This report covers June 1987-August 1988, a period when new specific and nonspecific ... 3Poisons Unit, Guy's Hospital, London, U.K.. Received January 19 ...
CLIN. CHEM. 35/7, 1309-1312 (1989)
The Continuing Need for Quality Assessment of Cyclosporine Measurement Athoil Johnston,1
Joanne T. Marsden,2 and David W. Holt3
The returns from the United Kingdom Cyclosporin Quality Assessment Scheme were analyzed for the period June 1987 to August 1988. During this time the number of laboratones in the Scheme increased from 102 to 124 and the proportion of laboratories using nonspecific assay methods declined, as did the proportion of them measuring cyclosporine in plasma. Seven different methods were used to measure the drug in blood, and the seven methods gave seven different results when used to measure patients’ samples. The results, from lowest to highest, differed by a factor of approximately 3.4. The within-assay coefficient of variation (CV) was acceptable for all methods, but the between-assay and between-center CVs were poor. HPLC gave higher CVs than did the immunoassays. The United Kingdom Cyclosporm Quality Assessment Scheme was established in 1984. Our first report involved only 25 centers, all of which were using the original, polyclonal, nonspecific radioimmunoassay (RIA) kit manufactured by Sandoz (1). Since then the Scheme has expanded considerably, and there have been substantial developments in the methods available for the measurement of the drug. This report covers June 1987-August 1988, a period when new specific and nonspecific RIAs based on monoclonal antibodies became available (2), as did two nonspecific assays based on other polyclonal antisera, which were configured as an RIA and a fluorescence polarization immunoassay (3). Using the data from the Scheme, we have attempted to describe the comparative performance of techniques currently available for measurement of cyclosporine, in terms of accuracy, reproducibility, and selectivity. Materials and Methods The participating laboratories received three samples each month and were expected to return the results of their measurements to the Scheme coordinators within three weeks. A summary of the results, in which the data were presented grouped by sample matrix and analytical technique, was circulated four weeks after the samples were dispatched. Centers were identified by a randomly allocated code number known only to the coordinators. The laboratories were sent liquid samples of either whole blood or plasma, depending on the matrix used by the participating center. The source of blood and plasma was either the National Blood Transfusion Service or aliquota of pooled blood or plasma from patients receiving cyclosporine (Sandimmun#{174};Sandoz AG, Basel, Switzerland) after renal transplantation. The former had been screened to exclude hepatitis and HIV antibodies; technical difficulties excluded the testing of pooled blood or plasma from
1 Clinical Pharmacology, St Bartholomew’s Hospital, London, EC1A 7BE, U.K. 2Re Unit, Dulwich Hospital, London, U.K. 3Poisons Unit, Guy’s Hospital, London, U.K. Received January 19, 1989; accepted March 27, 1989.
patients, but the patients were tested regularly for hepatitis infection. All samples were treated with sodium azide, 10 mg/L, to minimize microbial growth and beta-propiolactone, 2.5 mLIL, to inactivate HIV, as advised by the Advisory Committee on Dangerous Pathogens (4). In addition, whole-blood samples were treated with EDTA, 1 g/L, to inhibit clotting. Unlike most drug quality-assessment schemes, most of the samples circulated in this Scheme are actual samples from patients. They therefore contain both parent drug and metabolites. There is no “gold standard” for cyclosporine measurements, so these patients’ samples were not assayed before circulation and assigned a value. The range of sample types prepared was designed to testassay selectivity, withinand between-assay reproducibility, accuracy and betweenmethod differences. To do this we used cyclosporine-free samples, repeat samples, samples containing weighed-in concentrations of cyclosponne, and pooled patients’ samples. The Scheme is open to any laboratory measuring cyclosporine, and there is no charge for membership. Results Centers In June 1987 there were 102 laboratories participating in the Cyclosporin Quality Assessment Scheme. By September 1988 this number had increased to 124, drawn from 26 countries. Of the centers, 109 were in Europe, but there were participants in North and South America, the Middle East, and Australia. Choice of Matrix During the study period there was a continuing trend towards the use of blood, instead of plasma, as the matrix of choice for the measurement of cyclosponne. In July 1987, 23% of reported results were for plasma, but by August 1988 this percentage had fallen to 16%. The number of results reported for plasma were too few to make meaningful comparisons between all of the analytical methods. Consequently, the results reported below refer only to the measurement of cyclosporine in blood, and the limited plasma data are dealt with separately. Analytical Methods Seven major analytical methods were used by the centers to analyze cyclosporina: #{149} Polyclonal RIA, Ciclosporin RIA-Kit, Sandoz AG (Original). #{149} Polyclonal RIA, CYCLO-Trac#{174},INCStar Corp., Stillwater, MN 55082 (CycloTrac). #{149} Polyclonal FPIA, Abbott Laboratories, Diagnostic Division, North Chicago, IL 60064 (TD). #{149} Monoclonal RIA, Sandimmun-Kit, Sandoz AG (Nonspecific). #{149} Monoclonal RIA, Sandimmun-Kit, Sandoz AG (Specific). #{149} Monoclonal RIA, CYCLO-Trac SP#{174}, INCStar Corp. (CycloTrac-SP). CLINICAL CHEMISTRY, Vol. 35, No. 7, 1989 1309
liquid chromatography (HPLC). During the 14 months covered by this report, four of these methods-Original, CycloTrac, Specific, and HPLCwere used to report results for all 42 samples. The Nonspecific assay was used for only 39, TD for 36, and CycloTracSP for 18. During the study period there was a steady increase in the number of reported results obtained with specific assays (from 16 to 50) and a decrease in those obtained with nonspecific methods (from 49 to 40). In August 1988,57% of results reported for measurements in blood were obtained with specific techniques. #{149} “High-performance”
Cyclosporine-Free
Samples, False Positives
The two cyclosporine-free were reported as positive for of 151 reports. The highest ples was 84 tgfL. Table 1
samples that were circulated the drug 28 times out of a total reported result for these samsummarizes the false-positive
reports.
AnalyticalVariables
Within-assay variation. This was assessed by circulating identical samples, drawn from pooled patients’ samples, in the same month. Most of the centers, 56/89, obtained a coefficient of variation (CV) 5%. However, 14 of the 89 produced CVs >11%; of these, six had CVs >15%. Table 2 shows the distribution of precision by method. Between-assay variation. This was measured at three concentrations in three pooled patients’ samples in three consecutive months. As measured by the Original method, these pooled patients’ samples had cyclosporine concentrations of -200, 400, and 800 g/L for the low, medium, and high pools, respectively. The median CV for CYCLO-Trac assay was consistently better and that for HPLC was consistently worse than was the case for the other three methods (Table 3). Accuracy. Measurement accuracy was assessed by analyzing cyclosporine-free samples to which the drug had been added to give eight known concentrations in the range 25 to 1000 gJL. The results are shown graphically in Figure 1. Between-method agreement. Figure 2 shows the relative response of the seven assay methods for the analysis of 34 samples of pooled patients’ blood. The figure uses both the original and HPLC assays as the 100% references, and the results are presented as the mean and 95% confidence intervals relative to these methods. The relative response was calculated for each method and sample as: Relative response = (100 x method mean)/reference assay mean.
Table 2. WithIn-Assay Variation of Each Method for Analysis of a Single Sample, from a Patient
Threefold
Poola
Original CYCLO-Trac TDx Nonspecific
Specific CYCLO-Trac SP HPLC
by Each L
Original CYCLO-Trac TDx Nonspecific Specific CYCLO-Trac SP HPLC
Minimum
Median
32 2
62 6
84 18
-
-
-
-
-
-
13/54
5
Not available 6/15 25 ‘The median, minimum, and maximum are given
1310
-
Maximum
70 -
-
36
78
for the positive results.
CLINICAL CHEMISTRY, Vol. 35, No. 7, 1989
21
14
484
18
12
582 694 224
4 21
8
Table 3. Between-Assay
>15%
2
4 13
3 3 0 3
2
4
1
3
5
0
0
10
5
0
2
2
1
0
0
CV for Four Methods for the
Analysis of Three Cyclosporine Concentrationsa Concn. Method CSP,pgiL Lower qr Median Upper qr Low Original 177 12.0 18.5 26.0 CYCLO-Trac 211 3.5 7.5 11.5 Specific
80
10.5
17.0
23.5
HPLC
70 363
23.5 8.0 7.5
17.5 16.0 10.5
415
12.5 6.0 6.0
Specific
165
10.0
12.0
19.0
HPLC
143 815
10.5
20.5 10.5
39.0 14.0
Medium Original CYCLO-Trac
High
Original 5.0 CYCLO-Trac 861 5.0 7.0 431 8.0 12.0 Specific HPLC 396 10.0 18.5 numberofcenters was 18, 12, seven, and eight for the
10.0
27.0 26.0 Original,
CYCLO-Trac, Specific, and HPLC, respectively. The number of results was insufficient to analyze the data from the remaining methods, Nonspecific, CYCLO-Trac SP, and TD. (Lowerqr, lower quartile; Upper qr, upperquartile).
Between-center, within-method agreement. For each sample circulated, the agreement between centers using a particular method can be assessed by the method CV. Table 4 lists, by method, the median between center CV, and the interquartile range, for the 34 pooled patients’ samples circulated during the 14 months. Therefore, this table sununai-izes how well centers using the same assay agreed: the smaller the CV, the better the between-center agreement. It might be thought that precision varies with concentration, but plots of CV vs concentration did not support this assumption over the concentration range of the circulated samples. Plasma Measurements The small proportion measurement
Ofl,
Number
positive 6/24 3/40 0/10 0/8
415
Not available 193 17
11-15%
Number in each group
Total
‘The number of centers within each range of CV are given. CSP, cyclospoflne.
three
Table 1. Number of False PosItives Recorded Method of Analysisa
6-10%
0-5%
CSP, g/L
Method
of centers using plasma for the
of cyclosporine, in the context of the number
of analytical methods available, allows only the comparison of the specific Sandoz RIA and HPLC for the ninemonth period November 1987 to July 1988. During this time, five samples containing known amounts of cyclesporine, between 50 and 1000 g/L, were analyzed. For these samples the agreement between the specific RIA and HPLC (100 x RIA/HPLC) was 100%; the 95% confidence interval, 75%to125%. For the 20 pooled specimens circulated, the betweenmethod agreement was 108%; the 95% confidence interval,
50 iig/L
25 ijg/L Method
Method
014044.1
0,Igl,41
C70147,4.
r’
“
TO.
Cy.IOTI20
00
#{149}42
24
l
32
‘-4
C,IOT,00-IP
$0 0l
24
tO
.53
20
‘It
000.402
30
II.
“PLC
40
20
40
Measured
50
40
SO
45 4’
10910
:-‘
SO
SO
10
20
30
40
00
#{149}‘
00
72
20
00
1
Measured valu.
value
100 ijg/L
150 g/L Method
M.thod 0,151,41
50
r’
01101112
7 lao
CYGOT14O 01 ‘-.----‘
TOO
4 113
0714T,14IP
TOO
5
II?
T’
40
00
00
180
104
Meaaur.d
4017
1414
32
00.040.
100 1
HPLC
100 ‘T’
CTIOTIOO-UP
9’ 101
Cl
173
131 ‘r’ 120
N0.SpflIlI
05
29.040.
103 ‘-v-’
CYO14T,00
St
75 ‘-l---_
5.41100010.
09LC
120 value
140
22
T’
134 ‘T’ 140
so
ISO
110
ISO
130
170
ISO
010
Measured value
250 iig/L
500 iig/L
Method
Method
0li
000
‘-I-’
0,1010.1
055
800 ‘r’
CY4OTI0.
TO.
SIT
:100
I
I
CYIOT00-OP 252
HPLC 774 0
200
270
200
3 0
SOS
0
400
‘-r’
524 000
500
Measured value
750
I
447 ‘i.’ 550
0910 330
r’
570
.----.‘
r-’520
404 ‘
‘9004115
.083 810
III 550
534
COIOT,0..lP
$72 -‘
‘-T’
TOO 500*000100
10.0014
lOSS
080
C,00T,04
544.00.1110
570
000
10
Measured value
1000 pg/i-
pg/i-
Method 0,10411.1
00
#{149}44
-OP
COlor,..
20
,
IS4
10050012110
11 -r’
00.010.
‘‘t’ 204
TOO
Method 713‘T 720
$40
0,101041
C,12OT,..
3
‘r’
:
08.147,4.
I
TO.
TOO
04.101.040.
554
551
‘-r-4
020
1030
---,--.
1075
50400404l10
CyI,T14o-eP
10004 I
‘90040. 0910
7004
700
750
,
500
.008
ISO.
ISO
0910
coo
050
I
1031
2904100 NI
500
Measured value
000
‘-r-’
1020
1100
1000
Meuwed
41101
‘r’1000
100
value
Fig. 1. Accuracy of the assay methods
y.axlS: assay method; x-axls: measured value. Results are presented as the mean and 95% confIdence Interval. No results were avsllable for CYCLO-Trac SP, and complete results were not availablefor the Original,TD0,and Nonspecificassays
92% to 124%. The two cyclosporine-free samples that were analyzed resulted in no falsely positive results (0/8) being reported by the centers using the RIA and two falsely positive results (2/6, 40 and 92 tzg/L) reported by the centers using HPLC. The median (inter-quartile range) within-method CV for the measurement of the patients’ samples was 20% (13% to 28%) and 24% (12% to 37%) for RIA and HPLC, respectively.
Discussion During the study period the number of centers in the Cyclosporin Quality Assessment Scheme increased steadily. During this time the trend was, as suggested in the AACC guidelines (5), towards the use of blood instead of plasma as the matrix of choice, and towards a specific rather than a nonspecific assay for the therapeutic momtoring of cyclosporine. CLINICAL CHEMISTRY, Vol. 35, No. 7, 1989 1311
Original
of false-positive results continues to be unacceptable, the frequency and magnitude of the results were an improvement over previous findings (1,6). Poor accuracy in the range 0 to 100 tg/L may cause problems for some centers monitoring the drug in samples from patients number
100%
Assay 200
01010.1
100
cToloTr.o Too
“+“0
L
Noo.Sp
So.cj
receiving low-dose immunosuppressive
130±137
142
---
150
C211101,00.SP
NPLC
40S1
I 20
gauged. Although
transplant
43+40 40
00
20
100
9.12.0150.
HPLC
120
140
160
HPLC
-
100%
1001100 104115
±132
II?
CyoIoT,oo-0o
0,lln.1
-
211±235
Cy.IoT,00 .. -To.
Non-SO
2O3’--2T1 370
1
‘--f-.
io
3t7’-f---362 50
100
100
200 Prcen,ag.
200
300
300
400
r.,gon,.
Fig. 2. Relative response of the assay techniques The upper portion shows percentage response and 95% confidence intervals relative to the Original method. The lower portion shows those relative to HPLC
Table 4. MedIan Between-Center Upper Interquartlles
CV and the Lower and
for Each Method
Lowerqr
Median
Upper qr
Original
13.4
23.6
CYCLO-Trac TD0 Nonspecific Specific CYCLO-Trac SP HPLC
12.4
16.4 15.8
10.0 9.8
13.0
15.4
13.0
12.0
16.0 16.0 21.0
15.0 24.8 20.0 25.0
Method
11.5 15.3
19.0
5The data summarizethe within-method, between-center agreement for the 34 pooled patients samples measured during the study period.
Despite the number of new, professionally prepared say methods available for cyclosporine measurement, between-center variation remains high. The median tween-center CVs ranged from 13% to 21%, depending
as-
the beon
the assay method used. Although the within-assay CV was acceptable for all the assay methods, between-assay precision was poor. Of the four methods tested-Original, CYCLO-Trac, Specific, and HPLC-only for CYCLO-Trac did the majority of centers report CVs of TD> CYCLO-Trac > Original > CYCLO-Trac SP> Specific> HPLC. Although differences in results by the nonspecific methods could be expected, given that the spectrum of crossreactivity of the antibodies differs in each case, the lack of
ISO
tl220fllO
Assay
02.01110 -
other
The pooled patients’ samples distributed to the laboratories allowed the relative responses of the assays to be
110
I
cyclosporine, in combination with therapy, after transplantation.
agreement between the specific methods was unexpected, a finding that is at variance with published data from single centers (7-9). Since these data were collected, the INCStar Corp. have made known that they have recalibrated the standards in their CYCLOTrac SP kit against a USP standard. As a result, the assigned values for their standards will be decreased by approximately 15%, thus minimizing the differences between HPLC and this immunoassay noted in our quality-assessment data. Further study, involving
a wider
selection of patients’ material, will be whether the differences noted here RIAs and HPLC are a sustained
required, to establish between the specific finding and whether significant. However, need for continuing
they are statistically or clinically our observations do reinforce the quality-assessment
as methods
for
measuring cyclosporine become more diverse and its routine measurement more widespread. The Scheme
receives
the
financial support of Sandoz AG, Basel,
Switzerland. References 1. Johnston Cyclosporine
A, Marsden JT, Holt DW. The United Kingdom
Quality Assessment Scheme. Ther Drug Monit 1986;8:200-4. 2. HoltDW, JohnstonA, Marsden JT,etal.Monoclonalantibodies for radioimmunoassay of cyclosporine: multicenter comparison of their performance with Sandoz polyclonal radioimmunoassay kit. Clin Chem 1988;34:1091-6. 3. Plebarn M, Sciacovelli L, Paleari CD, Burlina A. Evaluation of fluorescence polarization immunoassay for the measurement of cyclosporine and its metabolites in whole blood [Abstract]. Clin Chem 1988;34:1183. 4. Advisory Committee on Dangerous Pathogens. LAV/HTLV Ill-the causative agent of AIDS and related conditions. Available from the Departments of Health, and Health and Safety Executive, London, U.K. 5. Shaw LM, Bowers L, Demers L, et al. Critical issues in cyclosporine monitoring: report of the Task Force on Cyclosporine Monitoring. Clin Chem 1988;33: 1269-88. 6. Holt DW, Marsden JT, Johnston A. Measurement of cyclosporine: methodological problems. Transplant Proc 1986;6(Suppl 5):101-10. 7. Ball PE, Munzer H, Keller HP, Abisch E, Rosenthaler J. Specific 3H radioimmunoassay with a monoclonal antibody for monitoring cyclosporine in blood. Clin Chem 1988;34:257-60. 8. Sewing K-Fr, Christians U. Monitoring of cyclosporin by monoclonal radioimmunoassay [Letter]. Lancet 1988;ii:687. 9. Plebani M, Burlina A. Monitoring of cyclosporin by monoclonal
radioiminunoassay [Letter]. Lancet 1988;ii:687-8.
