Relationship between leukopenia and mortality among patients with hematological malignancies Authors: Caibao Hu, Qian Li, Xinyuan Ding Published: 2024 To the editor, We read with great interest the recent study [1] published in Critical Care examining sepsis mortality among patients with hematological malignancies admitted to intensive care units between 2000 and 2022. The study’s findings highlighted a notable decline in in-hospital mortality rates for patients with hematological malignancies (from 55.6% in 2000 to 23.1% in 2021) and those without hematological malignancies (from 33.1 to 14.4%). Additionally, the authors reported that, after adjusting for covariates, leukopenia (white cell count < 1.0 × 10⁹ cells/L) was associated with increased mortality in patients without hematological malignancies (p < 0.001) but not in patients with hematological malignancies (p = 0.60). This is an important clinical topic; however, there may be a potential methodological concern regarding how the relationship between leukopenia and mortality was analyzed. Specifically, the study appears to have created two separate variables in the multivariate logistic regression analysis: one representing leukopenia in patients with hematological malignancies and the other in patients without hematological malignancies—to explore the interaction between leukopenia and hematological malignancy status. While this approach may provide group-specific insights, it can introduce bias and does not adhere fully to the principles of interaction effect analysis. Interaction effects occur when the impact of one independent variable on a dependent variable is modified by another independent variable. In this study, the association between leukopenia and mortality likely depends on the presence of hematological malignancies. A standard approach to testing this hypothesis is to include an interaction term (e.g., leukopenia × hematological malignancy status) in a unified regression model. This approach avoids splitting the sample or creating separate variables and offers a direct assessment of the interaction effect. Furthermore, leukopenia was separated into two variables in the current study, and the association between leukopenia and mortality was significant in the crude analysis (OR 2.37, p < 0.001) but became non-significant in the multivariate logistic regression analysis (OR 0.98, p = 0.60). This discrepancy is likely due to multicollinearity. Multicollinearity arises when independent variables are highly correlated, which can distort parameter estimates in regression models. In this context, leukopenia and hematological malignancy are likely strongly correlated, as a substantial proportion of patients with hematological malignancies also exhibit leukopenia. Conversely, leukopenia is far less common among patients without hematological malignancies. This underlying correlation may explain why leukopenia remains significant in one group while becoming non-significant when combined with hematological malignancy in the adjusted logistic model. Addressing these methodological considerations would enhance the study’s ability to clarify the complex relationship between leukopenia, hematological malignancy status, and mortality. Such clarity would provide more robust evidence to inform clinical decision-making in critical care settings. We commend Dr. MacPhail and colleagues for their valuable contributions to this field. Acknowledgements Not applicable. Author contributions Caibao Hu and Qian Li wrote the letter and Xinyuan Ding Raised the question. All authors have reviewed and approved the letter. Funding None. Availability of data and materials Not applicable. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Reference Associated Data Data Availability Statement Not applicable.