Anda di halaman 1dari 7

Hematopathology / PLATELET COUNTING

Platelet Counting by the Coulter LH 750, Sysmex XE 2100, and Advia 120

A Comparative Analysis Using the RBC/Platelet Ratio Reference Method

Linda M. Sandhaus, MD, 1 Ebenezer S. Osei, MS, 2 Neeta N. Agrawal, MD, 1 Christine A. Dillman, MT(ASCP), 2 and Howard J. Meyerson, MD 1

Key Words: Platelet counting; Optical platelet counts; Impedance platelet counts; Flow cytometry; Reference method; Automated hematology analyzers

Abstract

We compared the accuracy and precision of the impedance platelet counts generated by the Beckman Coulter LH 750 and the Sysmex XE 2100 and the optical platelet counts produced by the Advia 120 and the Sysmex XE 2100 with flow cytometric reference platelet counts. Samples analyzed had platelet counts less than 150 × 10 3 /µL (150 × 10 9 /L) with a platelet flag or less than 75 × 10 3 /µL (75 × 10 9 /L) on the Sysmex SE 9500. The 105 samples were run sequentially through each analyzer. Anti-CD41 and anti-CD61 monoclonal antibodies were used for flow cytometric determination of the reference platelet count by the RBC/platelet ratio method. The Beckman Coulter and the Sysmex impedance platelet counts showed better correlation with the reference method than the optical platelet counts by the Advia and the Sysmex. At platelet transfusion thresholds of 10 and 20 × 10 3 /µL (10 and 20 × 10 9 /L), the precision of the impedance methods was somewhat better than that of the optical methods. Current methods of optical platelet counting may not be superior to impedance platelet counts for all patient populations.

© American Society for Clinical Pathology

Despite tremendous advancements in automated hema- tology analyzers, the reporting of accurate platelet counts continues to pose difficulties for the hematology laboratory. Factors that contribute to problems with platelet count accu- racy and precision include the following: (1) limitations of instrument technologies for eliminating cellular and particu- late interference; (2) unsatisfactory reproducibility of manual platelet counting methods; and (3) preanalytic interference, such as platelet clumping, that precludes accurate platelet counts by automated or manual methods. The need for automated hematology analyzers that produce accurate platelet counts, particularly for thrombocy- topenic patients, is accentuated by the increasing use of inten- sive chemotherapeutic regimens that produce profound and sustained thrombocytopenia and the clinical imperative to optimize platelet transfusions. Clinical outcomes analyses have led to the development of practice guidelines that recom- mend that the threshold for prophylactic platelet transfusion can be lowered from 20 × 10 3 /µL (20 × 10 9 /L) to 10 × 10 3 /µL (10 × 10 9 /L) or less for most clinical conditions. 1-4 Recently, an immunoplatelet count by flow cytometry (the RBC/platelet ratio method) has been advanced by the International Council for Standardization in Haematology (ICSH) and the Interna- tional Society of Laboratory Hematology (ISLH) as the new reference method for platelet counting. 5,6 Automated platelet counts are being evaluated against this new “gold standard” to validate the accuracy and precision of automated hematology analyzers at the lower platelet count transfusion threshold. 7-12 The impedance principle, patented by Coulter in 1953, was the first automated method for platelet counting. 13 The limitations of the impedance method are well known and include inability to distinguish platelets from other particles

Am J Clin Pathol 2002;118:235-241

235

Sandhaus et al / PLATELET COUNTING

that overlap the size range of platelets, notably, microcytic and fragmented erythrocytes on the higher end and noncel- lular particulate interference and “electronic noise” on the lower end of the size spectrum. “Curve fitting” and “moving threshold” are computerized algorithms that have been applied to impedance methods to correct for interference and improve the accuracy of impedance-derived platelet counts. In recent years, instrument manufacturers have taken diverse approaches to solve the remaining problems with platelet count accuracy. Bayer Diagnostics, Tarrytown, NY (Advia 120), and Abbott Diagnostics, Santa Clara, CA (Cell-Dyn 4000), have developed optical platelet counting methods, which use laser light scatter to identify platelets. 12,14-17 Sysmex has introduced a fluorescence light scatter technique to identify platelets. 18 These methods take advantage of the distinctive optical properties of platelets to distinguish them from nonplatelet cellular elements or particles. Beckman Coulter, on the other hand, has adapted new data extraction techniques and computer algorithms to reduce interference with the impedance platelet count. 8,10,11 Some recent studies seem to indicate that optical platelet counting methods may be more accurate than impedance methods. 7,15-18 The purpose of the present study was to compare the accuracy and precision of platelet counts produced by 3 state-of-the art hematology analyzers on the patient popu- lation of the University Hospitals of Cleveland (UHC; Cleve- land, OH), a tertiary care hospital with a comprehensive cancer center. The study was undertaken as part of a multiparameter hematology analyzer evaluation. The analyzers evaluated were the Beckman Coulter LH 750 (Miami, FL; hereafter referred to as Coulter LH 750), Bayer Advia 120, and the Sysmex XE 2100 (Sysmex, Kobe, Japan). The Abbott Cell-Dyn 4000 was not evaluated. The immunoplatelet RBC/platelet ratio method was used as the reference method.

Materials and Methods

Blood Specimens All blood samples were drawn in K 3 EDTA-anticoagulated blood and run on a Sysmex SE 9500 within 8 hours of phle- botomy. The criteria for inclusion in the study were a platelet count of 150 × 10 3 /µL (150 × 10 9 /L) or less with an instrument- generated platelet flag, and all samples with a platelet count of 75 × 10 3 /µL (75 × 10 9 /L) or less. Samples were run successively through each analyzer in the auto-sampling mode within 1 hour of the initial CBC count. The sequence of the analyzers was randomized for each batch of 5 samples. Four samples contained insufficient volume to obtain results in the auto-sampling mode by all 3 analyzers (no results were given for 3 samples on the Advia 120 and for 1 on the Coulter LH 750). The samples then

236 Am J Clin Pathol 2002;118:235-241

were hand carried to the flow cytometry laboratory where they were stained and analyzed within 90 minutes. Peripheral blood smears were reviewed on all study samples for the presence of erythrocyte or leukocyte fragments, marked microcytosis, giant platelets, and the presence of platelet clumps or fibrin strands.

Patients A total of 105 specimens were obtained from 61 adult and pediatric patients during a 5-day period, October 13 to October 17, 2001. Twenty-five patients were included in the study more than once. Seventeen patients received platelet transfusions during the study period. Diagnoses in this patient population included hematologic malignant neoplasm, 18 cases; solid organ malignant neoplasm, 9 cases; solid organ transplantation, 6 cases; liver disease or end-stage renal disease, 7 cases; cardiovascular disease, 6 cases; and other diagnoses, 15 cases.

Instruments The Sysmex XE 2100 has the capability to produce both an impedance platelet count and a fluorescence-based optical platelet count when the sample is run in the reticulocyte mode. 18 The Coulter LH 750 uses impedance technology with enhanced data extraction techniques and algorithms to eliminate interference. 8,10,11 The Advia 120 uses an optical method in which the volume and refractive index of platelets are determined simultaneously by measuring 2 angles of laser light scatter. 15,16 All of these instruments use software-gener- ated flagging algorithms for identifying abnormal platelet distributions that may indicate the presence of platelet- specific interference or the presence of platelet clumping. In the present study, we evaluated the flagging efficiency of the platelet clumps flag only, because this is the most frequently generated instrument-defined flag that results in smear reviews for verification of platelet counts in our laboratory. The Sysmex XE 2100, Coulter LH 750, and the Advia 120 were calibrated according to manufacturers’ specifications by field service representatives, and precision studies were done. Three medical technologists from the UHC Core Labora- tory were trained to operate the instruments by the field service representatives. Three levels of quality control and required maintenance procedures were performed daily, according to manufacturers’ instructions. Calibrators were run at the conclu- sion of the study to verify that the instrument had maintained calibration, and all stored quality control and maintenance data were reviewed and verified by field service representatives. Study samples were run by UHC technical personnel only.

Immunoplatelet Counting by the ICSH/ISLH Reference Method Platelet counting by the RBC/platelet ratio method was performed by flow cytometry as previously described. 5,6

© American Society for Clinical Pathology

Briefly, 5 µL of well-mixed blood, 5 µL of anti-CD41 fluo- rescein isothiocyanate (Beckman Coulter, Brea, CA), and 5 µL of anti-CD61 fluorescein isothiocyanate (Becton Dick- inson, San Jose, CA) were pipetted into diluent to reach a final dilution of 1:1,000. The preparation was incubated in the dark for 15 minutes and was mixed by gentle inversion to ensure proper and equal distribution of RBCs and platelets before flow cytometric analysis on a flow cytometer (FACSCalibur, Becton Dickinson). The flow rate, discriminator, and photomultiplier settings were as specified elsewhere. 6 At least 50,000 total events were acquired. If a minimum of 1,000 events within the platelet gate was not reached after 50,000 total events, acquisition was continued until either 1,000 events were obtained in the platelet gate or 500,000 total events were acquired. Platelet and RBC counts were determined based on bitmap analysis using Cellquest software (Becton Dickinson). RBC-platelet coin- cident events also were determined as previously described. 5,6 The platelet count was calculated by using the following formula:

Reference Platelet Count = RBC * × Platelets/RBC where RBC * is the mean RBC count of the 3 analyzers; platelets, the number of events in the platelet gate; and RBC, the number of events in the RBC gate by flow cytometry. We calculated the reference platelet count both with and without correction for RBC-platelet coincidence events, using each instrument’s RBC count and the mean RBC count of the 3 analyzers. As there was no significant difference between any of these methods for calculating the platelet count (data not shown), we used the platelet count calculated from the simple ratio and the mean RBC count for all statistical analyses. Precision studies done on 3 to 6 replicates of 7 samples with platelet counts ranging from 26 to 90 × 10 3 /µL (26-90 × 10 9 /L) demonstrated a mean coefficient of variation of 4%.

Statistical Analysis All statistical analyses were performed in EP Evaluator (David G. Rhoads Associates, Kennett Square, PA). The platelet counts generated by each instrument were correlated with the reference platelet count. For the Sysmex XE 2100,

Table 1Regression Results for Analyzer vs Reference Platelet Counts *

Hematopathology / ORIGINAL ARTICLE

the impedance and optical platelet counts were analyzed independently. Regression statistics were determined by using Deming linear regression, with an assumption of equal variance between the methods.

Results

The regression results for the entire range of data are shown in Table 1and Figure 1. The regression results for platelet counts of 50 × 10 3 /µL (50 × 10 9 /L) or less are shown in Table 2and Figure 2. The results indicate that the corre- lation coefficients for the Coulter LH 750 and the Sysmex XE 2100 impedance counts were comparable and were higher than the correlation coefficients for either of the optical platelet counts. The differences between the impedance and optical methods became more apparent when the analysis was limited to very low platelet counts. An analysis of 17 samples that contained many RBC fragments and/or large platelets also was done. The correlation coefficients for this subset analysis Table 3also fail to demonstrate superior performance of the optical platelet counts. To examine the precision of platelet counts near the platelet transfusion thresholds of 10 and 20 × 10 3 /µL (10 and 20 × 10 9 /L), we compared the 95% confidence intervals of platelet counts at these decision points Table 4. Although there were no major differences among the instru- ments, the confidence intervals were slightly wider for the optical methods than for the impedance methods. Six samples were excluded from the correlation study owing to the presence of marked platelet clumping observed on the peripheral blood smear, resulting in pseudothrombocy- topenia in 4 samples. The platelet counts for these 6 speci- mens are shown in Table 5. An interesting observation is that the platelet counts generated by the Advia 120 tended to be higher than the platelet counts produced by the other instru- ments. We evaluated the flagging efficiency of the different analyzers for detecting marked platelet clumping Table 6. All of the analyzers had deficiencies in this discrimination, with the Coulter LH 750 seeming to have the poorest sensi- tivity and the Sysmex XE 2100 the poorest specificity.

Analyzer

N

Slope

Intercept (× 10 9 /L)

Bias (× 10 9 /L)

R 2

Coulter LH 750 Advia 120 Sysmex XE 2100 Impedance Optical

98

1.06

(1.03 to 1.10) (1.03 to 1.12)

–1.5

(–4.2 to 1.2) (–5.5 to 0.8)

2.4

0.968

96

1.08

–2.4

2.4

0.958

99

1.00

(0.97 to 1.04) (0.96 to 1.05)

–3.3

(–5.8 to 0.8) (–4.9 to 0.4)

–3.3

0.970

99

1.01

–1.8

–1.4

0.951

* The conversion from Système International units (× 10 9 /L) to conventional units (× 10 3 /µL) is 1 to 1 (ie, divide by 1). For proprietary information, see the text. The 95% confidence intervals are shown in parentheses.

© American Society for Clinical Pathology

Am J Clin Pathol 2002;118:235-241

237

Sandhaus et al / PLATELET COUNTING

A

A

B

B
180 200 135 150 90 100 45 50 0 0 0 45 90 135 180
180
200
135
150
90
100
45
50
0
0
0
45
90
135
180
0
45
90
135
180
9
9
RBC Ratio ( × 10
/L)
RBC Ratio ( × 10
/L)
C
D
220
200
165
150
110
100
55
50
0
0
0
45
90
135
180
0
45
90
135
180
RBC Ratio ( × 10 9 /L)
RBC Ratio ( × 10 9 /L)
LH ( × 10 9 /L) Sysmex XE ( × 10 9 /L)Coulter
( × 10 9 /L) Sysmex XE ( × 10 9 /L)Advia

Figure 1Regression results for platelet counts of 150 × 10 9 /L or less. Comparison of the immunoplatelet reference count (derived from the RBC/platelet ratio) with the platelet counts from the Sysmex XE 2100 (A, impedance, Y = X – 3.3, R 2 = 0.970; B, optical, Y = 1.01X – 1.8, R 2 = 0.951), Beckman Coulter LH 750 (C, Y = 1.06X – 1.5, R 2 = 0.968), and Advia 120 (D, Y = 1.08X – 2.4, R 2 = 0.958) over the entire range of data. Short dashes, 95% confidence intervals; long dashes, line of equality; solid line, regression line. The conversion factor to conventional units (× 10 3 /µL) is 1. For proprietary information, see the text.

A

A

B

B
60 60 45 45 30 30 15 15 0 0 0 15 30 45 60
60
60
45
45
30
30
15
15
0
0
0
15
30
45
60
0
15
30
45
60
9
9
RBC Ratio ( × 10
/L)
RBC Ratio ( × 10
/L)
C
D
60
60
45
45
30
30
15
15
0
0
0
15
30
45
60
0
15
30
45
60
RBC Ratio ( × 10 9 /L)
RBC Ratio ( × 10 9 /L)
Sysmex XE ( × 10 9 /L)Coulter
LH ( × 10 9 /L)
Sysmex XE ( × 10 9 /L)Advia
( × 10 9 /L)

Figure 2Regression results for platelet counts of 50 × 10 9 /L or less. Comparison of the immunoplatelet reference count (derived from the RBC/platelet ratio) with the platelet counts from the Sysmex XE 2100 (A, impedance, Y = 0.96X – 1.1, R 2 = 0.92; B, optical, Y = 0.95X + 1.7, R 2 = 0.72), Beckman Coulter LH 750 (C, Y = 0.99X + 1.3, R 2 = 0.88), and Advia 120 (D, Y = 0.96X + 2.0, R 2 = 0.81). Short dashes, 95% confidence intervals; long dashes, line of equality; solid line, regression line. The conversion factor to conventional units (× 10 3 /µL) is 1. For proprietary information, see the text.

Table 2Regression Results for Analyzer vs Reference Platelet Counts for Platelet Counts 50 × 10 9 /L *

Analyzer

N

Slope

Intercept (× 10 9 /L)

Bias (× 10 9 /L)

R 2

Coulter LH 750 Advia 120 Sysmex XE 2100 Impedance Optical

48

0.99

(0.89 to 1.09) (0.83 to 1.09)

1.3

(–1.9 to 4.5) (–2.1 to 6.1)

0.9

0.88

46

0.96

2.0

0.8

0.81

49

0.96

(0.88 to 1.04) (0.79 to 1.1)

–1.1

(–3.6 to 1.4) (–3.1 to 6.5)

–2.3

0.92

49

0.95

1.7

0.2

0.72

* The conversion from Système International units (× 10 9 /L) to conventional units (× 10 3 /µL) is 1 to 1 (ie, divide by 1). For proprietary information, see the text. The 95% confidence intervals are shown in parentheses.

Discussion

We used the ICSH/ISLH immunoplatelet reference method to evaluate the platelet count accuracy of the Sysmex XE 2100, the Coulter LH 750, and the Advia 120 on blood samples from patients with thrombocytopenia. We challenged

238 Am J Clin Pathol 2002;118:235-241

the reference method for potential sources of error and found the method to be robust. We used the simple RBC/platelet ratio for calculating the reference immunoplatelet count after demon- strating that platelet counts obtained using the correction for platelet-RBC coincidence events did not significantly alter the results. A potential source of error of the immunoplatelet

© American Society for Clinical Pathology

method could be that prolonged acquisition times for speci- mens with very low platelet counts would affect the RBC/platelet ratio owing to cell settling. However, we observed no differences between continuous, uninterrupted collections and interrupted, remixed sample collections (data not shown). We observed some limitations of the reference method that were related to the presence of large platelets or platelet clumps. In a few samples, large platelets may not have been included in the initial platelet gate, because their high fluorescence and forward scatter placed them outside the bitmap gate. Regating those samples so that all highly fluores- cent events were included resulted in up to a 9% increase in the platelet counts in 3 samples. These adjusted platelet counts resulted in a slight but insignificant improvement in the corre- lation results for the 2 optical platelet methods. This limitation can be surmounted by adjusting the photomultiplier settings on the flow cytometer, using quadrant statistics, or visually adjusting the bit map gates. A limitation that could not be corrected for, however, was the presence of marked platelet clumping that was observed in 6 samples. When there is marked platelet clumping in the sample, the flow cytometric determination of the RBC/platelet ratio cannot be considered reliable, because the platelet events may be falsely lowered. The major finding is that the impedance methods of platelet counting evaluated in the present study seemed to be more accurate than the optical methods of platelet counting for platelet counts of 150 × 10 3 /µL (150 × 10 9 /L) or less.

Table 4Precision of Platelet Counts at Transfusion Thresholds *

Hematopathology / ORIGINAL ARTICLE

Table 3Analyzer vs Reference Platelet Counts for Specimens With RBC Fragments or Giant Platelets (N = 17)

Analyzer *

R

R 2

Coulter LH 750 Advia 120 Sysmex XE 2100 Impedance Optical

0.98

0.96

0.96

0.92

0.97

0.94

0.95

0.90

* For proprietary information, see the text.

Both the Coulter LH 750 and the Sysmex XE 2100 imped- ance counts showed slightly better correlation with the refer- ence platelet counts than did the Advia 120 or the Sysmex XE 2100 optical counts. These differences became even more apparent at very low platelet counts. Furthermore, the 2 impedance methods correlated as well with each other (R 2 = 0.969) as they did with the reference method. The optical methods, on the other hand, correlated better with each other (R 2 = 0.971) than they did with the reference method. These results suggest that the sources of error for the 2 optical methods may be similar and that they differ from those of the impedance methods. Our results are in contrast with those of recently published studies that suggest the superior accuracy of optical platelet counting methods for thrombocytopenic specimens. 7,15-18 Harrison et al 7 demonstrated an R 2 of 0.96

Sysmex XE 2100

Platelet Count (× 10 9 /L)

Impedance

Optical

Coulter LH 750

Advia 120

10

4.5-8.9

(4.4)

5.4-11.2 (5.8)

6.7-11.6 (4.9)

5.5-11.3 (5.8)

20

14.7-18.7

(4.0)

15.8-20.9 (5.1)

17.6-21.9 (4.3)

16.7-21.7

(5.0)

* Data are given × 10 9 /L. Numbers in parentheses represent width of 95% confidence intervals (× 10 9 /L). The conversion from Système International units (× 10 9 /L) to conventional units (× 10 3 /µL) is 1 to 1 (ie, divide by 1).

For proprietary information, see the text.

Table 5Platelet Counts on Six Specimens With Platelet Clumping *

Sysmex XE 2100

Specimen No.

Impedance

Optical

Coulter LH 750

Advia 120

Reference Method

23

10

19

6

15

3

30

159

150

125

231

104

38

167

229

162

285

176

41

150

145

174

200

159

58

79

140

87

121

54

89

19

23

19

44

21

* Data are given × 10 9 /L. The conversion from Système International units (× 10 9 /L) to conventional units (× 10 3 /µL) is 1 to 1 (ie, divide by 1).

For proprietary information, see the text. When there is marked platelet clumping in the sample, the flow cytometric determination of the RBC/platelet ratio cannot be considered reliable, because the platelet events may be falsely lowered.

© American Society for Clinical Pathology

Am J Clin Pathol 2002;118:235-241

239

Sandhaus et al / PLATELET COUNTING

Table 6Analyzer Flagging of Platelet Clumps *

Platelet

Sysmex XE 2100

Coulter LH 750

Advia

Clumps Flag

120

Total No. of flags True-positive False-positive False-negative

20

3

4

4

23

16

1

1

2

4

3

* See the “Discussion” section for comments on the reliability of flagging platelet clumps. For proprietary information, see the text. A true-positive flag was defined as the presence of marked platelet clumping on the peripheral blood smear that precluded quantitative platelet estimation. See text for comments on false-negative rate.

for the Advia 120 with the RBC/platelet reference method for platelet counts less than 100 × 10 3 /µL (<100 × 10 9 L), which is consistent with our own results. In their multi- instrument comparison, the impedance counts from the 3 other instruments evaluated yielded poorer correlation results. However, the sample size for thrombocytopenic samples in that study 7 was less than in the present study, and the Coulter LH 750 was not among the instruments compared. Other published studies that seem to demonstrate superior performance of the Advia optical platelet count over impedance platelet counts have used manual platelet counting methods as the reference methods. 15-17 It also is important to note that the blood samples included in these studies were highly selected for potential platelet interfering substances. Similarly, an evaluation of the Sysmex XE 2100 platelet counts using the immunoplatelet reference method demonstrated significantly better performance of the optical platelet count when the samples were selected specifically for known platelet disorders or potential interfering substances. 18 Sample selection may partially explain the results of our study. The patient samples were selected primarily on the basis of low platelet count. Platelet flags, which might indi- cate the presence of interfering particles, such as RBC or leukocyte fragments, were used to select only samples with platelet counts between 75 and 150 × 10 3 /µL (75-150 × 10 9 /L; 21.0% [22/105] of the total samples). Forty-four percent (27/61) of the patients in the study had oncology diagnoses, and the majority of them were thrombocytopenic owing to bone marrow suppression. Several of these patients underwent sampling more than once. Therefore, oncology patients may be overrepresented in the present study in comparison with some other published studies. In addition, 17 patients had received platelet transfusions. Evaluation of platelet morphologic features by peripheral smear review suggested that the high frequency of patients with hypopro- liferative bone marrow states might have created a sampling bias for small platelets, which would tend to minimize any advantage of optical methods for discriminating large

240 Am J Clin Pathol 2002;118:235-241

platelets. Nevertheless, a subset analysis of 17 samples with RBC fragmentation and/or giant platelets identified on peripheral blood smears failed to demonstrate improved accuracy of the optical methods. For very low platelet counts at or near a prophylactic platelet transfusion threshold of 20 × 10 3 /µL (20 × 10 9 /L), all instruments showed comparable reliability, with confidence intervals ranging from 4.0 to 5.1 × 10 3 /µL (4.0-5.1 × 10 9 /L). Slightly wider confidence intervals were obtained for the optical methods than for the impedance methods, which might reflect the lower numbers of total events that are enumerated by optical methods. Based on the combined performance of these analyzers, the results indicate that a true platelet count of 20 × 10 3 /µL (20 × 10 9 /L) could range from 14.7 to 21.7 × 10 3 /µL (14.7-21.7 × 10 9 /L). If 10 × 10 3 /µL (10 × 10 9 /L) is used as the platelet transfusion threshold, the reported platelet counts could range from 4.5 to 11.6 × 10 3 /µL (4.5-11.6 × 10 9 /L). Clinicians should be aware of the limits of the analytic method for platelet counting in deter- mining the appropriateness of recommended platelet transfu- sion thresholds for different clinical conditions. 19 Platelet clumping in EDTA anticoagulants continues to pose procedural problems for the clinical hematology labo- ratory. 20 In our laboratory, when a platelet clumps flag is generated, the automated platelet count is removed from the CBC count results in the laboratory information system, a smear is reviewed for platelet count verification, and the final platelet count is entered manually and verified in the laboratory information system. Each of these steps influ- ences the turnaround time for CBC counts and introduces the potential for laboratory error. In our evaluation of the analyzers’ flagging efficiency for platelet clumps, we defined a true-positive flag as the presence of marked platelet clumping on the peripheral blood smear that precluded determination of a numeric platelet estimate. At the factory settings, the Sysmex XE 2100 had the highest sensitivity (67%) and the poorest specificity (84%). The platelet clumps flag on this analyzer is an adjustable flag, but adjustments to improve specificity might further reduce the sensitivity. The Coulter LH 750 demonstrated the poorest sensitivity for platelet clumps, with 4 of 6 clumped speci- mens not flagged. Failure to flag platelet clumps could lead to erroneous reporting of low platelet counts if laboratory algorithms for confirming low platelet counts are inadequate or not followed carefully. It is difficult to be conclusive about the flagging efficiency of the analyzers based on these limited data, because the possibility that the degree of platelet clumping varied during the performance of the sequential CBC counts and flow cytometric analysis cannot be excluded. In this regard, it is interesting that the platelet counts generated by the Advia 120 on the specimens with platelet clumping tended to be higher than the counts

© American Society for Clinical Pathology

obtained from the other 2 instruments. One possible explana- tion for this observation is that mechanical dispersion of platelet clumps occurs in the optical counting chamber. Mechanical dispersion of platelet clumps also might explain the lower flagging rate for platelet clumps by the Advia 120. This multi-instrument comparison did not demonstrate any advantages of optical platelet counting methods over current impedance methods when the RBC/platelet reference method was used as the gold standard. We suggest that patient population, sample selection, and reference method should be considered possible explanations for discordant results of evaluations of platelet count methods. However, it is important to recognize that the accuracy of automated platelet counts is a function not only of the underlying tech- nology that is used to detect platelets but also of the comput- erized algorithms that are designed to evaluate the platelet data. The improved performance of the Coulter LH 750 over that of the Coulter GenS (Beckman Coulter, Miami, FL) has been attributed largely to improvements in the analyzer’s data extraction process. 10,11 The Sysmex XE 2100 deter- mines which platelet count is reported, optical or impedance, based on a predefined algorithm. Optimally, the performance of the reported platelet count by the Sysmex XE 2100 actu- ally might exceed the performance of either the optical platelet count or the impedance count. The present study, however, was not designed to evaluate the “switching” algo- rithm. Finally, improvements in the ability to reliably detect (or overcome) EDTA-induced platelet clumping would contribute greatly to rapid and accurate platelet count reporting by clinical laboratories.

From the Departments of Pathology, 1 Case Western Reserve University School of Medicine and 2 University Hospitals of Cleveland, Cleveland, OH.

Address reprint requests to Dr Sandhaus: Dept of Pathology, University Hospitals of Cleveland, 11100 Euclid Ave, Cleveland OH 44106.

References

1. Ancliff, PJ, Machin SJ. Trigger factors for prophylactic platelet transfusion. Blood Rev. 1988;12:234-238.

2. Lawrence JB, Yomtovian RA, Dillman C, et al. Reliability of automated platelet counts: comparison with manual method and utility for prediction of clinical bleeding. Am J Hematol.

1995;48:244-250.

3. Lawrence JB, Yomtovian RA, Hammons T, et al. Lowering the prophylactic platelet transfusion threshold: a prospective analysis. Leuk Lymphoma. 2001;41:67-76.

© American Society for Clinical Pathology

Hematopathology / ORIGINAL ARTICLE

4.

Schiffer CA, Anderson KC, Bennett CL, et al. Platelet transfusion for patients with cancer: clinical practice guidelines of the American Society of Clinical Oncology. J Clin Oncol. 2001;19:1519-1538.

5.

International Council for Standardization in Haematology Expert Panel on Cytometry and International Society of Laboratory Hematology Task Force on Platelet Counting. Platelet counting by the RBC/platelet ratio method: a reference method. Am J Clin Pathol. 2001;115:460-464.

6.

Harrison P, Ault KA, Chapman S, et al. An interlaboratory study of a candidate reference method for platelet counting. Am J Clin Pathol. 2001;115:448-459.

7.

Harrison P, Horton A, Grant D, et al. Immunoplatelet counting: a proposed new reference procedure. Br J Haematol.

2000;108:228-235.

8.

Kaplan SS, Johnson K, Wolfe N. Validation of low platelet counts from Coulter LH 750 and Coulter GenS analyzers. Lab Hematol. 2001;7:198-203.

9.

Kunz D. Possibilities and limitations of automated platelet counting procedures in the thrombocytopenic range. Semin Thromb Hemost. 2001;27:229-235.

10.

Keeney M, Brown W, Yee IC. Platelet counts and flagging rates from the LH750 hematology analyzer compared with the ICSH/ISLH platelet reference method and the GenS hematology analyzer. Lab Hematol. 2001;7:204-210.

11.

Charie LA, Harrison P, Smith CU, et al. Accuracy in the low platelet count range: a comparison of automated platelet counts on Beckman Coulter high-volume hematology analyzers with the ISLH/ISCH platelet reference method. Lab Hematol. 2001;7:236-244.

12.

Kunz D, Kunz WS, Scott CS, et al. Automated CD61 immunoplatelet analysis of thrombocytopenic samples. Br J Haematol. 2001;112:584-592.

13.

Dalton WT, Bollinger P, Drewinko B. A side-by-side evaluation of four platelet-counting instruments. Am J Clin Pathol. 1980;74:119-134.

14.

Kickler TS. Clinical analyzers: advances in automated cell counting. Anal Chem. 1999;71:363R-365R.

15.

Chapman DH, Hardin J, Miers M, et al. Reduction of the platelet review rate using the two-dimensional platelet method. Am J Clin Pathol. 2001;115:894-898.

16.

Kunicka JE, Fischer G, Murphy J, et al. Improved platelet counting using two-dimensional laser light scatter. Am J Clin Pathol. 2000;114:283-289.

17.

Stanworth SJ, Denton K, Monteath J, et al. Automated counting of platelets on the Bayer ADVIA 120 analyser. Clin Lab Haematol. 1999;21:113-117.

18.

Briggs C, Harrison P, Grant D, et al. New quantitative parameters on a recently introduced automated blood cell counter: the XE 2100. Clin Lab Haematol. 2000;22:345-350.

19.

Hanseler E, Fehr J, Keller J. Estimation of the lower limits of manual and automated platelet counting. Am J Clin Pathol.

1996;105:782-787.

20.

Bartels PCM, Schoorl M, Lombarts AJ. Screening for EDTA- dependent deviations in platelet counts and abnormalities in platelet distribution histograms in pseudothrombocytopenia. Scand J Clin Lab Invest. 1997;57:629-636.

Am J Clin Pathol 2002;118:235-241

241