Use of pathology-specific peripheral blood CD34 thresholds to predict leukapheresis CD34 content with optimal accuracy: a bicentric analysis of 299 leukaphereses

CD34+ cell counts in peripheral blood (PB) and corresponding numbers of CD34+ cells and colony-forming units-granulocyte/macrophage (CFU-GM) in 299 leukapheresis products of 209 patients undergoing PB progenitor cell (PBPC) mobilization for autologous transplantation in two different centers were analyzed and compared according to diagnosis: non-Hodgkin lymphoma (NHL, 94 leukaphereses), multiple myeloma (MM, 75), Hodgkin's disease (HD, 37), solid tumors (35), and chronic myeloid leukemia (CML, 32). Without separating disease entities, correlations between PB CD34+ cell counts and leukapheresis content of CD34+ cells (r>0.83, P0.81, P 10(6) CD34/kg was determined. This threshold was higher in center 1 than in center 2, and its predictive accuracy (91.4%, i.e., prediction correct 91.4% of the time) was significantly lower than in center 2 (98.4%, P=0.02). When data were analyzed by pathology, PB CD34+ cell counts and leukapheresis content of CD34+ cells and CFU-GM remained well correlated, and in both centers PB CD34 thresholds predictive of a yield > 10(6) CD34/kg per leukapheresis could be determined for each pathology. For most patients, pathology-specific PB CD34 thresholds could be obtained directly from the equation of the PB CD34/leukapheresis CD34 correlation curve; they varied depending on both pathology and center (range: 7-20 x 10(6) CD34/l). Pathology-specific thresholds predicted a leukapheresis yield > or = 10(6) CD34/kg accurately 100% of the time for MM patients in center 2 and HD and solid tumor patients of both centers, resulting in overall rates of accurate prediction of sufficient graft CD34 content of 96.6% in center 1 and 98.9% in center 2.