6th ESACP Congress, Heidelberg, April 7-11, 1999

A144
AUTOMATED CLASSIFICATION OF BLOOD LEUKOCYTE AND BONE MARROW CELLS FROM FLOW CYTOMETRICALLY IMMUNOPHENOTYPED CHRONIC LYMPHATIC LEUKEMIAS (B-CLL
Valet G 1, Höffkes HG 2

1) AG Zellbiochemie, Max-Planck-Institut für Biochemie, Martinsried, 2) Division of Hematology, Internal Medicine, University of Magdeburg, Germany

Background:Thorough analysis of flow cytometrically determined light chain restriction in B-CLL is clinically required but not always easily obtained especially in cases of weakly expressed cell surface immunoglobulins. Aim & Methods: It was investigated whether the CLASSIF1 analysis (http://www.biochem.mpg.de/valet/classif1.html) of kappa/CD19/5, lambda/CD19/5, CD45/14/20 and CD8/4/3 immunophenotype list mode files of blood leukocytes (BL) and of bone marrow cells (BM) permitted automated data classification according to an earlier elaborated classification scheme for lymphomas (Cytometry(CCC) 30:275-288(1997)). Results: BL or BM cells of normal and APAAP/histologically proven kappa or lambda expressing B-CLL patients served as learning set. Upon completion of the learning phase, the reclassified learning set classified with predictive values of 100, 100, 96.3% for normal, kappa, lambda B-CLL for BM samples (n=22/26/27). Similarly BM samples classified with predictive values of 100, 96.2, 96.3% when BM cells were classifed against BL cells of normal individual as reference (n=58/26/27). BL classifed with predictive values of 91.8, 91.2, 94.1% (n=58/34/38). Clear, prediagnosed BM (n=12) and BL (n=10) samples which were unknown to the classifiers served as first test set. They were correctly classified in 91.2 and 100% of the cases. Unclassifiable BM (n=8) nd BL (n=7) samples served as a second test set. All samples were unambiguously classified by the CLASSIF1 classifier. The classification "truth" in these cases was provided by logically consistent classifications of e.g. BM and BL samples from the same patient taken on different days or by reprobing a patient on several days from the same cell compartment. Conclusion: CLASSIF1 analysis is a suitable automated and laboratory independent alternative to the current computer assisted manual analysis in B-CLL immunophenotyping. Clear classifiation results are obtained in manually unclassifiable cases, the analysis is signficantly faster than manual analysis and can be operated on-line with the flow cytometric measurements.