Comparison of blood cell counts in leukemoid reaction and chronic myeloid leukemia: A study using Scopio blood cell counter with statistical analysis

Abstract

Leukocytosis, defined as an increase in the number of white blood cells (WBC), is a common feature in hospitalized patients. The most common form of leukocytosis is the increase of neutrophilic granulocytes. A profound increase in neutrophilic granulocytes is also known as a leukemoid reaction (LR). Since the first introduction of the term “leukemoid reaction” by Krumbhaar in 1926,¹ its diagnosis has significantly improved when various laboratory methods were introduced to differentiate LR from malignant granulocytic proliferation. The prognosis for patients with LR relies primarily on their underlying causes; however, the mortality rate remains relatively high.² Therefore, a simple and reproducible method for the initial evaluation of blood, especially when ancillary studies are unavailable, is clinically paramount for improving patient care.

Different cutoffs have been reported in the literature for defining neutrophilic LR. Some authors have used a cutoff of 25.0 × 10⁹ leukocytes/L³, while others have applied higher cutoffs of 40.0 × 10⁹/L, or 50.0 × 10⁹/L.^{4, 5} Regardless of the cutoff, neutrophilic LR shares similar features with profound increases in neutrophils and precursors mimicking chronic leukemia of granulocytic lineage. The increase in immature forms in LR, also known as left-shift, however, is predominantly late forms (segmented neutrophils, band neutrophils) with a minority of intermediate forms (metamyelocytes, myelocytes) and no increase in early forms (promyelocytes, blasts). In addition, neutrophilic LR is usually associated with morphologic features of “toxic changes,” including heavy cytoplasmic granules (toxic granules), Dohle bodies, and cytoplasmic vacuoles. Clinical assessment of patients with potential LR relies heavily on these morphologic features.

One of the most important differential diagnoses of LR is chronic myeloid leukemia (CML), which may show similar findings when evaluating blood smears. CML is a stem cell neoplasm affecting all three lineages of hematopoietic cells, with the most profound proliferation in the granulocytic lineage. The patients typically present with marked granulocytosis with a predominance of neutrophilic granulocytes and precursors. Eosinophilic and basophilic lineages are also increased but to a lesser degree. CML can be differentiated from LR by several features when evaluating blood smears. CML shows a more prominent left-shift in neutrophilic granulocytes with the entire spectrum of immature forms. Eosinophils and basophils are also increased in CML. The granulocytic cells in CML typically lack the “toxic changes” commonly seen in LR.

The vast majority of CML and LR can be readily differentiated by a review of blood smears and clinical history. Occasionally, differentiating LR from CML can be challenging on blood smear review due to morphologic overlap. Additional tests can be used to confirm the diagnosis, but these tests require more time to perform. Leukocyte alkaline phosphatase activity, performed by staining the blood smears with a substrate, is normal or high in LR but low in CML.⁶ Cytogenetics for the Philadelphia chromosome, FISH analysis for BCR/ABL1, or rtPCR for BCR/ABL1 fusion transcript provides definitive diagnosis.⁷

The purpose of this study is to compare the percentage of each category of WBC in LR and CML more precisely and provide statistical information. Although it is generally known that the patterns of left-shift in LR and CML are different, upon review of the literature, we did not find that numerical data comparing WBC categories in LR and CML existed. Therefore, this study may provide new information on LR and CML blood morphology and contribute to the literature in this area.

The study was approved by the Thomas Jefferson University Institutional Review Board. The cases were obtained by searching the electronic medical record of Thomas Jefferson University Hospital over a 2-year period from July 2020 to June 2022. We used EDTA anticoagulated blood to prepare the blood smears with Sysmex Hematology Automated Analyzer (Sysmex America, Inc. Lincolnshire, IL). The slides were stained with Romanowsky stain after air drying. All patients' manual differential data were obtained using a hematology manual differential instrument. These manual counts replaced automated ones for all cases, and the results were documented in the patients' charts. As a result, automated differential counts from our Sysmex Hematology Analyzer were not accessible for these patients. We analyzed peripheral blood smears using the Scopio Hematology Analyzer, which is a manual differential count instrument that captures high-resolution images of stained blood cells and then analyzes the images using advanced artificial intelligence algorithms to classify and count different cell types. A pathologist then verifies the results. Microsoft Excel was used to organize data and perform descriptive analysis of ranges, medians, means, and statistical analysis using t-tests. A p-value equal to or below 0.01 is considered statistically significant.

Thirty-three cases of LR and 28 cases of CML were assessed. The patients with CML ranged from 21 to 83 years old, with a median age of 56.5. The male-to-female ratio was 2.1:1. All the CML patients presented as CML at the chronic phase. The patients with LR ranged from 18 to 90 years old, with a median age of 61. The male-to-female ratio was 1:1.4. The causes of LR were infection (45%), inflammation (43%), and drug (18%). The median count of WBC in CML was 86.75 × 10⁹/L (31.0 × 10⁹−365.1 × 10⁹), while the median count of WBC in LR was 33.5 × 10⁹/L (25.6 × 10⁹− 67.4 × 10⁹). The manual differential counts of CML and LR cases showed that CML had significantly higher percentages of metamyelocytes, myelocytes, promyelocytes, blasts, eosinophils, and basophils. However, LR had a significantly higher percentage of the neutrophil count. The band neutrophils and monocytes showed no significant differences between CML and LR (Table 1).

Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic performance of WBC counts in differentiating LR from chronic myelogenous leukemia. The area under the ROC curve (AUC) was 0.92 (95% CI: 0.85–0.99), indicating good discriminative ability. The optimal cutoff value of 45 × 10⁹/L provides a sensitivity of 0.93 and a specificity of 0.76, ensuring that most CML cases are identified. The cutoff value of 45 × 10⁹/L is recommended for clinical use to differentiate between LR and CML. However, due to the false negative and false positive results, this cutoff value cannot be used in isolation and should be used in combination with all available laboratory results for the most accurate assessment.

The results provide the percentages and ranges of individual granulocyte categories in CML and LR with statistical data. These results confirm that granulocytosis in CML tends to have a more profound left shift with an increase in intermediate and early forms. At the same time, LR primarily affects late forms with increased segmented neutrophilic granulocytes. In addition, blasts are more likely to be seen in CML and are consistently not detected in LR associated with infection and inflammation. There is no significant difference in band neutrophil count between CML and LR. As expected, eosinophils and basophils are increased in CML as part of the neoplastic population, while bacterial infection and chronic disease-associated inflammation typically do not affect eosinophils and basophils. Predictably, monocytes are normally unaffected by both entities, and there was no significant monocyte count difference between CML and LR. This study contributs to the literature with the numeric ranges and statistics in blood cell categories in CML and LR, which can serve as a helpful guide for practicing pathologists and can be included in textbooks for medical education. The study was limited by a small sample size and selection bias with parimarily hospitalized patients from a single institution. Larger, multi-institution studies in the future may be warranted to confirm our findings.