Reduction of variation in T-cell subset enumeration among 55 laboratories using single-platform, three or four-color flow cytometry based on CD45 and SSC-based gating of lymphocytes

Authors

  • Jan W. Gratama,

    Corresponding author
    1. Laboratory for Clinical and Tumor Immunology, Department of Internal Oncology, Erasmus Medical Center, Rotterdam, The Netherlands
    • Laboratory for Clinical and Tumor Immunology, Erasmus Medical Center - Daniel den Hoed Cancer Center, Groene Hilledijk 301, 3075 EA Rotterdam, The Netherlands
    Search for more papers by this author
  • Jaco Kraan,

    1. Laboratory for Clinical and Tumor Immunology, Department of Internal Oncology, Erasmus Medical Center, Rotterdam, The Netherlands
    Search for more papers by this author
  • Mike Keeney,

    1. Department of Hematology, London Health Sciences Centre, London, Ontario, Canada
    Search for more papers by this author
  • Viv Granger,

    1. UK NEQAS for Leukocyte Immunophenotyping, Department of Haematology, Royal Hallamshire Hospital, Sheffield, United Kingdom
    Search for more papers by this author
  • David Barnett

    1. UK NEQAS for Leukocyte Immunophenotyping, Department of Haematology, Royal Hallamshire Hospital, Sheffield, United Kingdom
    Search for more papers by this author

Abstract

BACKGROUND

Enumeration of CD4+ and CD8+ T-cell subsets provides relevant information for diagnosis and monitoring of patients with cellular immunodeficiencies. As a result, an external quality assurance scheme was implemented in Belgium, The Netherlands, and Luxemburg in 1995. A workshop was held to train the participants in state-of-the art technology for assessment of absolute T-cell subset counts (i.e., a three or four-color, single-platform assay with lymphocyte gating based on CD45 and sideward light scatter) with the aim to achieve between-site coefficients of variation (CVs) <10% and within-site CVs <5% for ≥75% of the participants. Methods: Three send-outs of stabilized blood from a healthy donor were distributed to 55 laboratories, each with the request to perform the standard assay on three occasions. For comparison, each laboratory performed its local technique in parallel. Results: With the standard technique, between-site CVs of ∼8% (CD3+ T cells), ∼9% (CD4+ T cells), and ∼10% (CD8+ T cells) were achieved. Within-site CVs were <5% for 82% (CD3+ T cells) and ∼70% (CD4+ and CD8+ subsets) of the participants. Local techniques yielded between-site CVs of 13%–17% for CD3+, CD4+, and CD8+ T cells. Conclusions: The state-of-the-art technology for T-cell subset enumeration was implemented successfully among 55 Belgian-Dutch laboratories and resulted in significant reductions of between-site variation of absolute CD3+, CD4+, and CD8+ T-cell counts. Cytometry (Clin. Cytometry) 50:92–101, 2002. © 2002 Wiley-Liss, Inc.

Enumeration of the major T-lymphocyte subsets, CD4+ and CD8+, yields important information for diagnosis and monitoring of immunocompromised patients. For the laboratory diagnosis of AIDS, the CD4+ T-cell threshold is 200/μl or 14% of total lymphocytes (1). Furthermore, CD4+ T-lymphopenia is also associated with opportunistic infections in individuals infected with the human immunodeficiency virus (HIV; 2) following hematopoietic stem cell transplants (3) and is also a risk facor for skin cancers in renal transplant recipients (4). CD8+ T-lymphocytosis is a hallmark of the immune response to primary cytomegalovirus (CMV) and Epstein-Barr virus infection in healthy carriers (5, 6) and is associated with recovery from reactivated CMV infection in renal and stem cell transplantation (SCT) recipients (7,8).

The last decade has witnessed significant advances in the flow cytometric technique of T-cell subset enumeration. In the early 1990s, dual-color assays using fluorescein-isothiocyanate (FITC) or phycoerythrin (PE)-labeled monoclonal antibodies (mAb) were used to stain density gradient-isolated mononuclear cell suspensions or white blood cells (WBC) in a stain-lyse-wash technique. During list mode data analyses, lymphocytes were distinguished from other WBC subsets on the basis of their bright expression of CD45 and negativity for CD14, followed by backgating on their forward (FSC) and sideward (SSC) light scatter characteristics (9). This approach performs well on samples with relatively normal proportions of lymphocytes, but light scatter gates on lymphopenic samples frequently suffer from heavy contamination with nonlymphoid cells (10). Nevertheless, the backgating technique formed, at that time, the cornerstone of guidelines for T-cell subset enumeration formulated by the Centers for Disease Control (CDC; 11) and the National Institute for Allergic and Infectious Diseases, Division of AIDS (NIAID-DAIDS; 12). The advent of new fluorochromes such as peridinin chlorophyllin (PerCP) and allophycocyanin (APC), as well as the tandem fluorochromes, PE-Cy5 and PE-Texas Red, made three and four-color assays accessible for the clinical laboratory. This advance prompted the development of improved strategies to select lymphocytes. The T-gating method was based on counterstaining CD4 and CD8 with CD3 as marker to select the T cells in combination with SSC (13). With the double-anchor approaches, lymphocytes are selected first on the basis of bright CD45 expression and low SSC, followed by the selection of T cells on the basis of CD3 positivity, counterstained with CD4 and/or CD8 (14–16). These gating strategies were adopted subsequently in the (updated) CDC (1), NIAID-DAIDS (17), and British Committee for Standards in Haematology (BCSH; 18) guidelines.

Absolute T-cell subset counts were assessed traditionally using a so-called dual-platform technique: (1) the flow cytometer provides the T-cell percentages as fractions of a denominator, i.e., WBC or lymphocytes, and (2) the hematology analyzer provides the absolute WBC count together with a differential count, which must include the denominator. The late 1990s saw the advent of single-platform techniques: the absolute T-cell counts are assessed directly on the flow cytometer in a precisely determined volume of blood sample. Single-platform techniques can either be volumetric (19) or based on counting beads (20, 21). Their advantage is the elimination of the need for the second instrument and the denominator. Side-by-side comparisons of single-platform techniques with a double-anchor gating strategy and dual-platform assays combined with a backgating strategy revealed that the first approach reduced interlaboratory and intralaboratory variation in absolute T-cell subset counting (22, 23).

Since 1995, a biannual scheme of external quality assurance (EQA) of lymphocyte subset enumeration has been operational in Belgium, The Netherlands, and Luxemburg (Benelux) countries, organized by the Flow Cytometry Working Party under the coordinated auspices of the Foundation for Quality Control in Medical Immunology (SKMI), the Foundation for Quality Control of Hospital Laboratories (SKZL), both in The Netherlands, and the Belgian Association for Cytometry (BVC/ABC). About 70 laboratories participate in this scheme. Over the past 5 years, between-laboratory coefficients of variation (CVs) of CD4+ and CD8+ T cells ranged between 15% and 30% for samples with (supra)normal T-cell counts and between 40% and 60% for T-lymphopenic samples (unpublished data). The majority of scheme participants still used dual-platform assays and the older gating strategies. To improve laboratory performance and to reduce between-laboratory variation in results, a workshop was held to introduce state-of-the-art methodology to the participants, i.e., single-platform absolute count assessment and double-anchor gating of T lymphocytes.

MATERIALS AND METHODS

Study Design

This study consisted of three send-outs (in March, April, and May 2001) of a stabilized blood sample to laboratories that participated in the biannual EQA scheme for lymphocyte subset enumeration organized in the Benelux countries by SKMI, SKZL, and BVC/ABC. The aims of the study were to introduce the current gold-standard technique for lymphocyte subset enumeration to the participants; to investigate whether the use of this technique would reduce the variation of CD3+, CD4+, and CD8+ T-cell enumeration between laboratories compared with the use of local techniques to levels of CV <10%; and to reach an intralaboratory CV of <5% for ≥75% of participants (both tested on samples with normal T-cell subset counts).

Prior to the main study, a pilot excercise was performed in which the participants stained and analyzed, in duplicate, short-term stabilized blood (StabilCyte buffer kit; BioErgonomics, St. Paul, MN) with the standard protocol. The data obtained, including list mode data, were reviewed by the coordinating laboratory (Daniel den Hoed Cancer Center).

The test material for Trials 1, 2, and 3 consisted of a single batch of blood from a healthy donor that had been stabilized long-term (24, 25) and divided into 180 aliquots of 2 ml each at the Royal Hallamshire Hospital (Sheffield, UK), shipped by overnight courier to the coordinating laboratory, and stored at 2–8°C until further distribution. The test material had normal T-cell subset counts (CD3+ T cells, 1,527/μl; CD4+ T cells, 869/μl; and CD8+ T cells, 612/μl). For each trial, the coordinating laboratory provided each participant by overnight courier with one vial of test material, the reagents needed to perform the standard technique, and reporting forms. The participants were only informed that the three test samples originated from the same batch of donor blood at the final debriefing of the study (see below). For each trial, each participant was requested to perform the standard technique in duplicate on three separate occasions and and its local technique in duplicate on a single occasion, all within a 1-week period. Each participant was required to return to the coordinating laboratory their results together with full methodological details within 2 weeks of issue. Full data analysis, including individual performance, was returned to the participants within 2 weeks of the closing date. All data were anonymous. Following completion of the study, an overall debriefing report was issued and discussed at a plenary participants meeting. The coordinating laboratory undertook remedial actions with the laboratories that had obtained outlying results.

Standard Technique

Reagents.

The standard technique consisted of a counting bead-based, single-platform, three or four-color method based on identification of lymphocytes on the basis of bright CD45 expression and low SSC signals, followed by the identification of the major two T-cell subsets on the basis of CD3 and CD4 or CD8 expression. The participants were offered the choice between three or four-color assay formats (Table 1). Each participant was provided with reagents originating from the same manufacturer as its flow cytometer; two users of FACStar cell sorters (Becton Dickinson Biosciences [BDB], San Jose, CA) and one user of a CytoronAbsolute instrument (Ortho Diagnostic, Raritan, NJ) were provided with Beckman-Coulter (BC; Miami, FL) reagents because of the incompatibility of these instruments with PerCP-conjugated mAb provided by BDB. In this way, BDB instrument users were provided with TruCount tubes for absolute count assessments, FACS Lysing Solution (10×) as the erythrocyte lysing reagent, and ready-to use mAb cocktails (TriTEST for three-color and MultiTEST for four-color assays). The BC and other instrument users were provided with a vial of Flow-Count bead suspension for absolute count assessments, IOTest lysing reagent (10×) as the erythrocyte lysing reagent, and ready-to use cocktails (OptiClone for three-color and tetraCHROME for four-color assays).

Table 1. Standard Technique: Assay Formats*
TubeFL1FL2FL3FL4
  • *

    ECD, energy-coupled dye. All other abbreviations as in the text.

  • a

    Number of instruments.

  • b

    mAb clones used: HLe-1 (CD45), SK7 (CD3), SK3 (CD4), and SK1 (CD8).

  • c

    Including two FACStar and 1 CytoronAbsolute instruments.

  • d

    mAb clones used: clones Immu19.2 (CD45), UCHT1 (CD3), 13B8.2 (CD4), and B9.11 (CD8).

  • e

    mAb clones used: clones 83821F4A (CD45), UCHT1 (CD3), SFCI12T4D11 (CD4), and SFCI21Thy2D3 (CD8).

Three-color, Becton Dickinson Biosciences (n = 24)ab
 1CD3 FITCCD4 PECD45 PerCP
 2CD3 FITCCD8 PECD45 PerCP
Three-color, Beckman-Coulter (n = 11)acd
 1CD45 FITCCD4 PECD3 PE-Cy5
 2CD45 FITCCD8 PECD3 PE-Cy5
Four-color, Becton Dickinson Biosciences (n = 15)ab
 1CD3 FITCCD8 PECD45 PerCPCD4 APC
Four-color, Beckman-Coulter (n = 8)ae
 1CD45 FITCCD4 PECD8 ECDCD3 PE-Cy5

Sample preparation.

The BDB users added 20 μl TriTEST or MultiTEST mAb cocktail to 50 μl of blood in TruCount tubes. After gently vortexing, the samples were incubated for 15 min at room temperature. Thereafter, 450 μl FACS Lysing Solution (1×) was added and vortexed gently followed by a further 15-min incubation. Flow cytometry was completed within 1 h. Samples were protected from light up to the point of flow cytometry. The BC users used an identical protocol with the exception that 20 μl OptiClone or 10 μl tetraCHROME mAb cocktail was used to 100 μl of blood, as well as 2 ml of IOTest lysing reagent (1×) when required. Between the second incubation step and flow cytometry, the BC users added 100 μl Flow-Count beads with the same pipette used for aliquoting the blood samples. Reverse pipetting was mandatory throughout (10).

Flow cytometry.

The instruments were set up according to local standard operating procedures. The BDB and BC users were provided with software-specific, three or four-color templates for data acquisition and analysis (CellQuest and System-2 software, respectively). The BDB users set a threshold on CD45 to exclude debris (i.e., CD45-negative events) while retaining the TruCount beads (being FL3bright but having very low FSC signals). In contrast, the BC users defined debris as events with very low FSC signals and thus set an FSC threshold accordingly.

A minimum of 2,000 lymphocyte events (defined as CD45bright,SSClow) was acquired with storage of ungated list mode data. For data analysis, lymphocytes were gated according to the above definition, and the CD4+ and CD8+ T-cell subsets were identified using quadrant statistics as CD3+,4+, and CD3+,8+, respectively. TruCount beads were identified as events with high SSC and very bright FL1, FL3, and FL2 (three-color) or FL4 (four-color) signals (doublet beads were included in the calculations). Flow-Count beads were identified in three-color analyses as events with high SSC and very bright FL3 signals, followed by the selection of singlet beads as the major FL3+ population in a time-versus-FL3 histogram; the minor population of doublet beads with brighter FL3 signals than the singlets was excluded. In four-color analyses, Flow-Count singlet beads were identified in an FL2-versus-count histogram as a single, tight peak of events with very bright FL2 signals.

Statistical Analyses

In the Results and Discussion, we refer to “site” as the unique combination of a laboratory performing a specific technique. Therefore, the standard technique was performed at 58 sites and local techniques at 55 sites. Each site was assigned a unique number (1–58) that is used consistently for referral purposes in this report.

Statistics.

The data were nearly always parametric and thus CVs were calculated from means and standard deviations (SD). Standard statistical tests were performed using SAS software (Statistical Analysis Systems, Cary, NC) as indicated in the text.

Calculation of test results and variations.

As per study protocol, each test consisted of a set of duplicate assay results. Each test result was computed as the mean of duplicate results. As a measure for the variation within a test for each site, the CV of duplicate results was taken. In case only a single assay had been performed, the single result was used as the test result. For the standard technique, each trial consisted of a set of three tests performed at different occasions. For the trial result, the median of this set of three test results was used. As a measure for the variation between tests within a trial for each site (termed “within-site variation”), the CV of the three test results was taken. As a measure for the variation between sites within a trial, the CV of the first assessment of the set of duplicates for the first of the three requested tests per trial per site was used.

RESULTS

Participants and Their Local Techniques

Participating in this study were 55 (19 Belgian, 36 Dutch) of the approximately 70 laboratories that regularly participated in the biannual EQA scheme for lymphocyte subset enumeration organized in the Benelux countries by SKMI, SKZL, and BVC/ABC. Thirty-nine operated flow cytometers manufactured by BDB (19 FACScalibur, 17 FACScan, 2 FACStar, and 1 FACStrak), 15 operated BC XL flow cytometers, and 1 an Ortho CytoronAbsolute instrument. Table 1 shows the reagent combinations used by the 55 laboratories. Three laboratories performed both three and four-color formats in parallel.

Table 2 is an overview of the local techniques in use by the 55 Belgian and Dutch laboratories. To establish absolute counts, only one third of the laboratories used single-platform assays. Among the two thirds assessing absolute T-cell counts using dual-platform assays, the use of hematology analyzer-derived lymphocyte counts as the denominator for the assay was predominant (78%). All 18 laboratories with single-platform techniques used lyse-no-wash methods for sample preparation. Interestingly, this method had also been adopted by a minority of laboratories using double-platform techniques (8 of 37 labs [22%]). The majority of laboratories (56%) used either a CD45 and SSC-based (14) or a CD3 and SSC-based (13) gating strategy, with the rest using either CD45 and CD14 backgating (9) or light scatter- based (FSC and SSC) gating (Table 2).

Table 2. Overview of Local Techniques*
Specification of techniqueLaboratories (n = 55) (%)
  • *

    HA, hematology analyzer. All other abbreviations as in the text.

  • a

    Sixteen of these laboratories (62%) used the single-platform assay.

  • b

    One of these laboratories (20%) used the single-platform assay.

  • c

    One of these laboratories (11%) used the single-platform assay.

Absolute counting technique
 Single-platform18 (33)
  TruCount10 (56)
  FACScount1 (6)
  Flow-Count7 (39)
  CytoronAbsolute0
 Dual-platform37 (67)
  HA, WBC count8 (22)
  HA, lymphocyte count29 (78)
Sample preparation technique
 Lyse-no-wash26 (47)
  Single-platform18 (69)
  Dual-platform8 (31)
 Lyse-and-wash29 (53)
  Single-platform0
  Dual-platform29 (100)
 Gating strategy
  CD45 and SSC26 (47)a
  CD3 and SSC (T-gating)5 (9)b
  CD45 and CD14 (backgating)15 (27)
  FSC and SSC9 (16)c

Pilot Trial

The main purpose of the pilot trial served to detect and correct flaws in performing the standard technique. Sufficient information for this purpose was submitted by 52 of the 58 sites (90%); problems were met by 20 sites (34%; Table 3). Several laboratories reported problems with obtaining a reliable WBC differential count on the stabilized blood. In response to these problems, a general recommendation was issued to users of dual-platform techniques to use the WBC rather than the lymphocyte counts generated by the hematology analyzers as the denominator for establishing absolute T-cell subset counts.

Table 3. Pilot Study: Technical Problems in Performing the Standard Technique
Specification of problem or commentNumber of sites
  • a

    One site had two problems with instrument set-up.

Sites with technical problems detected20 (34%)
 Problems with instrument set-upa8
  FSC or FL4 window of analysis placed suboptimally3
  FL2-% FL1 color compensation set too high or too low2
  CD45 or FSC threshold placed too high or too low3
 Sample preparation errors4
 Protocol violations3
  Dual instead of single-platform technique used2
  Aberrant analytical strategy of list mode data1
 Too restricted CD45 and SSC lymphocyte gate set2
 Errors in absolute cell count calculation3
No comments needed32 (55%)
No or insufficient information submitted to allow comments 6 (10%)

Within-Test Variation

Each participant could generate a maximum of three data points per trial and T-cell subset for the standard technique and one data point per trial and subset for its local technique. For the standard technique, the median results of the three data points per trial and subset are shown Table 4, as are the data obtained with the local techniques. The within-test variation obtained with standard and local techniques was similar and tended to be even lower for CD4+ T cells for the local techniques. For both standard and local techniques, the within-test variation was smallest for CD3+ T cells and slightly larger for CD4+ and CD8+ T cells.

Within-Site Variation

These data were only available for the standard technique (Table 4). The study target level (within-site CV < 5% by ≥75% of sites) was reached for CD3+ T cells in all three trials, for CD8+ T cells reached in Trial 2, and for CD4+ T cells almost reached in Trial 2. However, the performance of the group in performing reproducible CD4+ and CD8+ T cells dropped slightly in Trial 3 compared with Trial 2.

Table 4. Variation Within Tests and Sites
TrialNo. CD3+ T cells (%)No. CD4+ T cells (%)No. CD8+ T cells (%)
  • a

    The CV of duplicate assessments was taken as a measure for the intra-assay variation. Per trial, the median result of three tests (each consisting of a set of duplicate assessments) performed on different occasions was used. Shown are the number of sites with intra-assay variation <5%/total number of sites with these data available.

  • b

    The CV of three tests performed on different occasions was taken as a measure for the intra-site variation. Each test result was computed as the mean of duplicate assessments. Shown are the number of sites with intra-site variation <5%/total number of sites with these data available.

  • c

    The CV of duplicate assessments was taken as a measure for the intra-assay variation. Per trial, the result of a single set of duplicate assessments was used. Shown are the number of sites with intra-assay variation <5%/total number of sites with these data available.

Standard technique: sites with variation within tests <5%a
 150/58 (86)43/58 (74)43/58 (74)
 245/55 (82)39/55 (71)42/55 (76)
 349/56 (88)42/56 (75)40/56 (71)
Standard technique: sites with variation within site <5%b
 140/53 (75)36/54 (67)30/53 (57)
 245/55 (82)40/55 (73)41/55 (75)
 347/57 (82)39/57 (68)37/57 (65)
Local techniques: sites with variation within tests <5%c
 130/34 (88)24/34 (71)28/34 (82)
 234/35 (97)28/35 (80)30/35 (86)
 326/32 (81)28/32 (88)23/32 (72)

Between-Site Variation

The variation between sites using either the standard or the local techniques is summarized in Table 5. Using the standard technique, the between-site variation of all three T-cell subsets was reduced from at least 15% (Trial 1) to below 10% (Trial 3). In contrast, no trial, or parameter, achieved the target CV of <10% if the local techniques were employed. Even when the outliers from Site 39 were removed from the data, due to errors in absolute count calculation, the local techniques still failed to reach the target CV.

Table 5. Variation Between Sites: Standard and Local Techniques*
TrialNo. CD3+ T cellsNo. CD4+ T cellsNo. CD8+ T cells
  • *

    The CV was taken as a measure for the variation between sites. For each single test result, the first of duplicate tests of the first (standard technique) or only assay (local techniques) per trial was taken.

  • a

    CVs between brackets represent variation between sites after removal of the outlying results of Site 39 (i.e., standard technique: CD3, 73/μl; CD4, 41/μl; CD8, 29/μl; local technique: CD3, 76/μl; CD4, 57/μl; CD8, 36/μl).

Standard technique
 1a15%17%15%
(7.3%)(11%)(7.1%)
 28.0%8.7%10.8%
 38.1%9.4%9.7%
Local techniques
 1a20%21%22%
(14%)(15%)(16%)
 213%15%13%
 315%13%17%

Performance of Individual Sites

Figure 1 shows the distribution of results by site obtained during all three trials following exclusion of the outliers (Site 39). The coordinating laboratory took remedial action with Site 39 after Trial 1, whereupon its performance normalized. Figure 1 highlights that two sites (16 and 35) consistently overestimated CD3+, CD4+, and CD8+ T-cell counts (i.e., >90th centile), whereas two other sites underestimated these values (i.e., <10th centile). Three sites (34, 40, and 54) also underestimated these parameters using their local techniques only. No correlation was found in the performance of sites using standard versus local techniques.

Figure 1.

Performance of individual sites (left panels, standard technique; right panels, local techniques) by T-cell subset (upper panels, CD3+ T cells; middle panels, CD4+ T cells; lower panels, CD8+ T cells) in Trials 1, 2, and 3. Results of individual tests (each calculated as the mean of duplicate assay results) are shown by dots. Each site generated a maximum of nine data points. For each graph, the median values of each set of data points per site are connected by a line to highlight outliers. As a reference, the horizontal lines indicate the 10th and 90th centiles of the datasets consisting of the median results per site. Logarithmic scales were used for the y-axes in order to compress the figure. The outliers of Site 39 (Trial 1; standard and local techniques) were excluded in order to allow better visualization of the data of the other sites. T-cell subset counts between sites were significantly different for all subsets and techniques (Kruskal-Wallis test; P values are shown above each panel).

Sources of Between-Site Variation

Standard technique.

The major source of variation in results for absolute CD3+, CD4+, and CD8+ T-cell counts was constituted by differences among the individual sites (Fig. 1; P = 0.0001 for each T-cell subset using the Kruskal-Wallis test). Furthermore, stratification of the results by trial and then by the combination of (1) instrument manufacturer and type (i.e., BC XL; BDB FACScalibur, FACScan, or FACStrak; and FACStar or CytoronAbsolute) and (2) type of assay (i.e., three or four-color) yielded interesting information (Fig. 2). This stratification yielded five subgroups: BC three-color, BC four-color, BDB three-color, BDB four-color, and OTH(er) three-color. The greatest impact of these combinations was observed in Trial 3 (CD3+ T cells, P = 0.003; CD4+ T cells, P = 0.003; CD8+ T cells; P = 0.01).

Figure 2.

Combined effect of type of flow cytometer (BC, Beckman-Coulter XL; BDB, Becton Dickinson Biosciences FACScalibur, FACScan, FACSort or FACStrak; OTH, FACStar or CytoronAbsolute) and type of assay (3, three-color; 4, four-color) on T-cell subset enumeration results (standard technique). Per trial, the median results of triplicate tests (each consisting of the mean of duplicate assay results) per site were used for this computation. The results of the five groups were compared per trial and T-cell subset using the Kruskal-Wallis test. P values are shown above each panel. The width of the boxes are proportional to the numbers of observations within the respective groups. The line in the middle of each box represents the median of the data. The boxes extend from the 25th to the 75th centiles (i.e., the interquartile range). The lines emerging from the boxes (i.e., the whiskers) extend to the upper and lower adjacent values, which are defined as 1 ½ × the interquartile range rolled back to where there are data. Observed points more extreme than the adjacent values (i.e., outside values) are plotted individually (symbol, circles).

Local techniques.

As for the standard technique, differences among the laboratories constituted the major source of variation in results for the three T-cell subsets (P values ranging between 0.002 and 0.006; Kruskal-Wallis test). We then analyzed whether various aspects of the local techniques would significantly affect the outcome of T-cell subset enumeration. Dual-platform, lyse-and-wash techniques had a higher variation than lyse-no-wash approaches, yielding higher CD3+, CD4+, and CD8+ T-cell counts (Fig. 3, left panels). Furthermore, the choice of gating strategy also had a significant impact on the results obtained with CD45 and CD14 backgating and with FSC and SSC-based gating strategies, giving the highest variation and the lowest counts, respectively. A multivariate analysis revealed that sample preparation in combination with absolute cell count technique and gating strategy independently affected the results of T-cell subset enumeration.

Figure 3.

Effects of technical aspects of local techniques on the results of CD3+, CD4+, and CD8+ T-cell enumeration. Left panels: Sample preparation in combination with the technique for absolute cell count assessment (Single, single-platform, lyse-no-wash; Dual_LNW, dual-platform, lyse-no-wash; Dual_L&W, dual-platform, lyse-and-wash). Right panels: Gating strategy (CD45 &SSC, gating strategy based on selection of lymphocytes as CD45bright and SSClow events; CD3&SSC, gating strategy based on selection of T lymphocytes as CD3pos and SSClow events; CD45&14, selection of lymphocytes on the bases of FSC and SSC characteristics, guided through backgating on CD45bright, CD14neg events; FSC&SSC, classical selection of lymphocytes by gating on FSClow to intermediate, SSClow events. Upper panels: CD3+ T cells; middle panels, CD4+ T cells; lower panels, CD8+ T cells. See the legend to Figure 3 for a description of the box-and-whisker plots. Results of univariate statistics (Kruskal-Wallis test) are shown at the top of each panel.

DISCUSSION

Education is one of the cornerstones of an adequate scheme for EQA. When an EQA scheme for lymphocyte subset enumeration was set up for the Benelux laboratories in 1995, it was initiated with a workshop to introduce a technique considered by the organizers to be state-of-the-art and to have participants compare this method with their own techniques (26). The state-of-the-art technique at that time was a three-color, dual-platform assay based on CD45 and CD14 backgating of lymphocytes and on exclusion of debris by gating on nucleated cells using the nucleic acid stain, LDS-751. Ever since, the significant technical progress in the field and the lack of reduction of between-laboratory variation in the Benelux scheme (unpublished data) prompted us to organize a second workshop in 2001. Our aim was to introduce the current state-of-the-art methodology to the participants, i.e., single-platform absolute count assessment and double-anchor gating of T lymphocytes.

For the current workshop, we chose a study design based on the one that has been proven to be effective in standardizing CD34+ hematopoietic stem cell enumeration among a group of 24 European laboratories (27). The current workshop used stabilized test material, in order to exclude sample decay as a confounding factor for between-site variation; consisted of a pilot round, followed by three trials in which within-test, within-site, and between-site variation was studied; provided remedial action between rounds, i.e., troubleshooting and training of laboratories reporting aberrant results; and compared the performance of the standard, state-of-the-art methodology with that of the local techniques. The local techniques in the majority of laboratories were not current, i.e., they were based on dual-platform assay formats (67%) and used suboptimal gating strategies (44% of laboratories). Similar to the CD34 study (27), we set ourselves as study targets for absolute numbers of CD3+, CD4+, and CD8+ T cells: 1) between-site CV <10% and 2) within-site <5% for ≥75% of sites.

The key result of the current study was that use of the standard technique clearly reduced between-site variation compared with the local techniques. For between-site variation, the study target (CV <10%) was achieved for all three parameters, whereas the use of local techniques resulted in between-site CVs ranging between 13% and 17%. For the standard technique, the target for within-site variation was achieved for CD3+ T cells and was (almost) achieved for CD4+ and CD8+ T cells. Expectedly, within-test variation was slightly smaller than within-site variation (i.e., the variation of three replicate tests).

Our participants performed as well as 36–39 Canadian laboratories. In a workshop setting, they tested four samples of commercial stabilized blood products with normal T-cell subset counts using the same standard protocol as used in our workshop (15). In the Canadian study, between-laboratory CVs of 7%–10% (CD3+ T cells), 7%–9% (CD4+ T cells), and 10%–15% (CD8+ T cells) were achieved. Furthermore, the results of our workshop confirmed the results of two studies by the NIAID DAIDS New Technologies Evaluation Group, in which similar standard protocols as used in our workshop were compared with a dual-platform, CD45 and CD14 backgating-based method (22, 23). The combined results of these four studies unambiguously show that the method based on a single platform combined with double-anchor gating on T cells reduces between-laboratory variation. Supporting these findings are the results of the UK NEQAS Immune Monitoring Scheme, in which the variation of absolute CD4+ T-cell counts between laboratories using single-platform–based methods was significantly lower than that between laboratories using dual-platform–based methods (28, 29).

If the education provided by workshops or an EQA scheme is to be effective, a “learning effect” must be visible; between-laboratory variation is usually used as the marker for this learning effect. A learning effect was reported for the EQA scheme for T-cell subset enumeration set up by the U.S. Department of Army (30). This study comprised a group of 8–13 laboratories that performed dual-platform, CD45 and CD14 backgating as the standard method on monthly distributed specimens. Close monitoring of the participants' performance by the study coordinators resulted in a reduction of between-laboratory CVs from 15%–50% to 5%–10% four years later (30). The Canadian Quality Assurance Program for T-cell subset enumeration (15) is based on the use of a standard technique (i.e., single-platform–based double-anchor gating on T cells); bimonthly send-outs of blood specimens, with remedial actions of the coordinating laboratory to centers reporting outliers; “hands-on” training courses; annual consensus meetings; and workshops addressing the performance of instruments and laboratory methods. Expectedly, this intensive scheme of quality assurance was effective: the longer a laboratory participated in the program, the smaller was its variation from the group's mean result (15).

Most of the learning effect in our 2001 workshop occurred between the pilot trial and the study proper. Effective remedial actions were then undertaken in 20 sites (34%) experiencing problems in performing the standard technique (Table 3). As a result, participant performance was stable and sufficient during the subsequent three trials. More than one half of the local techniques were based on dual-platform methodology using lymphocytes as the denominator. As the use of stabilized blood does not allow reliable assessment of the WBC differential by a hematology analyzer, a general recommendation was issued to use WBC rather than lymphocytes as the denominator for the local techniques. Indeed, the between-laboratory variation for the local techniques (CV ∼15%) was better in the 2001 workshop than observed in the regular EQA scheme (i.e., between 15% and 30% for samples with [supra]normal T-cell counts and between 40% and 60% for T-lymphopenic samples [unpublished data]). The use of stabilized test material instead of fresh test material (such as in the EQA scheme) may also have contributed to the improved result of the 2001 workshop due to the avoidance of sample decay during transport (26).

We also addressed the source of between-site variation of T-cell subset enumeration results. Differences between individual sites constituted the major source of variation with both techniques. For the standard technique, part of this variation was due to the combination of instrument and assay format (i.e., three or four-color assays). In particular, T-cell subset counts obtained using four-color assays on BC XL instruments tended to be lower than those obtained with three-color assays on the same type of instrument. Of note, these differences were higher in Trial 2 and reduced for CD4+ and CD8+ T cells in Trial 3. Reviewing the percentage of CD3+ T cells and subsets revealed these differences to be statistically lower for the BC four-color than for the three-color assay, indicating that the phenomenon is not related to bias introduced by the addition of counting beads. The NAID-DAIDS study of four-color lymphocyte subset analysis on BC XL instruments also showed a slight negative bias for CD4 and CD8 in four of five laboratories when compared with the predicate method, and a slight positive bias in the fifth laboratory (23). However, on closer examination, it appeared that the bias was due mainly to overestimation or underestimation by the predicate method in at least two laboratories (23). One possible significant difference between the NIAID-DAIDS study and ours is that NIAID-DAIDS used a fully automated analysis protocol (i.e., TetraONE [BC]), whereas we used manual data analysis. As the use of automated software removes the subjectivity of gating from the analysis, variability among laboratories is expected to decrease. Instrument set-up may also be an issue in this context. Single-platform, four-color analyses on a single laser system require great attention to detail during compensation. In our study, compensation was left to the individual sites, allowing an increased potential for overcompensation or undercompensation. That said, for Trial 3, the actual negative bias between the four and three-color methods for CD4+ and CD8+ T cells was 2%, a figure that is statistically significant, but is unlikely to be of clinical significance.

Regarding the local techniques, the larger variation observed with dual-platform, lyse-and-wash applications compared with (single-platform) lyse-no-wash techniques confirms the observations made in the UK NEQAS Immune Monitoring Scheme (28, 29). The widely held view that CD45 and SSC-based gating of lymphocytes constitutes the most robust analytical method for lymphocyte immunophenotyping (16–18) is supported by our result that local methods employing this approach showed the lowest variation compared with CD45 and CD14 backgating and FSC and SSC-based gating strategies.

A comparison of the results of the 1995 (26) and 2001 workshops reveals that significant progress has been made. In 1995, only 7% of participants used a single-platform assay (all using volumetry, i.e., a CytoronAbsolute instrument) versus 33% in 2001 (all but one using counting beads). As for gating strategies, CD45 and SSC gating increased from 7% in 1995 to 47% in 2001, whereas the use of CD45 and CD14 backgating decreased from 69% to 27% and that of FSC and SSC-based gating strategies from 24% to 16%. In 2001, use of the standard technique reduced between-laboratory variation compared with that of local techniques, whereas the variation obtained with standard and local techniques was similar in 1995. The lack of effect of the 1995 standard technique may be explained by the fact that CD45 and CD14 backgating strategy of the standard technique was already used by about 70% of participants in their local protocols; the standard protocol introduced the use of the LDS-751 stain in a three-color technique, whereas most participants at that time had no experience with this stain or with three-color flow cytometry; and the workshop consisted of a single send-out, i.e., there was no opportunity to visualize a learning effect of remedial actions.

The major technical difficulties with T-cell subset enumeration identified in 1995 have now been resolved: unreliable WBC differentiation on leukopenic samples can now be avoided by using a single-platform technique; suboptimal selection of cells of interest is resolved with the use of double-anchor gating of T cells; and the variation in sample processing has been reduced by omission of washing steps. Although this result certainly is progress, its translation into improved laboratory performance using local three or four-color mAb cocktails on fresh patient blood samples must be awaited. Indeed, our EQA scheme makes use of patient specimens with varying degrees of CD4+ T-lymphopenia and should be able to give insight into this important question within the coming years.

Acknowledgements

We gratefully acknowledge the support of Beckman-Coulter (European headquarters, Nyon, China; Dr. A. Thews; Beckman-Coulter Nederland, Mijdrecht, The Netherlands; Mr. C. Heije; and Mrs. W. Harkema) and Becton Dickinson Biosciences (European headquarters, Erembodegem, Belgium; Drs. L. Ketele, R. Götte, and R. van den Beemd) for providing reagents (counting beads, monoclonal antibodies, and erythrocyte lysing buffers) to the users of Beckman-Coulter or Ortho, and Becton Dickinson Biosciences instruments, respectively. Supported in part by grant QLRI-2000-00436 awarded by Directorate General XII of the European Commission (Brussels, Belgium) in the setting of the Fifth Framework Programme “Quality of Life and Management of Living Resources” and the SKZL “Kalibratie 2000” project grant. This study was performed under the auspices of the Foundation for Quality Control in Medical Immunology (SKMI), the Foundation for Quality Control of Hospital Laboratories (SKZL), and the Belgian Association for Cytometry (BVC/ABC) with the participation of (alphabetical order): F. Bergkamp (Medial Medisch-Diagnostische Laboratoria, Haarlem); M. Bernier (Institut Bordet, Brussels); M. Blom (Unilever Research and Development Vlaardingen, Vlaardingen); N. Boeren (St. Elisabeth Ziekenhuis, Tilburg); M. Bon (Medisch Spectrum Twente, Enschede); X. Bossuyt (Universitair Ziekenhuis Gasthuisberg, Leuven); T. Braeckevelt (AZ Zusters van Barmhartigheid, Ronse); C. Bridts (Universiteit Antwerpen, Antwerp); G. Brouwer (CKHL, LUMC, Leiden); B. Catinieaux (St. Pieters Ziekenhuis, Brussels); B. Châtelain (Clinique Universitaire de Mont Godinne, Yvoir); A. Claessen (St. Antonius Ziekenhuis, Nieuwegein); J. Damoiseaux (Klinische Immunologie, AZM, Maastricht); M. de Metz (Canisius Wilhelmina Ziekenhuis, Nijmegen); V. Deneys (Cliniques Universitaires Saint-Luc, Brussels); W. de Vries (St. Franciscus Gasthuis, Rotterdam); M. De Waele (AZ-Vrije Universiteit Brussel, Brussels); J.L. D'Hautcourt (Hôpital de Warquignies, Boussu); O. Fagnart (Sint Etienne Kliniek, Brussels); A. Hendrix (Catharina Ziekenhuis, Eindhoven); P. Herbrink (Diagnostisch Centrum SSDZ, Delft); N. Hougardy (Cliniques du Sud Luxembourg, Arlon); W. Huisman (MCH Westeinde, Den Haag); B. Husson (Hôpital de Jolimont, Haine-Saint-Paul); G. Janssen (St. Joseph Ziekenhuis, Veldhoven); J. Kerckhaert (Ziekenhuis Rijnstate, Arnhem); R. Kester (Medische en Tumorimmunologie, Erasmus MC, Rotterdam); P. Kuiper (Isala Klinieken - Sophia, Zwolle); P. Lankheet (Immunohematologie en Bloedtransfusie, LUMC, Leiden); P. Limburg (Academisch Ziekenhuis Groningen, Groningen); R. Malfait (AZ Middelheim, Antwerp); A. Martens (Twenteborg Ziekenhuis, Almelo); P. Meeus (OL Vrouwziekenhuis, Aalst); H. Meisters (Hematologie, AZM, Maastricht); A. Mulder (Bosch Medicentrum, Den Bosch); W. Nooijen (Antoni van Leeuwenhoek Ziekenhuis, Amsterdam); T. Out (Academisch Medisch Centrum, Amsterdam); D. Park (Slotervaart Ziekenhuis, Amsterdam); P. Piro (CHU Vésale, Montigny le Tilleul); O. Pradier (Universite Libre de Bruxelles, Brussels); F. Preyers (UMC St. Radboud, Nijmegen); G. Rijkers (UMC - Wilhelmina Kinderziekenhuis, Utrecht); K. Roozendaal (OL Vrouwegasthuis, Amsterdam); J. Rummens (Virga Jesse Ziekenhuis, Hasselt); P. Scholten (VUMC, Amsterdam); W. Slieker (Medisch Centrum Alkmaar, Alkmaar); R. Slingerland (Isala Klinieken-Weezenlanden, Zwolle); H. Storm (Stichting Klinisch Chemisch Laboratorium, Leeuwarden); L. Vaessen (Transplantatie-Immunologie, Erasmus MC, Rotterdam); D. Van Bockstaele (Universitair Ziekenhuis Antwerpen, Edegem); R. van den Beemd (BD Biosciences Benelux, Erembodegem-Aalst); H. van Dijk (Ziekenhuis Eemland, Amersfoort); E. van Lochem (Immunologie, Erasmus MC, Rotterdam); J. van Wersch (Atrium, Heerlen); W. Veenendaal (Ziekenhuis Leyenburg, Den Haag); G. Vrielink (St. Lucas-Andreas Ziekenhuis, Amsterdam).

Ancillary