PDF
Citation
Print
INTRODUCTION
Differential white blood cell (WBC) count is a diagnostic test used for various diseases and is particularly significant in the diagnosis of acute leukemia. Confirming the presence of blasts in peripheral blood smears (PBS) of patients suspected of having acute leukemia is a crucial laboratory task. However, this method is time-consuming and requires experienced laboratory technicians. To overcome these challenges, numerous automated digital cell morphology analyzers have been developed [1]. One such device is the Mindray MC-80 automated digital cell morphology analyzer (Shenzhen, Guangdong, China). It functions as a digital morphology analyzer integrated into a cellular analysis line that performs complete blood count (CBC) and WBC differential. Several studies have shown that various digital morphology analyzers exhibit comparable performance to manual microscopy in distinguishing between normal cells and blasts [2-7]. However, to the best of our knowledge, no reports have been published on the capability of the MC-80 to analyze samples containing a considerable number of abnormal cells, such as promonocytes and abnormal promyelocytes, which differ from typical myeloblasts. These abnormal cells are considered blast equivalents and should be regarded as blasts [8]. Because the morphology of these cells differs from that of typical myeloblasts, even skilled experts may experience inconsistencies in interpreting their proportions. If inexperienced examiners or unvalidated morphology analyzers miss these cells, the consequences can be potentially fatal. This study assessed the degree of absolute reliability in blast equivalent proportions between the MC-80 and manual microscopy and investigated whether digital morphology analyzers, such as MC-80, could accurately diagnose acute leukemia in a clinical laboratory.
MATERIALS AND METHODS
1. MC-80 analyzer
Mindray MC-80 is a newly developed automated digital cell morphology analyzer that evaluates PBS images to differentiate and classify WBCs into cell types. The equipment can be integrated with Mindray’s CBC analyzers and slide makers, such as BC-6800, to form an integrated cellular analysis line capable of detecting abnormalities not only in WBCs but also in red blood cells and platelets. This study used pre-existing PBS samples for retrospective analysis; thus, the CBC evaluation and slide preparation capacities of the integrated system were not assessed. The focus was solely on evaluating the ability of the MC-80 to analyze cell morphology.
Two examiners from the laboratory participated in manual microscopy for comparison purposes. Examiner 1 (E1) was a resident of the clinical laboratory department who had clinically interpreted over 50,000 PBS samples in the past three years. Examiner 2 (E2), a clinical laboratory technologist, had 25 years of experience in hematological cell morphology reviews. These two examiners established a consensus on cell morphology identification through regular training and discussion in the past three years.
2. Sample collection and study design
Blood samples analyzed in this study were collected from patients who visited Yeungnam University Medical Center between January 2018 and February 2023 and underwent bone marrow examinations. Patients who tested positive for promonocytes or abnormal promyelocytes in the PBS cell morphology analysis were included in the study.
Patients in Group A had monoblasts or promonocytes in their PBS. All patients in this group tested positive for CD14 or CD64 antigen expression on bone marrow fiow cytometry. Group A included 24 patients (Table 1). Patients in Group B had abnormal promyelocytes in their PBS. All patients in this group tested positive for myeloid lineage antigen (CD13, CD33, CD117, and cMPO) expression and negative for HLA-DR expression in the bone marrow fiow cytometry analysis. Furthermore, these patients tested positive for the PML-RARA fusion gene on nested reverse-transcription polymerase chain reaction analysis of bone marrow samples. Group B included 16 patients (Table 2).
The PBS samples prepared for diagnostic evaluation were analyzed. All samples were prepared using the wedge method and stained with the Wright–Giemsa stain. The WBC differential index of the MC-80 includes blasts and abnormal promyelocytes, without distinguishing between myeloblasts, monoblasts, and promonocytes within blasts. Therefore, it was decided to consider both the blasts and abnormal promyelocytes identified by the MC-80 as leukemic cells. Manual examiners performed the same tasks to ensure uniformity.
Initially, the MC-80 was used to analyze 200 WBCs from each sample and calculate the proportion of leukemic cells. E1 and E2 sequentially analyzed 200 WBCs from the same samples and determined the proportion of leukemic cells. After all sample evaluations were completed, statistical analyses were conducted to assess the degree of absolute reliability of the proportion of leukemic cells between the examiners. Statistical analyses were performed to compare the results between the MC-80 and each manual examiner, and between the manual examiners themselves.
3. Statistical analyses
All statistical analyses were conducted using JASP (version 0.17.2.1; University of Amsterdam, Amsterdam, Netherlands). The intraclass correlation coefficient (ICC) was used to assess the absolute reliability of the proportion of leukemic cells between the MC-80 and each examiner. The ICC is a statistical measure used to evaluate the reliability and consistency of measurements made by different observers of the same quantity. It is often used in reliability studies, where repeated measurements are performed under varying conditions. The ICC values range from 0 to 1, where 1 indicates perfect reliability. The ICC estimates and their 95% confidence intervals (CIs) were calculated using a single rating (k=2), absolute reliability, and a two-way random-effects model. This is equivalent to the “ICC (2,1)” as defined by Shrout and Fleiss [9, 10]. The interpretation criteria used for the ICC values were based on the guidelines previously reported by Koo and Li (Table 3) [9].
RESULTS
1. Group A
For Group A samples, both the MC-80 and the two examiners were able to count 200 WBCs. The MC-80 failed to detect leukemic cells in one sample. Leukemic cells were detected in all the other samples, and the lowest percentage of leukemic cells detected in the total cell count for Groups A and B was 1.5% (Table 4).
The ICC value obtained between E1 and E2 was 0.962 (95% CI: 0.856–0.986). Therefore, based on statistical inference, the level of reliability between the two examiners was “good” to “excellent.” The ICC between MC-80 and E1 was 0.772 (95% CI: 0.306–0.915). The level of reliability was thus considered “good,” and based on statistical inference, ranged from “poor” to “excellent.” The ICC value obtained between MC-80 and E2 was 0.780 (95% CI: 0.544–0.900). Therefore, the level of reliability was considered “good,” and based on statistical inference, ranged from “moderate” to “excellent” (Table 5).
2. Group B
For Group B samples, both examiners counted 200 cells. However, of the six samples, MC-80 failed to count 200 cells. The WBC counts of these six samples ranged from 0 to 167. The MC-80 failed to detect WBCs in one sample. The sample (case no. B15) was excluded from final statistical analyses. Furthermore, the MC-80 failed to detect leukemic cells in three samples, including one excluded sample. Conversely, leukemic cells were detected in all samples, with the lowest value being 0.5% (Table 6).
The ICC between E1 and E2 was 0.996 (95% CI: 0.988–0.999). Therefore, based on statistical inference, the level of reliability was considered “excellent.” The ICC between MC-80 and E1 cells was 0.926 (95% CI: 0.660–0.978). Therefore, the level of reliability was considered “excellent,” and based on statistical inference, ranged from “moderate” to “excellent.” The ICC value between MC-80 and E2 was 0.927 (95% CI: 0.729 to 0.977). Therefore, the level of reliability was considered “excellent,” and based on statistical inference, ranged from “moderate” to “excellent” (Table 7).
DISCUSSION
A previous study indicated that the ability of MC-80 to detect blast cells in acute leukemia showed no significant difference compared with that of the manual method [11]. However, this study, which evaluated the ability to discriminate blast equivalents, demonstrated inferior performance compared with the reliability observed among manual examiners. The proportion of leukemic cells calculated using MC-80 in the samples containing promonocytes and abnormal promyelocytes was similar to that of the examiners. However, it had a lower level of reliability than a manual examination. This finding was particularly pronounced in samples containing promonocytes. While the examiners detected at least one blast equivalent in all 40 samples, the MC-80 failed to detect leukemic cells in four samples. These detection failures occurred in samples with a low proportion of leukemic cells or a low WBC count (<1,000/μL). These findings indicate the clear inferiority of the MC-80 for manual examination, especially in such cases. In particular, in the B15 sample, MC-80 failed to detect WBC, which appeared to be due to the low quality of the sample. However, because experienced laboratory personnel can review and validate images captured by such digital morphology analyzers, we consider the performance of MC-80 sufficient for initial screening, particularly in patients suspected of having conditions, such as acute leukemia. In cases where the automated system fails to differentiate an appropriate number of cells, setting a fiag and incorporating a system in which human inspectors verify the samples using the manual method are expected to help distinguish the cells in those samples. Thus, MC-80 can provide valuable assistance during the first-stage screening process.
This study has two limitations. First, the sample size was relatively small; only 40 samples were analyzed. Despite efforts to obtain a larger number of samples, the available samples were from before 2018 and were thus relatively old, presenting quality assurance challenges. Therefore, analyses were conducted with a limited sample size. Second, this study has significant technical limitations. The MC-80 was designed for integration with other Mindray equipment, which can provide additional useful information for cell differentiation. In this study, the analysis was conducted solely based on captured images without such additional data, creating a disparity from the actual clinical environment in which the MC-80 functions. Thus, better performance could be achieved if the samples were analyzed using the entire Mindray cellular analyzer line. Further investigations involving a larger dataset and improvements in research design may produce better outcomes that confirm the utility of the MC-80 in cell morphology analysis.
XML Download