Predicting the CD34 content of mobilized peripheral blood leukapheresis products: single institution experience over...

Since the standardization of CD34 measurement by flow cytometry, predictors of leukapheresis CD34 yield have played a pivotal role in planning donor leukaphereses. We describe here a single institution's experience with a multivariate predictor that was used for 2,929 products without alteration for 20 years.

The ordinary least squares regression model variables included log peripheral CD34 count, collection duration (3- versus 4-hours), collection number, donor sex, and transplant type.

During the study period we changed flow cytometers twice and leukapheresis instruments once. During the Cobe Spectra era the predictor explained 90% of the variability in CD34 collection yield for autologous transplants (r² = 0.90), and 70% for allogeneic transplants with an overall sensitivity to predict a CD34 yield of ≥ 1 × 10⁶/kg of 97.7%, and specificity of 81.4%.

Implemented prospectively with real-time result reporting, the model allowed us to predict CD34 yield with both 3- and 4-hour collection scenarios. Given this guidance, 3-hour collections were selected by the clinical team 25% of the time, saving patient leukapheresis time and resources. When faced with a prediction of < 1 × 10⁶ CD34/kg, the clinical team chose to defer collection 72% of the time. In instances where leukapheresis was performed despite a poor predicted outcome, 85% of patients collected on the Cobe Spectra, and 92% of patients collected on the Optia, failed to collect at least 1 × 10⁶ CD34/kg. A revised model is tested retrospectively on Optia data, and suggestions for further improvements are discussed.