Flow Cytometric Lymphocyte Subset Enumeration for Lymphopenia Grading and C-Reactive Protein-Defined Inflammation Discrimination.

BACKGROUND

To determine whether flow cytometric lymphocyte subset enumeration could reproduce conventional absolute lymphocyte count (ALC)-based lymphopenia grading and add information for discriminating C-reactive protein (CRP)-defined inflammation.

METHODS

We retrospectively analyzed 2896 hospitalized adults with concurrent complete blood count and flow cytometric lymphocyte subset analysis (T, CD4+ T, CD8+ T, B, and natural killer cells). CRP-defined inflammation was defined as high-sensitivity CRP ≥ 10 mg/L. The 974 patients with measured CRP constituted the prespecified primary discrimination cohort. Spearman correlations, receiver operating characteristic analyses with Youden-derived cutoffs, quadratic-weighted kappa, five-fold internal validation, and multivariable logistic regression were used.

RESULTS

T-cell absolute count correlated most strongly with ALC (rho = 0.886), largely reflecting its dominant contribution to the total lymphocyte pool. Internally optimized T-cell cutoffs of 700, 573, and 356 cells/μL showed substantial agreement with conventional grading (weighted kappa = 0.792). For CRP-defined inflammation (20.9%), conventional and T-cell grades showed only modest discrimination (areas under the curve, 0.611-0.622). B-cell count had the numerically highest single-marker area under the curve (0.667), although all single-marker values remained below clinically actionable thresholds. Parsimonious multivariable models incorporating B-cell count yielded modest cross-validated areas under the curve (0.628-0.642), whereas a full five-subset model exhibited severe collinearity and unstable coefficients.

CONCLUSION

T-cell count serves as an internally optimized surrogate for conventional lymphopenia grading rather than an independent classifier. For CRP-defined inflammation, lymphocyte subset models offer limited cross-sectional discrimination, with B-cell count providing the most reproducible incremental signal.