Multicenter evaluation of the Bayer Immuno 1(TM) CA 15-3(TM) assay

¹ Bayer Corporation, Business Group Diagnostics, Tarrytown, NY 10591.

² Department of Pathology, Hartford Hospital, Hartford, CT 06102.

³ Johns Hopkins University, Baltimore, MD 21287.

⁴ M.D. Anderson Cancer Center, Houston, TX 77030.

⁵ Memorial Sloan-Kettering Cancer Center, New York, NY 10021.
^a Address correspondence to this author at: The Institute for Diagnostic Research, Inc., 23 Business Park Dr., Brandford, CT 06405. Fax 203-315-4002; e-mail carol.cheli{at}diabetesdisc.com.

	Abstract

Top Abstract Introduction Materials and Methods Results Discussion References

 
We conducted a multicenter evaluation of the analytical and clinical features of the automated Bayer Immuno 1(TM) CA 15-3(TM) assay and compared assay performance to two manual tests. Results of the 10-day imprecision study of the Bayer Immuno 1 assay pooled across four evaluation sites and three lots of reagent produced total CV 4%. Lot-to-lot reproducibility for 26 different lots of reagents and calibrators manufactured over a 2-year period was demonstrated (CV, 1.1%). Results for the Bayer Immuno 1 assay correlated well with the Biomira TRUQUANT® BR(TM) 27.29 and Centocor® CA 15-3 RIAs (r 0.94). The upper limit of the reference interval for the Bayer Immuno 1 assay was 35.9 kilounits/L (35.9 units/mL); values were similar for all methods. Longitudinal monitoring of healthy women yielded assay values with an average CV of 11% and 21% for the Bayer Immuno 1 and Biomira assays, respectively. The Bayer Immuno 1 assay demonstrated the analytical features, intermethod correlation, and long-term performance characteristics that are essential for longitudinal monitoring of breast cancer patients.

	Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

 
Breast cancer is the second major cause of cancer death in American women today. Women have a one in eight chance of developing invasive breast carcinoma (1). The risk of death in breast cancer patients is related to the progression of metastatic disease. Measurement of serum CA 15-3(TM) (CA 15-3 is a trademark of Centocor, Inc.) appears to have a role in the management of metastatic breast cancer (2). CA 15-3 assays measure a heterogeneous high molecular weight mucin encoded by the MUC 1 gene (3). MUC 1 mucin is secreted by normal secretory epithelia, aberrantly overexpressed by tumors, and shed into the circulation in patients with breast cancer (4). CA 15-3 is not useful for early detection of breast cancer (5). However, numerous studies have documented that serum CA 15-3 correlates with the extent of disseminated mammary carcinoma (6)(7)(8). It is believed that monitoring serum CA 15-3 in patients with breast cancer can provide an indication of therapeutic response and give an early signal of disease progression or recurrence (5)(8)(9)(10).

Serum MUC 1 mucin bears epitopes that are recognized by two murine monoclonal antibodies, 115D8, generated against milk fat globulin membranes, and DF3, generated against a membrane extract from human metastatic breast carcinoma (11)(12). Several commercially available immunoassays use both of these monoclonal antibodies for quantifying CA 15-3 in serum; the antibody 115D8 is used as the capture antibody, whereas DF3 serves as the detection antibody. The first CA 15-3 assays were manual and used radioisotopic detection. Later, nonisotopic immunoenzymatic assays became available. Today, immunoassay automation can improve the precision of replicate results and increase the speed of generating results compared with nonautomated methods (13).

The clinical usefulness of longitudinal monitoring of individual patient serum CA 15-3 concentrations has been hampered by the analytical variability of CA 15-3 tests (14)(15)(16). At present, it is difficult for clinicians to distinguish between the variations in results due to tumor progression or regression from variations due to biological and analytical variability when monitoring their patients sequentially (15)(16). When monitoring a patient longitudinally, small increases or decreases in tumor marker concentration can be indicative of early recurrence of disease or response to therapy, respectively. Poor assay precision can mask these changes and confound the interpretation of disease status and, therefore, lead to poor medical decision-making. Suggestions to improve the long-term analytical performance of CA 15-3 immunoassays include: establishment of analytical reliability; comparability of CA 15-3 assay determinations between laboratories; comparability of lot-to-lot reagent performance; and automation of analytical procedures (14)(15)(16).

We have developed a fully automated random access immunoassay for the quantitative measurement of serum CA 15-3 on the Bayer Immuno 1(TM) immunoanalyzer. The aim of our studies was to evaluate the analytical and clinical performance of the Bayer Immuno 1 CA 15-3 assay. As part of this evaluation, we examined serial variability of the CA 15-3 assay values in healthy women, an important criterion often overlooked in the evaluation of serial tumor marker assays (14)(15). Our aim was to perform a realistic assessment of the variability of the method for monitoring longitudinal serial samples. Here we report our multicenter evaluation of the Bayer Immuno 1 CA 15-3 assay. We compared the Bayer Immuno 1 assay with two FDA-approved manual RIAs: the Biomira TRUQUANT® BR(TM) RIA (Biomira, Inc., Edmonton, Alberta, Canada) for the quantitation of serum CA 27.29, and the Centocor® CA 15-3(TM) RIA (Centocor, Malvern, PA) for the quantitation of serum CA 15-3. These assays have been reported to measure the same antigen encoded by the MUC 1 gene (17).

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

 
assay procedure
The Bayer Immuno 1 CA 15-3 assay is a magnetic particle separation immunoassay designed for the random access automated Bayer Immuno 1 immmunochemistry analyzer. The Bayer Immuno 1 CA 15-3 assay uses two monoclonal antibodies, DF3 and 115D8 (licensed from Centocor, Inc.), which both bind to MUC 1 mucin. Reagent 1 contains the monoclonal antibody 115D8, which is conjugated to fluorescein. Reagent 2 contains the second monoclonal antibody DF3, which is conjugated to alkaline phosphatase. Magnetic particles coated with anti-fluoroscein antibody, Reagent 1, and patient sample, calibrator, or control are incubated at 37 °C on the system for 20 min. Reagent 2 is then added, which binds to the immobilized antigen to form a sandwich immunocomplex. After 18 min, the immunocomplex is washed, a colorimetric substrate containing p-nitrophenylphosphate is added, and the resulting fluorescent signal is measured. A cubic-through-zero curve fitting algorithm is used to generate a calibration curve. The assay requires 4 µL of serum, and the first result is obtained in 38.5 min. Predilution of patient samples is not required. The operation of the Immuno 1 was performed according to manufacturer's instructions.

The Centocor CA 15-3 RIA system is a solid-phase immunoassay in which plastic beads coated with monoclonal antibody 115D8 form the solid phase and the radiolabeled monoclonal antibody DF3 is the detector. In the Biomira assay, polystyrene tubes are coated with human CA 27.29 antigen. The CA 27.29 antigen present in the sample competes with the solid-phase antigen for binding of the radiolabeled B27.29 tracer antibody. Tests were performed according to manufacturers' instructions.

calibrators and controls
The CA 15-3 assay controls used for the performance studies included two concentrations of Bio-Rad Lyphochek Tumor Marker Controls (BioRad Labs) and the Medical Decision Pool. The Medical Decision Pool is a serum-based control material manufactured at Bayer Corporation and used internally for routine quality control. This is a pool of healthy human serum to which antigen is added to a mean target concentration of 31 kilounits/L. The Bayer Immuno 1 CA 15-3 assay uses six calibrators, 0, 12.5, 25, 50, 100, and 200 kilounits/L of antigen added to a 60 g/L solution of bovine serum albumin. The antigen used in the Bayer Immuno 1 assay calibrators was obtained from Centocor, Inc.

specimen and patient selection
Human serum samples used for the correlation analysis and the determination of cutoff values were obtained from BioClinical Partners, Inc., and the in-house specimen collection at Bayer Corporation (Tarrytown, NY). There were ~500 serum samples from patients with healthy, malignant, and benign conditions used in these studies. Ages of the subjects ranged from 18 to 92 years (median age, 51). The samples from patients with malignant conditions examined included breast cancer (n = 150), lung cancer (n = 35), ovarian cancer (n = 35), and colorectal cancer (n = 35). Additionally, we analyzed sera from a total of 199 healthy subjects (100 premenopausal and 99 postmenopausal) and 46 patients with benign breast disease.

The majority of patient specimens for longitudinal monitoring of healthy individuals were obtained at the investigational sites or supplied by Western States Plasma Co., Inc. and the Bayer Corp. specimen bank. Both premenopausal (n = 21) and postmenopausal (n = 24) women were enrolled in the healthy subject longitudinal study. Ages of the subjects ranged from 22 to 68 years.

Healthy subjects included individuals with no fever or infections, who met the criteria for blood bank donation and had no known current or previous malignancy, and were not taking prescription medication at the time of sample collection. Healthy subjects were screened for their state of health by questionnaire, and patients with benign and malignant diseases were screened by physical examination. Preanalytical conditions of serum samples included age of specimen no greater than 10 years and storage within 24 h of collection at a temperature no higher than -20 °C. The serum samples used for the study were collected under a protocol reviewed and approved by the Institutional Review Board at each investigational site.

ca 15-3 performance protocol
The Bayer Immuno 1 CA 15-3 Assay was assessed for imprecision and lower limit of detection at four participating evaluation sites. These sites include the M. D. Anderson Cancer Center (Houston, TX), Johns Hopkins University (Baltimore, Maryland), Memorial Sloan-Kettering Cancer Center (New York, NY), and Hartford Hospital (Hartford, CT). Variability of longitudinal monitoring in healthy women was assessed at three of these sites by using the Bayer Immuno 1 CA 15-3 assay and the Biomira RIA. The remainder of the nonclinical performance evaluation was conducted at Bayer Corporation. For the correlation analysis and determination of the upper limit of the reference interval, serum samples were tested at Bayer Corporation using the Bayer Immuno 1 CA 15-3 assay and the Centocor CA 15-3 RIA and at Dianon Systems, Inc. (Stratford, CT) using the Biomira RIA.

lot-to-lot variability
To evaluate reproducibility in the manufacturing process of successive lots of reagents and calibrators, 26 lots of reagents and calibrators were used to assay one lot of the Medical Decision Pool. Assays were performed over a 2-year period in the manufacturing facility of Bayer Diagnostics (Bridgend, Wales). Mean assay values were determined from 20 replicate analyses per lot of reagent or calibrator, and the CV of the means was calculated.

imprecision
Imprecision of the CA 15-3 assay was evaluated by analysis of Bio-Rad Tumor Marker Controls (two concentrations), the CA 15-3 Medical Decision Pool, and three lots of the Bayer Immuno 1 CA 15-3 calibrators (levels 1–6). Samples were run in duplicate in 20 independent runs over 10 days by each of the four evaluation sites. At each site, three lots of Bayer Immuno 1 reagents were used. Site-to-site imprecision pooled across reagent lots and total imprecision pooled across sites and reagent lots were determined. Data were analyzed for variance components using the SAS (Version 6.07) NESTED procedure to give statistical estimates of within-run, run-to-run imprecision in the same laboratory, and total imprecision across laboratories.

lower limit of detection
The lower limit of detection of the assay was evaluated by determination of the minimum detectable concentration of CA 15-3 assay values that can be statistically distinguished from the concentration of the lowest calibrator, as calculated from a typical calibration curve. The minimum detectable concentration was calculated from the pooled within-run standard deviations of the zero calibrator response rates (mA/min). The rates were pooled over three calibrator lots for each of three reagent lots tested at each of the four evaluation sites. The overall minimum detectable concentration was calculated as the mean Bayer Immuno 1 CA 15-3 assay concentration corresponding to two pooled within-run SDs for each of the 12 reagent lot/site combinations.

linearity
To validate assay linearity over the dynamic range of the assay (0–200 kilounits/L), dilutions of four pools of human serum samples were analyzed over the entire calibration range. Each pool containing a high CA 15-3 assay concentration (180–200 kilounits/L) was diluted with the first calibrator (0 kilounit/L). Final concentrations representing 100%, 75%, 50%, 25%, 10%, and 0% of each high-sample pool were assayed with two lots of Bayer Immuno 1 reagents on two instruments. Test samples were assayed in triplicate. The data generated were subjected to regression analysis. Linearity was evaluated by comparing dilution-corrected values to the concentration of the undiluted sample.

high-dose hook effect
Very high concentrations of analyte in a sample can saturate both capture and tag antibodies, which decreases the assay signal, resulting in a falsely low result. To determine whether highly increased concentrations of CA 15-3 in patient serum cause a "hook effect", a pure preparation of antigen was diluted in the zero calibrator. Final concentrations ranged from 7 to 28 500 kilounits/L. Test samples were assayed in triplicate. The samples were tested in the Bayer Immuno 1 CA 15-3 assay with two lots of reagents. The assay response for the test samples was reported as the observed reaction rate vs the nominal CA 15-3 assay concentration.

correlation
To examine the agreement of the Bayer Immuno 1 CA 15-3 assay with the Centocor and Biomira RIAs, patient sample CA 15-3 or CA 27.29 assay values generated by the individual methods were compared by correlation analysis. Correlation studies were performed by using a panel of 500 serum specimens from healthy volunteers and from patients with malignant and benign diseases. The evaluation was conducted over an 8-week period. Patient results were obtained from the mean of duplicate determinations for the Biomira CA 27.29 RIA and the Centocor CA 15-3 RIA and from singlicate determinations by using the Bayer Immuno 1 CA 15-3 assay. CA 15-3 and CA 27.29 assay values generated from the three methods were compared by the linear least squares and Passing Bablok regression statistics.

upper limit of reference interval
The upper limit of the reference interval was determined for 199 healthy women by the Bayer Immuno 1 CA 15-3 assay and the Centocor and Biomira RIAs. Calculations of the mean, median, and range of CA 15-3 or CA 27.29 assay values were determined. The reference interval of CA 15-3 or CA 27.29 assay values was calculated by using a one-tailed nonparametric statistic as the assay value, which corresponds to the 97.5th percentile of all the measured values.

longitudinal variability of ca 15-3 or ca 27.29 assay values in healthy women
We prospectively studied total variability (biological and analytical) with the Bayer Immuno 1 CA 15-3 assay and the Biomira RIA in six samples from each of 15 women at each of three evaluation sites. Samples were obtained approximately once a month for 6 months. Specimens from any one individual were assayed by using one of four lots of Bayer Immuno 1 CA 15-3 reagent. Approximately equal numbers of patient specimens were assayed with each lot. The mean and CV for six serial assay determinations for both methods were calculated for each of the 45 healthy individuals entered into the study. t-tests for statistical significance between the two methods, individual means, grand means, individual variances, and average CV were calculated. The mean CV was calculated as 100 times the square root of the mean variance divided by the grand mean. The mean CA 15-3 and CA 27.29 longitudinal values for each patient were compared by the linear least squares and Passing Bablok linear regression statistics.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

 
Fig. 1 presents the CA 15-3 results for the Medical Decision Pool assayed with 26 lots of calibrators and reagents over a 2-year period. Each point represents the mean of 20 replicate assay measurements. Values ranged from 27.7 to 34.4 kilounits/L. The lot-to-lot CV of the Immuno 1 assay measurements was 1.1%.

View larger version (21K): [in this window] [in a new window]   Figure 1. Lot-to-lot variability in the manufacturing of the Bayer Immuno 1 CA 15-3 reagents and calibrators.

Imprecision data for two levels of Bio-Rad controls, Medical Decision Pool, and calibrators pooled across three reagent lots and four evaluation sites are presented in Table 1 . Pooled within-run CVs were 1.3%–3.4%. Total pooled CVs over the four evaluation sites and three reagent lots were 3.0%–4.0%.

In this study, a mean lower limit of detection of 0.13 kilounit/L was observed for the CA 15-3 assay. This value was determined based on multiple determinations (n = 1336) of the Level 1 calibrator (0 kilounit/L) at four evaluation sites using three lots of Immuno 1 calibrators and reagents. The detection limit supports a claim of 0.2 kilounit/L for the Bayer Immuno 1 CA 15-3 assay.

We examined linearity of this new method by analyzing five serial dilutions of four serum sample pools diluted with Bayer Immuno 1 CA 15-3 zero calibrator. The clinical samples diluted linearly as determined by linear regression analysis. The regression analysis yielded slopes ranging from 0.9938 to 1.01 (r = 0.99).

No hook effect was seen in samples containing 7 to 28 500 kilounits/L of antigen (data not shown).

Fig. 2 shows the distribution of 199 healthy patients and their Bayer Immuno 1, Biomira, and Centocor RIA values. The numbers of assay values obtained by the three methods at each assay concentration range were similar. The detailed results of the range analysis are presented in Table 2 . The 97.5^th percentile normal reference range cutoff for the Bayer Immuno 1 assay was 35.9 kilounits/L, 31.3 kilounits/L for the Biomira RIA, and 35.6 kilounits/L for the Centocor RIA.

View larger version (22K): [in this window] [in a new window]   Figure 2. Normal reference range distribution of 199 healthy females and their Bayer Immuno 1, Biomira TRUQUANT RIA, and Centocor RIA assay values.

Table 3 shows the correlation coefficients and the regression parameters for 500 samples over the range of the calibration curve analyzed by the Bayer Immuno 1 assay with the Biomira and Centocor RIAs. Graphs of the correlation between methods for the patient samples are shown in Fig. 3 . Test results obtained with the Bayer Immuno 1 CA 15-3 assay were similar to results obtained with the comparison methods.

View larger version (32K): [in this window] [in a new window]   Figure 3. Correlation of the Bayer Immuno 1 and the Centocor CA 15-3 (A) and Biomira BR 27.29 assay (B) results (n = 500). Linear regression was determined by the Ordinary Least Squares method.

The means and CVs for six serial measurements for each of 45 healthy women assayed using the Biomira RIA and the Immuno 1 CA 15-3 assay are summarized in Fig. 4 . The CV determined for the Bayer Immuno 1 assay for each of the 45 individuals was plotted against the corresponding CV determined from the Biomira RIA assay. The Biomira assay values were more variable than the assay values obtained with the Bayer Immuno 1 assay. The average CV was 11% for the Bayer Immuno 1 CA 15-3 assay and was 21% for the Biomira assay. In addition to the comparison of serial variability by using these two assays, we also evaluated the mean CA 15-3 assay values for both assays. The mean of six serial measurements using the Bayer Immuno 1 CA 15-3 assay for each individual was plotted against the mean values obtained with the Biomira RIA and results are presented as an inset in Fig. 4 . Linear regression analysis yielded: Immuno 1 = 1.27 x Biomira-5.56, S_yx = 3.37, r = 0.93. The difference between the individual means as well as the grand means for the two methods were not statistically significant (P = 0.22 and 0.67, respectively). However, the difference between the variances of the methods was significant (P = 0.0001), as was the difference in the average CV (P = 0.0001).

View larger version (14K): [in this window] [in a new window]   Figure 4. Comparison study of variability in longitudinal monitoring of healthy subjects (n = 45). Comparison of Bayer Immuno 1 CV vs Biomira TRUQUANT BR CV determined from individual serial assay values. Inset, Bayer Immuno 1 mean vs Biomira TRUQUANT BR mean from individual serial assay values.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

 
A major challenge facing the physician is to determine the response of breast cancer to therapy. Breast cancer metastasis is often difficult to monitor by conventional noninvasive imaging procedures. The clinical utility of CA 15-3 serum concentrations in monitoring response to therapy and for early detection of recurrence of metastatic breast disease has been described (10). Serial increases of CA 27.29 and CA 15-3 assay values have been reported to detect breast cancer recurrence with substantial lead times before clinical detection of recurrence (9)(18).

The Bayer Immuno 1 CA 15-3 automated method demonstrates advantages for the longitudinal monitoring of breast cancer patients. The imprecision across multiple sites and multiple lots of reagents and calibrators was 3%–4%. Imprecision (CV) with manual methods reportedly is >10% (13), whereas the CV of the Abbott CA 15-3 IMx automated method was 4.1%–5.8% in a multicenter evaluation (19). The Centocor and Biomira assays have reported imprecision data across multiple sites ranging from 8.4%–12.4% and 6.0%–15%, respectively (manufacturers product labeling). These findings indicate the improved performance of automation over manual methods and demonstrate that the testing of patient samples at different sites using the Bayer Immuno 1 CA 15-3 assay will produce comparable results.

Lot-to-lot variation is an important source of analytical variation within a method, but it is seldom quantified (16)(20). During the manufacturing of reagents and calibrators, considerable variation between lots can occur. van Dalen (14) emphasized the importance of quality control and standardization in the manufacture of immunoassay reagents and made the recommendation that manufacturers report lot-to-lot variation for their tumor marker assays (14)(20). Our analysis from tests using multiple lots of Bayer Immuno 1 reagents and calibrators performed over a 2-year duration (Fig. 1 ) document the precision (and quality control in the manufacturing process) of the Bayer Immuno 1 CA 15-3 assay.

The lower limit of detection of 0.13 kilounit/L appears acceptable because serum concentrations of MUC 1 mucin <3 kilounits/L are rare and of no known clinical importance. The linearity and freedom of the assay from a high-dose "hook effect" are important for monitoring because high values of the marker are to be expected, and dilutions are required. It has been reported that manual assays are less reliable in this respect and often give dilution values that are too high, possibly reflecting the higher imprecision of these tests (13).

Both the correlation analysis and the determination of the upper limit of the reference interval demonstrate agreement of patient sample assay values of the Bayer Immuno 1 CA 15-3 assay with the Biomira and Centocor RIAs. The differences noted in the y-intercept and the slopes of the regression lines may be due to differences in the statistical method used and may reflect the sensitivity of ordinary least squares regression to outlier results (21).

We found significant (P = 0.0001) discrepancy in the mean CV for serial measurements of healthy patient samples between the Biomira (manual) and automated methods. Variability during sequential testing is a combination of biological and analytical variability. In our study, the biological variability remained constant for both assays, because the same serum samples were analyzed. The improved analytical precision of the Bayer Immuno 1 CA 15-3 assay renders it attractive for serial monitoring. Moreover, the study demonstrates a realistic assessment of the method by analyzing its performance when monitoring clinical samples longitudinally over time, the clinical situation where the assay is most useful.

For CA15-3 immunoassays to provide reliable data in serial monitoring of patients, a high degree of long-term analytical performance of an assay must be maintained (15)(22). Variation in assay values because of poor analytical performance can mask small changes in assay values due to disease progression and compromise the clinical utility of the CA 15-3 assay. Long-term analytical precision of an assay will provide a more accurate interpretation of longitudinal results, including early detection of recurrence of disease, increasing the lead time for therapeutic intervention and a more accurate determination of therapeutic response.

Despite the strong linear relationship between the Biomira and Bayer Immuno 1 methods, we observed a bias in the slope of the regression line. This indicated that the two methods yield results that are related but not interchangeable. It has been reported that although there is good method correlation between assays in the quantitation of MUC 1 gene products, it is inappropriate to exchange patient results from one method to another (13)(20). A recent study has indicated that different antibodies used to detect the MUC 1 mucin differ in their affinity and epitope specificity for the mucin and, therefore, produce inexchangeable results (23). This may explain the difference observed in our study between the Biomira 27.29 assay and the Bayer Immuno 1 CA 15-3 assay. Other studies have demonstrated that even when using the same antibodies, the many different test systems available yield different assay results (14)(19)(24). International standardization has been proposed that may improve this situation (14)(24), but at present, it is prudent that only one test system be used during the follow-up of a given patient.

	Acknowledgments

 
We thank Ian Morgan and MaryAnne Acaster of Bayer Group Manufacturing in Bridgend, Wales for technical assistance; Dr. Paul Dillion for statistical analysis of the clinical trial data; Carol Smith, Deborah Bruzek, Deborah Witek, Dr. Saber Ahmed, and Robert Jensen for technical assistance in the clinical trials; and Dr. Glen Armstrong for critical review of the manuscript. Portions of this work were presented previously as a poster session at the 49^th AACC Annual Meeting in 1997 [Clin Chem 1997;43(Suppl 6):S233].

	References

Top Abstract Introduction Materials and Methods Results Discussion References

 

1. Cancer Facts and Figures. Washington, DC: American Cancer Society, 1997..
2. Tonodini C, Hayes DF, Kufe D. Circulating tumor markers in breast cancer. Henderson IC eds. Diagnosis and therapy of breast cancer 1989:653-674 W. B. Saunders Philadelphia. .
3. Gendler SJ, Spicer AP, Lalani EN, Duhig T, Peat N, Burchell J, et al. Structure and biology of a carcinoma-associated mucin, Muc1. Am Rev Respir Dis 1991;144:S42-S47. [ISI][Medline] [Order article via Infotrieve]
4. Hayes DF, Sekine H, Ohno T, Abe M, Keefe K, Kufe DW. Use of a mAb for detection of circulating plasma DF3 antigen levels in breast cancer patients. J Clin Invest 1985;75:1671-1677.
5. Hayes DF, Zurawski VR, Kufe DW. Comparison of circulating CA 15-3 and carcinoembryonic antigen levels in patients with breast cancer. J Clin Oncol 1986;4:1542-1550. [Abstract/Free Full Text]
6. van Dalen A. Tumor markers in breast cancer. Ann Chir Gynaecol 1989;78:54-64. [ISI][Medline] [Order article via Infotrieve]
7. Steiber P, Diergarten K, Eirmann W, Albiez M, Fateh-Moghadam A. CA 15-3: evaluation and clinical value in breast carcinomas compared with CEA and TPA. Klapdor R eds. New tumor markers and their monoclonal antibodies 1988:46-51 Thieme Verlag Stuttgart, Germany. .
8. Dnistran AM, Schwarz MK, Greenburg EJ, Smith CA, Schwarz DC. CA 15-3, carcinoembryonic antigen in the clinical evaluation of breast cancer. Clin Chim Acta 1991;200:81-94. [ISI][Medline] [Order article via Infotrieve]
9. Bombadieri E, Pizzichetta M, Veronisi P, Seregni E, Bogni A, Maffioli L, et al. CA 15-3 determination in patients with breast cancer. Eur J Cancer 1993;29A:144-146.
10. Safi F, Kohler I, Rottinger E, Beger H. The value of the tumor marker CA 15-3 in diagnosing and monitoring breast cancer. Cancer 1991;68:574-582. [ISI][Medline] [Order article via Infotrieve]
11. Hilkens J, Buijs F, Hilgers J, Hageman P, Calafat J, Sonnenberg A, van der Valk M. Monoclonal antibodies against human milk fat globule detecting differentiating antigens of the mammary gland and its tumours. Int J Cancer 1984;34:197-206. [ISI][Medline] [Order article via Infotrieve]
12. Kufe D, Inghirami G, Abe M, Hayes D, Justi-Wheller H, Schlom J. Differentiated reactivity of a novel monoclonal antibody (DF3) with human malignant versus benign breast tumours. Hybridoma 1984;3:223-232. [ISI][Medline] [Order article via Infotrieve]
13. Huber PR, Bischof P, Kretschmer R, Truschnig M, Halwachs G, Schmidt M. CA 15-3: a multicenter evaluation of automated and manual tests. Eur J Clin Chem Clin Biochem 1996;34:77-84. [ISI][Medline] [Order article via Infotrieve]
14. van Dalen A. Analytical requirements and standardization of CA 15-3. Scand J Clin Lab Invest 1995;55(Suppl 221):102-104.
15. Holzel WGE, Beer R, Deschner A, Greismacher A, Muller MM. Individual reference ranges of CA 15-3, MCA and CEA in recurrence of breast cancer. Scand J Clin Lab Invest 1995;55(Suppl 221):93-101.
16. Holzel W. Analytical variation in immunoassays and its importance for medical decision making. Scand J Clin Lab Invest 1991;51(Suppl 205):113-119.
17. Reddish MA, Helbrecht N, Almeida AF, Madiyalakan R, Suresh MR, Logenecker BM, et al. Epitope mapping of Mab B27.29 within the peptide core of the malignant breast carcinoma-associated mucin antigen coded for by the human Muc 1 gene. J Tumor Marker Oncol 1992;7:19-27.
18. Chan DW, Beveridge RA, Muss H, Fritsche HA, Hortobagy G, Theriault R, et al. Use of Truquant BR radioimmunoassay for early detection of breast cancer recurrence in patients with stage II and stage III disease. J Clin Oncol 1997;15:2322-2328. [Abstract/Free Full Text]
19. van Kamp GJ, Bon GG, Verstraeten RA, Lynch D, Krikau M, Fluckiger J, et al. Multicenter evaluation of the Abbott Imx® CA 15-3(TM) assay. Clin Chem 1996;42:28-33. [Abstract/Free Full Text]
20. van Dalen A. Quality control and standardization of tumor marker tests. Tumor Biol 1993;14:131-135.
21. Passing H, Bablok W. Comparison of several regression procedures for method comparison studies and determination of sample sizes: application of linear regression procedures for method comparison studies in clinical chemistry, Part II. J Clin Chem Clin Biochem 1984;22(6):431-445. [ISI][Medline] [Order article via Infotrieve]
22. Browning M. Biological variation of the breast marker CA 15-3: implications for necessary standards of performance and the interpretation of results. Ann Clin Biochem 1987;24:S1-35-S1-38.
23. Dai J, Allard WJ, Davis G, Yeung KK. Effect of desialylation on binding, affinity, and specificity of 56 monoclonal antibodies against MUC 1 mucin. Tumor Biol 1998;19(Suppl 1):100-110.
24. Schwarz MK. Current status of tumour markers. Scand J Clin Lab Invest 1995;55(Suppl 221):5-14.