It is well established that the 1–3% of cells in the bone marrow that express the CD34 antigen, a heavily glycosylated mucin-like structure, are capable of reconstituting long-term multilineage haemopoiesis following ablative therapy ( Berenson et al, 1988 ; Andrews et al, 1992 ). CD34+ cells are extremely rare in the peripheral blood of normal individuals (approximately 0.01–0.05%). However, current treatment regimes, including chemotherapy and/or haemopoietic growth factors, can significantly increase circulating CD34+ stem cell counts in patients and donors. Peripheral blood stem cells (PBSC) have now virtually replaced bone marrow as the primary source of stem cells for autologous transplantation after myeloablative therapy ( Gratwohl et al, 1996 ), and the procedure is also being used for allogeneic transplantation between HLA-identical siblings ( Russell et al, 1996 ). Advantages of PBSCs include the generally shorter engraftment time, reduced hospitalization costs ( Ager et al, 1995 ), the presence of large numbers of T lymphocytes and NK cells which may reduce post-transplant relapse ( Dreger et al, 1994 ), and the elimination of a general anaesthetic. In addition, the PBSC product is more suitable for ex vivo manipulation, including CD34+ cell selection ( Brugger et al, 1994 ), tumour purging ( Ross et al, 1995 ) and gene manipulation ( Bregni et al, 1992 ).
Transplant centres routinely rely upon the enumeration of CD34+ cells as an indicator for the optimal timing and adequacy of PBSC harvests ( Haas et al, 1994 ). Assessment of haemopoietic progenitors by colony-forming assays is laborious and time-consuming ( Appelbaum, 1979) and has the disadvantage of not enabling real-time planning of PBSC collections. A minimum threshold level of between 2 and 5 × 106 CD34+ cells/kg has been observed in multiple clinical settings to result in adequate engraftment ( Krause et al, 1996 ). However, the lack of assay standardization prevents a more exact definition of the threshold level ( Bender et al, 1992 ). For example, a variety of flow cytometric gating strategies for CD34+ cell enumeration have been developed based primarily upon the detection of total CD34+, or CD34+CD45dim cells ( Gratama et al, 1997 ; Siena et al, 1991 ; Sutherland et al, 1996 ; Verwer & Ward, 1997). In addition, there is a marked variation in the choice of monoclonal antibody, fluorochrome and lysing reagent used. Furthermore, the absolute enumeration of CD34+ cells can be determined using either a dual- or single-platform approach. The former derives the CD34+ value from the combination of a flow cytometrically determined percentage CD34+ count and an absolute nucleated cell count generated by a haematological analyser. In contrast, single-platform technology derives an absolute CD34+ cell count directly from the flow cytometer using either precision fluidics or micro-beads ( Mercolino et al, 1995 ; Verwer & Ward, 1997; Chin-Yee et al, 1997a ; Keeney et al, 1998 ). Such variation in methodologies and the requirement for precise CD34+ cell enumeration has made standardization difficult, and the best approach has been the subject of much recent discussion ( Johnsen, 1997; Sutherland et al, 1997 ).
In an attempt to address the problems of standardization, several national external quality assurance (EQA) programmes ( Gratama et al, 1997 ; Lowdell & Bainbridge, 1996; Lumley et al, 1996 ; Chang & Ma, 1996; Brecher et al, 1996 ; Chin-Yee et al, 1997b ) and international workshops ( Gee & Lamb, 1994; Johnsen, 1995; Wunder et al, 1992 ; Johnsen & Knudsen, 1996) have been set up during the last 5 years. The EQA programmes have reported widely varying interlaboratory coefficients of variations (CVs; summarized in Table I), the main cause of which has been attributed to the use of fresh, or cryopreserved, specimens ( Gratama et al, 1997 ). This view is supported by the findings of a recent Australian study which documented a marked reduction of interlaboratory CVs when only list mode data was analysed ( Chang & Ma, 1996). Chang & Ma (1996) additionally demonstrated that gating strategies were a major contributing factor to result variability and that only one gating strategy, the ISHAGE protocol, gave reproducible results from all centres of within ± 10% of the median CD34+ cell value on both peripheral blood (PB) and apheresis samples. In response to such findings, UK NEQAS initiated an EQA scheme, originally involving 64 participants in 12 countries and using whole blood stabilized in a manner previously described ( Barnett et al, 1995 ). Use of such material has previously been shown to circumvent analyte instability, facilitating a more accurate and detailed analysis of interlaboratory variation for CD4+ T-lymphocyte enumeration and, in addition, has enabled transportation of EQA specimens overseas by post, thus reducing transportation costs ( Barnett et al, 1996 ).
Table 1. Table I.
Summary of inter-laboratory QC surveys. Brecher et al, 1996
, Chang & Ma, 1996
, Gratama et al, 1997
, Johnsen & Knudsen, 1996
, Lowdell & Bainbridge, 1996
, Lumley et al, 1996
, Chin-Yee et al, 1997b
We report the findings from the first 18 months of the UK NEQAS CD34+ stem cell enumeration programme, currently the largest such EQA scheme described to date, using stabilized whole blood specifically prepared for PBSC enumeration. The use of such a preparation within the programme has resulted in reduced interlaboratory variation for CD34+ stem cell quantitation when compared to previously published studies ( Table I) and removed the sample variability seen when fresh or cryopreserved specimens are used. Thus, for the first time, in any EQA for CD34+ stem cell enumeration, a more accurate assessment of performance, both on a national and international scale, has been achieved. However, our data still highlights the need for further improvement in methodological approach and underlines the urgent requirement for national/international consensus guidelines.
Specimen collection and distribution
50 ml of peripheral blood was obtained from nine patients undergoing G-CSF stem cell mobilization and prior to peripheral blood stem cell harvesting. In addition, a single 50 ml sample of cord blood was obtained for one issue. All samples were stabilized using a procedure described previously ( Barnett et al, 1995 , 1996) and which is now used under licence to produce Ortho AbsoluteControl (Ortho Diagnostic Systems Inc., Raritan, U.S.A.). Longitudinal studies have previously shown that flow cytometric profiles are retained ( Barnett et al, 1996 ; Janossy et al, 1998 ) and confirmatory studies for the stability of the CD34 antigen were performed prior to the use of the stabilized material in the EQA programme (see below). All material used throughout this study was obtained following informed consent.
Each centre was issued with a 1 ml aliquot of stabilized peripheral blood, transported by either post or, if specifically requested, commercial courier. The UK NEQAS Scheme for Leucocyte Immunophenotyping has provided stabilized quality assessment specimens for many years to laboratories worldwide that perform leukaemia immunophenotyping and HIV lymphocyte subset analysis ( Barnett et al, 1994, 1996). The CD34+ Stem Cell Enumeration Scheme initially issued samples to 64 participants (38 U.K., 26 overseas), a number which increased to 91 (44 U.K., 47 overseas) after 18 months. Non-U.K. countries included Australia, Brazil, Canada, Denmark, Eire, Germany, Mexico, Netherlands, New Zealand, Portugal, Spain, Sweden, Switzerland and U.S.A.
Longitudinal specimen testing
In order to determine the degree to which the percentage value and mean channel fluorescence drifted with time, a 50 ml aliquot of peripheral blood was stabilized as previously described ( Barnett et al, 1995 ) and stored in 1 ml lots at 4°C until use. CD34+PBSC enumeration was undertaken according to the non-sequential gating strategy of Sienna et al (1991 ). This protocol was employed in order to identify any increase in non-specific staining due to debris and/or platelets. In addition, it was the most widely used approach by the participants (information obtained from a pre-survey questionnaire). Samples were analysed over a total of 228 d using a class III antibody (HPCA-2, Becton Dickinson, San Jose, U.S.A.), conjugated with phycoerythrin. An IgG1 fluorochrome-matched control antibody was used to correct for background staining and all antibodies were employed in accordance with the manufacturers' recommendations at saturating concentrations. Briefly, 20 μl of anti-CD34 PE, or control antibody as appropriate, was added to 100 μl of stabilized whole blood and incubated for 15 min at room temperature (RT). The erythrocytes were then lysed by incubating the sample with 1 ml of Ortho-mune lysing Solution (Ortho Clinical Diagnostics, Raritan, U.S.A.) for 15 min. After this period the sample was analysed without washing.
Flow cytometric calibration (FACScan, Becton Dickinson, San Jose, U.S.A.) was performed on a daily basis in accordance with manufacturers' instructions and the light scatter and immunostaining characteristics of stabilized samples compared with those of fresh samples as previously described ( Barnett et al, 1996 ; Janossy et al, 1998 ). Instrument settings remained constant for fresh and test samples.
Each centre was required to analyse the QC material within 3 weeks of issue, using local clinical laboratory procedures. A questionnaire, issued with each sample, requested details of total white cell count (WCC) and CD34+ cell count (both percentage and absolute values), as well as information on gating strategy, flow cytometer, sample preparation procedure (i.e. lyse-no-wash, lysis reagent, density centrifugation, etc.), antibody type (i.e. class I, II or III), antibody source/clone and fluorochrome, number of events analysed and whether an isotype control was employed (i.e. CD34 events corrected for IgG subclass binding).
Percentage and absolute CD34+ cell counts for a given laboratory were compared with the results obtained from other participants. Following the issue of the sixth specimen, a performance scoring system was introduced for absolute counts. This employed the use of stratified centiles as target ranges. Median and centile ranges were used for statistical evaluation due to the non-parametric nature of the data. Selected target ranges were the 5th, 10th, 25th, 75th, 90th and 95th centiles, criteria set by the Department of Medical Statistics and Evaluation (Royal Postgraduate Medical School, London) following 20 000 simulations using data from the first six issues. Development of such a system has enabled the identification of persistent unsatisfactory performers (PUPs), based on scoring criteria determined by the scheme Steering Committee. Laboratories scoring 100 points or greater, over a rolling three-sample window, were defined as a PUP. In brief, a score of 50 was awarded for laboratories whose absolute CD34+ value exceeded either the 5th or 95th centile, with 35 points for values between the 5th and 10th, or the 90th and 95th centile, 20 points for values between the 10th and 25th, or the 75th and 90th and zero points if the absolute value was between the 25th and 75th centile. Laboratories were therefore classified as a PUP if the results fell outside the central 80% on three consecutive samples (the chosen window of analysis). Failure to return a result (nil return) was also penalized under a separate scoring system (50 points for a nil return), such that a score of 100 points or greater, over a rolling three-specimen window, identified the participant as a PUP. The two scoring systems for performance and nil returns were mutually exclusive.
The determination of leucocyte antigens in EQA programmes, for example CD4+ lymphocyte quantitation, has previously involved the use of fresh whole blood, necessitating rapid distribution and incurring high transportation costs ( Edwards et al, 1989 ; Goguel et al, 1993 ; Homburger et al, 1993 ; Paxton et al, 1989 ; Schonwald & Jilch, 1994). However, there is no guarantee that the testing laboratory will receive the analyte in perfect condition, a fact which presents problems in identifying poor or inadequately performing laboratories. Indeed, the use of fresh or cryopreserved material in EQA schemes for CD4+ lymphocyte quantitation is known to have a significant effect on the results (high CV and SD values) ( Edwards et al, 1989 ; Goguel et al, 1993 ; Homburger et al, 1993 ; Paxton et al, 1989 ; Schonwald & Jilch, 1994; Barnett et al, 1996 ). Furthermore, recent EQA schemes for CD34+ stem cells have also identified analyte instability as a major contributing factor to the high CVs (CVs > 100%) ( Gratama et al, 1997 ). In an attempt to circumvent this problem, Lowdell & Bainbridge (1996) constructed an EQA programme which ‘clustered’ laboratories, based upon a single centre sending samples to four other institutions, who in turn repeated the sample issue process. However, even within such a complex system, it was recognized that sample stability was an issue. Recent data has also indicated that CD34 expression is affected by storage at room temperature, cryopreservation and fixatives used with specific cell labelling methods ( Macey et al, 1997 ). Furthermore, a recent Australian study has reported that when flow cytometric list mode data is issued, instead of samples, inter-laboratory CVs of < 20% are achieved, underlining the problem of using fresh specimens in EQA programmes ( Chang & Ma, 1996). Therefore, following the successful introduction of a novel stabilized whole blood preparation within the UK NEQAS Immune Monitoring scheme ( Barnett et al, 1996 ), we examined its role in an EQA scheme for CD34+ stem cell enumeration.
The current study demonstrates that stabilized whole blood, obtained from either patients undergoing peripheral blood stem cell mobilization or from cord blood, can be successfully introduced into such an EQA programme. Longitudinal studies have confirmed the stability of the CD34 antigen for over 200 d (independent analysis of a sample 615 d old on both Becton Dickinson and Coulter platforms gave equivalent results to day zero (R. Sutherland and M. Keeney, Ontario, Canada, personal communications). Furthermore, CD45 expression is preserved well enough to enable the use of such material in sequential gating strategies that employ CD45 detection ( Figs 2 and 3) such as the ISHAGE protocol. In addition, the material is suitable for use with strategies that employ nucleic acid dyes (i.e. LDS-751 and the ‘ProCount’, data not shown). We have previously reported the suitability of such material for use with a variety of flow cytometers (e.g. FACScan, Ortho CytoronAbsolute), lysing reagents (e.g. FACS Lysing solution, Becton Dickinson, San Jose, Mountain View, Calif., and Ortho-mune lysing reagent, Ortho Diagnostic Systems Inc., Raritan, N.J.) and no-wash–no-lyse techniques ( Barnett et al, 1996 ). The striking effect of the preparation's introduction as an EQA material in this programme was the low interlaboratory CVs obtained, when compared to other EQA schemes that lacked standardized protocols. Furthermore, when the stabilized material was used as a reference material, to optimize flow cytometer setup, acquisition and analysis, further reductions in interlaboratory CVs were noted (e.g. 22% and 24% for percentage and absolute CD34+ PBSC values respectively for sample 10).
In agreement with published data from other EQA schemes and workshops, the UK NEQAS CD34+ stem cell enumeration programme has observed a wide variation in methodology ( Chang & Ma, 1996; Brecher et al, 1996 ; Lumley et al, 1996 ; Gratama et al, 1997 ; Chin-Yee et al, 1997b ). The use of a stabilized material, in conjunction with the largest participant base reported to date (102 participants, March 1998), has enabled a detailed analysis of the various methodological approaches. The major factors affecting the results and thus requiring future standardization were identified as follows: the means of determining the total nucleated cell count, the lysing reagent, the class of anti-CD34 antibody and the fluorochrome used, the number of events collected and the gating strategy employed.
The absolute CD34+ count for a given specimen will vary depending on whether a dual or single platform system is used. The development of single-platform instrument systems resulted from the requirement for precise and accurate T-lymphocyte subset analysis ( Connelly et al, 1995 ; Strauss et al, 1996 ). Traditionally, absolute CD4+ T-cell counts were calculated using a dual-platform system; the percentage of CD4+ T lymphocytes being derived from the flow cytometer and the absolute lymphocyte, or total white cell count, from a haematological analyser. It became apparent, however, that such an approach can result in considerable variation ( Robinson et al, 1992 ; Goguel et al, 1993 ) and led to the development of the FACSCount and CytoronAbsolute, capable of producing data in both absolute and percentage format ( Mercolino et al, 1995 ; Connelly et al, 1995 ; Strauss et al, 1996 ). Further software developments have enabled single-platform approaches for CD34+ enumeration on instruments such as the FACScan and FACSCabilur ( Verwer & Ward, 1997). Furthermore, recent studies ( Chin-Yee et al, 1997a ; Keeney et al, 1998 ) have shown that the ISHAGE protocol can also be modified to a become a single-platform approach. The addition of a known number of Flow-Count fluorescent microspheres (Coulter Corporation, Miami, Fla., U.S.A.) to the sample, combined with a lyse–no-wash sample processing approach enables absolute CD34+ cell counts to be obtained directly from the flow cytometer. However, the majority of participants in the present study employed dual-platform instrumentation, a factor that contributed to the increased variation in absolute CD34+ counts. For example, the overall interlaboratory CV for single-platform analysers (sample 10) was 9.9% compared to 24% for laboratories using dual-platform technology.
A further major factor which contributed to the variation observed in absolute CD34+ count was the lysing solution employed. It is well documented that systems utilizing lyse–no-wash, or no-lyse–no-wash techniques have reduced variability and tighter CVs for CD4+T-lymphocyte enumeration ( Connelly et al, 1995 ; Strauss et al, 1996 ; Barnett et al, 1996 ). We have confirmed this observation for CD34+cell enumeration and found that laboratories using lyse–no-wash systems returned CD34 counts approximately 20% higher, suggesting that cells are lost during the washing process. It should also be stressed that certain lysing reagents may reduce antigen expression and therefore be a source of additional variability ( McCarthy et al, 1994 ; Macey et al, 1997 ). It is quite feasible that diminution of antigen density due to particular lysing reagents, coupled with the use of FITC and a lyse–wash technique, will result in PBSCs being significantly underestimated; therefore such an approach should not be used.
The CD34 antigen has been reviewed in detail elsewhere ( Sutherland & Keating, 1992; Sutherland et al, 1992 ). It is a heavily glycosylated mucin-like structure with three epitopes, defined by sensitivity to neuraminidase and O-sialo-glycoprotease from Pasteurella haemolytica. Epitopes recognized by class I antibodies are sensitive to both enzymes, class II antibodies are sensitive to glycoprotease only, whereas those detected by class III antibodies are insensitive to both enzymes ( Sutherland et al, 1992 ). Class I antibodies fail to detect all glycoforms of the CD34 antigen and may only weakly bind to CD34 expressed on some leukaemias and leukaemia-derived cell lines. An additional decrease in sensitivity is observed if class I and II antibodies are FITC-conjugated. As a result, only PE-conjugated class II and either FITC- or PE-conjugated class III antibodies are recommended for CD34 detection ( Sutherland et al, 1996 ). Not surprisingly, therefore, the single participant that used a class I antibody, variously conjugated to FITC or PE, was identified as a persistent unsatisfactory performer, returning consistently low CD34 counts, with performance only improving when the laboratory switched to a PE-conjugated class III anti-CD34. Initially, 10 laboratories used FITC-conjugated antibodies, although the numbers reduced during the programme to a single participant (class III, FITC-conjugated). Interestingly, the use of FITC-labelled antibodies generally resulted in higher CD34 counts (see Table III), when compared to centres using PE. This finding was most marked with the cord blood, the most likely cause being the different fluorochrome sensitivities and types of gating strategies employed. It is well established that sequential gating strategies exclude debris and define ‘true’ CD34+ CD45dim haemopoietic cells. However, the use of non-sequential gating methods, such as the Milan protocol ( Siena et al, 1991 ), will potentially include debris and non-viable cells which may bind anti-CD34 non-specifically ( Sutherland et al, 1997 ), resulting in a falsely high count. Such an approach, coupled with the use of FITC conjugates, will make the discrimination of debris and CD34+ cells more subjective, accounting for the observed difference in results, especially when analysing cord blood samples.
The present study has revealed a marked variation in the number of CD34+ cells routinely counted. Reduction of the number of events per analysis will reduce the reliability of the estimation to unacceptable levels. Given the fact that the standard error of the number of positive cells per analysis is given by the square root of the number of positive cells; the larger the acquisition, the less the coefficient of variation ( Wunder et al, 1992 ). As a result, to maintain precision and also to ensure a methodological CV of 10%, a minimum of 100 CD34+ events should be collected from at least 75 000 CD45 events ( Sutherland et al, 1996 ). This approach is supported by the CD34 Task Force on behalf of the European Working Group on Clinical Cell Analysis ( Gratama et al, 1998 ). Despite these views, 75% of laboratories in the current study collected fewer than 75 000 CD45+ events, with 8% of participants (sample 1) collecting 20 000 or fewer events. Surprisingly, two participants collected as few as 29 and 40 CD34+ events from a total of 14 500 and 10 500 CD45+ events respectively. Furthermore, nine centres using the Milan protocol failed to use an isotype control, despite this being intrinsic to the gating strategy.
In conclusion, we have demonstrated the advantage of utilizing stabilized whole blood samples for EQA of CD34+ cell enumeration. This, in turn, has facilitated the development of an international EQA programme, significantly larger than other schemes reported to date, involving 91 participants in 19 countries. More importantly, it has demonstrated that interlaboratory CVs can, on an international scale, be reduced to < 25%, without using specified gating criteria. However, our findings indicate that further improvements are possible if standardized protocols are adopted, such as those proposed by ISHAGE and EWGCCA ( Sutherland et al, 1996 ; Gratama et al, 1998 ), and highlights the urgent need for nationally and internationally agreed consensus guidelines.
We thank Professor G. Janossy (Royal Free Hospital, London) and Associate Professor D. R. Sutherland (Toronto Hospital, Toronto, Canada) for reviewing the manuscript and for the help and advice they have given to this UK NEQAS programme over the past 18 months. We thank Dr C. Doré, Department of Medical Statistics and Evaluation, Royal Postgraduate Medical School, London, and Professor G. Raab, Mathematics Department, Napier University, Edinburgh, for their statistical advice. We also extend our gratitude to the UK NEQAS Haematology Steering Committee and the medical and technical staff within the Department of Haematology, Royal Hallamshire Hospital, for their support of this programme.