Clinical Significance of CBC and WBC Morphology in the Diagnosis and Clinical Course of COVID-19 Infection

Olga Pozdnyakova, MD, PhD; Nathan T. Connell, MD, MPH; Elisabeth M. Battinelli, MD, PhD; Jean M. Connors, MD; Geoffrey Fell, MS; Annette S. Kim, MD, PhD

Disclosures

Am J Clin Pathol. 2021;155(3):364-375. 

In This Article

Materials and Methods

The study included 90 consecutive patients with COVID-19 admitted to our hospital between March 14, 2020, and April 14, 2020, as well as 30 ICU patients negative for COVID-19. The study was approved by the Institutional Human Research Committee. Clinical presentation of patients with COVID-19 varied from mild to severe disease: 51 patients were admitted to the ICU, and 39 patients were followed in the non-ICU settings. There were no significant differences in underlying comorbidities between COVID-19–positive ICU and non-ICU patients, with hypertension and cardiovascular disease being the most common (39% vs 36%, respectively), followed by diabetes (21% vs 15%, respectively). Nonhematologic malignancy was present in 15% in both groups, and hematologic malignancy was present in 12% of ICU patients (6% in remission) and 5% of non-ICU patients. Acute kidney injury was present in 6% and 5%, respectively. Approximately 15% of patients in both groups did not have significant underlying comorbidities.

The ICU patients negative for COVID-19 included age- and sex-matched patients who required ICU level of care as a marker for severity of clinical illness similar to COVID-19–positive patients and included patients with severe sepsis and/or acute respiratory distress syndrome to parallel the respiratory distress found in the COVID-19–positive ICU cohort. Overall, 70% of COVID-19–negative ICU patients had bacterial infections, with one patient with concurrent influenza B infection. The underlying comorbidities included cardiovascular disease (80% of patients), ischemic or nonischemic heart failure (65% of patients), diabetes (25% of patients), and atrial fibrillation (20% of patients). To ensure there were no other hospital-based confounding factors that would have skewed the results, we chose patients who were contemporaneously in the ICU with other patients with COVID-19 (but COVID-19 negative themselves).

Routine CBC with WBC differential was performed on or near the date of COVID-19 diagnosis (confirmed by SARS-CoV-2 reverse transcription polymerase chain reaction) and/or admission date (for transferred patients) on Sysmex XN-9000 hematology analyzers as a part of routine clinical care. An automated six-part WBC differential included absolute count of lymphocytes, monocytes, neutrophils, eosinophils, basophils, and immature granulocytes, the latter representing an automated count of promyelocytes, myelocytes, and metamyelocytes in peripheral blood. Research parameters associated with neutrophil, lymphocyte, and monocyte morphology (neutrophil lateral scatter light intensity [NE-SSC], neutrophil fluorescent light intensity [NE-SFL], neutrophil forward scatter light intensity [NE-FSC], lymphocyte lateral scattered light intensity [LY-X], lymphocyte fluorescent light intensity [LY-Y], lymphocyte forward scattered light intensity [LY-Z], monocyte lateral scattered light intensity [MO-X], monocyte fluorescent light intensity [MO-Y], monocyte forward scattered light intensity [MO-Z]), measured on Sysmex-XN hematology analyzers but not reported as a part of CBC, were collected in addition to the routine CBC parameters. WBC morphology was analyzed as changes from normal expected/baseline morphology by two independent board-certified hematopathologists (O.P. and A.S.K.) to review individual abnormal features not encapsulated in the differential count and/or the advanced research parameters. The morphologic changes for neutrophils included toxic granulation, cytoplasmic vacuolization, Howell-Jolly body-like inclusions, and Döhle bodies; for monocytes, the changes included the presence of large coalescing cytoplasmic vacuoles; for lymphocytes, the changes included the presence of cytoplasmic vacuoles, large granular lymphocytes, and atypical lymphocytes, including plasmacytoid forms; and for eosinophils, the changes included the presence of cytoplasmic vacuolization Image 1. Independent scoring of WBC morphology (4-point scale: 0, absent; 1, present in up to 10% of cell lineage; 2, present in 11%-25% of cell lineage; or 3, present in >25% of cell lineage) via Cellavision DM9600 was performed by two board-certified hematopathologists (O.P. and A.S.K.), and significant (>1-point) discrepancies were resolved by adjudication. In addition, we collected selected markers to benchmark systemic inflammation: ferritin and C-reactive protein (CRP) levels. These markers were measured on a cobas 8000 analyzer (Roche Diagnostics). For COVID-19–positive patients, CRP and ferritin results were collected at admission or closest to the time of the positive test (generally within 48 hours), in line with the institutional guidelines for patients with COVID-19.

Image 1.

A composite image of peripheral blood WBCs showing a spectrum of morphologic changes in coronavirus disease 2019 (COVID-19)–positive patients (A-C) and COVID-19–negative patients (D) (Wright-Giemsa, ×100). A, Segmented neutrophilia with vacuolization and toxic granulation (top left), Howell-Jolly body-like inclusions (top right), pseudo–Pelger-Huet nuclei (bottom left), and eosinophil with cytoplasmic vacuoles (bottom right). B, Large granular lymphocyte (top left), lymphocyte with cytoplasmic vacuolization (top right), atypical lymphocyte (bottom left), and plasmacytoid lymphocyte (bottom right). C, Atypical monocytes with large coalescing cytoplasmic vacuoles. In contrast, monocytes in COVID-19–negative patients (D) show only occasional small cytoplasmic vacuoles.

The impact of each parameter on the disease status (ICU vs non-ICU) was estimated using a univariate logistic regression model with α = 0.05. From the univariate analysis, a pool of significant candidate morphologic predictors adjusted for sex was then selected for the multivariate logistic regression. The candidate multivariate model was chosen, based on parsimony, from the five models with the smallest Akaike information criterion estimated from the backwards selection procedure. All modeling was done using the "glm package" from R version 4.0.0 (Free Software Foundation's GNU project). Overall, we examined 19 laboratory and morphology parameters associated with ICU status. The significance of research CBC parameters between ICU and non-ICU COVID-19–positive patients and COVID-19–positive and COVID-19–negative patients was assessed by Student t test using 2018 GraphPad Software.

processing....