Cytometry (Communications in Clinical Cytornetry) 26 :22 -31 (1996)

Standardizing Flow Cytometry: Construction of a Standardized Fluorescence Calibration Plot Using

Matching Spectral Calibrators A. Schwa-, E. F e h d e z Repollet, R Vogt, and Jan W. Gratama

Flow Cytometry Standards Corporation, San Juan, Puerto Rico 00919-4344 (AS.); School of Medicine, Department of Pharmacology, University of Yuerto Rico, San Juan, Puerto Rico 00936-5067 (E.F.R.); Centers for Disease Control, Atlanta,

Georgia 30333 (R.V.); Dr. Daniel den Hoed Cancer Center, Department Clinical and Tumor Immunology, 3008 AE Rotterdam, The Netherlands (J.W.G.)

Calibration of flow cytometers is becoming an increasingly important issue for both quality control of instrument performance and quantitation of antibody binding capacity of cells. Due to the numerous different instruments and analysis software currently available, a standardized method of calibration is necessary if interlaboratory comparison of instrument performance and antibody binding is to be achieved. This report describes a new methodology to obtain a standard calibration plot that can be derived from all instruments and from which specific instrument-independent performance parameters may be calculated that can be used to directly compare the performance and setup of these instruments. The requirements that the calibrated standards must meet are discussed, as well as the acceptable ranges proposed for the instrument- independent performance parameters. In addition, data are presented from standard calibration plots gen- erated by different flow cytometers in numerous laboratories. The corresponding Primary Performance Parameters calculated from these plots are presented and compared. It is expected that the use of this calibration method may help standardize flow cytometric measurements and will provide instrument-inde- pendent performance parameters to monitor quality control of instruments and reagents. 0 1996 Why-Liss, Inc.

Key terms: Calibration, standardization, molecules of equivalent soluble fluorochrome (MESF), antibody binding capacity (ABC), quality control

In recent years the need for Quality Control (QC) for flow cytometry has been extensively discussed ( 1-6). This has been partially the result of an increased use of flow cytometry in the diagnosis and monitoring of many diseases. At the same time, it has been recognized that in order to assure accurate and clinically relevant compar- isons of flow cytometric data, calibration procedures must be implemented to monitor instrument perfor- mance and sample preparation. Moreover, the implemen- tation of these procedures becomes essential when dif- ferent laboratories need to compare data that will eventually translate into patient care. The need for inter- laboratory flow cytometry data comparison is docu- mented by the increased number of interlaboratory pro- ficiency testing and quality assessment programs, as well as the numerous national and international surveys and ongoing clinical trials involving flow cytometry (7-10).

A first step in flow cytometry data comparison is the determination of adequate instrument performance ( 1- 6). Ruling out instrumentation as a source of variability is

0 1996 Wiley-Liss, Inc.

a basic key to minimize differences in research and clin- ical determinations. Thus, this discussion will focus on new methodology available to calibrate and monitor flow cytometer performance by using matching spectra Cali- brators and a correlation procedure based on linear re- gression equations.

Although determination of a correlation between a flu- orescence signal and a flow cytometer’s response via cal- ibrated standards may appear a rather straightforward and trivial accomplishment, this perception is not valid if multiple instruments are involved. In fact, such a task becomes extremely complex due to a variety of require- ments that need to be considered. For example, not only are the fluorescence units of the calibrated standards im- portant, but also the format in which the instrument re- ports its response to these calibrators. Calibration has

Received for publication April 1995; accepted October 1995.

STANDARDIZING FLOW CYTOMETRY 23

FL1 FL2

500 510 520 530 540 550 560 570 580 590 nm

FIG. 1. Emission spectra of a Type I l la standard without spectra matching and a Type I I Ib standard with spectra matching fluorescein labeled cells. Note, that under the FL1 and FL2 barrier filters, the ratios of the integrated signals from the calibrator standard and cell sample differ when the standards do not have matching spectra for FL2. Spe- cifically, since B = C in FL1, then A/B = A/C which is also equal DIE in FL2. However, since E C F in FL2, then DIE C D/F. Therefore, stan- dards with matching spectra preserve the standard to cell ratio across the complete spectrum and can be used as quantitative standards in differ- ent instruments which have different barrier filters, while standards that do not have matching spectra can not be used across instruments.

been defined as the process of evaluating and monitoring an instrument to provide a known relationship between the measurement response and the value of the substance that is being measured by the test procedure (Clinical Laboratory Improvement Act 1988, 11). In an attempt to provide such a calibration process for flow cytometry, we proposed a standardized calibration method for flow cytometric fluorescence quantitation from which Pri- mary Performance Parameters can be derived with prac- tical meaning in terms of instrument performance. This methodology is expected to improve the performance of multiple instruments at multiple sites for quantitative flu- orescence assessments.

REQUIREMENTS OF THE CALIBRATED STANDARDS The main requirement for a calibration standard for

flow cytometry is that its excitation and emission spectra will match those of the samples to be analyzed ( 11,12). Matching spectra of the calibrators and samples will nor- malize the responses between instruments that have dif- ferent barrier filters. As illustrated in Figure 1, differences in emission spectra can give rise to proportionally differ- ent levels of energy interacting with the calibrators and sample. Similar relationships are also found with the ex- citation spectra. If different instruments have different barrier filters, then these proportions will not remain constant, and thus the fluorescence intensity values of samples will vary among different instruments.

To illustrate this requirement, calibration standards that do not have spectral matching (Rainbow Calibration

Particles, Spherotech, Inc., Libertyville, IL) and calibra- tion standards that do have spectral matching (Quantum 24 fluorescein isothiocyanate (FITC) Microbeads, Flow Cytometry Standards Corp., San Juan, PR) were used to obtain comparative quantitative data across different re- gions of the spectrum. The classifications Type IIIa (non- matching standards) and Type IIIb (matching standards) (13-15) are used to refer to these calibration standards. The FL1 (-530 nm region) and FL2 (-565 nm region) channels of a single instrument (FACScan, BDIS) were used to simulate extreme differences of barrier filters among different instruments.

In addition to the calibration standards, samples of nor- mal peripheral whole blood were labeled with CD4-FITC antibody (BDIS), lysed, and washed (16). Both sets of calibration standards, as well as the blood samples, were also analyzed in two different flow cytometers (FACScan, BDIS). List mode files of 10,000 events were obtained using a forwadorthogonal scatter gate with compensa- tion off. Data were collected from both the FL1 and F12 channels.

A calibration plot was constructed for the FL1 channel from the assigned values and mean peak channels of the Quantum 24 fluorescent microbead populations. The mean peak channels of the Rainbow Calibration Particles were determined and assigned MESF values according to the Quantum 24 FITC microbead calibration plot in in- strument # 1. After this procedure, both the Type IIIa and the Type IIIb calibration standards described the same calibration plot for the FLl channel (Fig. 2a). Using this calibration plot, samples of CD4-FITC labeled lympho- cytes were found to have a fluorescence intensity of 67,000 MESF units.

Calibration plots were also constructed for the FL2 channel by plotting the assigned MESF values obtained from the FLl channel, for each of the two types of stan- dards, against the mean peak channels of the standards for the FL2 channel. As illustrated in Figure 2b, two different calibration plots were generated for the Type IIIa and Type IIIb standards. This is due to the fact that the FL2 channel covers the tail region of the FITC emission spec- tra, where even small differences in spectra between standard and sample can have a significant effect (Fig. 1 ).

In the FL2 channel, the same CD4-FITC labeled lym- phocytes were found to have an intensity of 59,000 MESF units (1 1% lower determination than for FL1) using the Type IIIb calibration plot, whereas the same lymphocytes had an intensity of 5,200 MESF units (91% lower deter- mination than for FL1) using the Type IIIa calibration plot. Clearly, differences in the measured fluorescence intensity of cells in the FL2 channel result from the much higher emission level of the Type IIIa standards in the FL2 channel relative to the cell sample (Fig. 1).

As previously stated, calibration plots for both sets of standards were also generated in a second instrument, also a BDIS FACScan. When CD4-FITC labeled lympho- cyte samples were analyzed in the FLl channel, using two different instruments, the measured fluorescence inten- sity of the cell samples was almost unchanged when cal-

24 SCHWARTZ ET AL.

106

2

3 (L m UI

I 1 0 4

103 300 400 500 6W roo 800

a FLl Channel

II WMESF REMESF

: 5,200MESF ’ Cell Mean Channcl

103 200 400 600 800 1000

b FLZ Channel

FIG. 2. a: Calibration plot from the FL1 fluorescence channel where Type Illa calibration standards (Rainbow Calibration Particles, Sphero- tech Inc.) are assigned MESF values from the the plot derived from Type l l lb calibration standards (Quantum 24 Calibration Standards, FCSC) such that both standards describe the Same line. Thus, both standards will yield the Same MESF intensity for the CD4-FITC labeled cell. b: Calibration plots derived from the Type l l la and Type l l l b calibration standards using their assigned MESF values and mean peak channels from the FL2 fluorescence channel. Note the significant difference (59,000 vs. 5,200) for the MESF values of the same labeled cells mea- sured against the Type l l la standards.

ibration standards that spectrally match the cell sample were used (instrument # 1, 67,000 MESF, vs. instrument #2, 64,400 MESF; A = -4% ). In contrast, when calibra- tion standards that do not spectrally match the cell sam- ple were used, the difference in the measured fluores- cence intensity of cell samples, analyzed in the FLl channel of two instruments, was much greater (instru- ment #1, 67,000 MESF, vs. instrument #2, 78,000 MESF; A = + 16%).

Taken together, these data clearly demonstrate the need for spectral matching between standards and sam- ples, especially when results from different instruments need to be compared.

A second requirement for a calibration standard is that its population be described by a representative peak channel. There has been discussion as to which statistical measurement to use, for example, mean, median, or mode, to define the peak channel. Such discussion as- sumes a normal distribution and, in the case of data from

log amplifiers, a log normal distribution. Most calibration standards meet this requirement to a reasonable degree. As to which statistical measurement to use, the mean is probably the best parameter, if only to reflect that its value was probably determined by some primary bulk measurement. Although low coefficients of variation (CV), for example, <2%, are desirable for alignment stan- dards, they are not necessary for fluorescent calibration standards. Calibrators with CVs as high as 15% may be used as long as the peak channels of the population(s) can be determined in a reliable and reproducible manner. It should be noted that most biological samples, with the exception of GO/Gl DNA measurements, also show high coefficients of variation (1 7).

A third requirement is that the standards be calibrated in some meaningful practical fluorescence unit. The most basic fluorescence unit would be the photon. However, in order to accurately perform fluorescence measure- ments in photons, sophisticated calibration hardware for each flow cytometer, as well as highly complex calcula- tions involving spectral and quenching corrections, are required. On the other hand, arbitrary units could be used, such as Relative Fluorescence Intensity (RFI), but such units lack any direct physical relationship to the fluorochromehntibody conjugate being measured. As re- cently discussed by Nicholson and Giorgi (lS), RFI is a relative measure, which does not provide the number of antigen molecules per cell and which requires careful standardization of the flow cytometer, saturating amounts of antibody, as well as the use of the same lot of mono- clonal antibody throughout a study. Two fluorescence units that do have physical relevance are MESF and ABC (19). MESF units represent Molecules of Equivalent Sol- uble Fluorochrome. This means that the standards are calibrated such that their fluorescence intensity is the same as the equivalent number of molecules of the same fluorochrome in solution. Note that this does not mean that the standards have that many fluorochrome mole- cules, but only that the standards have the same intensity as the solution of the particular fluorochrome at a given concentration. The fluorescence intensity units of MESF are well accepted, as evidenced by their use in Blind Panel of the 5th International Workshop on Human Leu- kocyte Differentiation antigens (20).

ABC units represent Antibody Binding Capacity. This means that the standards labeled with a specific conju- gated antibody have the same fluorescence intensity as that same number of specific conjugated antibodies bound to cells. In the case of ABC, the fluorescence per antibody, that is, the effective F/P ratio, does not need to be taken into consideration since both the cells and the standards are being labeled with the same fluorescently conjugated antibody. As noted by Vogt et al., this unit is not a direct measurement of cell receptors, but rather a measurement of the antibody binding capacity of the cell for the labeled reagent (21). The ABC may not reflect the true number of cell receptors due to for example, non- specific binding, multivalent antibody interactions, steric hindrance, and hidden or incomplete binding sites. The

STANDARDIZING FLOW CYTOMETRY 25

term binding capacity avoids these misinterpretations by reflecting the result of the actual chemical measurement, that is, antigen-antibody interaction (2 1 ). The ABC mea- surement is only valid when the standards are labeled to saturation with the same antibody. This concept has been used to quantitate antibody binding capacity in a variety of cells during different physiological and pathological conditions, including the quantitation of human anti- mouse antibody in serum (22), the amount of anti-CD3- FITC bound to peripheral blood lymphocytes (23), the number of EGF cell surface receptors in ocular cells (24), the antibody binding capacity of 12 lymphocyte surface antigens (i.e., CD3, CD4, CD20, CD57, HLA-DR) in 58 normal blood donors (25), and the identification of of B-cell growth factors in effusion fluids from patients with aggressive B-cell lymphomas (26).

REQUIREMENTS TO ESTABLISH A STANDARDIZED FLUORESCENCE CALIBRATION PLOT

It is generally accepted that the best method to estab- lish a linear calibration plot is through the least-squares regression line; thus, the relationship is expressed as

y=ax+b,

where a is the slope and b is the intercept (27). One must realize, however, that there are a large number of possi- ble combinations of scales and fluorescence units in flow cytometry that may be used to satisfy this simple rela- tionship.

To complicate the situation further, log amplifiers are used in most instruments to obtain the wide range of fluorescence intensities encountered in immunopheno- typing. However, some instruments, (e.g., EPICS XL) have linear amplifiers that can span this intensity range but still report the data on log scales. With such data, the linear regression statistics will not hold unless the fluo- rescence units of the calibrated standards are converted to their respective logarithms to the base 10. Therefore the regression equation for data from a log amplifier is

y = abg(x)+ b

An additional source of complexity arises from the nu- merous reporting scales that are used by the various pro- grams available for flow cytometry analysis (Table 1). These scale variations are extremely confusing and have been a major obstacle in obtaining standardized calibra- tion plots. To remedy this situation, it is necessary to convert the different reporting scales to a Common In- strument Performance Scale. The 256 histogram scale was chosen as the Instrument Performance Scale because all other scales can be accurately converted to this scale. List mode files are generally collected in such a histo- gram format.

After identifying the fluorescent units and normalizing the instrument reporting scales, we consider now the format of the regression calculation. To address this issue, we have applied the classic format of a dose-response plot. Thus, the calibrated values of the standards corre- spond to the independent variable (x-axis), while the

Table 1 Data Presentation and Conversion of Scales Available From

Different Now Cytometers

Conversion to instrument performance

Scale scale (0-255 HC) Histogram channels (HC)

0-225 HC 0-1023 H Cl4 0-4095 HC116

1-10,000 (4 decade) 0.1-1,000 (4 decade) 1-1000 (3 decade) 0.1-100 (3 decade)

Relative linear channels (RLC) 64.0 Log (RLC) 64.0 Log (10 x RLC) 85.33 Log (RLC) 85.33 Log (10 x RLC)

response of the instrument or the peak channel of the fluorescent standard corresponds to the dependent vari- able (y-axis). Under this classical approach, the MESF or ABC units would be represented on the x-axis and the instrument channels on the y-axis. However, we prefer to present flow cytometry calibration plots in an inverted format with the instrument scale on the x-axis and the fluorescence units on the y-axis (Fig. 3), because it re- sembles one-dimensional histograms in which instru- ment channels appear on the x-axis. Although, x (MESF/ ABC units) and y (peak channel) values are not mathematically interchangeable, the graphic representa- tion of the plot in this manner is justified as long as the regression calculation is performed correctly.

Certain parameters can be derived from the calibration plot with significant physical meaning. We refer to this group of parameters as the Primary Performance Param- eters for fluorescence intensity assessment ( 14). These parameters are instrument independent in that they can be directly compared across all instruments. The Primary Performance Parameters include: the Average Residual Percent, as an un-weighed measure of linearity; the De- tection Threshold, as a measure of detection sensitivity; the Coefficient of Response, as a measure of the dynamic range of the fluorescence channel; and the Zero Channel Value, as a measure of the position of the Window of Analysis (Fig. 3). Window of Analysis is defined as the range of fluorescence intensity measured by the flow cytometer at a particular instrument setting ( 13,15).

As shown in Table 2, the Average Residual Percent (AvRes% ) is calculated by averaging the absolute percent difference that the regression line is away from the actual data points (Fig. 3). The AvRes% is considered a good measure of linearity since it is not desensitized by the extensive range of the log amplifier, as in the case of the coefficient of determination, r2. The Detection Thresh- old (DT) is the level above which experimental fluores- cent signals can be discerned from the electronic and optical noise of the instrument. The DT is determined by finding the corresponding value of a nonfluorescent stan- dard on the calibration plot, as illustrated in Figure 4.

The Coefficient of Response (Coef. of Res.) is defined as the slope (a) of the calibration line when the linear

26 SCHWARTZ ET AL.

0 - $ 1

0 0 0 Calibration Plot

W d - z g

0 Blank QSC Microbead

I I I I I I I I I I I

4095

-0 60 120 180 240 4

Histogram Channel

Zero Chan Value Max Chan Value

43 ABC Units 362858 indow of Analvsis

FIG. 3. Calibration plot obtained from the QuickCaP v2.0 program (Flow Cytometry Standards Corp.) of data derived from Quantum Simply Celiular~ Microbead Standards labeled with CD4- PE (Sigma Bio- Sciences) and analyzed on a EPICS XL (Coulter) flow cytometer. The Primary Performance Parameters were AvRes% = 1.19, DT = 249, Coef. of Res. = 64.94, and ZCV = 43.

Table 2 Calculation of Average Residual Percent

A Absolute Assigned valuea Regressed valuea percent 4095 4063 0.78 15,493 15,469 0.15 32,636 33,419 2.40 107,795 106,252 1.43

aValues obtained from the calibration plot in Figure 3.

Average residual percent 1.19

regression is performed, as described earlier. The Coef. of Res. is also the number of 256 histogram scale channels per decade. Therefore, the Coef. of Res. of a four decade amplifier, should equal 64.0, whereas a three decade am- plifier should have a Coef. of Res. of 85.3. Conversely, the number of decades of the dynamic range for a particular fluorescence channel, for example, FL1 and FL2, can be determined by dividing the Coef. of Res. into the 256 channels of the Instrument Performance Scale. In prac- tice, the Coef. of Response differs somewhat (i.e., usually

Relationship between Zero Channel Value (ZCV) and Detection Threshold (DT)

/ Instrument Channels

FIG 4. This illustration depicts how the Detection Threshold (DT) and the Zero Channel Value (ZCV) are derived from a calibration plot. Note that when the DT is on scale, the ZCV will have a lower value than the DT, by definition, since the channel zero is set at the beginning of the scale.

Table 3 Standardized Calibration Plot Construction and Format

1.

2.

Take the logarithm of the calibrated fluorescence values of the standards. Convert the mean peak channels of each calibrated standard, from the instruments analysis scale, to the corresponding channels on a 256 histogram scale (Instrument Performance Scale). Perform the linear regression using the appropriate fluorescence units (MESF or ABC) of the calibrated standards on the x-axis, and the peak channel values of the Instrument Performance Scale on the y-axis. Calculate the Primary Performance Parameters (AvRes%, DT, Coef. of Res., and ZCV) as indicated in the text. For graphical illustration, the inverted graphical representation of the regression line with the fluorescence units on the y-axis and the Instrument Performance (256 histogram channel) Scale on the x-axis, is suggested.

3.

4.

5.

*5.0) , depending on how the log amplifier has been set up at the factory. This parameter provides the operator with an easy mechanism to independently measure the dynamic range of the instrument, that is, the range of fluorescence intensity measurable by the instrument.

The Zero Channel Value (ZCV) is the calculated MESF or ABC value for histogram channel zero. Again, note that this corresponds to the x-intercept, not the y-intercept, under this format. The ZCV is the point at which the Window of Analysis starts. If the instrument is set such that the noise level is on scale, then by definition, the ZCV will be lower than the noise level and may only be calculated rather than actually measured. To describe the full range or size of the Window of Analysis, the Max Channel Value may also be determined from the calibra- tion plot, as illustrated in Figure 3. Table 3 summarizes

STAN DAR Dl Z l NG FLOW CYTOM ETRY 27

Table 4 Acceptable Ranges for Primary Performance Parameters for Quantitative

Assessment of Fluorescence Intensity

Parameter Range Average residual percent

Detection threshold (DT) Coefficient of response

<3% for MESF

<lo00 MESF or ABC 64.0 2 5.0 for a 4 decade amp

(4.0 ? 0.3 decades) 85.3 ? 5.0 for a 3 decade amp

(3.0 2 0.3 decades) <200 MESF (ABC) for FlTC and 3rd-color dyes <lo0 MESF (ABC) for PE

MESF = Molecules of Equivalent Soluble Fluorochrome; ABS = Antibody

(AvRes%) <5% for ABC

(Coef. of res)

Zero channel value (ZCV)

Binding Capacity.

Autofluorescence ot Leukocytes Table 5 Acceptable Ranges for Primary Performance Parameters

Obtained From Interlaboratory Data (N = 35 Instruments)

Lymphocytes Monaeyles Granulocytes

C.ll Type

FIG. 5. Autofluorescence levels of monocytes, lymphocytes, and gran- ulocytes, as determined in a survey conducted in 61 laboratories across the United States.

the protocol and format to follow in constructing the Standardized Calibration Plot and calculating the four Pri- mary Performance Parameters.

ACCEPTABLE RANGES FOR PRIMARY PERFORMANCE PARAMETERS

It is important to consider the basis for the assignment of acceptable ranges for the Primary Performance Param- eters. The proposed values (Table 4 ) may serve as guide- lines and are based on preliminary data provided by var- ious clinical and research laboratories in the United States, Canada, and Europe, including those to be pre- sented in the next section.

The acceptable range for Average Residual % has been set at <3% when using MESF calibrator standards and at 1 5 % when using ABC calibrator standards. Since addi- tional sources of variation, such as the antibody-labeling procedure, may influence antibody binding, the accept- able AvRes% for ABC standards should therefore be more flexible than that of the MESF.

A study on the autofluorescence levels of human leu- kocytes conducted in over 60 laboratories across the United States (19,28) revealed that lymphocytes have a fluorescent intensity of approximately 657 * 270 MESF of fluorescein, with higher values reported for monocytes (928 k 373) and granulocytes (1200 * 697). These re- sults are illustrated in Figure 5. Based on these data, the

Parameter Mean SD %CV Average residual percent

(AvRes%) 2.24 0.73 32.6 Detection threshold (DT) 666 398 59.8 Coefficient of response

(Coef. of res) 61.5 2.4 3.9 Log decades 4.2 0.16 3.9 Zero channel value (ZCV) 33 1 244 73.7" Max. channel value 4,235,213 2,726,019 64.3a

"Large differences due to instruments not set up on a refer- ence standard with matching spectra.

quantitative acceptance level for the Detection Thresh- old was set at 1000 MESF/ABC, since the autofluores- cence of the most commonly used cells (i.e., lympho- cytes) is relatively close to this value.

The acceptable variation of 64 k 5 for the Coefficient of Response for a four-decade log amplifier, or its equiv- alent, is again based on observations that suggest this variation in range has minimal effect on the position of a variety of cell samples within the Window of Analysis. In addition, data currently available suggest that the large majority of flow cytometers do fall within this range, as discussed in the next section.

The Zero Channel Value has been set at 200 MESF (ABC) for fluorescein and third-color fluorochromes, and at 100 MESF (ABC) for phycoerythrin. These values appear to hold for the majority of instruments whose PMTs are set such that nonfluorescent particles, indicat- ing the Detection Threshold, appear on scale. Of note, the position of the Zero Channel Value is controlled by the PMT setting, and the Coefficient of Response of the in- strument is determined by the amplifier.

PRACTICAL APPLICATIONS OF THE STANDARDIZED FLUORESCENCE CALIBRATION PLOT TO COMPARE

INSTRUMENT PERFORMANCE Interlaboratory Application for Multi-Site Surveys

The value of a standardized fluorescence calibration plot that can provide standard performance parameters

28 SCHWARTZ ET AL.

Microbead Channel Numbers

0 8 16 24 32

a Sequence Number

Average Residual Percent 1 d

0 1 1 1 I l l I l l I l l J 8 1 6 24 32

b o Sequence Number

0 8 16 24 32

Sequence Number Sequence Number C

FIG. 6. Plots of the (a) microbead peak channel numbers, (b) Average Residual Percent, (c) Coefficient of Response and (d) Sensitivity Indicators (ZCV = a, DT = A) from 35 different instruments using Type l l lb calibration standards (Quantum 26, Flow Cytornetry Standards Corp.). The horizontal lines indicate ?1 SD from the mean.

can best be demonstrated by presenting data from profi- ciency testing of flow cytometers across multiple labora- tories. The following set of data were acquired to admit laboratories to a Quantitative Survey conducted by Fran- cis Mandy (Health and Welfare Canada) under the aus- pices of ISAC. The goal of this survey was to determine antibody binding capacity ranges for CD4, CD8, and CD3 antigens from sero-negative and HIV-positive subjects. To qualify, the participating laboratories had to show that their instruments met minimal performance criteria. Each laboratory was required to report data from a calibration standard (Quantum 26 FITC, Flow Cytometry Standards Corp.), which was analyzed with QuickCal@ software (Flow Cytometry Standards Corp.). The QuickCal@ pro-

gram was used to calculate and report the performance parameters previously described in this manuscript.

The results presented in Figure 6 were obtained from 35 different instruments using four different reporting scales. The instruments included FACScans and FACStars (BDIS), Profile 11s and Epics XIS (Coulter), as well as CytoronAbsolutes (Ortho Diagnostics System, Inc), which were located in nine different countries (Canada, France, Germany, Hungary, Italy, Spain, Sweden, Nether- lands, and the United States). Each data point (sequence number) in plots (Fig. 6) represents a different instru- ment. Values that were determined to be outside of the fixed acceptable ranges are "flagged" using asterisks. These plots provide a convenient and comprehensive

STAN DAR DlZl NG FLOW CYTOM ETRY 29

Table 6 Acceptable Ranges for Primary Performance Parameters

Obtained From lntralaboratory Data fN = 39 Calibrations)

Parameter Mean SD %CV Average residual percent

(AvRes%) 1.68 0.73 43.4 Detection threshold (DT) 873 128 14.7 Coefficient of response

(Coef. of res) 59.64 0.84 1.4 Log decades 4.29 0.06 1.4 Zero channel value (ZCV) 135 14 10.3 Max. channel value 2,550,285 306,318 12.0

way to directly compare the performance parameters of all instruments against each other, as well as to deter- mine acceptability against the proposed performance ranges. Table 5 summarizes the statistical results for the performance parameters presented in Figure 6.

From these data it is evident that all 35 instruments were reasonably comparable. Mean peak channels for all microbead populations were similar after normalization to the Instrument Performance Scale ( 2 5 6 histogram channels), indicating a relatively similar Window of Anal- ysis, except for instruments 4 , 5 , 9, 17, 18, and 24 (Fig. 6a). This is understandable since in this part of the study, the instruments were not 5et on a reference standard with matching spectra. In addition, with the exception of instruments 14 and 25, all of the instruments were found to have an Average Residual below 3% (Fig. 6b) meeting the proposed acceptable range for this parameter (Table 4). Only instrument 20 had a Coefficient of Response significantly out of the acceptable range for a four-decade amplifier (Fig. 6c). Even though instruments 17, 18, and 24 fell within the acceptable range for a four-decade log amplifier, they were flagged because the manufacturer rated them as having a 3.5 decade amplifier.

The Zero Channel Values in the Sensitivity Indicators plot (Fig. 6d) confirm that the Window of Analysis for instruments 4, 5, 9, 17, 18, and 24 began at different fluorescence intensities than that of the other instru- ments. The variation of these instruments appears to cor- relate to high Detection Thresholds, suggesting that if the blank standard is placed close to the same normalized channel for all instruments, as indicated in Figure 6a, then instruments with high Detection Thresholds will have lower peak channel positions for the labeled stan- dards than in the other instruments, and thus the calibra- tion plot will describe different Windows of Analysis. This observation is supported by the finding that the De- tection Threshold of all the instruments except 4 , 5, 9, 17, 18, and 24 fell within the proposed acceptable range (below 1,000 MESF).

lntralaboratory Application for Quality Control Although many laboratories do not participate in multi-

site surveys, use of a standardized calibration plot can provide comprehensive quantitative quality control against acceptance ranges, which can be used to monitor

daily instrument performance. Such quality control can provide instrument operators a much higher degree of confidence in the results of their assays because it indi- cates not only the linearity and sensitivity of their instru- mentation, but also determines the fluorescence inten- sity that the instrument covers at particular settings. The following is an example of this quality control applica- tion (Table 6, Fig. 7).

On a daily basis, an Epics Elite (Coulter) flow cytom- eter was set up by adjusting the PMT setting such that a reference standard (QC3, Flow Cytometry Standard Cor- poration, San Juan) fell in a specific Target Channel. Sub- sequently, the instrument was calibrated with calibration standards (Quantum Simply Cellular, Flow Cytometry Standards Corp.), which were labeled with the same lot of CD4-FITC monoclonal antibodies (Sigma Biosciences) over a 2 month period. The Levey-Jennings plots ob- tained for the daily measurements are presented in Figure 7 and the corresponding statistics in Table 6.

Mean peak channels for all microbead populations were very constant after normalization to the instrument performance scale ( 2 5 6 histogram channels), while the variation in PMT setting was less than 9 volts, confirming the stability of the instrument. The Average Residual re- mained below 4% throughout the study and fell within the proposed acceptable range for this parameter (Table 4 ) . For this particular instrument, the Coefficient of Re- sponse was consistently on the lower limit of the accept- able range, which may indicate the need for adjustment of the log amplifier. This is further indicated by the higher number of calculated log decades. The Zero Channel Value and the ABC Threshold (Sensitivity Indicators) are plotted on the lower panel of Figure 6. The consistency in the Zero Channel Value confirms that the reference stan- dard established a common Window of Analysis during this evaluation. The majority of the time the Detection Threshold of the instrument was within the acceptable range (below 1,000 ABC).

The uniformity of all four Primary Performance Param- eters indicates acceptable and stable performance of not only the instrument, but also the reagents and preparative procedures, since the Type IIIc standards require labeling with the antibody reagent.

SUMMARY AND CONCLUSIONS This report provides an overview of a novel concept in

flow cytometry standardization, namely, the construc- tion of a standardized fluorescence calibration plot using calibrators whose spectra match those of the sample. The four Primary Performance Parameters (Average Residual Percent, Detection Threshold, Coefficient of Response, and Zero Channel Values) that are derived from this cal- ibration plot may be used to evaluate instrument perfor- mance and setup. The protocol described in Table 3 pro- vides the step-by-step method of constructing this standardized calibration plot. This approach has the po- tential to identify and diminish instrument-related vari- ability in any flow cytometric determination, as evidence

30 SCHWARTZ ET AL

Microbead Channel Num

m 0 I

C 5 5 1

0 9 18 27 36 Sequence Number a

9 18 27 36

Sequence Number C

FIG 7. Levey Jennings plots of the (a) microbead peak channel num- bers, (b) Average Residual Percent, (c) Coefficient of Response and (d) Sensitivity Indicators (ZCV = 0, DT = A) generated from 39 consec- utive calibration runs from the same instrument (EPICS Elite) using Type

by the data provided on surveys conducted to evaluate interlaboratory and intralaboratory performance.

The proposed acceptable ranges for the Primary Per- formance Parameters emerge as an important criteria when assessing instrument setup and performance in quality control programs. As additional data become available and formal analysis of calibration plots derived from a much larger number of instruments is performed, these suggested guidelines may be modified until final guidelines can be fully documented and established. We are confident, however, that many of the suggested ac- ceptable values will be confirmed and validated.

Acceptance of this standardized calibration methodol- ogy will facilitate direct Comparison of flow cytometric

Average Residual Percent m

Sequence Number

f 3 0 9 18 27 36

d Sequence Number

Ilk calibration standards (Quantum Simply Cellular, Flow Cytometry Standards Corp.) labeled with the same antibody CD4-FITC (Sigma Bio- sciences) over a 2-month period. The horizontal lines indicate 21 SD from the mean.

data obtained from different flow cytometers and/or lab- oratories. The need for calibration and standardization methodology becomes extremely important when one takes into consideration the new regulatory requirements and guidelines established by the Centers for Disease Control (29), Clinical Laboratory Improvement Act 1988 (lo), the National Committee for Clinical Laboratory Standards (30), and the Dutch Foundation of Immu- nophenotyping in Hemato-Oncology (31), as well as the growing use of flow cytometric analysis in the clinical setting.

Recognizing the challenges and the potential clinical impact of flow cytometry in future years, we expect that the implementation of common standardization criteria

STAN DARDl ZI NG FLOW CYTOM ETRY 31

will maximize resources, minimize interlaboratory vari- ability, and assure accurate comparison of results. Flow cytometry standardization will certainly promote new health care-related applications by virtue of the availabil- ity of practical quality assurance programs.

ACKNOWLEDGMENTS This work was performed under a Cooperative Re-

search and Development Agreement (CRADA) between Flow Cytometry Standards Corporation and the Centers for Disease Control. Dr. Fernandez-Repollet's research is partially supported by National Institutes of Health grants RR-03051 and RR-08224.

LITERATURE CITED 1. Muirhead KA: Establishment of quality control procedures in clini-

cal flow cytometry. In: Annals of the NewYork Academy of Sciences, Vol 677, Landay AL, Ault KA, Bauer KD, Rabinovitch PS (eds). The New York Academy of Sciences, N e w York, 1993, pp 1-20.

2. Mayall BH: Quality control and quality assurance in flow cytometry. In: Annals of the New York Academy of Sciences, Vol677, Landay AL, Ault KA, Bauer KD, Rabinovitch PS (eds). The New York Academy of Sciences, New York, 1993, pp 78-8 1.

3. Muirhead KA. Quality control for clinical flow cytometry. In: Clin- ical Flow Cytometry: Principles and Applications, Bauer K, Duque R, ShankeyV (eds). Williams & Wilkins, Baltimore, 1993, pp 177-199.

4. Van? Veer MB, Kluin-Nelemans JC, Van der Schoot CE, Van Putten WLJ, Adriaansen HJ, Van Wering ER: Quality assessment of immu- nological marker analysis and the immunological diagnosis in leu- kemia and lymphoma: A multi-centre study. Br J Hematol 80:458- 465, 1992.

5. Cardwell CW, Maggi J, Henry I., Taylor H: Fluorescence intensity as a quality control parameter in clinical flow cytometry. Am J Clin Path01 188:447-456, 1987.

6. Waxdal MJ, Monical MC. Fleisher TA, Marti GE: Inter-laboratory sur- vey of lymphocyte immunophenotyping. Pathol lmmunopath Res 7545-356, 1988.

7. Kapn J, Gelman R. Waxdal M. Kidd P: NIAID division of AIDS flow cytometry quality assessment program. In: Annals of the New York Academy of Sciences, Vol677, Landay AL, Auk KA, Bauer KD, Rab- inovitch PS (eds). The New York Academy of Scicnces, New York,

8. Vogt RF, Cross GD, Phillips DL, Henderson 1.0, Hannon WH: Inter- laboratory study of cellular fluorescence intensity measurements with fluorescein-labeled microbead standards. Cytometry 12:525- 536, 1991.

9. Brando B, Sommaruga E: Nationwide quality Control trial on lym- phocyte phenotyping and flow cytometry performance in Italy. Cy- tometry 14:294-306, 1993.

10 CLlA'88. Federal Register, Feb. 28, 1992; Subpart K 493.1217. 1 1 . Schwartz A: Calibration method for flow cytometry using fluores-

cent microbeads and synthesis thereof. US patent 4,767,206, August 30, 1988.

12. Schwartz A: Method of calibrating a flow cytometer or fluorescence microscope for quantitating binding antibodies on a selected sample, and microbead calibration kit thcrefor. IJS patent 4,918,004, April 17, 1990.

13. Schwartz A, Fernandez-Repollet E: Development of clinical stan- dards for flow q'tometry. In: Annals of the New York Academy of

1993, pp 50-52.

Sciences, Vol 677, Landay AL, Ault KA, Bauer KD, Rabinovitch PS (eds). The New York Academy of Scicnccs, New York, 1993, pp

14. Schwartz A, Fernbndez-Repollet E: Technical aspects of fluorescence quantitative measurements by flow cytometry. Clin lmmunol News- lett 15:73-77, 1995.

15. Schwartz A, Fernindez-Repollct E: Standardization for flow cytom- etry. In: Methods in Cell Biology: Flow Cytometry, Vol 42, Danynk- icwics 2, Crissman HA, Robinson P (eds). Academic Press, New York, 1994, pp 605-626.

16. Robinson P Handbook of Flow Cytometry Methods. Wiley-Liss, New York, 1993, pp 64-67.

17. Karen DF. Hanson CA, Hurtubix PE: Flow cytometry and Clinical Diagnosis. American Society Clinical Pathology, Chicago. 1993, pp 174- 175.

18. Nicholson JKA, Giorgi JV: Quantitating cell surface antigens in hu- man immunodeficiency virus disease. Clin lmmunol Newslett 15:

19. Shapiro HM: Practical Flow Cytometry, 3rd ed. Alan R. Liss, New York, 1994, pp 173-174.

20. Poon R, Johnson R, Stanton T, Fernandez.Repollet E, Schwartz A: Calibration of fluorescence intensity microbead standards and as- signment of effective F/P ratios to the FITC-conjugated secondary antibody used in the Vth International Leukocyte Typing Workshop Leukocyte Typing V, Oxford University Press, Oxford, 1995, pp 13- 16.

21. Vogt RF, Marti GE, Schwartz A: Quantitative calibration of fluores- cence intensity for clinical and research applications of immunophe- notyping by flow cytometry. In: Reviews of Biotechnology and Bioengineering, Vol 1: Cytometry, HW Tyrer (ed) Ablex Publishing Corporation, Norwood, New Jersey, 1994.

22. Labus JM, Petersen BH: Quantitation of human antimouse antibody in serum by flow cytometry. Cytometry 15275-281, 1992.

23. Islam D, Lindberg AA, Christensson B: Peripheral blood cell prepa- ration influences the level of expression of leukocyte cell surface markers as assessed with quantitative multicolor fluorescence. Cy- tometry 22:128-134, 1995.

24. Lopez JG, Chew SJ, Thompson HW, Malter JS, lnsler MS, Beuerman RW: EGF cell surface receptor quantitation on ocular cells by an immunocytochemical flow cytometry technique. Invest Ophthalmol Vis Sci 352053-2062, 1992.

25. Lenkei R, Andersson B Determination of the antibody binding Cd- pacity of lymphocyte membrane antigens hy flow cytometry in 58 blood donors. J lmmunol Methods 183267-277, 1995.

26. Ford R, Tamayo A, Martin B, Niu K, Claypool K, Cabanillas F, Ambrus J. Identification of B-cell growth factors (interleukin-14; high-mo- lecular weight-B-cell growth factors) in effusion fluids from patients with aggressive B-cell lymphomas. Blood 86:283-293, 1995.

27. Bailar JC, Mosteller F: Medical Uses of Statistics. Murray Printing C h , Westford, Massachusetts, 1986.

28. Schwartz A, Fernandez-Repollet E: Autofluorescence of leukocytes: A survey. Flow Cytometry Standardization Forum 3:7-8, 1991

29. Centers for Disease Control and Prevention: Revised guidelines for the performance of CD4+ T-cell determinations in persons with human immunodeficiency virus (HIV) infections. MMWR 4 1 (No.

30. National Committee for Clinical Laboratory Standards: Clinical Ap- plications of Flow Cytometry: Quality Assurance and Immunophe- notyping of Peripheral Blood Lymphocytes, NCCLS, Villanova, PA, Publication H24-T, 1992.

3 1. Gratama JW: Quality control of flow cytometric irnmunophenotyp- ing of malignant blood diseases in the Netherlands. Flow Cytomctry Standardization Forum 4:l-5, 1992.

28-39.

77-82, 1995.

RR-3), 1994.