CA-125 automated assays

Am J Clin Pathol 2006;125:921-927 921921 DOI: 10.1309/NBA312W0LANRXYH9 921

American Society for Clinical Pathology

Clinical Chemistry / SEVEN CA 125 IMMUNOASSAYS

Performance Characteristics of Seven Automated CA 125 Assays

Shella K. Mongia, MD,1 Mindy L. Rawlins,2 William E. Owen, MT(ASCP),2

and William L. Roberts, MD, PhD1

Key Words: CA 125; Immunoassay; Imprecision; Linearity; Method comparison; Ovarian carcinoma; Tumor marker

DOI: 10.1309/NBA312W0LANRXYH9

A b s t r a c t

Cancer antigen 125 (CA 125) is a high-molecular-mass glycoprotein that is used as a tumor marker tomonitor disease progression and response to therapyand in early detection of recurrence after treatment forovarian cancer. The Access 2 (Beckman Coulter, Brea,CA), ADVIA Centaur (Bayer Diagnostics, Tarrytown,NY), ARCHITECT i2000 (Abbott Diagnostics, AbbottPark, IL), AxSYM (Abbott Diagnostics), Elecsys 2010(Roche Diagnostics, Indianapolis, IN), IMMULITE2000 (Diagnostic Products, Los Angeles, CA), andVITROS ECi (Ortho Clinical Diagnostics, Raritan, NJ)assays for CA 125 were evaluated for detection limit,dilution linearity, imprecision, correlation, andreference intervals. The maximum average deviationfrom target recoveries for dilution linearity studiesranged from 3.7% for the ADVIA Centaur to 18.2% forthe IMMULITE 2000. Imprecision studies yielded totalcoefficients of variation of 2.0% to 8.3% at CA 125concentrations of 35 and 114 U/mL (35 and 114 kU/L).Method comparison studies revealed good agreementwith the VITROS ECi comparison method, with slopesranging between 0.88 and 1.19 and correlationcoefficients of more than 0.95. All methods showacceptable performance characteristics and generallycompare well. However, for some samples, substantialdifferences exist between methods, necessitatingparallel testing when introducing a new method.

Cancer antigen 125 (CA 125) is a high-molecular-massglycoprotein (>200 kd) and initially was recognized by mono-clonal antibody OC125. The OC125 monoclonal antibodywas generated by immunization of BALB/c mice with theOVCA43 cell line isolated from ascites fluid of a patientwith serous papillary cystadenocarcinoma of the ovary.1 CA125 carries 2 major antigenic domains, and monoclonal anti-bodies against it can be classified as OC125-like or M11-like.2 CA 125 contains 24% carbohydrate and is expressednormally in fetal coelomic epithelium, by epithelial ovariantumors, normal and pathologic tissues of mllerian duct ori-gin, and mesothelial cells.3,4

The CA 125 assay is useful for determining the progno-sis of endometrial carcinoma, in evaluation of advancedendometriosis, and as an aid in monitoring the response totherapy for patients with epithelial ovarian cancer.5 Theassay is not useful for screening for ovarian cancer in asymp-tomatic women.6 An immunoradiometric assay for CA 125was first commercialized by Centocor (now FujirebioDiagnostics, Malvern, PA). That assay used the OC125 mono-clonal antibody for capture and tracer functions. To improveon the sensitivity and specificity of the first-generationassays, a double-determinant monoclonal antibodybased format was developed using 2 monoclonal antibodieswith specificities against the antigenic domains OC125 andM11. The second-generation assays have been reported toshow better agreement than the first-generation assays.7-13 Inthe present study, 7 automated second-generation immuno-assays for CA 125 were evaluated for detection limit, dilu-tion linearity, imprecision, method comparison, and refer-ence intervals.

922 Am J Clin Pathol 2006;125:921-927922 DOI: 10.1309/NBA312W0LANRXYH9


Mongia et al / SEVEN CA 125 IMMUNOASSAYS

Materials and Methods

The 7 assays that were evaluated were the Access 2(Beckman Coulter, Brea, CA), ADVIA Centaur (BayerDiagnostics, Tarrytown, NY), ARCHITECT i2000 (AbbottDiagnostics, Abbott Park, IL), AxSYM (Abbott), Elecsys2010 (Roche Diagnostics, Indianapolis, IN), IMMULITE2000 (Diagnostic Products, Los Angeles, CA), and VITROSECi (Ortho Clinical Diagnostics, Raritan, NJ). All assays wereperformed in 1 laboratory according to the manufacturersinstructions.

The limit of detection of each method was determined byanalyzing 10 replicates of the zero calibrator and 4 replicatesof the lowest nonzero calibrator (15-35 U/mL [15-35 kU/L])from each manufacturer. The data were analyzed using EPEvaluator Release 5 software (David G. Rhoads Associates,Kennett Square, PA). The results of 4 separate runs were aver-aged. For the Access 2, the Cal S0 (0 U/mL [0 kU/L]) andCalibrator S1 (25 U/mL [25 kU/L]) were used. For theADVIA Centaur, CA 125 MCM1 zero and Calibrator 1 (16.4U/mL [16.4 kU/L]) were used. For the ARCHITECT i2000,CA 125 Calibrator A (0 U/mL [0 kU/L]) and Calibrator B (20U/mL [20 kU/L]) were used. For the AxSYM, CA 125Calibrator A (0 U/mL [0 kU/L]) and Calibrator B (15 U/mL[15 kU/L]) were used. For the Elecsys 2010, CA 125CalCheck1 (0 U/mL [0 kU/L]) and Calibrator 1 (35 U/mL [35kU/L]) were used. For the IMMULITE 2000, the IMMULITEOV-MA diluent (0 U/mL [0 kU/L]) and Low Adjustor (16U/mL [16 kU/L]) were used. For the VITROS ECi, samplediluent B (0 U/mL [0 kU/L]) and CA 125 Calibrator 1 (24U/mL [24 kU/L]) were used.

Dilution linearity was assessed by pooling 2 or 3 patientsamples with similar high CA 125 concentrations for eachmethod, making serial dilutions (1:2 to 1:32), and analyzingeach sample in triplicate. The upper limits chosen were basedon the analytic measurement range stated in each assayspackage insert. For the Access 2 and Elecsys 2010 methods,the concentration of the high pool was approximately 5,000U/mL (5,000 kU/L); for the ARCHITECT i2000 and VITROSECi, the high pool was approximately 1,000 U/mL (1,000kU/L); and for the ADVIA Centaur, AxSYM, andIMMULITE 2000 methods, the high pool was approximately500 U/mL (500 kU/L). Whenever appropriate, the pooledpatient samples were first diluted with the manufacturers rec-ommended diluent until they were within the analytic mea-surement range of each method.

For the Access 2, the patient sample with high CA 125values was serially diluted with Diluent A (Beckman Coulter)to give final concentrations of 0.3%, 0.78%, 1.5%, 3.1%,6.25%, 12.5%, 25%, 50%, and 100% (range of CA 125 con-centrations, 23-5,087 U/mL [23-5,087 kU/L]). For the ADVIACentaur, serial dilutions were made with MD1 diluent (Bayer

Diagnostics) to give final concentrations of 1.5%, 3.1%,6.25%, 12.5%, 25%, 50%, and 90% (range of CA 125 con-centrations, 10-538 U/mL [10-538 kU/L]). For the ARCHI-TECT i2000, serial dilutions were made with ARCHITECTwash buffer (Abbott Diagnostics) to give final concentra-tions of 3.1%, 6.25%, 12.5%, 25%, 50%, and 90% (range ofCA 125 concentrations, 37-960 U/mL [37-960 kU/L]). Forthe AxSYM, serial dilutions were made with AxSYM CA125 diluent (Abbott Diagnostics) to give final concentrationsof 3.1%, 6.25%, 12.5%, 25%, 50%, and 90% (range of CA125 concentrations, 17-494 U/mL [17-494 kU/L]). For theElecsys 2010, linearity was assessed by serial dilution withUniversal diluent (Roche Diagnostics) to give final concen-trations of 0.39%, 0.78%, 1.5%, 3.1%, 6.25%, 12.5%, 25%,50%, and 100% (range of CA 125 concentrations, 22-4,720U/mL [22-4,720 kU/L]). For the IMMULITE 2000, the sam-ples were diluted with OV diluent (Diagnostic Products) togive final concentrations of 2.25%, 4.5%, 9%, 18%, 36%,73%, 81%, and 90% (range of CA 125 concentrations, 16-484U/mL [16-484 kU/L]). For the VITROS ECi, linearity wasassessed by diluting the sample with High sample diluent B(Ortho Clinical Diagnostics) to obtain final concentrations of1.5%, 3.1%, 6.25%, 12.5%, 25%, 50%, and 100% (range ofCA 125 concentrations, 13.5-975 U/mL [13.5-975 kU/L]).Assessment of dilution linearity was performed with EPEvaluator Release 5 software.

Imprecision studies were performed by assayingLyphocheck Tumor Marker Controls (Bio-Rad Laboratories,Hercules, CA) and the manufacturers high control for eachanalyzer, analyzed in replicates of 2 per run. Two runs per daywere conducted on each of 5 days with a minimum of 2 hoursseparating each run for the total of 20 replicates for each levelof quality control material. Assay imprecision data were ana-lyzed by using EP Evaluator Release 5 software.

Correlation studies were performed with 98 specimensfrom females between the ages of 14 and 87 years and withCA 125 concentrations from 0 to 1,000 U/mL (0-1,000 kU/L).The samples were tested according to the package insertinstructions for each method. The samples with measuredconcentrations of more than the analytic measurement rangefor each method were rerun after dilution according to themanufacturers instructions. The VITROS ECi was chosen asthe comparison method because it correlated the best with allof the other methods. Passing-Bablok regression analysis wasperformed using Analyse-it + Clinical Laboratory, version1.63 software (Analyse-It Software, Leeds, England).

Reference intervals were determined using samples from120 apparently healthy women between 20 and 65 years of agewho were not taking any prescription medications. One subjectwith the highest CA 125 concentration in this group was foundto have ovarian cancer; therefore, her samples were excludedfrom further analysis. Statistical analysis was performed using



EP Evaluator Release 5 software. The 97.5 percentile upperreference limit was determined using a transformed paramet-ric approach. All studies using samples from human subjectswere approved by the institutional review board of theUniversity of Utah, Salt Lake City.

Results

The limit of detection ranged from 0.05 U/mL (0.05kU/L) on the Access 2 to 1.45 U/mL (1.45 kU/L) on theAxSYM. The limit of detection was 0.30 U/mL (0.30 kU/L)for the ADVIA Centaur method, 0.35 U/mL (0.35 kU/L) forthe ARCHITECT i2000 method, 0.09 U/mL (0.09 kU/L) forthe Elecsys 2010 method, 0.10 U/mL (0.10 kU/L) for theIMMULITE 2000 method, and 0.06 U/mL (0.06 kU/L) for theVITROS ECi method. In all cases, the detection limit wemeasured was lower than the manufacturers claim.

For dilution linearity studies, the maximum average devi-ation from target recovery was 10.9% for the Access 2 at 162U/mL (162 kU/L), 3.7% for the ADVIA Centaur at 150 U/mL(150 kU/L), 8.2% for the ARCHITECT i2000 at 128 U/mL(128 kU/L), 7.7% for the AxSYM at 35 U/mL (35 kU/L),10.7% for the Elecsys 2010 at 151 U/mL (151 kU/L), 18.2%for the IMMULITE 2000 at 52 U/mL (52 kU/L), and 4.1% forthe VITROS ECi at 120 U/mL (120 kU/L). It is noteworthythat the upper limit of the analytic measurement range was5,000 U/mL (5,000 kU/L) for the Access 2 and Elecsys 2010

methods, 1,000 U/mL (1,000 kU/L) for the ARCHITECTi2000 and VITROS ECi methods, 600 U/mL (600 kU/L) forthe ADVIA Centaur method, 500 U/mL (500 kU/L) for theAxSYM method, and 400 U/mL (400 kU/L) for theIMMULITE 2000 method based on our studies with 1 lot ofreagent. Approximately 0.5% of the CA 125 results generatedin our laboratory are greater than 5,000 U/mL (5,000 kU/L),3.2% are greater than 1,000 U/mL (1,000 kU/L), and 5.6% aregreater than 500 U/mL (500 kU/L).

The imprecision of each assay was assessed by usingcommercial serum-based quality control materials. The coef-ficients of variation (CVs) for the low control (35 U/mL [35kU/L]) were 1.3% to 8.3% within run, 0% to 3.1% betweenruns, and 2.0% to 8.3% total Table 1. For the high control(114 U/mL [114 kU/L]), CVs were 1.4% to 4.3% within run,0% to 3.2% between runs, and 1.8% to 6.3% total.

Results from method comparison studies with Passing-Bablok analysis are shown Figure 1. Slopes ranged from0.87 for the IMMULITE 2000 assay to 1.19 for the ARCHI-TECT i2000. Difference plots for the same data are shownFigure 2. The mean differences ranged from 30 U/mL (30kU/L) for IMMULITE 2000 to 63 U/mL (63 kU/L) for theARCHITECT i2000 method. The IMMULITE 2000 had 2results that were lower than the VITROS ECi method by 250U/mL (250 kU/L) or more (Figure 2F). The analytic concor-dance of each method with the comparison method for thesesamples is shown using the manufacturers suggested upperreference limits Table 2.

Clinical Chemistry / ORIGINAL ARTICLE

Table 1Summary of Imprecision Data*

Coefficient of Variation (%)

Assay Control Material Mean Concentration (U/mL) Within-Run Between-Run Total

Access 2 Ly1 29.6 1.4 0.9 2.0Ly2 103 1.5 1.4 3.0

ADVIA Centaur Ly1 39.7 2.5 2.3 3.3Ly2 130 3.1 2.2 3.8Mfg3 237 2.2 0.0 2.4

ARCHITECT i 2000 Ly1 43.3 1.7 3.1 3.5Ly2 135 2.5 0.0 2.9Mfg2 306 2.5 0.4 2.6Mfg3 656 1.6 0.0 1.6

AxSYM Ly1 42.2 8.3 0.0 8.3Ly2 136 4.3 3.2 6.3Mfg3 208 7.1 0.0 7.1

Elecsys 2010 Ly1 34.1 2.0 0.0 2.4Ly2 101 1.4 0.9 1.8

IMMULITE 2000 Ly1 28.2 4.5 0.0 4.5Ly2 99.6 2.9 0.0 3.7Mfg3 126 4.1 0.0 4.3

VITROS ECi Ly1 27.9 1.3 1.1 2.1Ly2 95.7 2.1 0.6 2.2Mfg3 267 1.3 0.6 1.4

Ly, Lyphocheck control material (Bio-Rad Laboratories, Hercules, CA); Mfg, control material from the method manufacturer.* Values are given in conventional units; to convert to Systme International units (kU/L), multiply by 1. Access 2, Beckman Coulter, Brea, CA; ADVIA Centaur, Bayer

Diagnostics, Tarrytown, NY; ARCHITECT i2000, Abbott Diagnostics, Abbott Park, IL; AxSYM, Abbott Diagnostics; Elecsys 2010, Roche Diagnostics, Indianapolis, IN;IMMULITE 2000, Diagnostic Products, Los Angeles, CA; VITROS ECi, Ortho Clinical Diagnostics, Raritan, NJ.




Reference interval information obtained using samplesfrom apparently healthy women is shown Table 3. The medi-an CA 125 concentration varied from 6.3 U/mL (6.3 kU/L) forthe IMMULITE 2000 to 17.2 U/mL (17.2 kU/L) for theELECSYS 2010 method.

Discussion

The detection limit of all 7 immunoassays was less than1.5 U/mL (



The results of the present study show that all 7 second-generation immunoassays are highly comparable to eachother. The slope of IMMULITE 2000 compared with the VIT-ROS ECi was 0.87. The lower slope of the IMMULITE 2000method relative to the comparison method could be due part-ly to a deviation from linearity with underrecovery at CA 125concentrations between 400 and 500 U/mL (400-500 kU/L).However, results from reference interval studies also gave lowervalues with the IMMULITE 2000 assay, suggesting differences

in calibration. The ARCHITECT i2000 had a slope of 1.19relative to the VITROS ECi comparison method. Results fromreference interval studies demonstrated higher values as well.Relatively small differences in calibration likely explain theseregression and difference plot results. A previous report thatexamined 7 CA 125 immunoassays found slopes from regres-sion analysis that ranged between 0.61 and 1.80.13 Our resultssuggest that agreement among commercial methods hasimproved since that study was published.


500

250

0

250

500

Acce

ss 2

V

ITRO

S EC

i CA

125 (

U/mL)

0 250 500 750 1,000VITROS ECi CA 125 (U/mL)

500

250

0

250

500AD

VIA

Cent

aur

VIT

ROS

ECi C

A 12

5 (U/m

L)0 250 500 750 1,000

VITROS ECi CA 125 (U/mL)

500

250

0

250

500

ARCH

ITEC

T i2

000

VI

TROS

ECi

CA

125 (

U/mL)

0 250 500 750 1,000VITROS ECi CA 125 (U/mL)

500

250

0

250

500

AxSY

M V

ITRO

S EC

i CA

125 (

U/mL)

0 250 500 750 1,000VITROS ECi CA 125 (U/mL)

500

250

0

250

500

Elec

sys 2

010

VIT

ROS

ECi C

A 12

5 (U/m

L)

0 250 500 750 1,000VITROS ECi CA 125 (U/mL)

500

250

0

250

500

IMM

ULIT

E 20

00

VITR

OS E

Ci C

A 12

5 (U/m

L)

0 250 500 750 1,000VITROS ECi CA 125 (U/mL)

A B C

D E F

Figure 2 Comparison of CA 125 methods using difference plots. The solid line indicates the mean difference betweenmethods, and the dashed lines indicate the upper and lower 95% confidence limits of the difference between methods. Datafor 98 samples are shown in each panel. A, Mean difference, 36 U/mL (36 kU/L; 95% confidence interval [CI], 87 to 159 U/mL[87 to 159 kU/L]). B, Mean difference, 11 U/mL (11 kU/L; 95% CI, 98 to 76 U/mL [98 to 76 kU/L]). C, Mean difference, 63U/mL (63 kU/L; 95% CI, 69 to 194 U/mL [69 to 194 kU/L]). D, Mean difference, 24 U/mL (24 kU/L; 95% CI, 96 to 143 U/mL[96 to 143 kU/L]). E, Mean difference, 6 U/mL (6 kU/L; 95% CI, 94 to 105 U/mL [94 to 105 kU/L]). F, Mean difference, 30U/mL (30 kU/L; 95% CI, 158 to 99 U/mL [158 to 99 kU/L]). Access 2, Beckman Coulter, Brea, CA; ADVIA Centaur, BayerDiagnostics, Tarrytown, NY; ARCHITECT i2000, Abbott Diagnostics, Abbott Park, IL; AxSYM, Abbott Diagnostics; Elecsys 2010,Roche Diagnostics, Indianapolis, IN; IMMULITE 2000, Diagnostic Products, Los Angeles, CA; VITROS ECi, Ortho ClinicalDiagnostics, Raritan, NJ.




The large differences observed between methods forsome individual patient samples, which are best seen in Figure2, could be because of the use of different antibodies or uniquecomponents of these patient samples. Therefore, even thoughon average the results agree fairly well across assays, whenchanging assays for clinical use, we recommend parallel test-ing by the old and new methods to establish a new baseline foreach patient.

Although it is not recommended to use CA 125 for ovar-ian cancer population screening owing to a lack of adequatesensitivity and specificity for a disease with a relatively lowprevalence, it is interesting to examine the effects of assay dif-ferences on reference intervals. The 97.5 percentile upper ref-erence limits that were determined using the same group of120 apparently healthy female subjects ranged from 22 U/mL(22 kU/L) for IMMULITE 2000 to 45 U/mL (45 kU/L) for theARCHITECT i2000 method. These differences are larger thanwould be expected based on the method comparison studiesdiscussed previously. The percentage of results obtained forthese subjects that were less than 35 U/mL (



Supported by Abbott Diagnostics and the ARUP Institute forClinical & Experimental Pathology.

Address reprint requests to Dr Roberts: c/o ARUPLaboratories, 500 Chipeta Way, Salt Lake City, UT 84108.

References1. Bast RC Jr, Feeney M, Laarus H, et al. Reactivity of a

monoclonal antibody with human ovarian carcinoma. J ClinInvest. 1981;68:1331-1337.

2. Nustad K, Bast RC Jr, OBrien TJ, et al. Specificity and affinityof 26 monoclonal antibodies against the CA 125 antigen: firstreport from the ISOBM TD-1 workshop. Tumour Biol.1996;17:196-219.

3. Bischof P. What do we know about the origin of CA 125? EurJ Obstet Gynecol Reprod Biol. 1993;49:93-98.

4. Kabawat SE, Bast RC Jr, Bhan AK, et al. Tissue distribution ofa coelomic-epitheliumrelated antigen recognized by themonoclonal antibody OC125. Int J Gynecol Pathol.1983;2:275-285.

5. Chan DW, Booth RA, Diamandis EP. Tumor markers. In:Burtis CA, Ashwood ER, Bruns DE, eds. Tietz Textbook ofClinical Chemistry and Molecular Diagnostics. 4th ed.Philadelphia, PA: Saunders; 2005:745-795.

6. Carlson KJ, Skates SJ, Singer DE. Screening for ovariancancer. Ann Intern Med. 1994;121:124-132.

7. Van Kamp GJ, Verstraeten AA, Kenemans P. Discordantserum CA 125 values in commercial immunoassays. Eur JObstet Gynecol Reprod Biol. 1993;49:99-103.

8. Kenemans P, Bon GG, Kessler AC, et al. Multicentertechnical and clinical evaluation of a fully automated enzymeimmunoassay for CA 125. Clin Chem. 1992;38:1466-1471.

9. Kenemans P, van Kemp GJ, Oehr P, et al. Heterologous doubledeterminant immunoradiometric assay CA 125, II: reliablesecond-generation immunoassay for determining CA 125 inserum. Clin Chem. 1993;39:2509-2513.

10. Uhl W, Denk B. Improved CA 125 determinations using twodifferent monoclonal antibodies. In: Klapdor R, ed. CurrentTumor Diagnosis: Applications, Clinical Relevance, ResearchTrends. Munich, Germany: Zuckschwerdt Verlag; 1994:384-388.

11. Kenemans P, Verstraeten AA, van Kamp GJ, et al. The secondgeneration CA 125 assays. Ann Med. 1995;27:107-113.

12. Bonfrer JM, Baan AW, Jansen E, et al. Technical evaluation ofthree second generation CA 125 assays. Eur J Clin Chem ClinBiochem. 1994;32:201-207.

13. Davelaar EM, van Kamp GJ, Verstraeten AA, et al.Comparison of seven immunoassays for quantification of CA125 in serum. Clin Chem. 1998;44:1417-1412.


Documents

CA-125 automated assays