospital Sirio Libanes, Sao Paulo.

Canada: Hema-Quebec, R&D, Sainte-Foy, QC; Canadian Blood Services (CBS) British Columbia/Yukon, Vancouver, BC; CBS Calgary, Calgary, AB; CBS Edmonton, Edmonton, AB; CBS Regina/Saskatoon, Regina SK; CBS Winnipeg/Thunder Bay, Winnipeg, MB; CBS London, London, ON; CBS Hamilton, Hamilton, ON; CBS Toronto, Toronto, ON; CBS Ottawa, Ottawa, ON; CBS Sudbury, Sudbury, ON; CBS Halifax/Prince Edward Island, Halifax, NS; CBS New Brunswick, St. John, NB; CBS Newfoundland/Labrador, St John's, NL; CBS Centre for Blood Research, Vancouver, BC; McMaster University, Hamilton, ON.

Finland: Finnish Red Cross Blood Service, Helsinki.

France: Etablissement Francais du Sang (EFS) Bourgogne Franche Comte, Besancon; ESF Alsace, Strasbourg.

Germany: Institut fur Immunologie und Transfusionsmedizin, Thrombocytenlabor, Greifswald; DRK-Blutspendedienst Baden-Wurttemberg-Hessen, Institut Baden-Baden, Baden-Baden.

Italy: Ospedale Maggiore Policlinico, Centro Transfusionale e di Immunoematologia, Milano.

Netherlands: Sanquin Research, Department of Blood Cell Research, Amsterdam; Sanquin Blood Bank North East Region, QC and R&D Laboratory, Groningen; Sanquin Blood Bank North West Region, QC and R&D Laboratory, Amsterdam; Sanquin Blood Bank South East Region, QC Laboratory, Nijmegen; R&D Laboratory, Maastricht; Sanquin Blood Bank South West Region, QC Laboratory, Rotterdam; R&D Laboratory, Leiden.

Norway: Haukeland University Hospital, Blood Bank, Bergen.

Portugal: Lisbon Regional Blood Centre, Lisbon.

Spain: Hospital Clinic Provincial, Department Hemotherapy and Hemostasis; Hospital Universitario Aurna de Villanova, Banc de Sang i Teixits, Lleida.

Sweden: Karolinska University Hospital, Huddinge; Clinical Immunology and Transfusion Medicine, Stockholm.

United Kingdom: Blood Transfusion Service, Aberdeen, Scotland; North of Scotland Blood Transfusion Service, Raigmore Hospital, Inverness, Scotland; Churchill Hospital, Oxford Haemophilia Centre and Thrombosis Unit, Oxford; Scottish National Blood Transfusion Service, Quality Laboratory, Edinburgh, Scotland; Glasgow & West of Scotland Blood Transfusion Service Centre, Glasgow, Scotland; National Blood Service (NBS), Birmingham; NBS, Brentwood, Essex; NBS, Bristol; NBS, Colindale, London; NBS, Leeds; NBS, Manchester; NBS, Newcastle Upon Tyne; NBS, Sheffield; NBS, Southampton; NBS, Tooting, London; National External Quality Assessment Scheme for General Haematology, Watford.

United States: Haemonetics, Braintree, MA; American Red Cross, Holland Laboratory, Rockville, MD; University of Maryland Medical Center, Blood Bank, Baltimore, MD; American Red Cross, Penn-Jersey Region, Research Department, Philadelphia, PA; Hoxworth Blood Center, University of Cincinnati Medical Center, Cincinnati, OH; Dartmouth-Hitchcock Medical Center, Cell Labeling Lab, Department of Pathology, Lebanon, NH; Yale University School of Medicine, Department of Laboratory Medicine, New Haven, CT; Puget Sound Blood Center, Seattle, WA; LAC + USC Medical Center, Los Angeles, CA; Gambro BCT, Lakewood, CO; Navigant Biotechnologies, Lakewood, CO; Massachusetts General Hospital, Blood Transfusion Service, Boston, MA; Lifespan Academic Medical Center, The Miriam and Rhode Island Hospital, Providence, RI.