Predictive Classification of Chronic Gvhd Status by Immune Reconstitution Testing

BACKGROUND:  Many individual and treatment-related factors impact post-allogeneic hematopoietic cell transplant (HCT) immune reconstitution (IR). Our institution measures lymphocyte subsets/activation status with 8-color flow cytometry panels at important milestones (e.g., development of GVHD, day +365 evaluation), allowing for sensitive monitoring of GVHD and post-HCT IR. We sought to identify whether previously unrecognized patterns of CBC data and lymphocyte subsets are associated with GVHD status.

METHODS:  288 allogeneic HCT recipients have undergone clinical IR testing at our institution since 2010.  Included in this analysis are 43 HCT patients who underwent testing for IR with our updated antibody panel (CD3, CD4, CD8, CD8beta, CD19, CD25, CD27, CD45, CD45RO, CD56, CD69, IgD, and HLA-DR). Patients were classified according to clinical status: no GVHD (9), active acute (3), resolved acute (6), active chronic (18), and treated chronic GVHD (7). Differences in cell populations were determined by Kruskall-Wallis tests. Multivariate pattern recognition techniques of principal components/factor analysis (PCA/FA) and discriminant analysis (DA) were used for exploratory predictive modeling. 

RESULTS:  Patients with active chronic GVHD had a significant increase in CD4-CD8- T cells (p=0.01) and trended towards increased in NKT cells (p=0.06).  PCA/FA revealed 3 factors that account for 88% of the variability of the IR data (Factor 1=CD4+, CD8+, and naïve T cells, Factor 2=naïve and switched memory B cells, and Factor 3=absolute lymphocyte counts [ALC] and platelet counts). By applying all variables from the 3 factors in a DA model (Table), the ability to discriminate between active vs. no GVHD approached statistical significance (p=0.06). Adding CD4-CD8- T cells and NKT cells to the model did not allow for discrimination of active vs. treated chronic GVHD. Lymphocyte activation (e.g., by HLA-DR or CD69 expression) could be identified in a few patients who did not have overt GVHD or infections.

CONCLUSIONS: 3 independent factors – subsets of T cells, B cells, and total lymphocyte counts along with platelet counts – account for the majority of variability in IR in this cohort. CD4-CD8- T cells and NKT cells may be emerging biomarkers of chronic GVHD activity. Lymphocyte activation without overt GVHD adds to the complexity of the study of post-HCT IR. With more patient data, predictive classification of GVHD status by peripheral blood cell subset testing may become possible.

Table.  Factor and Discriminant Analysis

Factor	Percent Of Data Variance	Cumulative Percent of Variance	Cumulative Percent Misclassified on DA	Cumulative p
Factor 1 (CD4+, CD8+, CD45RO- T cells subsets)	33.91	33.91	41.9	0.54
Factor 2 (naïve B and switched memory B cell subsets)	28.19	62.10	35.0	0.10
Factor 3 (ALC + platelet count)	25.54	87.65	13.3	0.06