by Automated Cell Counters

Transcription

1 Cytometry (Communications in Clinical Cytometry) 22:26-34 (1995) Variability in Absolute Lymphocyte Counts Obtained by Automated Cell Counters E. Simson and W. Groner Department of Pathology, Long Island Jewish Medical Center, the Long Island campus for the Albert Einstein College of Medicine, New Hyde Park, New York (E.S.); Center for Laboratory Technology, Inc., Great Neck, New York (W.G.) There is increasing interest in the absolute lymphocyte count. This is partly driven by the need to obtain absolute values for lymphocyte subsets such as absolute CD4+ counts in human immunodeficiency virus (HIV)-infected persons. The absolute total lymphocyte count is usually determined in the routine hematology laboratory on a separate sample from the same patient specimen and then combined with percentage results from flow cytometry to obtain the absolute value of the lymphocyte subsets. We have studied analytic variability in the absolute lymphocyte determination and compared it to the variability of the total white blood count (WBC). In a series of 524 specimens, four different automated methods were compared to each other and to the traditional eye count differential. The automated methods were four widely used automated cell counters (Technicon H*l, TOA NE8, Coulter STKS, and Abbott CD3). The results indicate that analytic variability in the absolute lymphocyte counts, due, primarily, to method variability, is significant and is larger than the variability typically observed on interlaboratory trials of relative CD4 counts. These method biases cannot easily be reduced by calibration, since the cell classification algorithms are built-in features of the various cell counters. Analytic variability of the absolute lymphocyte counts was found to be 12.4% compared with analytic variability of only 4.9% for total WBC counts on the same samples. Our data suggest that more precise results would be obtained if flow cytometry results expressed each phenotype as a fraction of the leukocytes as well as total lymphocytes. Conversion to absolute values could then be accomplished through determination of the total WBC in the routine hematology laboratory Wiley-Liss, Inc. Key terms: Flow cytometry, hematology methods, lymphocyte counts, HIV, CD4+ lymphocytes The determination of the absolute value of CD4+ lymphocytes has been shown to be useful in monitoring persons with human immunodeficiency virus (HIV) infection (12). The U.S. Public Health Service has recommended monitoring CD4+ T lymphocytes every 3-6 months in all HIV-infected persons (5), and the level of + CD4 T lymphocytes has been included as a criterion in the classification system for HIV infection (2). It is therefore of great interest to consider the potential for and magnitude of error in this determination. The measurement of CD4+ T-cell levels presently involves the combination of at least two (sometimes three) independent laboratory determinations. The first, which is the counting of the relative number (percentage) of CD4 lymphocytes by flow cytometry, has received considerable attention. Stand ards for performance have been issued recently by the Centers for Disease Control (CDC), the National Committee for Clinical hbordtory Standards (NCCLS), and the National Institute for Allergy and Infectious Disease (NIAID; 6,19,4) in an effort to minimize the impact of technical and analytic factors (such as staining procedures and alignment and calibration of flow cytometers) on intralaboratory and interlaboratory variability in the measurement of CD4 cell number. Each of these documents also refers to the need for additional accurate hematology measurements to convert the relative values into the required absolute value (cells/ volume). The documents stipulate that the hematology counts be performed on fresh specimens drawn at the same time, and they suggest that automated counters be used (except for samples rejected or flagged by the automated counter) for the determination of the absolute lymphocyte count to reduce the imprecision resulting from the relatively low numbers of cells typically classified by manual microscopy. The NIAID guidelines state that, when manual counting is required, a minimum of Received for publication February 1, 1994; accepted June 14, Address reprint requests to Elkin Simson, Associate Chairman, Clinical Pathology, Long Island Jewish Medical Center, New Hyde Park, NY Wiley-Liss, Inc.

2 VARIABILITY IN ABSOLUTE LYMPHOCYTE COUNTS 27 4 cells must be classified, and all atypical lymphocytes and large unstained cells (LUCS) are to be included in the lymphocytes. Studies of the imprecision of the determination of absolute lymphocytes have also been reported ( 16,17). These studies demonstrate that the total variability in the computed CD4 counts can be substantially affected by specimen handling and counting imprecision. A recent comment has also indicated that significant bias may result from the use of different automated cell counters (24) and that flow cytometers may be better in determining the percentage of lymphocytes ( 13). In this study, an attempt was made to characterize the analytic variability of absolute lymphocyte counts by considering the variation across different automated cell counting instruments in addition to within-system imprecision. Results were obtained on the same samples with each of four different multiparameter cell counters. The percentage of total lymphocytes from each system was compared with the reference for manual (eye count) differential, per NCCLS H2A (22), to ascertain the performance of lymphocyte counting for each system. The results were also compared with each other for total white blood cell count (WBC) and for absolute lymphocyte count. Protocols were established to isolate the effects of specimen handling, method bias, and imprecision. The results were then analyzed to demonstrate the variability in absolute CD4 counts that can be expected as a result of the variability of the hematology measurements. MATERIALS AND METHODS Specimen Collection and Handling Samples of blood from 697 patients, including 12 normal individuals, were run on each of the four instruments over a period of 2 days. Patient specimens were part of the daily workload arriving in the hematology laboratory of the Department of Pathology at Long Island Jewish Medical Center, an 87-bed tertiary-care hospital complex. Specimens were drawn in tripotassium EDTA and were analyzed within 18 h from phlebotomy. Specimens were analyzed within 6 h from their receipt in the laboratory. Specimen handling errors due to aliquoting were reduced by running the identical samples sequentially on each analyzer. The analyzer sequence was varied in a random manner. Most samples were run in duplicate on each analyzer, but approximately 7 were only run singly on each analyzer due to insufficient sample volume for duplicate analysis. Reference Differential Wedge smears were made from each sample and analyzed by eye microscopy. A 4-cell eye differential was obtained from two independent 2-cell differentials as described in NCCLS H2A (22). Instruments Four instruments were used for the study: 1) Technicon H* 1 System (Miles Diagnostics, Tarrytown, NY), us- ing software version 1.3; 2) Coulter STKS (Coulter, Hialeah, FL), using software version 1D and ld.l; 3) Sysmex NE8 (TOA Instruments, Kobe, Japan), using software version 27; and 4) CD3 (Abbott Diagnostics, Chicago, IL), using software version Revision L. Each of the instruments was installed by the manufacturer and run according to the manufacturer s instructions. The instruments were calibrated independently according to the manufacturers specifications. Quality control materials recommended by each manufacturer were run throughout the study to verify constancy of calibration. Any sample on which any of the analyzers did not produce results for WBC, percentage lymphocytes, or absolute lymphocytes was eliminated. This left 524 specimens for comparative analysis. In this sample set, the WBC ranged from 2 X 1 Aiter to 6 X 1 Aiter. The absolute lymphocyte counts ranged kom.15 X lo /liter to 9 X lo /liter, and the proportional lymphocytes from 1% of WBC to 75% of WBC. Accuracy The accuracy of each analyzer in determining lymphocyte percentage was obtained by comparing the result from each specimen with the result from the 4-cell reference manual differential. When duplicate analyzer results were available, only the first result was used for comparison. Imprecision For each analyzer, duplicate imprecision was determined for both absolute lymphocytes and total WBC by analyzing the paired readings from each of the analyzers. For comparison, the duplicate imprecision of 2-cell manual counts multiplied by WBC was calculated as per NCCLS H2A. In addition, the duplicates for each method were compared to each other by regression analysis to determine the residual scatter around the regression line contributed by method imprecision. Analytic Variability To determine the analytic variability in the determinations, the first result for both absolute lymphocytes and total WBC from each analyzer on each specimen was used. As a measure of the analytic variability across the four analyzers, a coefficient of variation was calculated for each specimen. Specimen Aging A series of four specimens was drawn, and aliquots of each specimen were stored both at room temperature and refrigerated (-4 C) over a period of 24 h. These samples were then analyzed repeatedly by each of the four instruments over a period of 24 hours. RESULTS Accuracy Figure l a4 shows graphic summaries for each system compared to the reference method for percentage lym-

3 28 SIMSON AND GRONER phocytes. Table 1 summarizes the population and regression statistics for the accuracy study. The results show relatively good agreement between each method and the reference method, with the bulk of the results falling within the bounds of imprecision for the reference 4- cell manual. Imprecision Table 2 summarizes the results of the duplicate imprecision studies for absolute lymphocytes and for total WBC obtained from each instrument. Duplicates of the 2- cell microscopic eye count for lymphocytes are also included in Table 2 for comparison. The results in Table 2 demonstrate that the larger numbers of cells typically counted with the automated instruments have the expected effect of substantially reducing the duplicate imprecision as compared with the microscopic differential, even when 2 cells are classified by eye rather than the routine 1-cell differential. The imprecision for lymphocyte counts for the analyzers ranged from a standard deviation (S.D.) of.7 X 1 cells/liter to.18 X 1 ceils/liter yielding coefficients of variation (C.V.) from 3.95% to 1.45%. There were greater differences among systems for WBC counts where the S.D. varied from.14 to.47 X IO cells/liter. However, due to the larger denominator, C.V.s were smaller, ranging from 1.71% to 5.13%. The mean values for absolute lymphocytes and WBC varied somewhat among the analyzers as a result of the differences in samples included in the precision study. As expected, the standard error on regression analysis (Sy.x) was similar to the S.D. Analytic Variability In order to focus on the total analytic variability, the data were subjected to a one-way analysis of variance. The coefficient of variation (C.V.) across the four measurement systems for each of the 524 specimens was reviewed; one sample, whose C.V. was greater than 5% for WBC and greater than 1% for absolute lymphocyte count, was considered an outlier and was excluded from further analysis of variability, leaving 523 samples. The results were then analyzed in a manner analogous to the determination of comparability between two methods ( 15), with the C.V.s replacing the difference between methods. Figure 2 shows the coefficients of variation across the analyzers for absolute lymphocytes and for total WBC plotted vs. the level of each parameter. In order to test for any relationship between the variability and the level (the analogy for proportional bias), the data were subjected to regression analysis. Figure 2a demonstrates the results obtained for WBC. For a few samples, the C.V. was slightly greater than 2%, but, for most, it was less than 1%. The regression line was horizontal, indicating no relationship between the magnitude of the variability and the magnitude of the WBC. The mean value of the C.V. for the WBC count was 4.9%. Figure 2b shows the C.V. across the analyzers for absolute lymphocytes vs. the mean absolute lymphocyte count. The C.V. for several samples was as high as 5%, with many samples greater than 1%. The mean C.V. was 12.4%. When tested by regression variability analysis, variability also increased significantly as absolute lymphocyte count decreased. However, this increase could be the result of increased imprecision, because fewer cells are classified as the absolute lymphocytes count decreases. Figure 3 compares the distribution of the coefficient of variation for WBC with absolute lymphocytes. The variability for WBC is generally much less than that for absolute lymphocyte counts, with 99% of the WBC C.V. less than the average value for absolute lymphocyte counts ( 12.4% ). Specimen Handling Figures 4 and 5 show results obtained from specimens stored at room temperature for 1 hours from each method for absolute lymphocytes and total WBC respectively. This is a greater period than that stipulated in current standards for determining CD4+ counts. The dashed lines for each analyzer are derived from the duplicate imprecision 2 2 S.D. from the starting value. Thus, since each data point is a single determination, only results exceeding this boundary would indicate a shift from the initial value with greater than 95% confidence. It can be seen from the figures that no significant drifts (defined as two consecutive results exceeding the boundary) occurred within the 1 h postphlebotomy study period for either absolute lymphocytes or total WBC. Similar results were obtained from the samples stored at refrigerated temperatures. DISCUSSION The results demonstrate that, as was previously noted (25), increased precision in lymphocyte counting is achieved with the use of automated cell counters. Furthermore, the results demonstrate that there is generally good agreement between each analyzer and the traditional microscopic eye count, as has also been previously reported for individual analyzer comparisons (3,8,9,1 I, 23,26,27). However, these comparisons are typically limited by the imprecision of the microscopic eye count and, therefore, are not generally useful in detecting small method variances that may be significant. It was the intent of this study to examine the overall comparability of the various methods for producing absolute lymphocyte counts to be used to complement immunophenotyping performed on a flow cytometer. Thus, in our analysis, we attempt to characterize the total analytic variability in lymphocyte counts that may be encountered in, for instance, tracking an HIV-positive individual by obtaining CD4 and lymphocyte counts every 6 months, not necessarily from the same analyzer. We assumed only that the hematology results were obtained in a manner consistent with the current guidelines for immunophenotyping, i.e., that the determination was performed with a properly calibrated and quality controlled automated cell counter on a fresh blood sample.

4 VARIABILITY IN ABSOLUTE LYMPHOCYTE COUNTS 29 f 2. r P E, -1 D - C95%-LINE - C95%+LINE CLYMP4 ae a Manual 4 cell Lymphocyte% 1 I I b o Manual 4 cell Lymphocyte % N95%-LINE - N95%+LINE NLYMPA Manual 4 cell Lymphocyte% 1 7 f 2. c D 5 m Y I-m S95%-LINE - S95%+LINE SLYMP% d o Manual 4 cell Lymphocyte% FIG. 1. Comparison to reference method. Scatterplot and regression analysis for relative lymphocytes (%I vs. the 4 cell manual microscopic reference method. A single reading for each specimen is taken for comparison. Where duplicate readings were available, the first of the pair is taken. The oval boundary represents the 95% confidence limits for the reference method. a: CD3 vs. reference. b: H*l vs. reference. c: NE8 vs. reference. d: STKS vs. reference.

5 3 SIMSON AND GRONER Table 1 Summary of Accuracy Data: Percentage Lymphocytes Sample Population Regression Statistics Analyzer Mean Rangea Slope Intercept Corr. Coeff. H* CD NE STKS aentire range of specimens tested. The results of this study indicate that substantially greater analytic variability can be encountered in the absolute lymphocyte determination than in the determination of total WBC. The additional variability of absolute lymphocyte counts is due, in part, to increased imprecision but, to a larger degree, to both scatter and biases between the cell counters. These biases are most probably a result of differences in both the technologies and the cell classification algorithms used in the cell counter design. Thus, the differences are essentially built into the cell counter and cannot be adjusted out by calibration. These instrument algorithms can result in a bias for lymphocytes or granulocytes, even between two instruments from the same manufacturer (7). Although no results of interlaboratory trials across instruments are available for the determination of absolute lymphocytes, our data suggest that the analytic variability is greater than 1% or significantly greater than the C.V. reported in interlaboratory studies of relative CD4 counts (24). The data in this study show the analytic variability of total WBC to be approximately 5%. This is approximately one-third of the analytic variability of absolute lymphocytes. This result for WBC is quite consistent with results of interlaboratory trials for this determination ( 1). Further, each of the automated instruments contains a means for adjusting out the bias in total WBC by means of calibration. The absolute number of a lymphocyte subset is currently obtained from the product of two independent measurements; a relative measurement (Rel) of the percentage of lymphocytes isolated by one or more monoclonal antibodies, and the determination of the absolute lymphocyte counts (ABS lymph). It is likely that advances in technology will be made to overcome this situation for the special case of CD4+ counting. This may occur through better flow cytometric methods (2 1 ) or through the use of an alternate technology in which a direct determination by sensitive immunoassay is made of the concentration of CD4' (18). However, it is unlikely that the need to combine results from quantitative cell counters with flow cytometers will be totally eliminated for some time to come. Therefore, it remains of interest to consider how this combination can be made with minimal variability. The theory of the propagation of errors states that the coefficient of variation for a simple product of two independent determinations is given by the root sum square of the coefficients of variation for each measurement ( 1 ). Thus, the coefficient of variation for the absolute cell count can be expressed as: CVZ(ABS) = CV2(REL) + CV2(ABS LYMPH) (1) Interlaboratory trials (2) have shown the C.V. for the relative determination of lymphocyte subsets to be typically less than 1% and, for CD4, on the order of 5%. As an example, assume that 5% of lymphocytes are CD4+ at an absolute count of 1. X 1" cells/liter. The value from Figure 2b for the variability of absolute lymphocytes at 1. X 1" cellsfliter is 15%. Combining it with a C.V. for CD4 (Rel) of 5%, one obtains an estimate of analytic variation of absolute CD4+ counts on the order of 16%, or +.8 X lo9 cells/liter. This analytic variability will increase as the absolute lymphocyte count decreases as a result of both counting imprecision and method variability. Following the same theory, the coefficient of variation for absolute lymphocytes could alternatively be given in terms of the C.V. of the lymphocyte subset expressed as a percentage of total leukocytes (Rel* ) together with the C.V. of the total WBC (TOT WBC) as: CVZ(ABS) = CV2(REL*) + CV2(TOT WBC) (2) In this case, the relative count obtained on the flow cytometer is the fraction obtained by dividing the cells isolated by the monoclonal antibodies by total gated leukocytes. The data we have collected show that both of the components of the C.V. (imprecision and analytic variability) are much less for the total WBC than for absolute lymphocytes. Furthermore, the total WBC is a more robust parameter, is less subject to abnormal flagging, can be calibrated independently, and is regularly assessed in interlaboratory trials. Thus, the component of variability contributed by the hematology result should be reduced by using the total WBC. In addition, we have conducted a brief trial with five normal samples to confirm the implications of equations 1 and 2. Each of five normal specimens was analyzed for CD4+ lymphocytes 3 times on a flow cytometer (Becton Dickinson Facscan) with the total lymphocyte population isolated with CD45 and gated. The fraction of + CD4 lymphocytes was calculated. The samples were then reanalyzed 3 times with only the total leukocyte population gated (using forward vs. side scatter), and the CD4+ lymphocytes were calculated as a fraction of the total leukocytes. Isolation of + CD4 cells was by dual fluorescence (CD4/CD3) in each case, and Becton Dickinson reagents were used throughout the study. A sample of each specimen was also analyzed on a hematology cell counter (Miles Diagnostics H* 1) ten times. Results for the concentration of absolute + CD4 lymphocytes were then obtained by each of the methods described above, and the C.V. for the two methods was calculated. Concentrations of CD4+ were similar by both methods, and replication showed that the imprecision of the flow cytometry results were similar whether the

6 VARIABILITY IN ABSOLUTE LYMPHOCYTE COUNTS 31 Absolute lymphocyte count Table 2 Duplicate Imprecision" Total WBC Method N Mean Value S.D. C.V. (Yo) Sv.xb Mean Value S.D. C.V. (Yo) SV.Xb CD H*l NE STKS Manualc N /A N /A NIA - "All values and S.D. in 1O911iter. bsy.x, root mean square of residuals when a linear regression analysis is performed on the duplicate results. It is a measure of the scatter of observation s about the line of best fit. 'Two hundred cell count multiplied by the WBC for each sample to obtain an absolute count > 8 4 3o 2 1 a WBC MEAN Ln - El2 m 1 = a o WBC CV.% Y = -2.9X Mean CV = 12.4 Mean CV = 12.4% t bo I Abs. LYMP MEAN FIG. 2. Analytic variability. Each data point represents the coefficient of variation (C.V.) across four methods of 523 single specimens. The results of a regression analysis between the C.V. and level is also indicated. a: C.V. across methods for total WBC vs. total WBC x lo9 cellsil. b: C.V. across methods for absolute lymphocytes vs. absolute lymphocytes x 1Og/liter. CD4+ lymphocytes were expressed as a fraction of lymphocytes or as a fraction of leukocytes, as shown in Table 3. The average C.V. for CD4+ concentration, as determined using the leukocyte fraction (method 2), was 5.% bo abs LYMP. CV.% FIG. 3. Distribution of variability. The frequency of occurrence of the C.V. across methods for 523 specimens is plotted vs. the C.V. a: Frequency distribution for total WBC. b: Frequency distribution for absolute lymphocytes. compared with 5.3% for CD4 concentration using the lymphocyte fraction (method 1). Thus, we propose that the total variability of absolute lymphocyte subset counts may be reduced significantly if the bridge between the results obtained with the flow cytometer and those obtained in hematology is made through the total WBC. This would simnlifv the gating nrocedure on the flow

7 32 SIMSON AND GRONER lo T STKS T O I I CD T lo N E lo T H"l - 1 A Hours After Draw 8 1 FIG. 4. Specimen stabilityitotal WBC. Each of four specimens was aliquoted at phlebotomy to each of four analyzers. Samples were stored at room temperature and analyzed at the tlmes indicated. Each point represents a single sampling.

8 VARIABILITY IN ABSOLUTE LYMPHOCYTE COUNTS 33 2 J CD t Hours After Draw 8 1 FIG. 5. Specimen stabilityiabsolute lymphocytes. Each of four specimens was aliquoted at phlebotomy to each of four analyzers. Samples were stored at room temperature and analyzed at the times indicated. Each point represents a single sampling.

9 34 SlMSON AND GRONER Table 3 Imprecision Comparison on Normal Specimens: Flow Cytometry Gating of Lymphocytes (Method I) vs. flow Cvtometric Gating of WBC (Method 2) Measurement Average C.V. (%) CD4 divided by lymphocyte 4.1 Lymphocytes cellsiliter x lo9 3.1 Concentration of CD4 5.3 Method 1 (a x b) CD4 divided by WBC 4.7 WBC cells/liter x lo9 1.8 Concentration of CD4 5. Method 2 (d x e) cytometer and eliminate the need for reagents used specifically to isolate lymphocytes. This hypothesis will have to be confirmed by more extensive studies of the equivalence of results including studies of abnormal specimens and interlaboratory trials. These studies are being pursued in our laboratory and will be reported in due course LITERATURE CITED Beers Y: Introduction to the Theory of Errors. Addison-Wesley, Reading. MA, Brando B, Sommaruga E: Nationwide quality control trial on lymphocyte immunophenotyping and flow cytometer performance in Italy. Cytometry 14:294-36, Buttarello M. Gedotti M, Lorenz C, Toffalori E, Ceschini N, Valentini A, Rizzotti P Evaluation of four automated hematology analyzers. Am J Clin Pathol , Calvelli T, Denny TN, Paxton H, Gelman R, Kagan J: Guideline for flow cytometric immunophenotyping: A report from the National Institute of Allergy and Infectious Disease, Division of AIDS. Cytometry 14:72-715, CDC recommendations for prophylaxis against Pneumocysfis curinii pneumonia for adults and adolescents infected with HIV. MMWR RR4, Vol 41, Centers for Disease Control: Guidelines for the performance of CD4+ T cell determination in persons with HIV infection. MMWR RR8 4l:l-17, Cohen AJ, Peerschke E, Stegbigel RT: A comparison of the Coulter STKS, Coulter S + IV and manual analysis of white blood cell differential counts in a human immunodeficiency virus infected population. Am J Clin Pathol 1:611, Cornbleet J, Myrick D, Judkins S, Levy R: Evaluation of the Cell-Dyn 3 Differential. Am J Clin Pathol 98:63, Cornbleet JR, Myrick D, Levy R: Evaluation of the Coulter STKS five part differential. Am J Clin Pathol , Data recap CAP, Skokie, IL, 1981, p Devreese K, DeLogi E, Francart C, Heyndrickx B, Phillipe J, LeRoux- Roels J: Evaluation of the automated hematology analyser Sysmex NE8. Eur J Clin Chem Clin Biochem , Fahey JL, Taylor JMG, Detels R, Hofmann B, Melmed R, Nishanian P. Giorgi JV: Prognostic value of cellular and serologic markers in infection with human immunodeficiency virus type I. N En@ J Med 322: , Gave HB, Henry K Measuring percent lymphocytes by flow cytometry to calculate absolute lymphocytes subset counts for HIV specimens. Cytometry 13: , Homburger HA, McCarthy R, Deodhar S: Assessment of interlaboratory variability in analytic cytology. Arch Pathol Lab Med I13: , International Committee for Standardization in Hematology: Protocol for evaluation of automated blood cell analyzers. Clin Lab Haematol , Koepke JA, Jones MS: Lymphocyte counting in HIV positive individuals. Sysmex J Int 2:71-74, Koepke JA, Landay AL: Precision and accuracy of absolute lymphocyte counts. Clin Immunol Immunopathol 52:19-27, Landay A: Putting alternative CD4 T cell enumeration methodologies in perspective. Symposium at the IX International Conference on AIDS, June, National Committee for Clinical Laboratory Standards: Clinical applications of flow cytometry: Quality assurance and immunophenotyping of peripheral blood lymphocytes. Tentative Guideline H42T. NCCLS, Villanova, PA, Revised classification system for HIV infection and expanded AIDS surveillance case definition for adolescents and adults. MMWR RR8, Vol 41, Prince HE, Lesar W: Simultaneous determination of absolute total lymphocyte and CD4+ lymphocytes in peripheral blood by flow cytometry. Am J Clin Pathol 92:26-29, National Committee for Clinical Laboratory Standards: Reference leukocyte differential count (proportional) and evaluation of instrumental methods. Approved Standard. NCCLS document H2A, Villanova, PA, Robertson EP, Lai NW, Wei DC: Analysis of leukocyte analysis on the Coulter STKS. Clin Lab Hematol 14:53-68, Robinson G, Robinson G, Morgan L, Evans M, McDermott S, Pereira S, Wainsbrough-Jones M, Griffin G: Effect of type of haematology analyser on CD4 count. Lancet 2:485, Rumke CL: The statistically expected variability in differential leukocyte counting. In: Differential leukocyte counting. Koepke JA (ed). Coll Am Pathol, Skokie, IL, pp 39-45, Van Lieuwen L, Eggels PH, Bullen JA: A short evaluation of a new hematological cell counter. Eur J Clin Chem Clin Biochem 29:15-11, Wen2 B, Burns E: The H'1 hematology analyzer: Its performance characteristics and value in the diagnosis of infectious disease. Am J Pathol Lab Med 111: , 1987.