Decision-tree algorithm for optimized hematopoietic progenitor cell–based predictions in peripheral blood stem cell mobilization

Chia Yun Wu, Tzeon Jye Chiou, Chun Yu Liu, Feng Chang Lin, Jeong Shi Lin, Man Hsin Hung, Liang Tsai Hsiao, Chueh Chuan Yen, Jyh Pyng Gau, Hsiu Ju Yen, Giun Yi Hung, Hui Chi Hsu, Cheng Hwai Tzeng, Jing Hwang Liu, Yuan Bin Yu

Research output: Contribution to journalArticlepeer-review

8 Citations (Scopus)


BACKGROUND: Enumerating hematopoietic progenitor cells (HPCs) by using an automated hematology analyzer is a rapid, inexpensive, and simple method for predicting a successful harvest compared with enumerating circulating CD34+ cells. However, the optimal HPC cutoff count and the indicating factors to be considered for improved predicting have not yet been determined. STUDY DESIGN AND METHODS: Between 2007 and 2012, a total of 189 consecutive patients who proceeded to peripheral blood stem cell (PBSC) harvesting were retrospectively recruited. Baseline characteristics were analyzed to identify the risk factors for a failed harvest, which were defined as less than 2 × 106 CD34+ cells/kg. Variables identified by multivariate logistic regression and correlation analysis for predicting a successful harvest were subjected to classification and regression tree (CART) analysis. RESULTS: PBSCs were successfully harvested in 154 (81.5%) patients. An age of at least 60 years, a diagnosis of a solid tumor, at least five prior chemotherapy cycles, prior radiotherapy, and mobilization with granulocyte–colony-stimulating factor alone or high-dose cyclophosphamide were independent baseline predictors of poor mobilization. In CART analysis, patients with zero to two host risk factors and either higher HPC (≥28 × 106/L) or mononuclear cell (MNC; ≥3.5 × 109/L) counts were categorized as good mobilizers and their harvest success rate was 92.3%. By contrast, 30.3% of harvests were adequate in the patients with three to five host risk factors and lower HPC and MNC counts. CONCLUSION: A CART algorithm incorporating host predictors and HPC and MNC counts improves predictions in a successful harvest and might reduce the necessity of monitoring peripheral CD34+ cells.

Original languageEnglish
Pages (from-to)2042-2051
Number of pages10
Issue number8
Publication statusPublished - Aug 1 2016

ASJC Scopus subject areas

  • Immunology and Allergy
  • Immunology
  • Hematology


Dive into the research topics of 'Decision-tree algorithm for optimized hematopoietic progenitor cell–based predictions in peripheral blood stem cell mobilization'. Together they form a unique fingerprint.

Cite this