PRINCIPLES OF AUTOMATED BLOOD CELL ANALYSIS
Automated blood cell analysis is the cornerstone of the modern hematology laboratory, allowing rapid, cost-effective, and accurate analysis of the cells of the blood, including new parameters with diagnostic utility. The morphologic and functional complexity of blood cells requires direct microscopic examination of a stained blood film by a trained observer. However, it is possible to use automated techniques to analyze and report on the majority of samples, using defined criteria (“flags”) to select those that need further microscopic review. Automated hematology analyzers typically incorporate multiple proprietary software flags based on acceptability criteria related to pattern recognition in the multiparameter displays or comparison of different detection modes for the same cell type. These are frequently updated in software or when new models are introduced to improve sensitivity and specificity. In this way, instruments identify samples that contain cells or abnormalities the instrument cannot definitively identify, so that a skilled morphologist can visually evaluate that specimen. Some of these flags can be adjusted or suppressed by the user to achieve an appropriate balance that minimizes both false positives and false negatives. The optimum balance is dependent on the patient population examined. Guidelines for manual smear review based on comparative data have been published, based on instruments then in common use.1 Protocols for evaluating and adjusting flagging criteria within an individual laboratory have been described.2 Manual review may consist of a scan of the blood film, a more detailed blood film examination including leukocyte differential count, or a physician’s review, based on laboratory defined criteria.3 The proportion of samples requiring manual blood film review differs among instruments and the type of patient population tested. Studies show a 10 to 30 percent manual review rate,4,5,6 with a false negative rate (i.e., abnormal samples that were not flagged for review) varying from approximately 3 percent1 to 10–14 percent.4 Most of the false negatives with current instrumentation are related to red cell and platelet morphology with relatively limited diagnostic significance.4 Continued improvement in methodology and increased sophistication of data analysis will result in further reduction of unnecessary manual smear reviews. Depending on workload and space considerations, laboratories may choose to link automated hematology analyzers with automated blood film preparation and automated image analyzers to facilitate manual morphologic review of cells by traditional light microscopy or online review of digitized images.7 These instruments can provide a provisional differential count with good accuracy,8 although typically final classification of problematic cells is performed by a technologist or physician.
The characteristics of automated hematology analyzer systems have been reviewed.9 A detailed description of individual instruments is beyond the scope of this chapter, but the general principles employed by state-of-the-art instrumentation are summarized below. The major analytical challenges are the frequency of the different cell types, which vary over many orders of magnitude, from red cells (millions per μL) to basophils (dozens per μL), and the complexity of the structure of normal and abnormal blood cells. Over the past several decades, instruments have become increasingly sophisticated with the use of multiple parameters to produce more precise results in the great majority of patient samples. In a typical automated hematology analyzer, the blood sample is aspirated and separated into different fluidic streams. The streams are mixed with various buffers that accomplish specific purposes in the analysis, for instance, using differential lysis to distinguish subsets of leukocytes, reagents to measure hemoglobin or detect myeloperoxidase containing leukocytes, and various fluorescent dyes. Measurements of each fluidic stream are made in flow as the sample passes through a series of detectors in what are essentially modified flow cytometers (Chap. 3). Commonly used principles include light scatter at various angles, electrical impedance and conductivity, and fluorescence or light absorption of cells stained in flow. Light scatter yields information about cell size (using scatter at low-incident angles), nuclear lobulation, and cytoplasmic granularity (using high-angle light scatter) and refractive index, with polarization of the scattered light as an additional parameter. If red cells are converted to spherocytes by the buffer solution to eliminate the variability of cell shape, light scatter at different angles can provide information about hemoglobin content, as well as size of individual red cells. Cell size is also estimated by measuring change in electrical resistance, which is proportional to cell size as cells enter a narrow orifice through which a direct current is maintained, the original Coulter principle, named for Wallace Coulter who developed the electronic particle counter.10 Radiofrequency capacitance measurement yields additional intracellular structural information that complements the direct current measurement. Differential lysis with detergents of varying strength or pH is used to separate certain leukocyte types, such as basophils and immature granulocytic cells, from the major normal blood cell types. In addition, nucleic-acid-binding fluorescent dyes incorporated into the lysis buffer measure total RNA plus DNA in the cells and are used in some analyzers to help differentiate leukocyte types. Fluorescence measurements after staining with RNA binding dyes are commonly used to detect and subclassify reticulocytes and platelets. Light absorption is the principle used for hemoglobin measurement and in some instruments for identifying peroxidase-positive granulocytes. Instruments rely on a combination of techniques for accuracy and precision11 (Fig. 2–1). Complex algorithms are invoked to determine whether the distribution of variables for a specific result or for the specimen as a whole fit sufficiently within an expected variable space so that the results can be reported with high confidence, or whether the specimen should be “flagged” for further analysis or manual blood film review (Fig. 2–2). There is significant overlap in methodology between automated hematology analyzers and flow cytometers (flow cytometers are discussed in Chap. 3). The latter are distinguished by extensive use of fluorochrome tagged antibodies to identify cell subtypes. These instruments have replaced laborious manual work, but also demand increasing interpretation skills on the part of laboratory technologists. Automated blood analyzers have been adapted to accurately count the smaller numbers of blood cells typically found in body fluids,12 but accurate differential counts13 and detection of blast cells in fluids of patients14 remains a challenge.
Schematic of multiparameter cell discrimination in an automated hematology analyzer. The Sysmex XE-2100 is used as an example, in which leukocytes are discriminated by (A) DNA/RNA fluorescence using a polymethine dye versus high-angle (side) light scatter in lysed blood; (B) side scatter versus low-angle (forward) light scatter after acidic lysis in a separate aliquot that preserves basophil structure; and (C) direct current (DC) impedance versus radio frequency (RF) capacitance of cells subjected to a lysis reagent that relatively preserves immature cells with lower membrane lipid content. Nucleated red blood cells (NRBC) are distinguished (D) in a lysed sample stained with nucleic acid dye where leukocyte nuclei have detectably higher DNA/RNA content than red cell nuclei. Atyp Lymph, atypical lymphocytes; Baso, basophils; Blasts, blast cells; Diff Channel, differential count channel; Eos, eosinophils; HPC, hematopoietic progenitor cells; Imm Gran, immature granulocytes; Lymph, lymphocytes; Mono, monocytes; Neut + Baso, neutrophils + basophils; Plt Clumps, platelet clumps; WBC, white blood cells.
Examples of how samples containing various abnormal findings are flagged for manual review. A. Normal sample showing how the major variables and results are displayed. B. Immature granulocytes appearing on the DIFF (leukocyte differential count) and IMI (immature myeloid) histograms, as well as a dimorphic red cell population. C. Multiple flags, including cells in the area of atypical lymphocytes, and platelet clumps with abnormal platelet volume distribution. D. Appearance of nucleated red blood cells (NRBCs), reticulocytes, and reticulated platelets on a different set of parameters. This figure is not intended as a comprehensive illustration of the technical details, but serves to demonstrate that differential lysing reactions coupled with multiparameter light-scatter, impedance, capacitance, and fluorescence measurements are used to analyze blood cells in current high-throughput instruments.
Point of care “bedside” testing is far more challenging in hematology than for typical clinical chemistry analytes for many of the reasons described above. Instruments have been described for bedside measurement of hemoglobin, total leukocytes, three-part leukocyte differential count, malaria parasitemia, and CD4+ T-cell count, mainly targeting clinical settings with limited access to standard laboratory testing. More work remains to be done to demonstrate the reliability and clinical impact of such testing strategies.15
AUTOMATED ANALYSIS OF RED CELLS
Some red cell parameters (for instance, mean cell volume [MCV], red cell number, hemoglobin concentration, red cell distribution width [RDW]) are directly measured, while others (for instance, hematocrit, mean cell hemoglobin [MCH], mean cell hemoglobin concentration [MCHC]) are derived from these primary measurements.
Measurement of the Red Cell Count and Hematocrit
In electronic instruments, the hematocrit (Hct; fractional volume of blood occupied by erythrocytes) is calculated from the product of direct measurements of the erythrocyte count and the MCV: (Hct [μL/100 μL] = RBC [× 10−6/μL] × MCV [fl]/10). Falsely elevated MCV and decreased red cell counts can be observed when red cell autoantibodies are present and retain binding capability at room temperature (cold agglutinins and some cases of autoimmune hemolytic anemia). This causes red cells to clump and affects the accuracy of both the red blood cell (RBC) count and MCV, as well as the resultant hematocrit.
The hematocrit may also be determined by subjecting the blood to sufficient centrifugal force to pack the cells while minimizing trapped extracellular fluid. This approach was traditionally done in capillary tubes filled with blood and centrifuged at very high speed in a small tabletop centrifuge, and the technique was referred to as the “microhematocrit” or informally as a “spun crit.” Before standardized methods for hemoglobin quantification were available, the hematocrit was the simplest and most accurate method for determining the fractional volume of red cells in blood and by inference the hemoglobin. However, this is a manual procedure not well adapted to routine processing in a high-volume clinical laboratory, and is affected by varying amounts of plasma trapped between red cells in the packed cell volume,16 typically about 2 to 3 percent of the packed volume.17 The hematocrit from polycythemic samples or blood containing abnormal erythrocytes (sickle cells, thalassemic red cells, iron-deficient red cells, spherocytes, macrocytes) is increased because of enhanced plasma trapping associated with increased red cell rigidity.17 Therefore, although automated hematocrit values are adjusted to be equivalent to spun hematocrit for normal samples, in abnormal samples, the spun hematocrit may be spuriously elevated (up to 6 percent in microcytosis).18 The hemoglobin determination now is preferred to the hematocrit, because it is measured directly and is the best indicator of the oxygen-carrying capacity of the blood.
Measurement of Hemoglobin
Hemoglobin is intensely colored, and this property has been used in methods for estimating its concentration in blood. Erythrocytes contain a mixture of hemoglobin, oxyhemoglobin, carboxyhemoglobin, methemoglobin, and minor amounts of other forms of hemoglobin. To determine hemoglobin concentration in the blood, red cells are lysed and hemoglobin variants are converted to the stable compound cyanmethemoglobin for quantification by absorption at 540 nm. All forms of hemoglobin are readily converted to cyanmethemoglobin except sulfhemoglobin, which is rarely present in significant amounts. In automated blood cell counters, hemoglobin is usually measured by a modified cyanmethemoglobin or an alternate lauryl sulphate method. In practice, the major interference with this measurement is chylomicronemia, but newer instruments identify and minimize this interference. Noninvasive transcutaneous monitoring of total hemoglobin concentration, as well as methemoglobin and carboxyhemoglobin, using multiwavelength pulse oximetry has become available.19 Although these instruments offer the opportunity to track hemoglobin concentration trends in patients subject to blood loss and fluid shifts,20 it is not yet clear that they have sufficient precision to guide transfusion decisions.21,22 Such hemoglobin measurements may be unreliable under conditions of peripheral circulatory hypoperfusion.
The hemoglobin level varies with age (Table 2–1). Chapter 7 discusses changes in hemoglobin in the neonatal period. After the first week or two of extrauterine life, the hemoglobin falls from levels of approximately 17 g/dL to levels of approximately 12 g/dL by 2 months of age. Thereafter, the levels remain relatively constant throughout the first year of life. Any child with a hemoglobin level below 11 g/dL should be considered anemic.23 Chapter 8 discusses changes in hemoglobin concentration with pregnancy and Chap. 9 discusses changes in hemoglobin levels in older persons.
Table 2–1.Reference Ranges for Leukocyte Count, Differential Count, and Hemoglobin Concentration in Children* ||Download (.pdf) Table 2–1. Reference Ranges for Leukocyte Count, Differential Count, and Hemoglobin Concentration in Children*
|Age ||Leukocytes Total (× 103/μL) ||Neutrophils ||Eosinophils ||Basophils ||Lymphocytes ||Monocytes ||Hemoglobin g/dL Blood |
|Total ||Band ||Segmented |
|12 mo ||11.4(6.0–17.5) || |
|4 yr ||9.1(5.5–15.5) || |
|0.25(0.02–0.65) ||0.05(0–0.2) ||4.5(2.0–8.0) ||0.45(0–0.8) ||12.7(11.2–14.3) |
|2.8 ||0.6 ||50 ||5.0 || |
|6 yr ||8.5(5.0–14.5) || |
|10 yr ||8.1(4.5–13.5) || |
|21 yr ||7.4(4.5–11.0) || |
Standard Red Cell Indices
The size and hemoglobin content of erythrocytes (red cell indices), based on population averages, have traditionally been used to assist in the differential diagnosis of anemia.24 A variety of newer indices based on size and hemoglobinization characteristics of red cell subpopulations are discussed in the section “Novel Red Cell and Reticulocyte Indices”.
Automated blood counters measure the MCV directly by either electrical impedance or light scatter measurements of individual red cells. The MCV has been used to guide the diagnostic workup in patients with anemia; for example, testing patients with microcytic anemia for iron deficiency or thalassemia, and those with macrocytic anemia for folate or vitamin B12 deficiency. This approach has practical value, but also limitations25; for instance, MCV may be normal in some older patients with pernicious anemia,26 or in advanced pernicious anemia with severe red cell fragmentation,27 while one-third of older patients have an elevated MCV without an evident cause.28 Mathematical manipulations of various red cell indices take advantage of the trend toward relatively more severe microcytosis than hypochromia in thalassemia trait versus iron-deficiency anemia to assist in the differential diagnosis of these disorders,29 particularly in high-prevalence populations where laboratory resources are limited,30 but their usefulness has been questioned.31
The MCH, the amount of hemoglobin per red cell, increases or decreases in parallel with the red cell volume (i.e., MCV) and generally provides similar diagnostic information, although because this parameter is affected by both hypochromia and microcytosis, it is as least sensitive as the MCV in detecting iron-deficiency states.32 Another advantage of the MCH is the consistency across different analyzer types, as it is derived from two of the most accurately measured parameters: hemoglobin and red cell count.33 The MCHC is not used much diagnostically, and is primarily useful for quality control purposes, such as detecting sample turbidity. These red cell indices are average quantities and, therefore, may not detect abnormalities in blood with mixed-cell populations. In situations such as sideroblastic anemia, recently transfused patients, patients with severe pernicious anemia with red cell fragmentation, and folate plus iron deficiency, both large and small red cells are present, diminishing the value of the MCV.
Red Cell Distribution Width
The RDW is an estimate of the variance in volume within the population of red cells, expressed as 1 SD of red cell volume measurements divided by the MCV. Instrument manufacturers calculate RDW using different algorithms, so that reference ranges vary according to analyzer model. The RDW can be used in the laboratory as a flag to select those samples that should have manual review of blood films for red cell morphology. More significantly, a large literature has now developed around the evidence that the RDW is a biomarker predicting morbidity and mortality in a broad variety of clinical settings,34 such as angina/myocardial infarction,35 heart failure, trauma, pneumonia, sepsis, intensive care treatment, renal and liver disease, and in the general population.36 Most of these studies are retrospective, observational, or cohort-based studies, often using databases of routinely collected data gathered for other purposes, but prospectively designed studies have arrived at similar conclusions.37,38 The RDW retains its association with poor clinical outcomes whether or not anemia is present,39 and it adds predictive power to more established predictive risk models.40 RDW may be a surrogate for systemic inflammation41 and/or oxidative stress, but the predictive value of RDW is independent of other inflammatory markers,40 suggesting that this biomarker is tracking other mechanistic processes as well. Identification of physiologic mechanisms linking RDW to adverse clinical outcomes will be important in using this predictive biomarker to inform therapeutic decisions.34
Reticulocyte Count and RNA Content
The reticulocyte is a newly released anucleate red cell that enters the blood with residual detectable amounts of RNA (Chaps. 31 and 32). The number of reticulocytes in a volume of blood permits an estimate of marrow erythrocyte production and is thus useful in evaluating the pathogenesis of anemia by distinguishing inadequate production from accelerated destruction (Chap. 32). The manual method for enumerating reticulocytes by placing a sample of blood in a tube containing new methylene blue and preparing a blood film to enumerate the proportion of cells that show blue beaded precipitates (residual ribosomes) has largely been replaced by automated methods, which are incorporated into high-volume hematology analyzers.42 Reticulocytes are identified by direct fluorescence measurement after staining with RNA-binding dyes or light scatter measurements to detect staining if nonfluorescent RNA-binding dyes are used. Various proprietary combinations of light scatter and other parameters are used to minimize interferences such as nucleated red cells, nuclear remnants (Howell-Jolly bodies), malaria parasites, or platelet clumps.
Automated reticulocyte counts are typically reported in absolute numbers (reticulocytes per μL or per L of blood), obviating the need to correct for a reduced red cell count (anemia), if present. However, one may still consider the effect of elevated erythropoietin levels secondary to severe anemia, which results in premature release of reticulocytes persisting in the circulation for more than the usual 1 day, correspondingly inflating estimates of daily marrow reticulocyte production based on the reticulocyte count (Chap. 32). The correlation between manual and automated methods of reticulocyte enumeration is good, but reference ranges differ slightly among the methods, given the different dyes and conditions used and the continuous nature of the variables separating reticulocytes from mature red cells.
Many hematology analyzers now report some quantitative measure of reticulocyte RNA content. Increase in the immature (highest RNA content) reticulocyte fraction is an early sign of marrow recovery from cytotoxic therapy43 or treatment for nutritional anemias, usually preceding the rise in total reticulocyte count. A limitation at present is that the methods lack standardization and reference ranges for these parameters are instrument dependent.44
Additional Red Cell and Reticulocyte Indices
Current high-end automated cell counters measure unique properties of mature red cells and reticulocytes on a cell-by-cell basis, not just as population averages. The result is a plethora of new indices that are in many cases specific to an instrument manufacturer, presenting new diagnostic opportunities but also a confusing nomenclature and a potential lack of comparability. Some examples of parameters that have been studied include %HypoHe, %MicroR, RET-He (available on Sysmex instruments), CHr, HYPO% (Siemens), RSf, LHD% (Beckman-Coulter), and FRC (fragmented red cells; Sysmex and Siemens).
New formulas for distinguishing causes of microcytosis based on several novel red cell indices function about as well45 or somewhat better46 than traditional formulas for differentiating iron deficiency from thalassemia trait. More sophisticated mathematical modeling of individual cell-based volume and hemoglobin content data available in current analyzers has been used in a systems biology approach to demonstrate latent iron deficiency and to distinguish causes of microcytosis.47,48 The ability of new automated analyzers to measure parameters specifically in reticulocytes on a cell-by-cell basis also opens up the possibility of reticulocyte-specific indices. The theoretical advantage is that acute changes in red cell function would be detected more rapidly and reliably in the reticulocyte fraction as opposed to the total red cell population.
Estimates of reticulocyte-specific hemoglobin content (CHr and RET-He, which are comparable) by light-scatter measurements of reticulocytes are closely related to adequacy of iron availability to erythroid precursors during the preceding 24 to 48 hours, and have been described as diagnostically useful in detecting functional iron deficiency in complex clinical settings, such as chronic inflammation49 and chronic renal disease.50 The increase in serum ferritin as an acute phase reactant combined with the physiologic variation of serum iron and iron-binding capacity limits the value of conventional parameters in these settings. The CHr may be a better predictor of depleted marrow iron stores than traditional serum iron parameters in nonmacrocytic patients,51 and is a more sensitive predictor of iron deficiency than hemoglobin for screening infants52 and adolescents for iron deficiency. Estimates of percentages of red cells falling below a cutoff for hemoglobin concentration (HYPO%) or hemoglobin content (%HypoHe) may provide greater sensitivity than the corresponding mean values averaged over all red cells, for instance with respect to iron deficiency in renal disease.53 Four of the newer parameters (HYPO%, %HypoHe, CHr, RetHe) similarly outperformed transferrin saturation and ferritin in hemodialysis patients54 for diagnosis of iron deficiency. However, both the CHr and RET-He are less effective than the MCH in screening elderly patients for iron-deficiency anemia.55 The RSf (square root of MCV times MRV [mean reticulocyte volume]) and LHD% (a mathematical transformation of the MCHC) have similar diagnostic utility as RET-He.56 Fragmented red cell (FRC) counts by automated analyzers, based on better methods of separating small red cells from platelets, appear to lack specificity and their clinical role is not yet defined.
These parameters have the advantage of ready access in the context of an automated blood count, but the availability of differently derived and calculated parameters from various instrument makers is a challenge to remember and compare across laboratories.
Nucleated red cells are present in newborns, particularly if physiologically stressed, and in a variety of disorders, including hypoxic states (congestive heart failure), severe hemolytic anemia, primary myelofibrosis, and infiltrative disease of the marrow (Chap. 45). Most modern automated hematology analyzers are capable of detecting and quantitating nucleated red blood cells, which were a source of spuriously elevated leukocyte counts in earlier instruments, at a level of 1 to 2 nucleated red cells per 100 leukocytes.
Malarial parasites can also be detected by some current analyzers, based on detecting parasite infected red cells or neutrophils containing ingested hemozoin in regions of the multiparameter display that are not characteristically populated in normal blood (sometimes causing spurious eosinophilia57). Some reports indicate high sensitivity and specificity with certain instrumentation,58 a useful consideration in endemic areas where access to technologists with morphologic expertise may not be consistent. Careful attention to instrument characteristics and limitations as well as the relative prevalence of disorders causing instrument flags in the laboratory’s patient population is essential in fine tuning instrument review criteria to provide reasonable sensitivity and specificity.
Other Abnormalities Not Detected by Automation
Some disorders, such as immune and hereditary spherocytosis (Chaps. 46 and 54), hemoglobin C disease (Chap. 49), elliptocytosis (Chap. 46), inherited granule abnormalities (Chap. 66), and malaria and other parasitic diseases (Chap. 53), may not be reliably detected by the various flagging strategies on automated analyzers, and morphologic findings such as basophilic stippling (Chap. 31), toxic granulation (Chap. 60), siderocytes (Chap. 31), and pathologic rouleaux (Chap. 109) are only detectable by microscopic examination of the blood film.
AUTOMATED ANALYSIS OF LEUKOCYTES
Leukocyte counts are performed by automated cell counters on blood samples appropriately diluted with a solution that lyses the erythrocytes (e.g., an acid or a detergent), but preserves leukocyte integrity. Manual counting of leukocytes is used only when the instrument reports a potential interference or the count is beyond instrument linearity limits. Manual counts are subject to much greater technical variation than automated counts because of technical and statistical factors, and with modern instrumentation, need to be done infrequently. Instruments that perform an automated 5-part differential can measure absolute neutrophil counts accurately down to 100/μL.59 Automated leukocyte counts may be falsely elevated as a result of cryoglobulins or cryofibrinogen, clumped platelets or fibrin from an inadequately anticoagulated or mixed sample, ethylenediaminetetraacetic acid (EDTA)–induced platelet aggregation, nucleated red blood cells, or nonlysed red cells, and falsely decreased because of EDTA-induced neutrophil aggregation. This potential interference is instrument dependent, and current analyzers use a variety of algorithms to minimize their effect and flag those rare samples on which accurate automated analysis cannot be performed.
Leukocytes in the blood serve different functions and arise from different hematopoietic lineages, so it is important to evaluate each of the major leukocyte types separately. Modern automated instruments use multiple parameters to identify and enumerate the five major morphologic leukocyte types in blood: neutrophils, basophils, eosinophils, lymphocytes, and monocytes, as well as indicate the possible presence of immature or abnormal cell. Customarily, both absolute (cells per μL) and relative (percent of leukocytes) counts are reported in the leukocyte differential. It is the absolute values that relate to pathologic states, and percentages are sometimes misleading (e.g., absolute neutropenia appearing as a relative lymphocytosis) if the absolute values are not carefully examined. Some have proposed to eliminate the reporting of differential count percentages entirely for this reason.60 “Band” neutrophils cannot be identified as such by automated analyzers, although they will usually trigger a manual review flag if present in increased numbers. Current high-throughput instruments can perform an accurate automated “five-part” differential count with a false-positive rate (i.e., unnecessarily flagged for review) of 2 to 15 percent in samples from a medical center patient population.61 Eosinophils are accurately counted by current state-of-the-art instruments, but automated basophil counts remain imprecise.11 Small numbers of abnormal cells can escape detection by either automated or manual methods. The false-negative rate for detection of abnormal cells varies from 1 to 20 percent, depending on the instrument, type of abnormal cell examined, and the detection limit desired (1–5 percent abnormal cells).62,63,64 Careful attention to use of flagging criteria designed to prompt manual review, which are linked to instrument-specific methodology, is essential to insure that optimum workflow strategies are used to detect samples containing abnormal cells with a manageable rate of manual review. Many instruments have “blast” flags designed to pick up leukemic blasts, but the sensitivity of such flags alone varied from 65 to 94 percent in a recent study,11 and is lower in leukopenic patients.65 One must rely not only on the specifically designed “blast” flags, but also on other abnormalities identified in the automated blood count, including other flags, to select samples for manual morphologic smear review. Lymphoma cells and reactive lymphocytes are the most difficult for both automated instruments and the human observer to identify. If one needs to search for infrequent abnormal cells or evaluate leukocyte morphology, there is still no substitute for microscopic examination of a properly stained blood film by a trained observer. The variability of morphologic quantification of band neutrophils is so high that some have advocated ceasing quantitative reporting of band cells.66 In spite of instrumentation that permits automated analysis of a majority of clinical samples, the leukocyte differential count is still labor intensive relative to other high-volume laboratory tests, and its value as a cause-finding tool in screening of asymptomatic patients has been questioned.67
The normal differential leukocyte count varies with age. As described in Chap. 7, polymorphonuclear neutrophils are predominant in the first few days after birth, but thereafter lymphocytes account for the majority of leukocytes. This pattern persists up to approximately 4 to 5 years of age, when the polymorphonuclear leukocyte again becomes the predominant cell and remains so throughout the rest of childhood and adult life. Chapter 9 discusses the leukocyte count in older persons. The leukocyte count may decrease slightly in older subjects because of a fall in the lymphocyte count with age. Neutrophil counts are lower in individuals of African descent, and in some Middle Eastern populations than in persons of European descent.68
AUTOMATED ANALYSIS OF PLATELETS
Platelets are usually counted electronically by enumerating particles in the unlysed sample within a specified volume window (e.g., 2–20 fl), where volume may be measured by electrical impedance or light scatter.69 The platelet count was more difficult to automate than the red cell count because of the small size, tendency to aggregate, and potential overlap of platelets with more numerous smaller red cells and cellular debris. Current instruments typically construct a platelet volume histogram based on platelet size within a reliably measured platelet volume window and mathematically extrapolate this histogram to account for platelets whose size overlaps with debris (smaller) or small red cells (larger). This works because platelet volumes in health or disease follow a log-normal distribution. Some analyzers compare platelet counts determined by different methods (e.g., impedance, light scatter, or fluorescence staining) to improve accuracy, especially useful for low platelet counts. Based on analysis of volume-distribution histograms of platelets and red cells and comparison of optical and impedance-based platelet counts, suspect samples are flagged for microscopic review. Automated platelet counting by current instrumentation is accurate and far more precise than manual methods. At very low platelet counts (less than 20 × 109/L), results are less precise70 and there is a method-dependent tendency to overestimate platelet counts.71 Conversely, platelet activation in disorders such as disseminated intravascular coagulation (DIC) and acute leukemia may result in systematic slight undercounting of platelets.72 Advances in instrumentation, such as fluorescent dyes to more specifically identify platelets in thrombocytopenic73 and microcytic74 samples, should improve accuracy. When reviewing the blood film, platelet count may be roughly estimated as 2000 times the number of platelets in 10 consecutive oil immersion (1000×) fields.75
Falsely Decreased Platelet Counts
Causes of falsely decreased platelet counts include incomplete anticoagulation of the sample (sometimes accompanied by small clots in the specimen or fibrin strands on the stained film) and platelet clumping (pseudothrombocytopenia) or “satellitism” (adherence of platelets to neutrophils), caused by aggregation induced by nonpathogenic antibodies recognizing platelet adhesion molecule epitopes exposed as a result of chelation of divalent cations in the anticoagulated sample.69 Platelet clumping occurs in approximately 0.1 percent of hospitalized patients.76 The same phenomenon may occur to a lesser degree in citrate, which is often used to obtain platelet count in such cases. Magnesium EDTA, as compared to sodium EDTA, anticoagulant is reported to more effectively inhibit platelet aggregation in these patients and provide an accurate platelet count.77 Classical causes of falsely elevated platelet count include severe microcytosis, cryoglobulins, and leukocyte cytoplasmic fragmentation.69 Infrequently, it may be necessary to confirm automated results by a microscopic (phase contrast) platelet count or platelet estimate from the blood film, bearing in mind that these methods are imprecise.
The mean platelet volume (MPV) has been proposed as a useful clinical tool in the differential diagnosis of thrombocytopenias, and is associated with cardiovascular risk, stroke, and metabolic disease. Increased MPV may be related in a complex way to thrombopoietic stimuli that affect megakaryocyte ploidy, and not platelet age per se. A platelet volume distribution width (PDW) can be calculated just as the RDW, and is correlated with platelet count and MPV.78 However, platelet size parameters are difficult to accurately quantify and use diagnostically because of the wide physiologic variation of the MPV in normal subjects, lack of standardization of automated measurement techniques and instability of platelet size parameters in the presence of commonly used anticoagulants.79
Newly Released (Reticulated) Platelets
Newly released platelets contain RNA, as do newly released red cells, and are functionally more active, with enhanced expression of adhesion molecules and bound coagulation factors.80 The number of platelets with high RNA content (sometimes termed reticulated platelets or immature platelet fraction, measured by flow cytometry with RNA-binding fluorescent dyes, or by certain automated analyzers81) is a marker of marrow megakaryocytopoiesis and is proposed as a way of differentiating decreased production of platelets from circulatory destruction or removal as a cause of thrombocytopenia, in an analogous fashion to the use of the reticulocyte count. The percentage of reticulated platelets is increased in destructive thrombocytopenias, but remains within the reference range in hypoproductive states.82 Reticulated platelet number or RNA content correlates with imminent platelet recovery after chemotherapy.83 Reticulated platelet number is correlated with risk of death in patients with acute coronary syndrome84 and DIC,85 and with hyporesponsiveness to platelet function inhibitors86 or aspirin.87
The use of reference ranges for quantitative hematology measurements deserves some additional comment. The physiologic variation of certain blood cell counts is notably higher than usually found in blood chemistry analytes. This is a reflection of the adaptive responsiveness of the marrow and other tissues to cytokine and hormonal signaling. For instance, the leukocyte and differential counts are affected by stress, diurnal variation, tobacco smoking, and ethnic origin. With increasing globalization of clinical research and therapy, ethnic characterization of populations used for reference ranges is critical to data interpretation of clinical studies.88 Platelet count and MPV show substantial ethnic variation.89 The platelet and absolute neutrophil counts are lower in individuals of African ethnic origin.68 American men and women of African descent have lower hemoglobin concentrations than do men and women of European descent, a difference that is reduced by half, but still significant, when subjects with iron deficiency, thalassemia, sickle trait, and renal disease are excluded.90 Important clinical consequences may result from these differences; for instance, reduced neutrophil counts in Americans of African descent result in lower-dose intensity of treatment in early stage breast cancer, which may be related to survival outcome disparities.91 Beutler and West90 summarize the situation well: “The problem cannot be solved by simply establishing different ranges for different ethnic groups, especially since all represent some degree of admixture. Thus, it is basically information that the physician must possess that becomes one of the many factors that we designate as clinical judgment.” With these caveats in mind, reference ranges for children, and African American, Hispanic, and white adults are presented in Tables 2–1 and 2–2. As with all laboratory parameters, clinical interpretation of patient results should be based on laboratory specific reference ranges. Therefore, these tables are not presented to guide interpretation of specific laboratory results, but to indicate the challenges facing laboratories and physicians in constructing and interpreting reference ranges of even standard and traditional assays.
Table 2–2.Published Reference Ranges for Key Blood Variables ||Download (.pdf) Table 2–2. Published Reference Ranges for Key Blood Variables
| ||NORIP107 ||Wakeman92 ||Cheng93 || || ||Bain106 || |
|Date ||2003 ||2004 ||1994 ||1994 ||1994 ||1996 ||1996 |
|Ethnicity ||Nordic ||U.K. ||U.S. European descent ||U.S. African descent ||U.S. Mexican descent ||U.K. European descent ||U.K. African descent |
|No. ||1800 ||250 ||3125 ||1712 ||1735 ||200 ||115 |
|Hgb (g/dL) (M) ||13.4–17.0 ||13.7–17.2 ||13.2–16.9 ||12.0–16.2 ||13.1–16.7 ||NA ||NA |
|(F) ||11.7–15.3 ||12.0–15.2 ||10.7–15.1 ||10.2–14.4 ||11.4–15.0 || || |
|Hct (%) (M) ||40–50 ||40–50 ||39–50 ||36–48 ||39–50 ||NA ||NA |
|(F) ||35–46 ||37–46 ||34–45 ||32–43 ||33–45 || || |
|MCV (fl) ||82–98 ||83–98 (M) ||79–97 (M) ||75–97 (M) ||83–96 (M) ||NA ||NA |
| || ||85–98 (F) ||77–97 (F) ||75–97 (F) ||81–98 (F) || || |
|WBC (× 109/L) ||3.5–8.8 ||3.6–9.2 ||4.1–11.7 (M) ||3.5–9.5 (M) ||4.6–10.6 (M) ||3.6–9.2 (M) ||2.8–7.2 (M) |
| || || ||4.3–12.0 (F) ||3.4–10.5 (F) ||4.3–11.3 (F) ||3.5–10.8 (F) ||3.2–7.8 (F) |
|Neutrophils (× 109/L) ||NA ||1.7–6.2 ||2.7–8.1 (M) ||1.5–7.4 (M) ||2.2–6.6 (M) ||1.7–6.1 (M) ||0.9–4.2 (M) |
| || || ||2.5–6.9 (F) ||1.5–8.4 (F) ||2.5–7.9 (F) ||1.7–7.5 (F) ||1.3–4.2 (F) |
|Lymphocytes (× 109/L) ||NA ||1.0–3.4 ||1.1–3.7 (M) ||1.1–3.6 (M) ||1.3–3.4 (M) ||1.0–2.9 (M) ||1.0–3.2 (M) |
| || || ||1.2–3.7 (F) ||1.3–3.9 (F) ||1.3–3.9 (F) ||1.0–3.5 (F) ||1.1–3.6 (F) |
|Monocytes (× 109/L) ||NA ||0.2–0.8 ||0.13–0.86 (M) ||0.11–0.72 (M) ||0.14–0.70 (M) ||0.18–0.62 (M) ||0.15–0.58 (M) |
| || || ||0.11–0.78 (F) ||0.12–0.83 (F) ||0.12–0.79 (F) ||0.14–0.61 (F) ||0.15–0.39 (F) |
|Platelets (× 109/L) (M) ||145–348 ||140–320 ||161–385 ||161–381 ||166–388 ||143–332 ||115–290 |
|(F) ||165–387 ||180–380 ||178–434 ||178–452 ||171–411 ||169–358 ||125–342 |
Note the variation in reference ranges obtained from different studies. The major variability is likely population selection, especially the degree to which chronic illness or asymptomatic iron deficiency are excluded, and physiologic factors, such as diurnal variation, are considered. For example, the Wakeman study92 exclusively used early morning samples, hence the upper limit of leukocyte count is lower because of diurnal physiologic variation. The National Health and Nutritional Examination Surveys (NHANES) III national database has the advantage of being a very large broad nationwide sampling, which, as used by Cheng and colleagues,93 excluded any subjects with history of smoking, alcohol consumption, contraceptive use, and a variety of chronic diseases (excluding 60 percent of the tested subjects). However, those with asymptomatic iron deficiency were not excluded, so hemoglobin tends to be lower than in studies that may have been weighted toward groups of individuals in which undiagnosed iron deficiency and other asymptomatic disorders are less common. α- and β-thalassemia traits are also quite common in healthy individuals of certain ethnic groups, and inclusion of subjects with these disorders will also affect reference ranges. Normal lower limits for hemoglobin have been determined in U.S. subjects of different ethnic backgrounds carefully screened for occult disease.94 Such considerations also affect determination of the upper (97.5th percentile) limit of normal hematocrit and hemoglobin in relation to a possible diagnosis of polycythemia, where one has to carefully weigh the likelihood that a “normal-range” study has adequately excluded iron-deficient subjects.95,96 Biomedical parameters are also subject to historical trends, such as the observed improvement in hemoglobin levels in the post–folic-acid-fortification era.97 Finally, when one observes significant changes in reference ranges based on age (e.g., glomerular filtration rate, lipid parameters, hemoglobin), there is the question of whether this is physiologic or a result of increased prevalence of undiagnosed occult disease.
Most hematologic variables show more stability within an individual than between individuals, illustrating one reason for the lack of sensitivity and specificity of any test “cutoff,” which is typically designed for a population rather than for an individual person. A study of repeated analyses of blood variables from older subjects98 graphically demonstrates this phenomenon. Some normal subjects have a normal steady-state platelet count between 170 × 109/L and 200 × 109/L, whereas others have one between 280 × 109/L and 310 × 109/L (Fig. 2–3). For the latter group, a progressive fall in platelet count because of marrow failure may not be detected as quickly as the former group. The same observations are shown for absolute neutrophil count, hemoglobin, and MCV, among others. In normal subjects, the ratio of between subject to within subject variation ranges from about two times for absolute neutrophil count99 to four to six times for hemoglobin, platelet count, absolute reticulocyte count, and MCV.100 Data from a large clinical trial’s central laboratory show similar findings for hemoglobin in study subjects with various disease states. In this report bayesian methods were used to construct a (narrower) personalized reference range using progressive accumulation of baseline measurements to achieve greater sensitivity to perturbations following treatment.101 Circadian variations in hematologic laboratory values, including hematocrit, total leukocyte count, serum iron, and serum folate have been described.102 Some have proposed to customize reference ranges for time of sample collection, so that reference ranges aren’t inflated by the need to accommodate circadian variation.103 Genetic loci affect quantitative hematologic traits (such as hemoglobin, MCH, platelet count, leukocyte count, etc.) in normal subjects of European, African, and South Asian ancestry.104,105 The loci contain many candidate genes known to be involved in hematopoiesis, but known genetic influences identified in such studies only explain a small proportion (4–9 percent) of the observed phenotypic heterogeneity of these variables.
Absolute neutrophil count, hemoglobin, mean cell volume (MCV), and platelet count determined repeatedly by automated hematology analyzer on 24 healthy elderly subjects. Fasting (7–9 am) blood samples were obtained 9 to 10 times at 14-day intervals from seated elderly subjects with minimal stasis by the same phlebotomist and performed in duplicate on the morning specimen collection. Subjects had no chronic medical conditions requiring therapy and were not taking drugs. The mean and range for each patient is shown separately for each assay. This is an illustration of the relatively narrow range within which most variables are maintained in an individual, whereas there are striking differences in both mean and variance between subjects. Reference ranges need to encompass at least 95% of values from all healthy individuals, placing limits on diagnostic sensitivity in detecting progressive change in a hematologic variable, previously maintained in a homeostatic range. (Adapted with permission from Fraser CG, Wilkinson SP, Neville RG, et al: Biologic variation of common hematologic laboratory quantities in the elderly. Am J Clin Pathol 92(4):465–470. 1989.)