Using 106 samples from patients with an absolute neutrophil count (ANC) less than 2.0 × 109/L, two 5-part differential hematology instruments (Sysmex XE-2100, Sysmex, Kobe, Japan, and Advia 2120i, Siemens Healthcare Diagnostics, Deerfield, IL), two 3-part differential hematology instruments (Sysmex K4500, Sysmex, and Advia 60, Siemens Healthcare Diagnostics), and an automated system for examination of microscopic slides (CellaVision DM96, CellaVision, Lund, Sweden) were compared with a flow cytometric (FCM) neutrophil count using monoclonal antibodies for cell classification. The precision and accuracy of the 5-part differential instrument ANC was very good at more than 0.1 × 109/L, although a small systematic difference (10.3%) was found between the 2 instruments. The ANC of the 3-part differential instruments was less reliable, but the WBC count correlated very well with the WBC count from the 5-part differential instruments. Also, the neutrophil count from the CellaVision DM96 compared very well with FCM. When used in the correct laboratory setting, all of the evaluated instruments provide ANCs and WBCs with adequate accuracy and precision.
Neutropenia is associated with an increased risk of severe infection, which carries a high risk of mortality if untreated.1,2 Guidelines for treatment of neutropenic fever define neutropenia as an absolute neutrophil count (ANC) less than 0.5 × 109/L or 1.0 × 109/L and profound neutropenia as ANC less than 0.1 × 109/L.3–5 ANCs are closely monitored in patients with or at risk of developing neutropenia.
The automated hematology instruments used for this monitoring are well validated for ANCs within and more than the reference interval.6,7 A thorough validation using samples with low ANCs is more rarely performed, although some instruments have been evaluated comparing the automated neutrophil count with a manual count of 100 leukocytes.8–10 The reference method for the neutrophil count is by counting 400 leukocytes by microscopic examination in 2 blood smears (Clinical and Laboratory Standards Institute, H20-A2).11 However, in samples with a very low ANC, there are insufficient leukocytes to count 100 cells, let alone 400 cells. This means that the manual count is unlikely to be adequately precise at very low neutrophil counts.8 Flow cytometry (FCM) with monoclonal antibodies for cell identification has been suggested as an alternative reference method for the leukocyte differential count.12
The aim of this study was to evaluate the accuracy and precision of the ANC at the clinical decision limits and at very low neutrophil counts. Combining elements from previously published flow cytometric counting methods, neutrophils were counted by FCM in 106 patient samples with ANCs less than 2.0 × 109/L.12,13 The results from FCM were compared with the count of 2 commonly used 5-part differential hematology instruments (Sysmex XE-2100, Sysmex, Kobe, Japan, and Advia 2120i, Siemens Healthcare Diagnostics, Deerfield, IL), two 3-part differential hematology instruments (Sysmex K4500, Sysmex, and Advia 60, Siemens Healthcare Diagnostics), and a microscopic count using the automated microscopic image recognition system CellaVision DM96 (CellaVision, Lund, Sweden).
Materials and Methods
Residuals from K2EDTA-anticoagulated tubes (Medkjemi, Asker, Norway) containing samples from adult patients submitted for WBC differential count were obtained from January to April 2011 at Oslo University Hospital, Ullevaal, Oslo, Norway. By using the ANC from Sysmex XE-2100, 106 samples with an ANC less than 2.0 × 109/L were included. Our laboratory serves a large hospital including hematology and oncology wards. Hematology instruments routinely flag samples when the presence of abnormal leukocytes is suspected. Some of these flags (“immature granulocytes [?],” “blasts?,” “white cell abnormal scattergram,” “monocytosis,” and “nucleated red blood cells?”) have previously been considered to affect the correctness of the neutrophil count.9,14,15 Of the samples included, 32 and 44 samples were reported with 1 or more of these flags by the Sysmex XE-2100 and Advia 2120i, respectively. The ANCs of these samples were included regardless of flags as long as the instrument provided a neutrophil count. The project was approved by the regional ethical committee.
The automated differential counts were performed by Sysmex XE-2100, Sysmex K4500, Advia 2120i, and Advia 60. The instruments, except the Advia 60, are in routine use and undergo assessment in internal and external quality control programs. The Advia 60 is used regularly at our research laboratory and is used for analysis of routine patient samples when needed at the isolation center for infectious diseases. It is controlled against one of our Sysmex XE-2100 instruments with a patient sample when used. The leukocyte differential count was performed within 8 hours from blood sampling and within a time frame of 2 hours for all hematology instruments. Within-run analytic coefficients of variation (CV) were determined by running samples in duplicate (Sysmex XE-2100, n = 103; Advia, 2120i, n = 105; SysmexK4500 and Advia 60, n = 20).
The Sysmex XE-2100 and Advia 2120i perform a 5-part differential count by lysing erythrocytes and analyzing the light scatter/fluorescence or size/myeloperoxidase content of the leukocytes, respectively. The Sysmex K4500 and Advia 60 perform a 3-part differential count by impedance analysis after erythrocyte lysis. The Sysmex K4500 differentiates among neutrophils, lymphocytes, and a third group consisting of monocytes, eosinophils, and basophils, and the Advia 60 performs a 3-part differential count consisting of granulocytes, monocytes, and lymphocytes.
Within 6 hours after blood sampling, 2 blood smears were manually prepared on SP Slides (Sysmex) and stained with May-Grünwald Giemsa from the last 80 of the 106 samples included. By using the automated image recognition system, CellaVision DM96, the percentage of neutrophils for each sample was determined in the 2 slides. The ANC was calculated by multiplying the mean neutrophil percentage from the 2 slides with the WBC count obtained by FCM. The CellaVision DM96 is an automated microscope with software that provides a preclassification of leukocytes based on morphologic features.16,17 The preclassification was reviewed and confirmed or reclassified by 1 skilled operator (T.A.H.) blinded to the results from the hematology instruments and FCM.
Within 10 hours after collection of the samples, 50 μL of whole blood was incubated for 30 minutes at room temperature with 9.5 μL of a mixture of monoclonal antibodies. The following antibodies were used: CD45 allophycocyanin (APC; clone J33) and CD16 phycoerythrin-cyanine 7 (PE-Cy7; clone 3G8) from Beckman Coulter (Fullerton, CA), CD14 Horizon v450 (clone MfP9) from Becton Dickinson (BD, Franklin Lakes, NJ), and CD193 phycoerythrin (PE; clone eBio5E8-G9-B4) from ebioscience (San Diego, CA). Antibodies were titrated for an optimal signal/noise ratio in 50 μL of whole blood. A mixture of antibodies was prepared for each day.
After incubation of whole blood with antibodies, 500 μL of ammonium chloride lysis buffer (8.26 g/L NH4Cl, 1 g/L KHCO3, and 37 mg/L EDTA) was added, and the samples were incubated for another 15 minutes at room temperature. Countbright beads (50 μL, Molecular Probes, Eugene, OR) were added to the samples to enable absolute cell counts, and the tubes were then placed on ice and analyzed within 1 hour.
The samples were acquired on a Cyan ADP flow cytometer (DAKO, Glostrup, Denmark). Sphero Ultra rainbow particles, 3.0 to 3.4 μm (Spherotech, Lake Forest, IL), were used to ensure optimal instrument performance, and photomultiplier tube settings were standardized by Flow set pro fluorospheres (Beckman Coulter). Excitation was by 405 nm (UV laser), 488 nm (Argon laser), and 635 nm (red laser). Compensation was set automatically using BD CompBeads (BD). By gating away noise and cellular debris, CD45+ events were acquired at a maximum of 500 events per second using an acquisition trigger on CD45 APC with the aim of acquiring 5,000 CD16+ neutrophil events. In samples with very low numbers of neutrophils, this was not obtainable. However, at least 1,000 CD16+ neutrophil events were acquired in all samples with an ANC of more than 0.1 × 109/L. The data were analyzed using Summit Software, version 4.3 (Summit Software, Little Rock, AR). Within-run CVs were determined by labeling and analyzing 20 samples in duplicate.
Statistical analysis was performed with GraphPad Prism 5 software (GraphPad Software, La Jolla, CA). Deming regression was performed with analytic CVs determined from duplicate runs. Correlation coefficients were determined by using a 2-tailed Pearson test. The mean ANC and WBC of the instruments were compared by paired t test.
Image 1 shows the gating strategy for the FCM neutrophil count. In sequence, cellular debris (non-R1) was removed (Image 1A), and leukocytes (R2) were identified using CD45 (Image 1B). Subsequently, by using CD14 and side scatter, neutrophils and eosinophils (non-R3) were separated from lymphocytes, monocytes, and basophils (R3, Image 1C). The non-R3 cells were plotted in Image 1E, and mature neutrophils (R4) were identified as CD16+. In Image 1F, non-R4 cells were identified as CD193+ eosinophils (R5) or CD193– immature neutrophils (R6). Furthermore, the events from R3 in Image 1C could be identified as basophils (R7), monocytes (R8), or lymphocytes (R9) using CD193 and CD14 (Image 1D).
Gating strategy for classification of leukocytes; 1 representative patient sample. A, Forward scatter (FS) vs side scatter (SS). Countbright beads seen in top left corner (black) were counted in the Pacific Orange channel (not shown) and excluded from the rest of the gating strategy. Debris from lysed erythrocytes was removed, R1 to B. B, CD45 vs SS. CD45+ leukocytes (R2) were separated from nonleukocyte events, R2 to C. C, CD14 vs SS. Neutrophils and eosinophils (non-R3) were separated from monocytes, lymphocytes, and basophils in R3. R3 to D, not R3 to E. D, CD193 vs CD14. Events from R3 were classified as basophils (R7), monocytes (R8), and lymphocytes (R9). E, CD16 vs SS. Events located outside R3 in C. Neutrophils classified as CD16+ (R4), cells located outside R4 to F. F, CD193 vs SS. Eosinophils were classified as CD193+ (R5) and separated from CD193– immature neutrophils (R6). Cyan, mature neutrophils; red, eosinophils; blue, immature neutrophils; magenta, monocytes; yellow, lymphocytes; green, basophils.
Doublets of cells and doublets of beads were identified by using the pulse width and corrected for in the determination of the cell counts. The ANC was estimated by relating the sum of mature and immature neutrophils to the acquired number of Countbright beads as described by the manufacturer of the beads. The within-run analytic CV of the FCM neutrophil count as determined by 20 duplicates was 7.3%. The WBC count was derived by relating the number of CD45+ events (R2) to the number of Countbright beads.
Neutrophil Count of 5-Part Differential Hematology Instruments
The Advia 2120i was unable to perform a differential count in 1 sample owing to myeloperoxidase deficiency. The Sysmex XE-2100 was not able to count neutrophils in another sample with an FCM count of 0.37 × 109/L. In addition, the Sysmex did not return results in duplicate for 2 samples with ANCs less than 0.1 × 109/L. Comparison of the neutrophil counts of the 104 samples with at least 1 result for the Sysmex XE-2100 and Advia 2120i are shown in Figure 1. The ANCs of the Advia 2120i and Sysmex XE-2100 showed excellent correlation with the FCM ANC (r2 = 0.97 and r2 = 0.98, respectively). The Advia 2120i gave 10.3% higher ANCs than Sysmex XE-2100 (95% confidence interval [CI], 7.6%–13.0%). The bias seemed to be proportional and present at all measured neutrophil counts Figure 2. There was only a small difference in WBC counts between the Advia 2120i and Sysmex XE-2100 (1.2%; 95% CI, 0.1%–2.3%; Advia higher than Sysmex). The FCM WBC count differed somewhat from the WBC count from the hematology instruments, indicating a small bias in the FCM leukocyte count (4.4% higher than the mean of the WBC count from the 4 hematology instruments included; 95% CI, 2.7%–6.1%).
Bland-Altman plot comparing the results of the Sysmex XE-2100 and Advia 2120i. x-axis, mean neutrophil count of the Sysmex XE-2100 and Advia 2120i; y-axis, absolute difference in neutrophil count between the Sysmex XE-2100 and Advia 2120i.
Neutrophil Count of 3-Part Differential Hematology Instruments
The Sysmex K4500 was unable to count neutrophils in 24 of 32 samples with ANCs less than 0.5 × 109/L as measured by FCM and 4 of 74 samples with ANCs in the range 0.5 to 2.0 × 109/L. In contrast, the Advia 60, which performs a 3-part differential consisting of granulocytes, lymphocytes, and monocytes, was able to give a granulocyte count in all samples. Comparison between the ANCs of FCM and the Sysmex K4500 (n = 78) and of FCM and the Advia 60 (n = 104) are shown in Figure 3A and Figure 3B, respectively.
The neutrophil counts of the Sysmex K4500 correlated well with FCM (r2 = 0.91); the accuracy of the Sysmex K4500 was good and there was no sign of bias. The analytic within-run CV was estimated at 5.7%. Although the Sysmex K4500 did not give a neutrophil count for most of the samples with ANCs less than 0.5 × 109/L, the instrument seemed to give reliable results when the neutrophils in the samples were countable.
The granulocyte count of the Advia 60 was compared with the neutrophil count of FCM. The correlation between the granulocyte count by the Advia 60 and the neutrophil count by FCM was satisfactory (r2 = 0.85). However, the regression line showed a large intercept indicating that the Advia 60 overestimated the neutrophil count in samples with few neutrophils. The analytic CV was estimated at 6.0%.
We also compared the WBC counts of the Sysmex K4500 and the Advia 60 with the mean WBC counts obtained from the Sysmex XE-2100 and the Advia 2120i. An excellent correlation and good regression fit were found for both instruments (Sysmex K4500, r2 = 0.99, regression y = ax + b, a = 0.99, b = 0.07; Advia 60, r2 = 0.99, a = 0.97, b = 0.05).
Neutrophil Count of the CellaVision DM96 Compared With FCM
The CellaVision DM96 identified and counted on average a total of 128 leukocytes (5th and 95th percentiles, 53 and 239, respectively) in 2 slides for each sample, whereas FCM counted on average 21,094 leukocytes (5th and 95th percentiles, 6,383 and 35,596, respectively). To match the results to the differential count performed by the automated hematology instruments, band cells and the few myelocytes were added to the segmented neutrophil count. Compared with FCM, the CellaVision DM96 showed acceptable accuracy, no significant bias, and very good correlation Figure 4.
By using samples from 106 patients with neutropenia, we evaluated the precision and accuracy of the neutrophil count of 4 automated hematology instruments and an automated microscopic image recognition system (CellaVision DM96). The neutrophil counts were compared with a flow cytometric neutrophil count in which monoclonal antibodies were used for leukocyte classification.
Neutrophil count for 80 patient samples obtained by the CellaVision DM96 and flow cytometry (FCM). y = 1.01x + 0.02. r2 = 0.958.
The two 5-part differential hematology instruments compared very well with the FCM count, giving accurate results down to less than 0.1 × 109/L. Within-series precision was satisfactory above this level, but poor counting statistics owing to few neutrophils probably made the neutrophil count less reliable at less than 0.1 × 109/L.
A relatively small but statistically significant bias in the neutrophil count (10.3%) was found between the Sysmex XE-2100 and Advia-2120i in this study. This was not due to difference in the accuracy of the WBC of the 2 instruments because the WBC differed by only 1.2%. It is therefore unlikely that the difference in ANC found in this study is caused by a calibration error. When the ANCs were compared with FCM, it seems that the Sysmex XE-2100 underestimated the ANC, whereas the Advia 2120i did not. However, since FCM overestimated the WBC count by 4.4%, it is likely that the ANC from FCM is also slightly overestimated. It is therefore difficult to draw any conclusions as to which of the instruments has the correct level for the ANC based on the ANC from FCM. The difference between the hematology instruments is probably not large enough to make a difference in the outcome for an individual patient. However, if the decision limits are followed categorically, the bias will lead to small differences in the practice of administration of antibiotics and chemotherapy between clinics depending on which instrument is in use.
The two 3-part differential hematology instruments also performed well, although less well than the 5-part differential instruments. The Sysmex K4500 was often unable to count neutrophils when the FCM count was around or less than 0.5 × 109/L, but if a neutrophil count was reported, it was reliable. The Advia 60 reports granulocytes rather than neutrophils, thus including eosinophils and basophils in the granulocyte count. This probably explains the positive bias and the moderately less good correlation found between the Advia 60 and FCM in this study. In summary, the smaller hematology instruments performed well when counting samples with moderately low concentrations of neutrophils, and these instruments could be useful in a general practice setting or in selected outpatient clinics. Because the WBC count was very reliable for both instruments, it should be possible to use an algorithm in which samples with a WBC count less than some defined cutoff can be reanalyzed on a 5-part differential instrument or the patient referred to a hospital. However, the 5-part hematology instruments should be preferred when samples with leukopenia are frequent, such as in samples from hematology or oncology wards or when rapid identification of neutropenia is important, such as in emergency departments.
The 5-part hematology instruments included in this study are capable of precisely counting very low numbers of neutrophils, which, for clinical reasons, is much wanted. Patients with neutropenia are at increased risk of infection. Current guidelines for the treatments of patients with neutropenia usually focus on cutoffs at 0.5 and 1.0 × 109/L. The ability to reliably count neutrophils at even lower concentrations might be of interest because patients with profound neutropenia (ANC <0.1 × 109/L) are at greater risk of infection than patients with an ANC between 0.1 and 0.5 × 109/L.1,2,18 Furthermore, it is likely that the inverse relationship between the ANC and the risk of infection is continuous rather than changed in steps at defined cutoffs. Based on this study, it should be possible for laboratories using the Sysmex XE-2100 or Advia 2120i to report ANCs down to 0.1 × 109/L. If desired, the result could be reported to 2 decimal places.
The results clearly indicate that the neutrophil counts performed by the CellaVision DM96 correlate very well with the FCM count, even at very low cell counts (r2 = 0.96). The CellaVision DM96 neutrophil count has previously been shown to correlate well with the manual count, with WBCs ranging from 0.51 to 73.89 × 109/L.16,17
The present study differs from previous studies in that the correlations are based on samples with only low ANCs (<2.0 × 109/L). These data thus confirm that neutrophil counts can be performed with the CellaVision DM96 even when the ANC is very low and support the view that automated blood film analysis to some extent may replace the manual differential count.
It is common practice for laboratories to perform slide review when the automated neutrophil count is less than a defined cutoff value. The consensus guideline for slide review from the International Society of Laboratory Hematology advocates slide review when the first ANC is less than 1.0 × 109/L or the WBC count is less than 4,000/μL (4.0 × 109/L).19 However, some authors suggest reviewing only samples flagged by the hematology instrument for the possible presence of pathologic cells, such as immature granulocytes, blasts, or monocytes, or when flags indicate possible failure in the classification of leukocytes (eg, “abnormal scattergram” on the Sysmex XE-2100).9,14,15
In our study, approximately one third of the samples were flagged for immature granulocytes and/or blasts and 3 samples for monocytosis. The samples with an ANC less than 0.1 × 109/L were often flagged as abnormal scattergrams by the Sysmex XE-2100; only 1 sample significantly more than 0.1 × 109/L had this flag. We, however, did not observe any outliers when we compared the ANC of the Sysmex XE-2100 or Advia 2120i with FCM. This finding indicates that it may not be necessary to validate the ANC by microscopic counting, with the exception of samples flagged with the abnormal scattergram flag from the Sysmex XE-2100, for which our data are insufficient to draw any conclusions.
Rapid administration of antibiotics is important when a patient has neutropenic fever. In a report from 2009, The National Chemotherapy Advisory Group (United Kingdom) recommend a maximum of 1 hour for door-to-needle time.20 It is therefore important to not unnecessarily delay the ANC by performing a slide review before the ANC is released. In our laboratory, we report the ANC (Sysmex XE-2100 is the instrument in use) without slide review regardless of flags as long as the neutrophil population in the plot from the hematology instrument is well defined and well separated from the other leukocyte populations. Slide review is performed later if it is needed for diagnostic purposes. To this end, the CellaVision DM96 might be useful as it is 10% to 25% faster than a manual differential, even after manual reclassification.16,17
The accuracy and precision of the ANC for the Sysmex XE-2100 and Advia 2120i were satisfactory, even at very low neutrophil counts. The accuracy and precision of the Sysmex K4500 and Advia 60 ANC and WBC count were sufficient for use in most primary care settings and in selected outpatient clinics.
We are very grateful to Hanne E. Lunde and Anne Marie S. Trøseid for technical assistance.
Clinical practice guideline for the use of antimicrobial agents in neutropenic patients with cancer: 2010 update by the Infectious Diseases Society of America. Clin Infect Dis. 2011;52:e56–e93.doi:10.1093/cid/cir073.