engleski

Decision tree approach to the immunophenotype based prognosis of B-CLL

Chronic lymphocytic leukemia (CLL) is a lymphoproliferative disorder characterized by a clonal expansion of a small long-lived immature CD5+ B-cells. Appearance of a malignant B-cell clone and development of late immunodeficiency sequels supports the assumption that impaired T-cell regulatory functions might be important triggering and perpetuating pathogenic mechanisms responsible for CLL. In this study an attempt was made to elucidate how quantitatively imbalanced and/or functionally impaired B- and T-cell subsets may influence the natural evolution and prognosis of CLL. Three-color immunophenotyping with a panel consisting of different combinations of CD19, CD5, CD72, CD23, CD21, CD4, CD45RA and CD45RO monoclonal antibodies was performed on whole blood samples in forty-two nontreated B-CLL patients. In addition, serum level of soluble CD23 (sCD23) was measured using an immunoenzyme assay. The patients were classified according to the modified Rai staging system into low, intermediate and high risk groups. In order to evaluate the degree of association between laboratory results and clinical stages in patients with B-CLL we applied machine learning approach. The C4.5 algorithm generates a classifier in the form of decission tree consisting of branching nodes, specifying test values (attributes), and the leafs, showing a class reached after satisfying pertinent attribute. In B-CLL patients, beside sCD23, three linked phenotypic features: % of CD23+ lymphocytes, CD4+CD45RA+RO- cells as % of CD4+ lymphocytes, and the absolute number of CD4+CD45RA+RO+ cells, made the nodes. Three different combinations of these features were characteristic for the high risk, distinguishing them from low and intermediate risk form of CLL. This indicate the important role of T-cell subset in pathogenic mechanisms responsible for CLL.

chronic lymphocytic leukemia; immunophenotyping; sCD23; machine learning

nije evidentirano