Body Fluid Cell Counts by Automated Methods
Body Fluid Cell Counts by Automated Methods
Body Fluid Cell Counts by Automated Methods
Automated Methods
Linda M. Sandhaus, MD, MS
KEYWORDS
Cell counting Body fluids Automated body fluid analysis Hematology analyzers
KEY POINTS
The introduction of body fluid modes on automated hematology analyzers is an important
advance in automated body fluid analysis.
The current limits of precision for low white and red blood cell counts pose barriers to
wider acceptance of automated cell counts for cerebrospinal fluid.
Improved analyzer flagging algorithms for abnormal cell distributions in body fluid samples
will help define criteria for microscopic review.
The impact of automated body fluid analysis on laboratory efficiency depends on many
factors and will vary among laboratories.
Department of Pathology, University Hospitals Case Medical Center, 11100 Euclid Avenue,
Cleveland, OH 44106-3205, USA
E-mail address: [email protected]
Abbreviations
BF Body fluid
CBC Complete blood count
CNS Central nervous system
CSF Cerebrospinal fluid
CV Coefficient of variation
ED Emergency Department
LOQ Limit of quantitation
MN Mononuclear cells
PMN Neutrophils
QC Quality control
RBC Red blood cell
TAT Turnaround time
TNC Total nucleated cell count
WBC White blood cell
Table 1
Features of body fluid modes on some hematology analyzers
Feature Effect
Short time to transition between modes Improves turnaround times
No sample dilution requirement Improves accuracy and efficiency
Reports TNC and WBC Improves accuracy of WBC
Differential count excludes tissue cells Improves accuracy of differential counts
Flags samples with abnormal cellular Improves detection of malignant cells
distributions
Reports RBC counts to more significant Improves accuracy for low RBC counts and
figures than CBC mode assessment of traumatic taps
Extended counting for low cell counts Improves accuracy for low cell counts
Abbreviations: CBC, complete blood count; RBC, red blood cell; TNC, total nucleated cell count;
WBC, white blood cell count.
Accuracy
Accuracy is typically evaluated by demonstrating comparability of results obtained by
the new method and the reference method using split patient samples. However, this
approach is fundamentally flawed for the evaluation of automated BF methods
because the new (automated) method is being compared with a reference (manual)
method that is very imprecise. Recent College of American Pathologists proficiency
surveys for manual BF cell counts have demonstrated coefficients of variation (CVs)
of 20% to 40% over the range of WBC and red blood cell (RBC) counts assayed,
with the highest CVs being obtained at the lower cell counts.30 For this reason, the
manual method cannot be considered a gold standard in the statistical analysis.
Large CVs for either method will yield lower correlation coefficients in method com-
parison studies.31 The measures of comparability that are commonly used in labora-
tory correlation studies are the Pearson correlation coefficient and ordinary linear
regression. These parametric statistical methods are based on assumptions that the
data and the error in each method have normal (bell-shaped) distributions. Use of
these parametric statistical methods should be discouraged for BF method compar-
isons, because often the data do not meet these assumptions. Passing-Bablok
regression, Deming regression, Spearman correlation, and Bland-Altman difference
plots are more appropriate measures of correlation and methodological bias, because
these nonparametric statistical methods are not based on assumptions about the dis-
tribution of the data or the error in each method. There is no well-defined value for the
correlation coefficient that indicates an acceptable correlation, as this depends on
96 Sandhaus
how the test will actually be used in clinical practice. Each laboratory must decide if
the correlation is satisfactory for their use.
Another challenge in performing correlation studies is that sample integrity deterio-
rates over time. Therefore, it is important for the split samples to be analyzed in close
time proximity to minimize differences caused by cellular degeneration. Assay of com-
mercial control material and recovery of expected values within predefined limits of
precision can also be used to assess accuracy. However, commercial control mate-
rials can only supplement, not replace patient samples, as they do not have the
same matrix effects or other variability that occurs in actual patient samples.
Analysis of the correlation data may indicate statistically significant differences
between the 2 methods for some BF types, certain cell types, or a particular range
of values. It is important to ask whether these differences are also clinically significant
and likely to affect patient care. For example, if the 2 methods produce statistically
significant differences in the range of cell counts that are near the upper limit of the
existing reference ranges for cerebrospinal fluid (CSF), some samples might be clas-
sified as normal by the reference method and abnormal by the automated method.
The apparent misclassifications by the automated method could affect patient diag-
nosis and management.12
WBC differential counts should only be performed on hematology analyzers that have
a dedicated BF mode for this purpose.
Automated differential counts must be validated against manual differential counts,
preferably those performed on cytospin smears.33–35 At present, BF modes offer
limited differential counts, consisting of PMN and MN cells. Depending on the analyzer
software, other cell types might be included in the PMN and MN categories (eg, eosin-
ophils and basophils might be included in the PMN category and lymphocytes in the
MN category). To obtain a valid measure of correlation, these combined cell cate-
gories must be taken into account when correlating automated 2-part BF differential
counts with the manual differential counts performed on Wright-Giemsa–stained cyto-
spin smears. In addition to the usual limitations of manual differential counts (impreci-
sion, subjectivity), time delays and cytocentrifugation will affect cell recovery and may
affect the proportion of cell types that are present in the split samples.21,36 Recent
studies have shown that for BFs with WBC greater than 10 cells/mL, automated
2-part differential counts correlate well with manual differential counts.3,10,14,15,37
Some of these studies showed an apparent positive bias for PMNs, especially when
the WBC is less than 20 cells/mL. A possible explanation is that there is dispropor-
tionate loss of the more labile PMNs over time and with cytocentrifugation, which
will decrease the proportion of PMNs on the cytospin smear. Different laboratories
may reach different conclusions about which BF samples are appropriate for auto-
mated differential counts.
Although current technology may be satisfactory for WBC differential counting in
some BF samples, microscopic examination is still indicated for the detection of
malignant cells. Leukemic blasts, lymphoma cells, and metastatic carcinoma may
be present in low numbers in CSF, and even samples with normal cell counts may
contain malignant cells. Some recent studies suggest that flags generated by
abnormal cell distributions in the BF mode might be useful in screening samples for
malignant cells.21,36,38 Therefore, method validation studies should also include an
evaluation of the flagging capabilities of the BF mode. Most laboratory technologists,
though skilled in WBC differential counts, have limited training in the detection of non-
hematopoietic malignant cells.39 By contrast, cytopathology laboratories are staffed
by cytotechnologists whose primary concern is the detection of malignant cells. How-
ever, when malignancy is not clinically suspected and the samples are not submitted
to cytopathology, the hematology laboratory may have the first opportunity to detect
malignant cells. Therefore, reliable flagging to signify the presence of atypical cells
could potentially improve the detection of malignancies in hematology laboratories.
The different BF types pose different challenges for automated analysis. CSF and
synovial fluid are normally present in healthy persons, whereas serous cavity effusions
represent pathologic conditions. Therefore, reference ranges are important in the
interpretation of CSF and synovial fluid cell counts, but are of less value in the interpre-
tation of pathologic effusions.
Cerebrospinal Fluid
CSF poses the greatest challenges for automated cellular analysis. CSF is examined
for a variety of medical indications, and most patient samples are normal. Therefore,
method correlation studies tend to be skewed toward extremely low cell counts.
Imprecision is higher for both manual and automated methods when cell counts are
low. As shown in Fig. 1A, a range of automated WBC values may be obtained for
98 Sandhaus
Fig. 1. Manual hemocytometer white blood cell (WBC) counts versus automated WBC
counts on 200 clear cerebrospinal fluid samples. (A) Expanded scale for manual cell counts
ranging from 0 to 10 cells/mL. Spearman correlation coefficient: r 5 0.47. (B) Complete data
set. Spearman correlation coefficient: r 5 0.77. (Adapted from Sandhaus L, Ciarlini P,
Kidric D, et al. Automated cerebrospinal fluid cell counts using the Sysmex XE-5000. Am J
Clin Pathol 2010;134:736; with permission.)
each increment in the manual cell count. This nonlinear relationship between the 2 sets
of values will yield a weak correlation coefficient (Spearman r 5 0.47), even though a
strong correlation (Spearman r 5 0.77) may be obtained when the full range of data is
included in the analysis (see Fig. 1B). The reason for this is that a few samples with
high cell counts can exert a disproportionate effect on the correlation coefficient.
Body Fluid Cell Counts by Automated Methods 99
Furthermore, as RBCs and WBCs are not normal constituents of CSF, the accepted
reference ranges are extremely low and generally exceed the limits for which the
analyzers were originally intended. Published reference ranges were established by
manual methods and are widely accepted, despite poor documentation of how the
reference range studies were done and uncertainty about pediatric reference
ranges.35 As mentioned earlier, several correlation studies with automated analyzers
have shown a positive bias in CSF samples when WBC counts are extremely low.
The reasons for the observed positive bias by automated methods are not clear,
but various explanations have been suggested, including “electronic noise” and inclu-
sion of cellular debris or other particulate debris in the cell counts.10,15 Therefore,
some samples that have normal manual cell counts may appear abnormal by auto-
mated methods. For this reason, some investigators have recommended that samples
with low cell counts should be excluded from automated analysis,12 whereas others
have suggested that new reference ranges for automated methods may need to be
determined.22 Until the causes of the discordant results are better understood, it re-
mains unclear as to which results are more accurate.
Enumeration of RBCs by automated analyzers poses another set of technical and
interpretive difficulties for CSF. RBC counts are useful in the diagnosis of intracranial
hemorrhage, and for determining whether a traumatic tap has contaminated the
sample with peripheral blood and has potentially introduced blood into the CNS.
For an acute intracranial hemorrhage, the RBC count is generally high enough to
be enumerated by the analyzer with adequate precision for clinical interpretation.
However, most hematology analyzers do not report RBC counts at the low concen-
trations that might exist with mildly traumatic tap. The clinical significance is that
peripheral blood introduced into the sample by a traumatic tap might lead to misin-
terpretation of the source of the leukocytes, such as neutrophils or leukemic blasts.
The traditional approach has been to calculate the ratio of RBC/WBC, with a ratio
greater than 500/mL favoring peripheral blood contamination.35,40,41 However, if the
RBC count falls below the lowest reportable value of the analyzer, either no result
or a result of zero might be produced. There is no widely accepted agreement on
the number of RBCs that signify a traumatic tap. Some pediatric oncology protocols
use 10 RBCs/mL as an indicator of peripheral blood contamination,42–44 although
most laboratorians would consider this value to be unrealistically low. Another pedi-
atric study used 1000 cells/mL as the cutoff for traumatic tap and determined that for
CSF profiles with RBC counts of less than 10,000 cells/mL, adjustments to the CSF
WBC count were not warranted and that blood in the CSF was not sufficient to
explain an increased WBC.45 Establishing criteria for defining traumatic tap by auto-
mated RBC counts would help promote clinical acceptance of this methodology,
especially among pediatric oncologists, who rely on standardized protocols for
defining CNS involvement by leukemia.
Serous Fluids (Pleural, Peritoneal, Pericardial, Peritoneal Lavage, and Dialysate Fluids)
The reporting of cell counts on serous BFs has been inconsistent and confusing. The
inconsistency arises from different approaches to the mesothelial cells that are
normally present in these fluids and can be numerous. Some laboratories include
these cells in their hemocytometer cell counts and report a total nucleated cell count
(TNC), whereas others exclude them. Some laboratories include mesothelial cells in
the differential count as MNs, whereas others exclude them. There is similar inconsis-
tency in the reporting of macrophages. Automated analyzers with BF modes use
gating strategies to exclude tissue cells from the WBC count. The WBC differential
count produced by the BF channel will include only those cells that fall within the
100 Sandhaus
Synovial Fluid
Few studies have evaluated automated methods for cell counting in synovial fluid
samples. Cell counts are generally higher, and there is better agreement between
WBC and PMN counts when samples are pretreated with hyaluronidase.11,48 Pretreat-
ment with hyaluronidase might also help prevent clogging of the flow cells in some
analyzers. There are no reference ranges for RBCs in synovial fluid, and a WBC count
of less than 150 cells/mL is generally considered normal. The potential for interference
with automated cell counts by the presence of crystals, fat globules, or microorgan-
isms should be carefully evaluated.
QUALITY CONTROL
Quality control (QC) material should be analyzed in the same manner whereby the BFs
will be processed and analyzed. For automated analysis, this means that the QC
material must follow the same fluidics pathways and be analyzed with the same
software algorithms used for actual BF samples. QC should also include performance
of background counts of the fluidic system, and any additional diluents or reagents
that are used for BF analysis that are not used in the analysis of blood samples.
multiple analyzers, and a high BF sample volume, the option to designate one analyzer
for BF analysis might improve efficiency without interrupting the CBC workflow.
The potential cost savings with automated BF analysis are largely attributable to
reduced labor costs. One study calculated a mean cost of V1.22 for the automated
method compared with V6.74 for the manual method, largely attributable to savings
in technologist time.25
Some hematology analyzer manufacturers have clearly taken the lead in automated
BF cellular analysis. This discussion has attempted to highlight both the advantages
and the limitations of current automated methodology. The introduction of BF modes
offers the potential to improve accuracy and reporting of BF cell counts and differential
counts, and to standardize performance, which will make these results more clinically
useful. However, there is a need to validate reference ranges for BF cell counts
performed by automated methods, especially for CSF. Multi-institutional studies
would be helpful to determine whether reference ranges can be standardized across
multiple analyzer platforms.
Improvements in BF mode flagging capabilities may lead to improved detection of
malignant cells in the hematology laboratory.
REFERENCES
30. College of American Pathologists. Hemocytometer fluid count (HFC) and auto-
mated body fluid (ABF2) surveys. Participant Summaries; 2007-2009.
31. Chinchilli VM, Gruemer HD. The correlation coefficient in the interpretation of
laboratory data. Clin Chim Acta 1994;229:1–3.
32. Omuse G, Makau P. Interference of cerebrospinal fluid white blood cell counts
performed on the Sysmex XT-4000i by yeast and bacteria [letter to the editor].
Int J Lab Hematol 2013;35:e5–7.
33. Galagan KA, Blomberg D, Cornbleet PJ, et al. Color atlas of body fluids: an illus-
trated field guide based on proficiency testing. Northfield (IL): College of Amer-
ican Pathologists; 2006.
34. Stokes BO. Principles of cytocentrifugation. Lab Med 2004;35(7):434–7.
35. Kjeldsberg C, Knight J. Body fluids: laboratory examination of cerebrospinal,
seminal, serous, and synovial fluids. 3rd edition. Chicago: ASCP Press; 1993.
36. Zimmermann M, Otto O, Gonzalex J, et al. Cellular origin and diagnostic signifi-
cance of high-fluorescent cells in cerebrospinal fluid detected by the XE-5000
hematology analyzer. Int J Lab Hematol 2013;35:580–8.
37. Labaere D, Boeckx N, Geerts I, et al. Validation and implementation of the
Sysmex XN 2000 body fluid module for white blood cell differential count in
serous fluids. Int J Lab Hematol 2014;36(Suppl 1):1–136.
38. Labaere D, Boeckx N, Geerts I, et al. Detection of malignant cells in serous body
fluids by counting high-fluorescent cells on the Sysmex XN2000. Int J Lab Hem-
atol 2014;36(Suppl 1):1–136.
39. Jerz J, Donohue R, Mody R, et al. Detection of malignancy in body fluids: a com-
parison of the hematology and cytology laboratories. Arch Pathol Lab Med 2014;
138(5):651–7.
40. Bonadio WA, Smith DS, Goddard S, et al. Distinguishing cerebrospinal fluid
abnormalities in children with bacterial meningitis and traumatic lumbar puncture.
J Infect Dis 1990;162:251–4.
41. Novak RW. Lack of validity of standard corrections for white blood cell counts of
blood-contaminated cerebrospinal fluid in infants. Am J Clin Pathol 1984;82:95–7.
42. Burger B, Zimmermann M, Mann G, et al. Diagnostic cerebrospinal fluid examina-
tion in children with acute lymphoblastic leukemia: significance of low leukocyte
counts with blasts or traumatic lumbar puncture. J Clin Oncol 2003;21(2):184–8.
43. Gajjar A, Harrison P, Sandlund J, et al. Traumatic lumbar puncture at diagnosis
adversely affects outcome in childhood acute lymphoblastic leukemia. Blood
2000;96(10):3381–4.
44. te Loo DM, Kamps W, van der Does-van den Berg A, et al. Prognostic signifi-
cance of blasts in the cerebrospinal fluid without pleiocytosis or a traumatic
lumbar puncture in children with acute lymphoblastic leukemia: experience of
the Dutch Childhood Oncology Group. J Clin Oncol 2006;24(15):2332–6.
45. Byington CL, Kendrick J, Sheng X. Normative cerebrospinal fluid profiles in
febrile infants. J Pediatr 2011;158(1):130–4.
46. Piraino B, Ballie GR, Bermardini J, et al. Peritoneal dialysis-related infection
recommendations; 2005 update. Perit Dial Int 2005;2005:107–31.
47. Runyon B. Management of adult patients with ascites due to cirrhosis: an update.
Hepatology 2009;49:2087–107.
48. Sugiuchi H, Ando Y, Manabe M, et al. Measurement of total and differential white
blood cell counts in synovial fluid by means of an automated hematology
analyzer. J Lab Clin Med 2005;146:36–42.