A Comparison of Differential White Blood Cell Counts Using
Manual Technic and the Coulter® S-Plus IV
ROBERT M. GREENDYKE, M.D., DONALD R. KANTER, PH.D., LOUISE D E B O O V E R , B.S., LINDA SAVAGE, B.S.,
AND SUSAN VANGELDER, B.S.
The results of 100-cell and 500-cell manual differential white
blood cell counts were compared with those obtained using a
Coulter® S-Plus IV electronic particle counter. Significant
(P < 0.001) correlations were observed between the manual
and instrument-derived data for percentages of granulocytes
and lymphocytes. (Key words: Differential white blood cell
counts; Coulter® S-Plus IV) Am J Clin Pathol 1985;
84: 348-350
TO VALIDATE the routine use of quantifying three
white blood cell (WBC) subpopulations using a Coulter®
S-Plus IV electronic blood cell counter as a replacement
for manual performance of differential WBC counts, a
study has been conducted to compare data produced by
the instrument and by conventional technic. Little has
been published to document the acceptability of the
substitution.
Methods
One hundred consecutive single Wright's-stained peripheral blood smears were prepared from routine specimens submitted from patients hospitalized in this institution. The blood samples came from patients with
psychiatric and/or generally chronic medical problems
representative of the hospital population as a whole,
and included no cases of primary hematologic disorder.
One hundred-cell and 500-cell manual differential WBC
counts were independently performed on the same
smears by each of three of the authors (LDB, LS and
SVG) who all are certified MT (ASCP) and have several
years experience as laboratory technologists. Differential
counts were performed in duplicate on the same blood
specimens on the Coulter S-Plus IV cell counter according to the manufacturer's directions.1 The instrument
distinguishes three WBC subpopulations (segmented and
band granulocytes, small lymphocytes, and large cells
including monocytes, large lymphocytes, and myeloid
cells younger than bands). To permit comparisons,
values derived from the manual counts on eosinophils,
basophils, band neutrophils, and segmented neutrophils
Laboratory Service, U.S. Veterans Administration
Medical
Center, Canandaigua, New York
were combined for comparison with instrument-generated granulocyte data. Manually determined monocyte
percentages were compared to instrument generated
mononuclear (large cell) values. The data were then
subjected to statistical study using the Pearson ProductMoment Correlation.2
Results
Granulocytes
100-Cell Count Determinations. The results of the
Pearson Product-Moment Correlation on granulocyte
data derived from 100-count manual determinations
and the Coulter Counter instrument are indicated in
Table 1. Each of the three manual counts showed highly
significant correlations to each other and to the Coulter
determination.
500-Cell Count Determinations. As indicated in Table
2, correlational analyses of 500-count manual determinations also revealed highly significant correlations between each of the manual counts, as well as between
manual counts and Coulter counts. In every case, the
500-cell count manual determinations showed higher
correlations than 100-cell count values.
Received November 1, 1984; received revised manuscript and
accepted for publication January 16, 1985.
Address reprint requests to Dr. Greendyke: Chief Laboratory Service,
VA Medical Center, Canandaigua, New York 14424.
348
Lymphocytes
100-Cell Count Determinations. Correlational analyses
of manual 100-cell count lymphocyte determinations
and the Coulter instrument count resulted in highly
statistically significant correlations, which are summarized in Table 3.
500-Cell Count Determinations. Again, 500-count determinations of lymphocyte percentages produced higher
correlations than did 100-count determinations. These
analyses are summarized in Table 4.
Mononuclear Cells
100-Cell Count Determinations. Analyses of manual
counts indicated significant correlations between the
BRIEF SCIENTIFIC REPORTS
Vol. 84 • No. 3
Table 3. 100-Cell Count Manual Determinations
and Coulter Instrument Count Percentages
for Lymphocytes
Table 1. 100-Cell Count Manual Determinations
and Coulter Instrument Count Percentages
for Granulocytes
Tech 1
Tech 1 versus Tech 2
Tech 1 versus Tech 3
Tech 1 versus Coulter
Tech 2
Tech 2 versus Tech 3
Tech 2 versus Coulter
Tech 3
Tech 3 versus Coulter
Coulter
Mean
SD
67.35
11.77
r
0.87*
0.88*
0.81*
68.54
11.38
0.88*
0.85*
69.48
11.83
0.86*
73.19
349
9.43
• /'<o.obi.
Tech 1
Tech 1 versus Tech 2
Tech 1 versus Tech 3
Tech 1 versus Coulter
Tech 2
Tech 2 versus Tech 3
Tech 2 versus Coulter
Tech 3
Tech 3 versus Coulter
Coulter
Mean
SD
27.83
10.97
r
0.87*
0.85*
0.83*
27.38
10.73
0.86*
0.88*
25.96
11.14
21.90
8.73
0.86*
* P< 0.001.
three technologists (see Table 5); however, only one of
the three technologists' determinations showed a significant correlation with the Coulter instrument count.
500-Cell Count Determinations. All correlations between the technologists and between technologists and
machine counts were significant for 500-cell count determinations. However, the technologists demonstrated
a significantly higher level of agreement with each other
than with the instrument count (see Table 6).
Discussion
The difficulty in attempting to evaluate the validity
of differential WBC counts produced by an electronic
particle counter by comparing them with manually
generated results is readily apparent. The two methods
use different criteria for leukocyte identification, and the
Coulter instrument categorizes the cells into only three
groups. The problems associated with manual WBC
differential counts are well known, and include nonrandom distribution of cell types on the smear, possibly
Table 2. 500-Cell Count Manual Determinations
and Coulter Instrument Count Percentages
for Granulocytes
Mean
SD
Tech 1
Tech 1 versus Tech 2
Tech I versus Tech 3
Tech 1 versus Coulter
68.25
10.71
Tech 2
Tech 2 versus Tech 3
Tech 2 versus Coulter
68.65
Tech 3
Tech 3 versus Coulter
68.56
10.50
10.81
0.97*
0 97*
0.89*
0 97*
0.91*
0.90*
inadequate sample size counted, and generally minor
difficulties with cell identification.3 The Coulter S-Plus
IV counts 20,000 cells, thereby minimizing error variance
associated with sampling limitations, but substitutes a
potentially larger problem in cell identification by virtue
of using only cell size as a criterion.
The present study indicated highly significant (P
< 0.001) correlations among the data produced manually
by three technologists and also between manual and
instrument-generated granulocyte and lymphocyte percentages. Five hundred-cell differential counts yielded
higher correlation coefficients between manual and instrument data for granulocyte and lymphocyte percentages than did 100-cell counts, suggesting that sample
size may have contributed to the differences. Not unexpectedly, in light of the different cell identification
criteria, the manual monocyte vs. instrument mononuclear cell percentages correlated poorly.
There are obvious limitations to the use of three-
Table 4. 500-Cell Count Manual Determinations
and Coulter Instrument Count Percentages
for Lymphocytes
Tech 1
Tech 1 versus Tech 2
Tech 1 versus Tech 3
Tech 1 versus Coulter
Tech 2
Tech 2 versus Tech 3
Tech 2 versus Coulter
Tech 3
Tech 3 versus Coulter
Coulter
*/>< 0.001.
Mean
SD
27.19
9.95
27.31
9.85
27.78
10.34
0.96*
0.96*
0.90*
0.95*
0.91*
0.89*
21.90
8.73
350
GREENDYKE ET AL.
Table 7. Criteria Based upon Instrument Printout for
Manual Repetition of Automated WBC Differential
Counts and Frequency of Occurrence of Each
in 500 Consecutive Tests*
Table 5. 100-Cell Count Manual Determinations
and Coulter Instrument Count Percentages
for Mononuclear Cells
Tech 1
Tech 1 versus Tech 2
Tech 1 versus Tech 3
Tech 1 versus Coulter
Tech 2
Tech 2 versus Tech 3
Tech 2 versus Coulter
Tech 3
Tech 3 versus Coulter
Coulter
Mean
SD
4.8
2.49
r
2.3
0.38t
0.16 NS
4.61
3.0
4.82
1.59
0.17 NS
• /' < 0.05.
t/"< 0.001.
population differential WBC counts as produced by the
Coulter S-Plus IV instrument, especially for hospital
inpatients. It may be argued that stained peripheral
blood smears should be manually scanned in all cases,
and this currently is routine in the authors' laboratory.
Such a practice, however, vitiates some of the advantage
of automated blood cell counting. It is beyond the scope
of this paper to discuss the relative merits of data
produced by the Coulter instrument relative to red blood
cell indices and relative distribution width vs. traditional
estimates of hemoglobinization and anisocytosis, or of
platelet count, mean platelet volume, and platelet distribution width vs. estimates from blood smears. That
there are WBC distribution abnormalities that are missed
without manual examination is readily apparent, two of
Table 6. 500-Cell Count Manual Determinations
and Coulter Instrument Count Percentages
for Mononuclear Cells
Tech 1
Tech 1 versus Tech 2
Tech 1 versus Tech 3
Tech 1 versus Coulter
Tech 2
Tech 2 versus Tech 3
Tech 2 versus Coulter
Tech 3
Tech 3 versus Coulter
Coulter
• P < 0.001.
t/'<0.0l.
t P < 0.05.
Mean
SD
4.44
2.06
r
0.69*
0.72*
0.26t
3.99
Percentage Frequency
of Occurrence
Criteria
0.42t
0.34f
0:25*
4.08
AJ.C.P. .September 1985
Increased total WBC
Decreased total WBC
Increased granulocyte percentage
Decreased granulocyte percentage
Increased lymphocyte percentage
Increased mononuclear cell percentage
"Valley failure"
43
6
44
6
16
13
14
• Thirty-two percent of counts requiring manual repetition met two or more criteria. For
discussion of "valley failure" see text. In addition, a very small number of additional differential
counts were repeated as a result of eosinophilia, basophilia, or band neutrophilia as isolated
findings on routine scanning of peripheral blood smears from samples not displaying any of the
other abnormalities listed above.
the most significant being the presence of band neutrophilia and eosinophilia, although it is acknowledged that
the former usually is associated with granulocytosis. The
cost effectiveness of routinely screening for such abnormalities has not been assessed.
The criteria established in this institution for manual
repetition of the automated WBC differential count are
listed in Table 7, together with the frequency of occurrence of each in 500 consecutive blood samples. The
phenomenon of "valley failure" with the use of the
Coulter instrument requires comment. In 14% of the
present cases, the instrument was unable to discriminate
between the mononuclear and lymphocyte populations
because of a continuous cell size gradient resulting from
the presence of numerous large lymphocytes. The absence
of a "valley" between the cell size distribution peaks
(whence the term "valley failure") precluded enumeration
of the two types of cells and required manual counting.
A sampling of 1,591 random differential WBC counts
performed in the 12 months during which the above
criteria have been used indicates that 58.6% have required
manual repetition. Clinical acceptance of data so generated has been excellent.
Acknowledgment. Roy Johnston, Ph.D., and B. W. Lundy, Ph.D.,
performed the initial statistical analysis of the data presented.
1.75
0.68*
0.27t
3.85
2.37
4.82
1.59
0.21J
References
1. Coulter Counter Model S-Plus IV with three-population differential
product reference manual. Hialeah, FL, Coulter Electronics,
October 1983
2. Hays W: Statistics for the Social Sciences. Second edition. New
York, Holt, Rhinehart and Winston, 1973, p 616
3. Nelson DA, Morris MW: Basic methodology, Clinical Diagnosis
and Management by Laboratory Methods. Edited by JB Henry.
Edition 17. Philadelphia, Saunders, 1984, pp 611-612