A comparative study of four serological tumor markers for the detection of breast cancer.

Shawn R. Clinton (1)

Kevin L. Beason (1)

Sabrina Bryant (1)

James T. Johnson (1)

Margaret Jackson (1)

Cynthia Wilson (2)

Kay Holifield (3)

Charlton Vincent (3)

Margot Hall (1,4)

Breast cancer is currently the third most common cause of cancer in the world. Circulating tumor antigens are often used as a minimally invasive tool for noting breast cancer progression. The objective of this study was to compare four tumor antigens (CA 15-3, CA 27.29, alpha-fetoprotein [AFP], and carcinoembryonic antigen [CEA]) for their diagnostic efficacy in breast cancer patients. It was hypothesized that CA 15-3 would prove to be superior to CA 27.29, CEA, and AFP in assay performance. Tumor marker assays were performed according to the manufacturers' directions. Assays used in this study were CA 15-3 and CA 27.29 (Fujirebio Diagnostics/Centocor Inc.), AFP (Abbott Inc.), and CEA (Hybritech Inc.). A total of 554 patient samples were obtained from an area hospital, plus 200 healthy adult samples which were used for the determination of normal reference intervals. The patients included patients with no disease (184), with non-malignant disease (11), with breast cancer (87), and with other types of cancer ( 272). Diagnostic percent sensitivities for each marker were: CA 15-3 (63%), CA 27.29 (39%), CEA (22%), and AFP (22%). Diagnostic specificities for each marker were comparable, ranging from 80-88%. Analytical parameters were evaluated for the assays and compared favorably. We concluded that CA 15-3 was the best tumor antigen for use as a diagnostic aid and monitoring agent.

Keywords: cancer, breast cancer, carcinoembryonic antigen, alpha-fetoprotein, CEA, AFP, CA 15-3, CA 27.29, tumor marker.

Cancer is a leading cause of death in the United States. Contributing 22% of total deaths, it ranks second only to cardiac disease. With 182,000 new cases in the year 2000, breast cancer incidence ranks first among US females (Cancer Facts and Figures, 2000) and third in the world for cancer occurrence (Key et al., 1998). It affects one in ten women and its financial and emotional costs are huge.

Traditional methods of breast cancer diagnosis have included biopsy, radiographs, ultrasonography, computed tomography (CT), and magnetic resonance imaging. A monthly self examination is recommended for all women and it is suggested that women have a baseline mammogram when they are between the ages of 35 and 40, with a follow-up mammogram approximately every 2 years thereafter (Hossfeld and Sherman, 1990). Additionally, tests for serum tumor antigens have proven useful as an aid in the assessment of tumor burden, and therapeutic monitoring of breast cancer (Wu and Nakamura, 1997). Among the tumor markers that have been reported to have some clinical utility in establishing the diagnosis and prognosis of breast cancer are CEA, CA15-3, CA27.29, and CA549 (Wu and Nakamura, 1997; Hubbard, 1990). Elevated circulating alpha-fetoprotein has not been reported in breast cancer patients and therefore could serve as a good negative control.

CEA is a 150-300 kDa. heterogeneous glyco-protein whose serum concentration is abnormally increased in patients with colorectal cancer, breast cancer, and a variety of other carcinomas (Reynoso et al., 1972; Cooper et al., 1979). Additionally, CEA levels are elevated in heavy smokers and nonmalignant pathologies (Clarke et al., 1982). For this reason CEA is not recommended as a screening procedure but has been used in combination with other methods for diagnosis and is often used in therapeutic monitoring.

CA 15-3 is a glycoprotein with a molecular weight of 300-450 kDa. Two monoclonal antibodies have been developed against it: DF3 which was developed against a membrane enhanced breast cancer extract and 115D8 which was developed against the membrane of human milk fat globule (Kufe et al., 1984; Hilkens et al. 1984). It has been reported in cases of breast, ovarian, pancreatic, lung, and colorectal cancer (Wu and Nakamura, 1997).

CA 27.29 is a mucin antigen detected by the monoclonal antibody B27.29 which was developed against an antigen found in the ascites of patients with metastatic breast cancer. CA 27.29 has an epitope that is shared with the DF3 antibody of CAl5-3. (Burtis and Ashwood, 1996). It is currently being marketed as a specific test for breast cancer.

CA 549 is a glycoprotein with two identified species of 400 and 512 kDa. It is detected by two monoclonal antibodies developed against the T417 human breast cancer cell line, and human milk fat globule membranes (Bray et al., 1987). It has been reported in breast, prostate, ovary, and lung cancer (Beveridge et al., 1988).

Alpha-fetoprotein is a 70,000 kDa glycoprotein which was first described in murine sera (Abelev et al., 1963) and subsequently described in human patients with hepatocellular carcinomas and germ cell tumors (Chan et al., 1986). Synthesized by the yolk sac and the liver during fetal development, AFP is present in both maternal and fetal sera and is elevated in maternal sera and amniotic fluid in cases of open neural tube disease and gastroschisis, and decreased in cases of Down's syndrome (Milunsky, 1987; Knight et al., 1988). Alpha-fetoprotein has been reported to be useful in screening for hepato-cellular carcinoma in high incidence areas such as Asia, and for classifying and staging germ cell tumors (Chan et al., 1986).

When comparing different assay methods one evaluates their specific performance characteristics (precision, linearity, analytical sensitivity and specificity) and their clinical performance (normal reference interval and predictive values). Precision is determined by assaying replicate samples and determining the mean, standard deviation, and coefficient of variation. Linearity is determined by assaying dilutions of an elevated serum sample and plotting the results and/or performing regression analysis. Analytical sensitivity represents the minimum detectable concentration of analyte and is determined by assaying replicate samples lacking the analyte (e.g., diluent) and calculating the mean plus two standard deviations. Values which fall below this cutoff are presumed to have no analyte. The analytical specificity represents the degree of interference in the assay from drugs and other chemicals such as bilirubin or lipids. This is not always reported but can be determined by spiking samples with varying conce ntrations of the suspected interfering analytes.

In clinical chemistry, the normal or healthy adult reference interval is established by calculating the mean plus or minus the standard deviation (95% confidence interval) for a population set of adults known to be in good health. Thus any patient result which is within this interval is considered to be "normal" or healthy; whereas, patient results that fall outside (above or below) the limits of this interval are considered to be abnormally elevated or decreased respectively. Since for tumor markers a low result would have no clinical significance, one establishes the cutoff between normal (presumed negative for disease) and abnormal (presumed positive for disease) results by using the mean plus two standard deviations. Predictive validity compares the ability of a new test method to accurately diagnose/predict the presence or absence of disease with that of an established method. Predictive value results include diagnostic sensitivity and specificity, diagnostic efficiency, and positive and negative predict ive values. For the calculation of predictive values, one compares the test results with the "true results" as defined by an external test method considered to be the reference test method. For example one could compare the results of a tumor antigen test with those obtained by the physician with histologic analysis of biopsy material. Predictive values include: (a) diagnostic sensitivity (percent of individuals with the disease who test positive by the assay), (b) diagnostic specificity (percent of individuals without the disease who test negative by the assay), (c) diagnostic efficiency (percent of all test results that are either true positives or true negatives), (d) positive predictive value (percent of all positive test results that are true positives), and (e) negative predictive value (percent of all negative test results that are true negatives).

The purpose of this study was to compare the analytical and clinical performances of four serologic tumor marker tests (CAl5-3, CA 27.29, AFP, and CEA) for the detection of breast cancer. Test kits for CA 549 were not available and so this marker was not evaluated in this study. A working hypothesis that CA 15-3 would prove to be superior to the other three tumor markers was developed based on reports in the literature of its superiority and the fact that it has been evaluated on much larger groups of patients than CA 27.29 which shares an overlapping epitope and therefore could be expected to perform equivalently (Safi et al., 1991; Pannall and Kotasek, 1997; Wu and Nakamura, 1997).

MATERIALS AND METHODS

Assays--All assays were performed according to the directions supplied by the manufacturers. The Tandem [R]-E CEA assay (Hybritech, Inc) is a solid phase two-site immunoenzymometric assay (ELIZA) utilizing two monoclonal IgG antibodies directed against unique sites on the CEA antigen. This assay was quantitated spectrophotometrically using the Photon Immunoassay Analyzer [TM] from Hybritech, Inc. The Centocor [R] CA 15-3[R] assay (Fujirebio Diagnostics, Inc./Centocor, Inc.) is a solid phase radioimmunoassay (RIA) using the 115D8 murine monoclonal antibody as the capture antibody and the [I.sup.125] labeled DF3 murine monoclonal antibody as the tracer. This assay was quantitated using an Iso Data [R] gamma counter. The Truquant [R] BR [TM] assay (Fujirebio Diagnostics, Inc./Centocor, Inc) is a solid phase competitive inhibition radioimmunoassay (competitive RIA) using polystyrene tubes coated with CA 27.29 antigen and [I.sup.125] labeled murine monoclonal B27.29 antibody. This assay was quantitated using an Is o Data [R] gamma counter. The IMx [R] AFP assay (Abbott Laboratories, Inc.) is a microparticle enzyme immunoassay (MEIA) utilizing two monoclonal antibodies directed against unique sites on the AFP antigen. This assay was quantitated using the IMx [R] Automated Analyzer from Abbott Laboratories, Inc. Statistical analysis was performed using SPSS software.

Patients--Procedures used in this study were in accord with ethical standards established by the University of Southern Mississippi (USM). Permission for the study was granted by the USM Human Subjects Protection Review Committee (HSPRC/IRB).

All study participants were selected from patients seen in an area hospital. Five hundred fifty-four patients were randomly chosen and the assays were run blind. Blood samples were collected using appropriate aseptic technique. Following serum separation aliquots were coded and frozen at -20[degrees]C. Subsequently, aliquots were thawed at 37[degrees]C and assayed in duplicate for the tumor antigens. The diagnoses were obtained from the attending physicians and were based on pathological examination. Patient classifications included (a) no known disease, (b) nonmalignant disease, (c) non mammary cancer, and (d) breast cancer. Cancer patients were classified according to the primary site of the tumor, regardless of the presence or absence of metastases. Since available information on patient therapy was incomplete, statistical analyses were performed on the total patient pool without reference to this.

The normal control subjects were healthy males (100) and females (100) ranging from 18-65 years of age. Their blood samples were collected and processed in the same manner as the patient samples.

RESULTS

Precision and linearity--Quality control samples analyzed over a 3 month period were used to determine intra- and inter-assay precision. The within-run coefficient of variation (%CV) was less than 10% for all but the CA 15-3 assay which was somewhat higher (Table 1). Similarly the between-run coefficient of variation was less than 15% for all the assays (Table 2). Serial dilutions of abnormal pool (sera combined from multiple human donors and spiked with an elevated concentration of tumor antigen) samples exhibited good linearity (Fig.1) with [R.sup.2] values equal to or greater than 99% for all the assays.

Reference intervals--The minimum detectable concentration was determined by analyzing approximately 20 replicates of the zero calibrator/diluent and establishing the mean + 2SD as the cut-off value (Table 3). The normal adult reference intervals were established by determining the 95% confidence intervals for healthy control male and female subjects. The intervals (Table 3) were broader than those reported by the manufacturer for both CEA and CA 15-3 and somewhat narrower for the other two assays. There was no significant difference between healthy adults by gender for any of the assays.

Table 1

Within-run precision for CA 15-3, CA 27.29, AFP, and CEA.

Sample               n   Mean          SD    %CV

CA 15-3 Control      50  46.83 U/mL    9.60  20.50
CA 27.29 Control I   42  75.36 U/mL    6.61  8.77
CA 27.29 Control II  37  106.51 U/mL   9.93  9.32
AFP Low Control      10  20.36 ng/mL   2.22  10.90
AFP Medium Control   10  77.87 ng/mL   3.16  4.06
AFP High Control     10  171.22 ng/mL  4.96  2.90
CEA Low Control      43  4.28 ng/mL    0.29  6.78
CEA High Control     40  64.04 ng/mL   2.79  4.36
Table 2

Between-run precision for CA 15-3, CA 27.29, AFP, and CEA.

Sample               n   Mean          SD     % CV

CA 15-3              67  45.21 U/mL    6.61   14.62
CA 27.29 Control I   73  74.99 U/mL    6.95   9.27
CA 27.29 Control II  68  117.76 U/mL   16.38  13.91
AFP Low Control      38  19.60 ng/mL   1.44   7.35
AFP Medium Control   38  78.15 ng/mL   3.88   4.96
AFP High Control     38  167.01 ng/mL  6.28   3.76
CEA Low Control      76  4.44 ng/mL    0.37   8.33
CEA High Control     72  62.64 ng/mL   3.40   5.43
Table 3

Reference intervals for CA 15-3, CA 27.29, AFP, and CEA.

Sample                 n    Mean        SD     Range

Zero/Diluent Controls

CA 15-3                21   0.02 U/mL   0.08   0.00-0.18
CA 27.29               24   0.24 U/mL   1.16   0.00-2.56
AFP                    13   0.00 ng/mL  0.01   0.00-0.02
CEA                    20   0.00 ng/mL  0.35   0.00-0.70

Healthy Adult Males

CA 15-3                106  25.36 U/mL  13.92  0.00-53.20
CA 27.29               100  18.94 U/mL  8.28   2.38-35.50
AFP                    107  3.47 ng/mL  1.79   0.00-7.05
CEA                    133  3.08 ng/mL  2.36   0.00-7.80

Healthy Adult Females

CA 15-3                108  24.08 U/mL  14.12  0.00-52.32
CA 27.29               100  16.54 U/mL  6.28   3.98-29.10
AFP                    107  3.73 ng/mL  2.06   0.00-7.85
CEA                    131  2.55 ng/mL  2.89   0.00-8.33

Healthy Adults

CA 15-3                214  24.71 U/mL  14.00  0.00-52.72
CA 27.29               200  17.74 U/mL  7.42   2.90-32.58
AFP                    214  3.60 ng/mL  1.93   0.00-7.46
CEA                    264  2.82 ng/mL  2.64   0.00-8.10
Table 4

Comparison of predictive values of CA 15-3, CA27.29, AFP, and
CEA for breast cancer.

Parameter                 CA 15-3    CA 27.29   AFP        CEA
                          (n = 515)  (n = 494)  (n = 418)  (n = 554)

Sensitivity (%)           63.38      39.29      21.82      22.35
Specificity (%)           80.64      83.56      88.01      79.66
Predictive Value (%) (-)  93.25      91.50      88.92      84.93
Predictive Value (%) (+)  34.35      23.40      20.34      16.67
Efficiency (%)            78.30      78.54      79.87      70.83
Cutoff Value              35.0       37.7       8.9        5.0
                          U/mL       U/mL       ng/mL      ng/mL

Diagnostic parameters--In this study there were 184 patients without disease, 11 patients with nonmalignant disease, 87 patients with breast cancer, and 272 patients with other types of cancer including: pancreatic, gastric, small intestinal, esophageal, lung, ovarian, prostatic, renal, colorectal, gallbladder, hepatic, cecal, uterine, testicular, head and neck, leukemia, lymphoma, and all other types. For purposes of this study, patients with breast cancer were considered to be positive for disease. Patients without disease, with non-malignant disease, and with non-mammary cancer were considered to be negative for disease. Similarly, cutoffs between normal (negative) and abnormal (positive) test results used were those listed by the manufacturers and are cited in Table 4. The predictive values shown in Table 4 compare the ability of four test methods (CA 15-3, CA 27.29, AFP, and CEA) to accurately diagnose/predict the presence or absence of disease (breast cancer). In Table 4, the diagnostic sensitivity of CA 15-3 (63%) is superior to that of the other three markers (CA 27.29 39%, AFP 22%, and CEA 22%). The diagnostic specificities of the four assays range from 80-88% with AFP having the highest value. The negative predictive and positive predictive values range from 85-93% and 17-34% respectively with CA 15-3 having the best values in each case. The efficiencies of the CA 15-3, CA 27.29, and AFP assays (78-80%) were somewhat superior to that of CEA (71%) but were not significantly different from each other.

DISCUSSION

Breast cancer is a serious medical problem both nationally and world wide. Clearly, a minimally invasive, inexpensive, early diagnostic test would be beneficial to the medical community. In general, tumor markers have not proven useful as screening tests because they are sometimes present in increased concentrations in patients with benign disease and absent in patients with in situ cancer (Roulston and Leonard, 1993; Wu and Nakamura, 1997). However, many tumor antigens have been useful for diagnosis and for therapeutic monitoring and the detection of recurrent disease. In this study we compared four serologic assays (CA 15-3, CA 27.29, AFP, and CEA) for their efficacy at detecting breast cancer. The AFP assay was automated and the other three assays were performed manually. Despite this difference the within-run and between run precision was comparable for all but the CA 15-3 assay which was slightly elevated, and the linearity of all four assays was excellent. The minimum detectable concentration of analyte was established by assaying the zero calibrator/diluent and determining the mean plus two standard deviations. The results for the CA 27.29 assay were slightly more elevated than for the other three assays. The normal (healthy) adult reference intervals were determined by calculating a 95% confidence interval and shown to be somewhat broader (CA 15-3 and CEA) and narrower (CA 27.29 and AFP) than those reported by the manufacturers. There were no significant gender differences for any of the assays. The assays compared favorably for cost and availability of instrumentation. Two of the assays (CA 15-3 and CA 27.29) were radiolabeled ([I.sup.125]) and therefore had shorter shelflives. The turnaround time varied from 1 hour (AFP) to approximately 7 hours (CA 15-3 and CA 27.29) for the assays.

To compare the diagnostic parameters of these assays, 554 patients were randomly selected and their sera tested. The cutoffs suggested by the manufacturers and the diagnoses given by the attending physicians were used to assign them to the categories of true or false positives and negatives. Predictive values were calculated for breast cancer. The most important finding of this study was the observation that CA 15-3 was superior to CA 27.29 for the detection of breast cancer, exhibiting a diagnostic sensitivity of 63% as compared to 39%. This was an important finding since CA 27.29 is more commonly used than CA 15-3 for the serological detection and monitoring of patients with breast cancer. Indeed, in many labs it has replaced CA 15-3. In a similar study, the results of Lauro and his colleagues (Lauro et al., 1999) supported our findings using a large population of patients with metastatic breast cancer. In a study by Gion and Minone (Gion and Minone, 2001) the authors reported CA 15-3 to be comparable to CA 27.29 at higher concentrations of tumor marker and slightly less sensitive than CA 27.29 at lower concentrations of tumor marker. Frenette and co-workers (Frenette et al., 1994 reported that in their study CA 27.29 was more sensitive than CA 15-3 for a large population of breast cancer patients. Discrepancies between their results and ours could be the result of genetic differences in the patient populations, the stage of the tumors, the presence of pathologic complications and/or the use and type(s) of therapies. The diagnostic specificities were comparable for the four assays in our study and the CA 15-3 assay had the best positive and negative predictive values.

In conclusion, four assays (CA 15-3, CA 27.29, AFP, and CEA) were evaluated using 554 patients seen in a local hospital. CA 15-3 proved to be superior to the other three assays for the diagnosis of breast cancer.

ACKNOWLEDGMENTS

The authors thank Ms Jan Oglesby (Kessler AFB) and Mr. Allen Keely (Fujirebio Diagnostics Inc.) for their assistance. This study was partially supported by the Aubrey Keith and Ella Ginn Lucas Research Award. Fujirebio Diagnostics Inc., Hybritech Inc., and Abbott Laboratories donated the reagents. The gift of sera from Keesler AFB Medical Center and Forrest General Hospital is gratefully acknowledged.

[Graph omitted]

(1.) University of Southern Mississippi, Hattiesburg, MS 39406-5134;

(2.) University of Mississippi Medical Center, Jackson, MS 39216;

(3.) Laurel Clinic for Women, Laurel, MS 39442

(4.) Author for correspondence.

LITERATURE CITED

Abelev, G.I., S.D. Perova, and N.I. Khramkova. 1963. Production of embryonal alpha globulins by transplantable mouse hepatomas. Transplantation 1:174-180.

Beveridge, R.A., D.W. Chan, D. Bruzek, D. Damron, K.R. Bray, P.K. Gaur, D.S. Ettinger, R.C. Rock, M.S. Shurbaji, and F.P. Kuhajda. 1988. A new biomarker in monitoring breast cancer: CA549. J. of Clin. Oncol. 6(12): 1815-1821.

Bray, K.R., J.E. Koda, and P.K. Gaur. 1987. Serum levels and biochemical characteristics of cancer-associated antigen CA 549, a circulating breast cancer marker. Cancer Res. 47:5853-5860.

Burtis, C.A., and E.R. Ashwood. 1996. Tumor markers. Pages 336-350 in C.A. Burtis and E.R. Ashwood, eds. Tietz Fundamentals of Clinical Chemistry, 4th Ed., W.B.Saunders Co., Philadelphia, PA.

Cancer Facts and Figures 2000: Graphical Data. Internet, 2000

Chan, D.W., M. Kelsten, R. Rock, and D. Bruzek. 1986. Evaluation of a monoclonal immunoenzymometric assay for alpha-fetoprotein. Clin Chem 32:1318-1322.

Clarke, C.A., T.P. Whitehead, and A.G.W Whitfield. 1982. Carcinoembryonic antigen and smoking. J Royal College of Phys. London 16 (2):112-113.

Cooper, M.J., C.R. Mackie, D.B. Skinner, and A.R. Moossa. 1979. A reappraisal of the value of carcinoembryonic antigen in the management of patients with various neoplasms. Br. J. Surg. 66:120-123.

Frenette, P.S., M.P. Thirlwell, M. Trudeau, D.M. Thomson, L. Joseph, and J.S. Shuster. 1994. The diagnostic value of CA 27.29, CA 15-3, mucin-like carcinoma antigen, carcino-embryonic antigen, and CA 19-9 in breast cancer and gastrointestinal malignancies. Tumour Biology 15 (5):247-254.

Gion, M., and R. Minone. 2001. CA 27.29: a valuable marker for breast cancer management; a confirmatory multicentric study on 603 cases. Eur. J. Cancer 37 (3):355-63.

Hilkens, J., F. Buijs, and J. Hilgers. 1984. Monoclonal antibodies against human milk-fat globule membranes detecting differentiation antigens of the mammary gland and its tumor. International J. of Cancer 34:197-206.

Hossfeld, D.K., and C.D. Sherman. 1990. International Union Against Cancer: Manual of Clinical Oncology, 5th ed. Springer-Verlag, New York, NY. 389 pp.

Hubbard, E. 1990. Tumor markers. Diagnostic and Clinical Testing 28:13-26.

Key, T., D. Forman, and M.C. Pike. 1998. Epidemiology of cancer. Pages 34-59 in L.M. Franks and N.M. Teich, eds. Introduction to the Cellular and Molecular Biology of Cancer, 3rd ed., Oxford University Press, Oxford, England.

Knight, G.J., G.E. Palomaki, and J.E. Haddow. 1988. Use of maternal serum alpha-fetoprotein measurements to screen for Down's syndrome. Clin. Obstet. Gynecol. 31:306-327.

Kufe, D., G. Inghirami, and M. Abe. 1984. Differential reactivity of a monoclonal antibody (DF3) with human malignant versus benign breast tumors. Hybridoma 3:223-232.

Lauro, S., L. Trasatti, and F. Bordin. 1999. Comparison of CEA, MCA, CA 15-3 and CA 27.29 in follow-up and monitoring therapeutic response in breast cancer patients. Anticancer Res. 19(4C):3511-3515.

Milunsky, A. 1987. The prenatal diagnosis of neural tube and other congenital defects. Pages 300-306 in A. Milunsky, ed. Genetic Disorders in the Fetus: Diagnosis, Prevention, and Treatment. Plenum Press, New York, NY.

Pannall, P., and D. Kotasek. 1997. Cancer and clinical biochemistry, ACB Venture Publications, London. 136 pp.

Reynoso, G., T.M. Chu, D. Holyoke, E. Cohen, T. Nemoto, and J.J. Wang. 1972. Carcinoembryonic antigen in patients with different cancers. JAMA 220 (3):361-365.

Roulston, J.E., and R.C.F Leonard. 1993. Applications to cancer. Pages 37-51 in J.E. Roulston and R.C.F. Leonard, eds. Serological Tumour Markers. Churchill Livingstone, New York, NY.

Safi, F., I. Kohler, and E. Rottinger. 1991. The value of the tumor marker CA 15-3 in diagnosis and monitoring breast cancer: a comparative study with carcinoembryonic antigen. Cancer 68:574-580.

Wu, J., and R. Nakamura. 1997. Human circulating tumor markers. American Society of Clinical Pathologists, Chicago, IL. 263 pp.