| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
1
Institut für Laboratoriumsmedizin und Pathobiochemie, Universitätsklinikum Charité, Campus Charité Mitte, Medizinische Fakultät der Humboldt Universität zu Berlin, Schumannstrasse 20-21, 10117 Berlin, Germany.
a Author for correspondence. Fax 49-030-2802-8422; e-mail reinhard.ziebig{at}charite.de
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: We tested the hematology analyzer CellDyn 3500 (CD) and the urine flow cytometer UF-100 (UF), which are not designed for CSF analysis. We studied >104 samples with both analyzers, and the counts obtained were compared with the reference method (Fuchs-Rosenthal chamber).
Results: Good linearity in the medically relevant range of 15 x 106 to 1000 x 106 leukocytes/L and a high degree of within-run accuracy were seen for both analyzers. Cell counting on the UF was excellent, especially when low cell counts were encountered (CV, 4.9% compared with 28% observed for the CD). Method comparison showed that identical results could be detected for a majority of the count pairs. For a few samples, there was a discrepancy between the results from the analyzers and the counting chamber. In most cases, these were CSF samples containing a high proportion of lymphocytes. For these samples, the CD result led to a false-positive high leukocyte count, and on the UF these cells were not allocated to the leukocyte population, thus leading to false-negative counts.
Conclusions: Both analyzers should not be used for CSF cell counting in all cases at present. However, once the technical and software problems have been solved, routine use of the two analyzers for CSF analysis should be seriously contemplated.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The reason for the lack of more specific CSF diagnoses, including, e.g., a cell differentiation, often is insufficient sample volume and/or too few cells in a sample. Analyses performed with the modified sedimentation chamber technique according to Sayk (1) or with centrifugation (2) have contributed to the optimization of CSF cytology. They do not, however, solve the primary problem of the accurate determination of the number of cells.
Specific automated systems for CSF cell counting are not available at present. The following requirements, among others, would be necessary for such a system: (a) the ability to count small numbers of cells (e.g., 10 x 106 cells/L); (b) the ability to differentiate leukocytes into polymorphonuclear and mononuclear cell populations; and (c) the use of small sample volumes. On the basis of these requirements, we considered two analyzers appropriate for a test to determine cells in the CSF: the Abbott CellDyn 3500 (CD) and the Sysmex UF-100 (UF).
The automated hematology analyzer CD permits, according to the manufacturer, the counting and differentiation of leukocytes in <50 x 106 cells/L. This analyzer requires a sample volume of 150–200 µL. The urine flow cytometer UF was designed for the analysis of particles in urine and thus also of small numbers of leukocytes and erythrocytes. The sample volume required by this analyzer is 800 µL.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Routine CSF samples were brought from the ward on ice-water and were stored in the laboratory until the analyses were performed. The samples were obtained from patients after brain surgery (n = 76), patients suffering from viral and bacterial meningitis (n = 12) and/or encephalitis (n = 5), and patients in remission who had undergone control examinations (n = 11). Because only the material that was leftover from routine analyses was used, special consent from the patients was not required.
Initial analyses in defined cell suspensions prepared in our laboratory were used to check the accuracy and linearity. For this purpose, cell-rich plasma (buffy coat) was obtained from EDTA-blood samples after their spontaneous sedimentation, and its composition was checked on the H3 hematology analyzer (Bayer Diagnostics) in repeated tests. Thereafter, these samples were diluted with physiological saline solution to cell concentrations typical in the CSF in patients suffering from encephalitis and meningitis.
A special CSF pool was prepared to test the within-run imprecision in CSF. Ten samples were mixed and analyzed.
analyzers
The counting principles, specifications, and evaluations of the CD
and the UF have been published previously (3)(4)(5)(6)(7)(8). For
counting and differentiation of leukocytes, the CD uses a multiangle
polarization scatter separation technology in the optical channel
[white blood cell optical count (WOC)], combined with a second
channel with impedance count (white blood cell impedance count). The
erythrocyte counting is based on the impedance principle
(3)(4)(5). On the UF, cells in the urine are
determined by light scatter (small-angle and wide-angle scattering) and
the fluorescence of the cell membrane and the chromatin after staining
with phenanthridine and carbocyanine as well as by impedance
(6)(7)(8).
In preparation for CSF cell counting, the CD was flushed three times with saline to obtain cell counts of <3 x 106 cells/L. Only the leukocytes counted in the WOC channel were taken into consideration for the evaluation. All scattergrams including lobularity of 90 degrees and complexity of 10 degrees that showed the separation of polymorphonuclear cells from mononuclear cells were scrutinized carefully. In addition, the counts rate summary for leukocytes in the optical channel was printed. This indicated the stability of leukocytes.
As with the CD, the UF was flushed three times to obtain cell counts of <3 x 106 cells/L. The CSF samples were measured either directly or were first diluted with a physiological saline solution when the sample volumes were <800 µL. Samples were also prediluted when the number of cells (particles) counted was flagged by the analyzer as >40 000 particles.
interpretation of counting protocols
When interpreting the findings, it was helpful to take the
following alarms as well as some other results into account: For the
CellDyn 3500, "VARIANT LYM" indicates atypical lymphocytes and/or
pathological cells; "FWBC" indicates fragile white blood cells
(WBCs); "KWOC" indicates a kinetically corrected value of the
optical counting channel (WOC) caused by a nonlinear counting pulse
rate. For the UF-100, the only alarm that was considered was "Other
Particles", which indicates particles that can not be allocated to
any population in the scattergram.
statistical evaluation
The test results were evaluated by the regression method of
Passing and Bablok (9) as well as by demonstrating the count
difference of two methods in relation to the mean value of the two
value pairs, according to the Bland-Altman method (10).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
. The UF had surprisingly low within-run CVs (4.9–3.5%) for
the investigated cell suspensions (
= 28.9
x 106, 82.8 x 106,
and 387.4 x 106 WBC/L). The measurement in
the CSF pool produced a CV of 6.8% ( =
37.4 x 106 WBC/L). With the CD, the cell
suspensions with low leukocyte counts ( =
28.1 x 106 and 86.6 x
106 WBC/L) showed acceptable CVs (28% and 16%),
and the CV of the third test series was 3.3% (
= 363.7 x 106 WBC/L), identical to that of
the UF.
|
The linearity of leukocyte counting was carried out in serial dilutions
prepared from a defined cell suspension. Fig. 1
shows that of the leukocyte counts on the UF were linear even
in the lower range. Furthermore, the CD also produced leukocyte counts
with acceptable linearity in this range. The Cusum test produced no
significant deviation in both cases.
|
The results of the comparison of the methods used for the determination
of leukocytes in the CSF are shown in Fig. 2
. Leukocyte counts >1000 x 106/L
were not taken into account for statistical evaluation so that the
examinations were narrowed to the medically critical range for
decisionmaking.
|
The comparison of the CD, UF, and microscopic chamber counting methods
yielded correlation coefficients (r) between 0.87 and 0.91,
and the slope of the regression line in the comparison between CD and
UF was almost identical with the identity line (Fig. 2a
). In contrast,
the regression lines between the chamber and CD (Fig. 2b
) and the
chamber and UF (Fig. 2c
) methods show larger deviations from the
identity line. In Fig. 2c
, it is the slope of 1.35 (intercept =
-1.39 x 106 WBC/L) that is primarily
responsible for the difference in the leukocyte values between chamber
and UF. It should be critically stated that the extended range of
leukocyte counts (0 to 980 x 106 WBC/L)
makes the correlation look more favorable than it is.
A simple graphic presentation of pair differences in method comparison
(difference plot after Bland and Altman) clarifies this fact. Paired
values between CD and UF were chosen for this presentation. In Fig. 3
, the differences between the value pairs are depicted on the
y-axis and the mean is depicted on the x-axis.
The mean difference of the comparative counting as well as the standard
deviation (± 2 SD) of the differences is shown parallel to the
x-axis. From Fig. 3
, it is apparent that some of the value
pairs show larger differences in different directions. Based on these
differences, medical acceptance is impossible. Because the results
obtained with microscopic chamber counting were similar to the results
obtained with CD or UF, they are not shown here.
|
The following results obtained for one sample illustrate the problems with the validity of the cell counts and the impact of interfering factors on CSF cell counting with automated analyzers. The macroscopic appearance of this sample was without any apparent abnormalities. Cell counts yielded the following results for leukocytes: chamber counting, 150 x 106 WBC/L; CellDyn 3500, 215 x 106 WBC/L; UF-100, 24 x 106 WBC/L.
The protocols of the CD indicated an abnormal lymphocyte population with KWOC, VARIANT LYM, and FWBC flags. KWOC in this case means the mathematical extrapolation of the number of leukocytes with assumed cell degeneration during the count period. This mathematical correction algorithm leads to a higher number of leukocytes. The percentage of the lymphocyte population registered here was 97%.
The same sample yielded a lower leukocyte count (24 x
106 WBC/L) on the UF. The scattergram (forward
scattered light intensity vs fluorescent light intensity; Fig. 4
) shows in addition to the blue cluster representing the
leukocytes, a second cluster marked in yellow. These cells (169 x
106 WBC/L) are not assigned to any of the
populations defined in the UF scattergram. The addition of these
169 x 106 cells to the leukocyte count
would produce a corrected value of 193 x
106 WBC/L. A strictly defined yellow cluster in
the scattergram such as the one shown for this sample was always found
in the presence of atypical lymphocytes.
|
Six additional samples yielded similar results. Method comparisons
between the leukocyte value obtained with the chamber and the UF
revealed poor correlation (r = 0.20). A comparison of
the corrected leukocyte value and chamber counting showed good
correlation (r = 0.94). Fig. 5
shows the corresponding regression line for seven CSF samples
(circles).
|
To study the influence of lymphocytes on the leukocyte count obtained
with the UF, defined cell suspensions whose leukocyte concentration and
composition had been determined earlier on the H3 were analyzed on the
UF (Fig. 6
). We can demonstrate that an increase in the lymphocyte
population leads to reduced leukocyte counts. Typically, these
lymphocytes are not allocated to any defined cell population.
|
If one correlates leukocyte differences (between CD and UF, CD and
chamber, or UF and chamber) with the percentage of lymphocytes in the
CSF sample, the correlation coefficients (r) are between
0.42 and 0.71 (Table 2
). Evidence of the same correlation was found when cell
suspensions were used instead of CSF (Table 2
, values in parentheses).
|
Better correlations between CD and the counting chamber for CSF samples
high in lymphocytes were achieved when raw uncorrected values measured
on the CD were used. Fig. 5
shows the regression line resulting from
the comparison (squares) of CD (raw leukocyte values) and counting
chamber results for 20 CSF samples high in lymphocytes (50–97%). It
is evident from these results that in this way, it is possible to
considerably minimize the differences of the counts. This is, however,
a tedious way of obtaining correct results.
The evaluation of the erythrocyte counts obtained with the UF (y) and by chamber counting (x) showed a correlation coefficient (r) of 0.83 (n = 48) and the following regression line: y = 0.873x + 1.518. Erythrocyte values of >1000 x 106/L were not included in the analysis. Because the CD does not print erythrocyte values <1000 x 106/L, no correlation in the relevant erythrocyte range could be determined between the CD and the UF as well as between the CD and the counting chamber.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The high degree of the scattering of the residuals around the identity line and/or the high degree of the scattering of the count differences with a still acceptable mean difference (bias) manifests the fault liability of the automatic cell counting. One possible fault was indicated during analyses of CSF samples with high lymphocyte populations and was confirmed in analyses of defined cell suspensions obtained from sedimented EDTA samples.
As far as CD is concerned, there is a spurious extrapolation of the number of leukocytes in the WOC channel with a recorded high lymphocyte percentage of the CSF sample in most cases indicated by the alarms VARIANT LYM and KWOC. The CD performed this erroneous extrapolation of the leukocyte count when certain samples containing atypical lymphocytes were measured repeatedly.
Thus, it must be assumed that the algorithm of the CD is sensitive to certain conditions (CSF) and/or that is susceptible to certain coincidences. In some subsequent examinations, it was possible to show that in counting atypical lymphocytes, the rough counts of the WOC channel lead to a better conformity with the chamber counts.
The UF does not classify certain lymphocyte populations in CSF and allocates them to a different particle population ("Other"). The number of leukocytes that is determined is falsely low. In some cases, it was possible to correct the leukocytes count by simple addition (leukocytes x 106/L + number of cells x 106/L in the "Other" population). This method of correction, however, requires additional confirmation before being used.
If the leukocytes count determined for sample mentioned in the Results is considered from a medical point of view, the possible diagnoses may range from suspected herpes simplex encephalitis (24 x 106 WBC/L on the UF) to meningitis (215 x 106 WBC/L on the CD).
The difference shown in this specific case is reflected in the
described series of tests by a considerable number of clinically
relevant differences in the number of leukocytes. The critical
difference between two counts may be derived roughly from three times
the time-dependent standard deviations
(dk 
3 SDT).
With 100 x 106 WBC/L and a
SDT of 15 x 106
WBC/L, the resulting difference of almost 50 x
106 WBC/L would still be acceptable.
The use of automated analyzers for cell counting in CSF is certainly feasible. The material to be analyzed is handled by the fully automated analyzers, meeting time and quality requirements, and in an objective manner. There is a good correlation between the three methods studied (Fuchs-Rosenthal chamber, CD, and UF) in most cases. In low ranges, the UF performs better than the CD (CV, 4.9% as opposed to 28%). Difficulties will be encountered when fragile cells, especially lymphocytes, are present in the sample. In these cases, the Fuchs-Rosenthal chamber is regarded as the reference method (although it is not known how many cells are destroyed during the preparation procedure). Here, the leukocyte cell counts on the CD lead to falsely high values. In contrast, the UF leads to falsely low values. In both analyzers, the presence of the atypical fragile lymphocyte population can easily be suspected.
Regression analyses showed from the high degree of scattering of the data points around the line of identity that both analyzers do not offer a sufficient degree of safety for analyzing CSF as yet.
The high degree of accuracy and linearity that is offered by both analyzers should prompt us and the manufacturers to remedy the interfering factors as described by improving the algorithms these analyzers have to offer. Once this is done, these analyzers may be very useful for cell counts in CSF.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
1 Nonstandard abbreviations: CSF, cerebrospinal fluid; CD, CellDyn 3500; UF, UF-100 urine flow cytometer; WOC, white blood cell optical count; FWBC, fragile white blood cell; and KWOC, kinetically corrected white blood cells in optical channel.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  B. D. Conner, Y. C. G. Lee, P. Branca, J. T. Rogers, R. M. Rodriguez, and R. W. Light Variations in Pleural Fluid WBC Count and Differential Counts With Different Sample Containers and Different Methods Chest, April 1, 2003; 123(4): 1181 - 1187. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. T. Van Acker, J. R. Delanghe, M. R. Langlois, Y. E. Taes, M. L. De Buyzere, and A. G. Verstraete Automated Flow Cytometric Analysis of Cerebrospinal Fluid Clin. Chem., March 1, 2001; 47(3): 556 - 560. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
, theoretical leukocyte value vs leukocyte counts on the CD 3500
(y = 0.93x + 0.19); •, theoretical
leukocyte value vs leukocyte counts on the UF-100
(y = 0.99x + 1.31).
, value pairs of chamber and UF-100; (——–), regression line
between chamber and UF-100: y =
0.929x + 29.515; r = 0.94; n =
7. 
, value pairs of chamber and CD 3500; (- - - - -),
regression line between chamber and CD 3500: y =
1.005x + 6.572; r = 0.95; n =
20. (· · · · · · · ·), line of identity. n, number of
value pairs.