Assessing agreement between CD34 enumeration by flow cytometry and volumetric analysis


Prior to replacement of an established method for CD34 enumeration by an alternative approach, evaluation of the agreement between the methods is essential. In this study, the comparison of two assays was evaluated according to the recommendation of Bland and Altman describing the agreement between two methods where the true value is not known. CD34 enumeration was performed on blood or leukapheresis product from 105 patients by flow cytometry (dual platform assay) and volumetric analysis (single platform assay). Both the flow cytometric and the volumetric analysis showed poor reproducibility for measures lower than approximately 9 CD34+ cells/mm3. For values higher than 29 CD34+ cells/mm3, evaluation of the agreement demonstrated a difference between the single and dual platform assay, where CD34 enumeration by the volumetric analysis demonstrated values 73–80% of the flow cytometric value. The difference between the two assays could be due to several technical pitfalls which are discussed.


Quality assessment of haematopoietic stem and progenitor cell grafts includes enumeration of CD34+ cells in accordance with the ‘Standards for Blood and Marrow Progenitor Cell Collection, Processing and Transplantation’ reviewed by Serke and Johnsen.1 Rapid and accurate measurement of CD34+ cells has increasing clinical significance as the use of stem cell mobilisation is widening. Several protocols for flow cytometry enumeration of CD34+ cells have been proposed2,3,4,5 and kits for absolute CD34+ cell enumeration are commercially available (Biometric Imaging; Becton Dickinson Biosciences, San Jose, CA, USA). Prior to replacement of an established method by an alternative approach, evaluation of the agreement between the two methods is essential.

We present a comparison of two methods for CD34+ enumeration: A new semi-automated flow cytometry method based on volumetric analysis, in which the number of CD34+ cells is expressed per unit of volume (single platform assay), is compared with standard flow cytometry, in which counting the percentage of CD34+ cells is multiplied with a number of cells per unit of volume obtained by automated analysis (dual platform technology).

Often, such comparisons of two different ways of measuring the same parameter are performed by analysis of correlation coefficients. As both methods are indirect, the true value remains unknown and the situation is different from calibration, in which one compares measured values with a known standard. The use of correlation in comparing two indirect methods is inappropriate and can be misleading, as a high correlation does not necessarily mean that the two methods agree. Bland and Altman6 have discussed this matter and proposed an alternative approach. The Bland and Altman method, which is applied here, is based on graphical techniques, simple calculations and assessment of reproducibility.

Materials and methods

Duplicate analysis of identical samples was performed at two different laboratories, The Stem Cell Laboratory and The Department of Clinical Immunology at Herlev Hospital. The samples comprised blood (n = 78) and leukapheresis products (n = 27) from a consecutive series of patients primed and mobilised as candidates for high-dose therapy and stem cell transplantation.

Patient samples

Blood samples or leukapheresis products were obtained from patients with either malignant lymphoma or multiple myeloma. Blood progenitor mobilisation was performed by priming with high-dose cyclophosphamide 4 g/m2 in combination with s.c. rhG-CSF 10 μg/kg/day. Stem cell harvest was initiated during marrow regeneration guided by the blood CD34+ cell level.7 Leukapheresis was performed on a Fenwal CS-3000 Plus separator according to the manufacturer's instructions.

The Nordic standard for CD34+ cell enumeration

Enumeration of CD34+ cells was performed according to a modified Nordic standard5 on fresh (<4 h after harvest) samples of 50 μl blood cell suspension containing 0.5–1.0 × 106 cells per test. The samples were incubated for 15 min at room temperature with 50 μl CD45 FITC from Becton Dickinson (BD) diluted 1:10, and 50 μl HPCA-2 PE (recognizes the class III epitope of the CD34 antigen) (BD) diluted 1:5 or IgG1 PE (BD) diluted 1:10. Erythrocyte lysis was performed after antibody staining for 10 min using ortho lysing solution, which is no longer available. The reagent is a 10× dilution of an aqueous ammonium chloride based lysing solution containing 1.5 M NH4Cl, 100 mM NaHCO3 and 10 mM disodium EDTA at pH 7.4. The samples were washed once in PBS and analysed immediately. Duplicate analysis was performed to test the reproducibility.

Data acquisition was performed on a FACScan instrument (BD) and analysis was performed using CellQuest software. An arbitrary forward scatter (FSC) threshold to exclude erythrocytes, platelets, subcellular particles, cell aggregates, and debris was set in data acquisition. A minimum of 50 000 events was acquired in each sample. In data analysis, the FSC threshold was adjusted to exclude debris. The CD34+ cell population was defined as a single population of CD45-positive events expressing CD34 bright and low side scatter. A minimum of 50 events within the positive population was considered significant for counting accuracy. The percentage of CD34+ cells was calculated as the number of CD34+ events as defined above, divided by the total number of CD45+ cells aquired, excluding debris.

Leukocyte counts were performed in duplicates on an Automatic Cell Counter MD II (Coulter, Miami, FL, USA). The total number of CD34+ cells was calculated from the CD34 percentage by multiplication with the mean leukocyte count.

The volumetric capillary CD34 assay

This assay is intended for use with the IMAGN 2000 Microvolume Fluorimeter for determination of the concentration of CD34+ cells in peripheral blood and leukapheresis products. This instrument enables the imaging and quantification of cells in precisely defined volumes and the STELLer CD34 assay has been developed to measure such numbers.

The sample is incubated with an anti-human mouse monoclonal antibody (mAb) recognizing a class III epitope of the CD34 antigen. The mAb is labelled with a fluorescent dye, Cy5 (Amersham Life Sciences, Cleveland, OH, USA) with an emission peak at 665 nm, which is excited at 633 nm by a low-power helium neon laser. The use of this laser minimises the interference of the autofluorescence of the red cells, making it possible to perform the measurement in unlysed blood samples. The sample is drawn into a glass capillary of a precisely known volume. The laser beam scans the capillary and fluorescent emissions from the labelled target cells are captured by the system's optics as peaks above the background fluorescence. Image analysis software evaluates the electronic images on the basis of size, fluorescence intensity, colour, and shape. Results are reported as the number of labelled cells per unit of volume.

Fifty μl of sample were added to the reaction vial, agitated, and incubated for 30 min at room temperature. After incubation, 90 μl of reaction diluent were added to the reaction vial and mixed. After mixing, 40 μl of the diluted sample were pipetted into the volumetric capillary. Each cartridge had two volumetric capillaries; thus, each sample was analysed in duplicate and the mean value taken as the final value.

Data sampling, presentation and statistics

Statistical analysis of the dataset was performed in accordance with the Bland and Altman recommendation.6,8 This analysis is based on graphical techniques, simple calculations, and transformations. In brief, comparison of the two methods is divided into two steps. The first step is an analysis of the reproducibility. Reproducibility is the method's ability to generate the same counts on the same blood sample. This analysis was performed for each of the two methods. The differences of the duplicates were plotted against the mean of the duplicate values which made it possible to determine a certain value of enumerated CD34+ cells/mm3 that is the lower limit of acceptable reproducibility. This applies when the differences are very high compared to the mean of the two measurements. The comparison of different methods becomes uncertain when the reproducibility of the methods is poor. The second step is an analysis of agreement. Again the difference was plotted against their mean. However, the values used for this analysis were the mean of duplicate values obtained by the two methods. For both microvolume fluorimetry and flow cytometry, plotting differences against mean values reflects a clear systematic pattern. For both methods, the difference between the first and second measurement seems to increase with the number of cells/mm3 and further, there is a clustering of observations for low values of cells/mm3. This systematic pattern is particularly present for the STELLer CD34 assay. By the use of a logarithmic transformation of data this relationship is reduced and the reproducibility of the method and agreement between the two methods are therefore based on logarithmic transformed data.

In the agreement analysis of microvolume fluorimetry and flow cytometry, a non-uniform distribution of data persists after the logarithmic transformation. A model of the relationship between the difference and the average is therefore estimated. The estimated function is given by the simple regression function  = b0 + b1A, where D is the difference, A the average and b1 is the slope of the line.8 The limits of agreement are given by  ± √\(\overline{π/2SD}\), where SD is the standard error of regression.



A log transformation was applied and the reproducibility in each of the two methods was evaluated by plots showing the difference of duplicate values against their mean. The 95% limits of agreement are given by  ± 1.96SD.8 The 95% limits of agreement for flow cytometry (Figure 1) are [In(−0.49):In(0.71)] or back transformed [0.61:2.03]. The 95% limits of agreement for STELLer CD34 assay (Figure 2) are [In(−0.49):In(0.59)] or back transformed [0.61:1.80]. The limits of agreement give the range of difference in which 95% of the observations are expected to be included. The ranges are wide for both methods, ie measurements for flow cytometry and the STELLer assay could vary up to 103% and 80%, respectively. Both methods have reproducibility problems for values below 9 cells/mm3 (e2.2). Poor reproducibility by low counts of CD34+ cells in both methods limits the possible amount of agreement in this measuring interval. Further, a comparison of Figures 1 and 2 reveals that the reproducibility of flow cytometry is relatively poor at low CD34 values, but better at high values compared to volumetric analysis.

Figure 1

Reproducibility analysis: flow cytometry. Difference of CD34+ cells obtained by duplicate flow cytometry analysis against their mean. Log transformed data. Poor reproducibility for values lower than ≈2.2 (≈9 CD34+ cells/mm3) which exceed two standard deviations (full lines). The truncated line indicates mean. n = 105.

Figure 2

Reproducibility analysis: STELLer CD34 assay. Difference of CD34+ cells obtained by the STELLer CD34 assay by duplicate analysis against their mean. Log transformed data. Acceptable repeatability for values higher than ≈2.2 (≈9 CD34+ cells/mm3) which are less than two standard deviations (full lines). The truncated line indicates mean. n = 105.


For both methods the difference of log transformed data was plotted against their mean (Figure 3). The plot displays a lack of agreement between measurements. For values below ≈9 CD34+ cells/mm3, the STELLer CD34 assay measures higher levels of CD34+ cells/mm3 than flow cytometry. As described above, both methods show poor reproducibility below this level which complicates any conclusions on agreement for counts lower than ≈9 CD34+ cells/mm3.

Figure 3

Agreement analysis: flow cytometry vs STELLer CD34 assay. Difference of CD34+ cells enumerated by the STELLer CD34 assay (IMAGN: mean of duplicate measurements) and flow cytometry (FLOW: mean of duplicate measurements) against their mean. Log transformed data. For high values (>2.2) the STELLer CD34 assay measures lower values than flow cytometry, whereas for low values it measures higher values. Full lines: Two standard deviations. Truncated line: mean. n = 105.

A closer look at Figure 3 reveals that the differences seem to decrease as the average of CD34+ cells/mm3 increases. This relationship is estimated by simple regression analysis.8 The estimated function is given by  = 1.44 − 0.36A, where D is the estimated difference and A is the average and the standard errors of estimation is SD = 0.63. The estimated parameters are highly significant (P < 0.01) ie the difference between the method is a decreasing function of the number of CD34+ cells. Further, Figure 4 clarifies that the STELLer assay counts are higher than flow at low CD34 numbers and vice verse on high numbers of CD34+ cells. This difference was independent of the number of cells enumerated as well as sample type (blood or leukapheresis product).

Figure 4

Agreement analysis: flow cytometry vs STELLer CD34 assay. Difference of CD34+ cells enumerated by the STELLer CD34 assay (IMAGN: Mean of duplicate measurements) and flow cytometry (FLOW: Mean of duplicate measurements) against their mean. Log transformed data. Full lines: Two standard deviations. Truncated line: regression line. n = 105.


In the comparison of quantitative methods neither the correlation coefficient nor regression analysis are appropriate. We present a comparison of two techniques for CD34 enumeration by the method described by Bland and Altman,6 which is simple to do and to interpret.

The outcome may not have clinical importance; however, it is important to explain the observed differences to be able to understand technical factors of importance for quantitative analysis of CD34+ cells.

Comparisons of first- and second-generation techniques or single and dual platform analysis for CD34 enumeration have been reported. Most of these studies have concluded that new methods are useful in clinical practice for blood level prediction and harvest including quality assessment of the autografts.9,10,11,12 Such usefulness may depend upon the statistical analysis as described.

A division of the untransformed data into quartilies (determined by the average of flow cytometry measurements) clarifies this further. Actually, the microvolume fluorimetry counts are on average more than 500% [313–761; 95% CI] the counts by flow cytometry when the measures belong to the first quartile (<5.4 CD34+ cells/mm3), but 27% [22–31; 95% CI] less when the measures belong to the upper quartile (>163 CD34+ cells/mm3) of CD34+ cells enumerated. Poor reproducibility exists for both methods below ≈9 CD34 cells/mm3, probably due to simple counting inaccuracy. In our haematology ward apheresis is initiated when the number of CD34+ cells is more than 20/mm3, a counting inaccuracy below 9 CD34+ cells/mm3 would not be considered a problem in clinical practice.

It must be remembered that the true value is not known, but different explanations for the observed differences are discussed below. The diminution found in the absolute number of CD34+ cells per unit of volume by the STELLer CD34 assay vs flow cytometry could be due to several technical pitfalls: In flow cytometry, gating excluding CD45-negative events and events below the FSC threshold would exclude apoptotic cells, subcellular particles, platelets, and erythrocytes. Apoptotic cells and subcellular particles may count as leukocytes in automatic cell counting which would cause an overestimation of the total number of CD34-positive cells. Erythrocyte lysis and a washing procedure may cause a selective loss of leukocytes, which could increase the percentage of CD34+ cells. Another possible explanation may be a systematic high cell count on the Coulter Analyser. In volumetric scanning two cells close to each other may be indistinguishable as the emission may be recognized as just one peak. Another explanation could be an insufficient amount of anti-human CD34 antibody in the manufacturer's reagent, which has not been evaluated here. The use of different monoclonal antibodies in the single and dual platform assay, although recognizing the class III epitope of the CD34 antigen, may also be of importance.

Why the STELLer CD34 assay seems to measure higher levels of CD34+ cells than flow cytometry, applying to values below 9 CD34 cells/mm3, is not evident, but may be because of simple counting inaccuracy.

The STELLer CD34 assay is easy to perform and semi-automatic, which makes it an attractive alternative procedure to flow cytometry for enumeration of CD34+ cells. Laboratories which intend to introduce this method should be aware of its poor reproducibility in enumerations lower than ≈9 CD34+ cells/mm3, and that above this level the method tends to measure values lower (20–27%) than flow cytometric values, as this may influence clinical procedures like CD34+ cell enumeration performed to decide when to start leukapheresis. The IMAGN2000 instrument has been withdrawn from the market by BD Biosciences in order to be improved. When the instrument is released for the market, laboratories which intend to replace an established method for CD34 enumeration by the STELLer CD34 assay have to compare the two methods to ensure the improvement of the instrument.


  1. 1

    Serke S, Johnsen HE . A European reference protocol for quality assessment and clinical validation of autologous haematopoietic blood progenitor and stem cell grafts Bone Marrow Transplant 2001 27: 463 470

    CAS  Article  Google Scholar 

  2. 2

    Wunder E, Sovalat H, Fritsch G et al. Report on the European Workshop on Peripheral Blood Stem Cell Determination and Standardization – Mulhouse, France, February 6–8 and 14–15. 1992 J Hematother 1992 1: 131 142

    CAS  Article  Google Scholar 

  3. 3

    Sutherland DR, Anderson L, Keeney M et al. The ISHAGE guidelines for CD34+ cell determination by flow cytometry. International Society of Hematotherapy and Graft Engineering J Hematother 1996 5: 213 226

    CAS  Article  Google Scholar 

  4. 4

    Serke S, Sauberlich S, Huhn D . Multiparameter flow-cytometrical quantitation of circulating CD34(+)-cells: correlation to the quantitation of circulating haemopoietic progenitor cells by in vitro colony-assay Br J Haematol 1991 77: 453 459

    CAS  Article  Google Scholar 

  5. 5

    Johnsen HE, Baech J, Nicolaisen K et al. Validation of the Nordic Flow Cytometry Standard for CD34+ Cell Enumeration in Blood and Autografts: Report from the Third Workshop J Hematother 1999 8: 15 28

    CAS  Article  Google Scholar 

  6. 6

    Bland JM, Altman DG . Statistical methods for assessing agreement between two methods of clinical measurement Lancet 1986 1: 307 310

    CAS  Article  Google Scholar 

  7. 7

    Knudsen LM, Gaarsdal E, Jensen L et al. Evaluation of mobilized CD34+ cell counts to guide timing and yield of large-scale collection by leukapheresis J Hematother 1998 7: 45 52

    CAS  Article  Google Scholar 

  8. 8

    Bland JM, Altman DG . Measuring agreement in method comparison studies Stat Meth Med Res 1999 8: 135 160

    CAS  Article  Google Scholar 

  9. 9

    Read EJ, Kunitake ST, Carter CS et al. Enumeration of CD34+ hematopoietic progenitor cells in peripheral blood and leukapheresis products by microvolume fluorimetry: a comparison with flow cytometry J Hematother 1997 6: 291 301

    CAS  Article  Google Scholar 

  10. 10

    Cabezudo E, Querol S, Cancelas JA et al. Comparison of volumetric capillary cytometry with standard flow cytometry for routine enumeration of CD34+ cells Transfusion 1999 39: 864 872

    CAS  Article  Google Scholar 

  11. 11

    Gratama JW, Orfao A, Barnett D et al. Flow cytometric enumeration of CD34+ hematopoietic stem and progenitor cells. European Working Group on Clinical Cell Analysis Cytometry 1998 34: 128 142

    CAS  Article  Google Scholar 

  12. 12

    Keeney M, Chin-Yee I, Weir K et al. Single platform flow cytometric absolute CD34+ cell counts based on the ISHAGE guidelines. International Society of Hematotherapy and Graft Engineering (see comments) Cytometry 1998 34: 61 70

    CAS  Article  Google Scholar 

Download references


This study was in part aided by grants from The Danish Blood Donors Research Foundation.

Author information



Corresponding author

Correspondence to HE Johnsen.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Gisselø, C., Roer, O., Hoffmann, M. et al. Assessing agreement between CD34 enumeration by flow cytometry and volumetric analysis. Bone Marrow Transplant 29, 699–703 (2002).

Download citation


  • CD34 enumeration
  • flow cytometry
  • volumetric analysis

Further reading


Quick links