Estimation of the allowable total error of the absolute CD34+ cell count by flow cytometry using data from UK NEQAS exercises 2004–2024

Introduction

The ISO15189:2022-12 international standard for clinical laboratories at section 7.3.4 states for any analyte or measurand the need to define the degree of measurement uncertainty, the biological reference intervals, and the values that constitute thresholds and limits for clinical decisions. The ISO/TS 20914:2019 document provides the practical guidance to calculate and establish such important laboratory parameters [1], [2], [3].

The knowledge of the measurement uncertainty (MU) of a diagnostic laboratory test is essential for the laboratory to keep the reliability of its results under control, is requested by regulatory bodies, and for the ordering clinician to be aware of the ‘grey zone’ of variability around the reported values.

A major contributor of the MU is the total analytical error (TE) related to the measurement of a certain analyte, a technical parameter generated by the summation of all the involved analytical and extra-analytical variables. Another large component of the MU is the biological variability (BV) of the analyte itself. While the components of TE are amenable of a continuous improvement through practice and technological advancements, within-subject and between-subject BV cannot be analytically improved [4], 5].

Therefore, for every analyte of clinical significance efforts have to be made to define the acceptable limits of the variation around the ‘true’ target value, calculating the allowable Total Error (aTE), in order to prevent the misdiagnosis or the misclassification of patients and provide confidence when taking necessary clinical decisions [6], 7].

Hundreds of common clinical laboratory measurands have been already extensively studied to define the MU, BV and aTE parameters, since their value ranges are reasonably stable and predictable. Open-source reference tables have been published by CLIA and other expert groups [8], 9], with several cases also taking into account the BV values of percent and absolute lymphocyte subset measurements [10], 11].

Percent aTE calculations are accomplished using the classical formulae devised by C.G. Fraser, defining the minimum, desirable and optimum ranges of total allowable imprecision and the goal of analytical bias [7], 12], 13]. Such formulae require however the previous knowledge of both within- and between-subject BV for the measurand of interest [5].

Published data on CD34+ hematopoietic precursor cells (HPC) BV to be applied to the Fraser’s formulae are unfortunately not yet available, and just a few studies have been published on the baseline CD34+ cell levels in the peripheral blood of healthy subjects by flow cytometry (FCM) [14]. In normal subjects the peripheral blood level of CD34+ HPCs ranges typically from 0 to about 10/μL or <0.1 % of leukocytes, and when the level raises stably above 10 cells/μL, the presence of some derangement in the bone marrow becomes likely [15].

International efforts have been made during the last three decades to standardize the CD34+ measurements and to increase their robustness in the clinical practice, thanks to the advent of FCM boolean gating, the development of the ISHAGE guidelines in 1996 and the worldwide availability of external quality assurance systems [16].

Defining the %aTE for CD34+ HPC enumeration in the autologous or allogeneic transplantation setting is of the utmost importance, but is particularly challenging, because the HPC absolute levels are not stable and predictable variables, but may vary by a factor of several hundreds during the various phases of the mobilization process and cell collection. In the absence of published BV data, as for newly devised analytes or drug dosages, alternative or in house methods to define the aTE must be sought for, which can be difficult [1]. Moreover, the various matrices in which CD34+ HPCs are measured, such as peripheral blood (PB), bone marrow, fresh and thawed cord blood, fresh and thawed apheresis require the application of specific technicalities and further complicate these assessments [16], [17], [18], [19], [20], [21].

At present only one valuable study has been published on the evaluation of %aTE in CD34+ HPC enumeration using data from a Spanish external quality assessment program and employing the ‘state of the art’ (SOTA) methodology [6] with the distribution of fresh samples [22].

Taking inspiration from this pioneer study we have made an extensive evaluation of %aTE of the measurement of CD34+ cells absolute levels in stabilized PB preparations distributed worldwide by the United Kingdom for National External Quality Assessment – UK NEQAS – CD34+ Stem Cell Enumeration EQA/PT Programme over the last 20 years, from 2004 to 2024. This involved calculating the SOTA-based %aTE values, defined as the desirable performance level that can be achieved by 80 % of participating laboratories and the optimal performance displayed by the best performers, as clustered in the upper quartile of results distribution, with a clear trend of improvement with time.

Materials and methods

Results

Comparison of single platform and dual platform techniques

Figure 1 represents all the %aTE cumulative distributions of the four absolute CD34+ cells/μL levels (A, B, C and D, respectively) in the three more significant groups of data. The respective %aTE values obtained by intercepting the curves with Y=80 % are indicated. The SOTA-based %aTE showed a trend to the reduction with the increase of CD34+ cell levels. The lowest %aTE values are associated with SP analyses, and improved during the two observation periods.

Figure 1:

The %aTE cumulative distributions of the four absolute CD34+ cells/μL levels (A, B, C and D, respectively, represented in colors) in the three main groups of data are shown. The respective %aTE values obtained by intercepting the curves with Y=80 % are indicated on the X axis. The lowest CD34+ cells/μL levels show the highest %aTE values in all instances.

Table 2 summarizes the quantitative results, including the median and interquartiles of the cumulative distributions, the SOTA-Based Desirable %aTE achievable by 80 % of the laboratories and the SOTA-Based Optimal %aTE, representing the highest level of analytical performance theoretically achievable.

Table 2:

The medians and interquartiles of all cumulative distributions, the SOTA-based desirable %aTE achievable by 80 % of the laboratories and the SOTA-based optimal %aTE values are summarized for DP 2004–2014 and SP 2004–2014/2014–2024. All the %aTE parameters progressively decrease with the increasing of the CD34+ cells/μL levels.

	Median CD34+ cells/μL		(Exercises, n) entries, n	Median %aTE (interquartiles)	SOTA-based desirable %aTE (achievable by 80 % of labs)	SOTA-based optimal %aTE (upper quartile of labs)
Dual platform 2004–2014	A	0–10.0	(19) 1,712	12.47 % (5.91–22.50)	± 24.37 %	± 2.50 %
	B	10.1–25.0	(37) 3,258	9.81 % (4.54–18.78)	± 20.61 %	± 2.12 %
	C	25.1–50.0	(29) 2,057	8.31 % (3.70–15.03)	± 16.55 %	± 1.73 %
	D	>50.1	(35) 2,429	8.15 % (3.74–15.29)	± 16.39 %	± 1.84 %
Single platform 2004–2014	A	0–10.0	(19) 2,820	11.57 % (5.26–21.97)	± 24.52 %	± 2.27 %
	B	10.1–25.0	(37) 4,857	8.94 % (4.12–16.91)	± 18.49 %	± 1.87 %
	C	25.1–50.0	(29) 5,193	7.50 % (3.46–13.78)	± 14.72 %	± 1.56 %
	D	>50.1	(35) 6,196	6.77 % (2.96–12.26)	± 12.86 %	± 1.43 %
Single platform 2014–2024	A	0–10.0	(26) 7,148	10.53 % (4.40–19.54)	± 21.76 %	± 1.68 %
	B	10.1–25.0	(53) 14,401	7.88 % (3.64–14.37)	± 15.34 %	± 1.68 %
	C	25.1–50.0	(28) 8,140	7.09 % (3.25–13.20)	± 14.23 %	± 1.63 %
	D	>50.1	(19) 4,217	6.57 % (2.94–12.50)	± 13.59 %	± 1.42 %

The DP and SP %aTE for the lowest CD34+ cell levels was nearly the same, while DP appeared to perform progressively worse with the increasing of the CD34+ cell levels. The SP %aTE in the 2014–2024 decade showed a further improvement.

As expected, the overall %aTE progressively reduces with the increase of the CD34+ cell levels in all instances. The marked difference between desirable and optimal %aTE levels summarized in Table 2 shows that a reduction of about 10 times of the measurement uncertainty is potentially achievable, as demonstrated by the best performing laboratories.

Supplementary Figure 1 illustrates more clearly the direct comparison between DP and SP %aTE in the 2004–2014 period. The overall performance of the two techniques was remarkably similar in the A level group, while for increasing CD34+ cells/μL levels the DP laboratories showed a progressively worsening performance.

Supplementary Figure 2 shows the comparison between SP %aTE in the 2004–2014 and 2014–2024 periods. The overall SP performance seemed to improve only slightly in the more recent period, especially for the two lowest CD34+ cells/μL levels, while no significant improvements could be noted for the two highest levels.

Supplementary Table 1 shows the performance of the residual DP users in the 2014–2024 decade. The %aTE parameters showed no significant improvement as compared to the 2004–2014 decade, and were significantly worse than the corresponding SP data during the same observation period.

Figure 2 shows the trend of the median %aTE with time in the three main groups of analyzed data. A progressive and parallel reduction of median %aTE was observed for both DP and SP in the 2004–2014 period, with just slightly higher figures for DP. Conversely, the overall performance of SP laboratories in the 2014–2024 period did not further improve with time, with aTE remaining stably around about 7 % over the years.

Figure 2:

Trend of the median %aTE with time in the three main groups of analyzed data. While performance improvements could be observed during the 2004–2014 decade, after 2014 the overall performance remained stable with a median aTE around 7 %.

No instrument- or manufacturer-related bias could be noted with %aTE values, and this is in line with the remarkably overlapping performance data between the two major providers of CE-IVD certified FCM systems that can be inferred from the UKNEQAS scheme reports. No results from participants using SP volumetric instruments in CD34+ counting are recorded at present.

According to our results, a reliable estimate of the ‘grey zone’ around a given reported result of absolute CD34+ cells/μL using the current SP technique can be defined. The grey zone varies according to the acceptability criteria used: the SOTA-Based desirable %aTE may range from 21.76 % for levels around 10 CD34+ cells/μL to 13.59 % for levels around 50 CD34+ cells/μL. Conversely, the SOTA-Based optimal %aTE, reflecting the results of highest quality achievable by the best performing laboratories may range in a much narrower spectrum, from 1.68% to 1.42 % for the same CD34+ cells/μL levels, as summarized in Table 3.

Table 3:

The ‘grey zones’ around a certain reported value of CD34+ cells/μL obtained by SP and calculated according to two different criteria of allowable imprecision are indicated. The reported levels of 10, 20 and 50 CD34+ cells/μL refer to the most commonly used thresholds to trigger an apheresis procedure.

What a given reported result by SP may mean
Reported CD34+ cells/μL	According to SOTA-desirable %aTE (80 % of labs)	According to SOTA-optimal %aTE (upper quartile of labs)
10	7.83–12.17	9.84–10.16
20	16.94–23.06	19.68–20.32
50	43.25–56.67	49.3–50.7

Discussion

The absolute level of CD34+ cells in a biological matrix represents an example of a peculiar, multi-faceted FCM analyte with several different applications in multiple clinical settings, and endowed with a great decision-making impact.

The CD34+ cell enumeration by FCM is a fully mature technology, taking advantage from internationally established guidelines [21], the availability of standard reference materials and calibration curves. An evaluation of the laboratory practice should nowadays focus on desirable and optimal performance levels only, leaving acceptable levels aside.

The measurement of CD34+ cells in the PB of steady-state individuals is a typical rare event FCM analysis, aimed at identifying bone marrow abnormalities leading to myelodysplastic syndrome, myelofibrosis and myeloproliferative disorders when levels exceed 10 CD34+ cells/μL [26], [27], [28]. Circulating and bone marrow CD34+ cells have long been assumed as ‘blast’ equivalents, and despite several biological limitations, the FCM enumeration of CD34+ cells proved to be a practical surrogate tool to estimate the number of myeloid blasts in hematologic malignancies, and especially to classify different disease states and to monitor patients [32], [33], [34]. The semi-quantitative evaluation of CD34+ blasts in the peripheral blood or bone marrow of patients with acute leukemias remains however outside the scope of true quantitative assessments.

In the setting of autologous and allogeneic transplantation of CD34+ HPCs the trigger PB values to start the apheresis collection in mobilized patients and donors vary greatly according to local policies, ranging typically from 10 to 50 CD34+ cells/μL. The concentrated fresh apheresis cell suspension may contain viable CD34+ HPCs ranging from several hundreds to a few thousands per μL, whereas the respective cryopreserved preparations introduce a great deal of measurement variability after thawing, related to the high cell mortality. Similar problems can be faced in the analysis of CD34+ HPCs in the cord blood.

With such a wide range of absolute levels and diverse biological matrices the evaluation of the allowable total error of the absolute CD34+ cell enumeration is therefore particularly challenging, also due to the total lack of published data on biological variability.

In this scenario, accuracy-based EQA/PT programs have the unique ability to provide laboratorians with large datasets of unbiased measurements across laboratories and manufacturers, that can be used to generate the necessary %aTE parameters to guide the routine practice [5]. The only study so far published on the %aTE evaluation of CD34+ cells has shown a global %aTE of 27.8 % for the best performing laboratories, using fresh unstabilized samples distributed to 40 laboratories in Spain [22].

In our study we aimed at evaluating the SOTA-Based %aTE for a large number of international laboratories participating to the UK NEQAS – CD34+ Stem Cell Enumeration Programme over two consecutive 10-years periods from 2004 to 2024, using highly consistent stabilized PB preparations. We tried to calculate the SOTA-Based %aTE levels for the enumeration of CD34+ cells in hematological patients’ PB (i.e., ≤10 cells/μL) and for the main relevant checkpoints of the mobilization procedure (i.e., 20, 50 and >50 cells/μL, respectively). In our analysis the highest levels provided during the exercises were between 141 and 184 CD34+ cells/μL, while the unavailability of concentrated apheresis suspensions prevented us from making evaluations for higher values.

In this context, the term ‘SOTA’ does not denote the use of a specific antibody mixture, instrumentation or software, rather it defines the %aTE values that can be achieved as current procedure standards by 80 % of laboratories [1], while the median %aTE of the upper quartile of participants represents the performance of the highest quality that can be virtually achieved by all as an improvement goal [6].

As compared to the study by Fernández-Luis, our results showed tighter %aTE ranges for all the considered absolute CD34+ cell levels, probably due to the type of distributed samples, that ensure stable and homogeneous levels of both total white cells and CD34+ precursors, despite the long delivery distances and times. Our findings therefore may reflect a more accurate picture of the purely analytical imprecision related to the FCM absolute CD34+ cell measurements, devoid of the unavoidable preanalytical decay related to the long-range distribution of fresh samples [35]. Our %aTE values showed a progressive narrowing with the increase of the absolute CD34+ cell level, in line with the statistical principles described by Horwitz [36]. Conversely, we were challenged by a globally higher number of outliers than in the study by Fernández-Luis (10.43 % vs. 6.4 %, respectively). This was likely due to the larger number of participants (up to 370 vs. 40), dispersed in up to 59 countries worldwide, with a frequent in-and-out turnover of laboratories, and consequently a lower degree of control and direct advice by the scheme managers.

The 2004–2014 decade showed a constant reduction of the %aTE with time for both DP and SP users, while during the subsequent decade the %aTE of SP users remained rather flat without showing further improvements. Almost all participants declared the usage of analysis protocols related to the ISHAGE, although some deviations from the originally recommended protocol have been highlighted [25]. The higher frequency of samples with very low absolute CD34+ cell levels during the 2014–2024 decade may explain the presence of some cases with an unexpectedly higher %aTE and the slight positive trend of the regression line.

Dual platform analysis seems more or less able to be on target with the expected central result values of exercises using stabilized PB, although with wider CVs [25] and higher %aTE. Conversely, DP is in trouble with real patient samples, both on the FCM and the automated blood counter sides, it is highly inaccurate with fresh concentrated apheresis and cord blood, and cannot be applied at all with thawed samples [21]. Scientific evidence should put pressure on the residual DP users to prompt an universal change to the far more reliable SP technique.

An excess variability of the measurement of the absolute level of CD34+ cells may have clinical consequences. For instance, the misclassification of patients with hematological disorders and the under- or over-estimation of the critical checkpoints of the CD34+ HPC transplantation procedure. The CD34+ PB trigger values for starting the HPC apheresis are strong predictors of the overall collection [37], as well as the post-apheresis PB level that are used to calculate the cell yield [38]. The worst error in this context is the overestimation of absolute CD34+ HPC levels that may trigger long and inefficient collections, with the need of additional procedures, or the reinfusion of too few viable CD34+ HPCs with the thawed apheresis products. As mentioned, the absolute CD34+ levels in the concentrated apheresis suspension are still not evaluable with EQA schemes.

An attempt should be made to translate our findings into daily practice. Commercially available reference materials for internal quality control are useful and mandatory for the clinical laboratory, but they invariably come with just an approximate target level of CD34+ cells/μL and a very wide acceptable range of results [39]. This makes such preparations not very useful to assess the desirable or optimal imprecision levels of the CD34 assay. As a matter of fact, a too wide distribution of the expected results does not provide the necessary indications to strive for goals of minimized assay imprecision. As such the improvement of existing materials by commercial providers to develop a widely available, stable control material with a clinically relevant acceptable range is crucial for laboratories to maintain confidence in their results. Performing replicate measurements of the stabilized UKNEQAS samples shortly after the vial opening, thanks to their high stability, the low intrinsic variability and the known consensus results from exercise reports, are useful to refine laboratory procedures, to reach SOTA-desirable and possibly optimal counting performance.

It seems however likely that the tight %aTE values obtained with stabilized samples would not be achievable with real-world patient samples, due to the variable underlying pathological conditions, the individually unpredictable response to mobilizers, the high number of interfering accompanying cells and the CD34 antigen modulation induced by apheresis. Optimizing the analysis procedures as discussed is however a fundamental prerequisite to improve patients’ diagnoses and transplant outcomes.

A major limitation of our study is the lack of a parallel analysis with concentrated leukapheresis preparations, that entail a number of delicate preparatory and preanalytical steps and are characterized by a much larger chance of generating abnormal results with unfavorable clinical impact. A similar study of %aTE in apheresis products, cord blood and thawed apheresis samples seems not yet possible at present, given that these materials are hardly available for distribution in large-scale EQA/PT programs.

Some large studies have been already published on BV of lymphocyte subsets [40], 41] and tables reporting BV parameters for normal subpopulations levels are publicly available. However, BV parameters for absolute CD34+ cell levels in PB may be generated at present for healthy steady-state subjects only (i.e., <10 cells/μL), thanks to large cooperative studies. The complex dynamics of absolute CD34+ cell levels in mobilized subjects and in the transplantation procedures seem at present hardly compatible with the generation of reliable BV tables, as for other ordinary analytes.

To conclude, the evaluation of the total allowable imprecision of CD34+ cell levels and other FCM analyses providing quantitative results of clinical relevance is now an issue of the utmost importance. Laboratorians should be fully aware of the current performance status of their measurement processes and of the improvement goals that have to be achieved with time and practice.

The analysis of %aTE of absolute CD34+ cell levels is still in its infancy. Since %aTE is a clinical parameter influenced by technology, practice and geography, scientific societies with the help of EQA/PT scheme providers should be able to establish and recommend the desirable and optimal aTE% levels in all the clinical settings, taking into account the necessary continuous improvement of this parameter with time and technical progress.