Machine Learning Models Improve the Diagnostic Yield of Peripheral Blood Flow Cytometry

OBJECTIVES

Peripheral blood flow cytometry (PBFC) is useful for evaluating circulating hematologic malignancies (HM) but has limited diagnostic value for screening. We used machine learning to evaluate whether clinical history and CBC/differential parameters could improve PBFC utilization.

METHODS

PBFC cases with concurrent/recent CBC/differential were split into training (n = 626) and test (n = 159) cohorts. We classified PBFC results with abnormal blast/lymphoid populations as positive and used two models to predict results.

RESULTS

Positive PBFC results were seen in 58% and 21% of training cases with and without prior HM (P < .001). % neutrophils, absolute lymphocyte count, and % blasts/other cells differed significantly between positive and negative PBFC groups (areas under the curve [AUC] > 0.7). Among test cases, a decision tree model achieved 98% sensitivity and 65% specificity (AUC = 0.906). A logistic regression model achieved 100% sensitivity and 54% specificity (AUC = 0.919).

CONCLUSIONS

We outline machine learning-based triaging strategies to decrease unnecessary utilization of PBFC by 35% to 40%.

Incidence and potential causative factors associated with chronic benign neutropenia in the Kingdom of Saudi Arabia

Background

Benign neutropenia often presents in certain populations without any genotype nor phenotype. Middle East countries are among the regions where endemic cases of chronic benign neutropenia are reported in the general population with an incidence of approximately between 10-15%. Not many studies have been performed to ascertain the cause or burden associated with this condition. The objective of the current study was to identify the frequency and characterize the consequences of chronic benign neutropenia in the country of Saudi Arabia.

Results

Benign neutropenia was found to be high in the Saudi Arabia general population (up to 20%), with an average neutrophil count of 1.48 (range 0.99 – 1.95 × 10⁹cells/L), with Saudis having a higher incidence of chronic benign neutropenia compared to non-Saudis (p = <0.05). Complete blood count analyses showed significant difference in the total white cell count of neutrophils (p < 0.0001), WBC (p < 0.0001), lymphocytes (p < 0.001), monocytes (p < 0.001), eosinophils (p = 0.013) as well as the CD19 B cells (p = 0.008).

Conclusions

Our study is the first to carefully quantitate benign neutropenia in Saudi Arabia. We identified that this condition is prevalent in the middle aged population (18 years to 55 years). These individuals not only had lower neutrophil counts, but also reduced peripheral blood cells types, especially the B-lymphocyte population (CD19 subset). As B-lymphocytes are involved in antibody production and antigen recognition, a decrease might easily predispose the individuals to infectious agents. As such more mechanistic studies need to be undertaken to understand the cause and potential long-term consequences of benign neutropenia.

Value-based flow testing of chronic lymphoproliferative disorders: a quality improvement project to develop an algorithm to streamline testing and reduce costs

OBJECTIVES

Flow cytometry is essential for the evaluation of lymphoproliferative disorders (LPDs) and their classification. Flow panels routinely incorporate a large array of antibodies, making testing complex and expensive; such panels are likely unnecessary in benign cases or those with straightforward diagnoses. Our aim was to develop a more cost-effective testing strategy based on a retrospective analysis of flow studies for possible LPDs in blood.

METHODS

We identified LPD frequencies and types, as well as associated results with patient age and absolute lymphocyte count.

RESULTS

We found that the likelihood of LPDs increased with patient age and absolute lymphocyte count and that CD5-positive LPD was the most common LPD diagnosed in our institution (71% of LPDs). Using these data, we devised flow-testing algorithms with a screening test for patients at low risk of disease and a focus on CD5-positive LPD detection, with reflexing as needed.

CONCLUSIONS

We project this approach will result in a 40% decrease in antibody utilization.