Original Source
..
Tumor Markers in Breast and Colorectal Cancer

POSTED 05.17.1996
..
Overview
..
Purpose

The primary objective was to determine clinical practice guidelines for the use of tumor marker tests in the prevention, screening, treatment, and surveillance of breast and colorectal cancers. These guidelines are intended for use in the care of patients outside of clinical trials.
..
Methods

Six tumor markers for colorectal cancer and seven for breast cancer were considered. They could be recommended or not for routine use or for special circumstances. In general, the significant health outcomes identified for use in making clinical practice guidelines (overall survival, disease-free survival, quality of life, lesser toxicity, and cost-effectiveness) were used when data were available. Expert consensus was used to inform the recommendations for use of tumor markers when published evidence was insufficient. A computerized literature search was performed using Medline. In addition to reports collected by individual panel members, all articles published in the English-speaking literature from January 1989 to April 1994 were collected for review and distributed to all members of the Panel. Values for use, utility, and levels of evidence were assigned by the expert reviewers and approved by the Panel. Tumor markers were assigned benefit if they had prognostic or predictive value that would lead to the favorable outcomes listed above. Harms considered were inappropriate disease management, and excess cost without definable benefit. Costs were considered but were never the sole determinant of a recommendation.
..
Results and Conclusion

For colorectal cancer, it is recommended that carcinoembryonic antigen (CEA) levels be measured preoperatively if it would change surgical management. It is recommended that CEA levels be monitored every 2 to 3 months for 32 years, if resection of liver metastasis would be clinically indicated. The data are insufficient to recommend the routine use of lipid-associated sialic acid (LASA), CA 19-9, DNA index, DNA flow cytometric proliferation analysis, p53 tumor suppressor gene, and ras oncogene. For breast cancer, estrogen receptor and progesterone receptor are recommended to be measured on every primary specimen, but on subsequent specimens only if it would lead to a change in management. The data are insufficient to recommend the routine use of DNA index, DNA flow cytometric proliferation analysis, CA 15-3, CEA, c-erbB-2, p53 or cathepsin-D. In the absence of readily measurable disease, CA 15-3 and CEA levels can be used to document treatment failure. New markers and new evidence will be evaluated by annual update of these guidelines.
..
1997 Update of Recommendations for the Use of Tumor Markers in Breast and Colorectal Cancer
..
J Clin Oncol 14:2843-2877 1996 by American Society of Clinical Oncology.
..
Tumor Markers may be used to indicate the risk, presence, status, or future behavior of cancer. New markers are frequently introduced into clinical practice without rigorous analysis, with the assumption that any information available to the clinician will help the patient. As is the case for pharmaceutical agents, however, inappropriate use of tumor markers to make clinical decisions may result in worse outcomes than if the clinician had made the decision in the absence of tumor marker information. Few guidelines or criteria have been established to determine if/and/or when a tumor marker should be used routinely in the clinic.
..
The American Society of Clinical Oncology (ASCO) recognized the ambiguity that exists between available data and clinical use of tumor markers, and convened a Tumor Marker Panel that used standard methods1 to establish practice guidelines for the use of tumor markers in breast and colon cancer. In this report, the Panel presents these guidelines (Table 1. and the methods used to develop them.
..
These guidelines are not intended to replace physician judgment with respect to particular patients or special clinical situations. The Panel recognizes that there will be certain instances in which the use of a tumor marker in a manner not defined or recommended by this Panel may be helpful in clinical decision-making. ASCO considers adherence to these guidelines to be voluntary. The ultimate determination regarding application of these guidelines should be made by the physician in light of the individual circumstances prescribed by a particular patient's case. Furthermore, these guidelines describe the use of tumor markers in routine clinical practice. These guidelines should not be applied in the context of clinical trials, in which the use of tumor markers may be prospectively dictated or may even be the subject of the investigation.
..
Process and Methods

Panel Composition
..
The Panel was composed of experts in medical oncology, surgical oncology, laboratory medicine, medical economics, and medical decision-making/clinical guideline development Appendix. The Panel included established investigators who had studied the application of tumor markers to the management of breast and colon cancer. Two patient representatives were also included on the Panel.
..
Guideline and Conflict of Interest Review

The content of the guidelines and the manuscript were reviewed and approved by the Health Services Research Committee and by the ASCO Board. All members of the Panel disclosed potential conflicts of interest.
..
Revision Dates

The Panel chose to review the following tumor markers with data through April 1994, with inclusion of subsequent published data; the guidelines will be updated annually by the Panel chairman and Panel members designated by the chairman beginning in 1997, with a re- examination of all the primary evidence every 3 years.
..
Tumor Markers Evaluated During Guideline Preparation

All commonly used circulating and tissue-based markers in the care of breast and colorectal cancer patients were evaluated Table 2).
..
Literature Review and Tumor Marker Utility Grading System

A computerized literature search was performed using Medline. Key words included the disease (colorectal or breast carcinoma) and the marker in question. In addition to reports collected by individual Panel members, all articles published in the English-speaking literature from January 1989 to April 1994 were collected for review and distributed to all members of the Panel.
..
The Panel made every effort to develop these guidelines using evidence-based deliberations. Wide variation in established practice, with limited data for or against the use of a particular marker, made some guidelines difficult to formulate. In the absence of data from well-performed studies, current use in clinical practice was also considered. The latter was, however, never regarded as the sole criterion for a recommendation if published data contradicted current practice patterns.
..
As in all issues of medicine, guidelines could not always be made with certainty. Indeed, in the case of tumor markers, uncertainty arose from two separate sources: (1) the prognostic values based on the technical and correlative aspects of the marker itself; and (2) the therapeutic value based on the efficacy of therapeutic options. For example, a marker might be reliable in predicting a high risk of relapse after primary treatment; however, if data demonstrating the value of adjuvant systemic therapy in that situation were not available, knowledge of prognosis was considered to be of limited value in patient management.
..
To initiate the process, the Panel considered several issues that reflected the sources of uncertainty. These included definitions of the marker, assay(s) to detect the marker, potential utilities for the marker in either colon or breast cancer, correlation of the marker with these utilities, and whether knowledge of the marker data results in a therapeutic decision that has a favorable impact on one of five outcomes: overall survival, prolongation of the time during which the patient is free of detectable disease (disease-free survival), toxicity, improvement in quality of life, and/or improvement in cost-effectiveness of care.2
..
Uses

The criteria to accept or reject a tumor marker for clinical purposes depend on whether knowledge of marker levels in an individual patient can be used reliably to make therapeutic decisions that will improve outcome. Detection and/or monitoring of neoplastic changes might be useful in several separate situations, which the Panel designated "uses." These uses include screening for a new primary cancer, differential diagnosis, predicting prognosis (of either newly diagnosed primary or metastatic disease), and monitoring clinical course.
..
Markers that predict prognosis can be subdivided into those that predict outcome independent of treatment effects (prognostic factors), and those that predict response to or benefit from therapy (predictive factors). Clinical course can be monitored for two reasons: (1) to determine if a patient free of detectable disease after primary and/or adjuvant therapy has evidence of impending relapse; or (2) to determine the status (progression, regression, or stable disease) of a patient with detectable disease, particularly in response to a specific therapy. Ultimately, markers were recommended for use only if published data confirmed a favorable impact (or provided utility) on one of the outcomes relevant to one of the uses.
..
Level of Evidence

A modification of the scale developed by the Canadian Task Force on the Periodic Health Examination was used.3 The highest level of evidence (level I) was from meta-analysis or large, high-powered concurrently controlled studies in which the primary objective of the trial design was to test the utility of the marker. Such trials and analyses have, however, rarely been performed to evaluate tumor markers. Level II evidence was obtained from prospective clinical trials designed to test a therapeutic hypothesis in which tumor marker evaluation was a secondary, but prospectively described objective. Level III studies were retrospective, but characterized by large size (> 200 patients per subgroup) and/or by inclusion of multivariate analysis. Level IV evidence was considered less reliable than level III evidence, either because the study was smaller or a multivariate analysis was not provided. Finally, level V evidence was derived from studies that were small, retrospective, and not designed to correlate marker results with clinical outcome.
..
Consensus Development Based on Evidence

Following specific formats (see below), Panel members analyzed the available literature and drafted the recommended guidelines. Each guideline was then very carefully evaluated by the entire group, and votes were taken on each proposed guideline. The guidelines were circulated in draft form, and all members had an opportunity to comment on the final wording.


..
Guideline Presentations

Statement of Guideline

For each marker evaluated, the practice guideline has been presented as concisely as possible. The guideline statement reflects the consensus of the Panel, which was developed from a literature-based review and lengthy discussions of the implications of both favorable and unfavorable statements.
..
Marker Definition

Following the guideline statement, the marker in question was precisely defined, including a description of the molecule or process of interest (eg, DNA, RNA, protein, antibody, etc) and the alteration that was detected in that molecule or process (eg, amplification, mutation, deletion, overexpression, elevated soluble levels, etc). These descriptions include an explanation of the normal function of the marker, if known.
..
Methodology

It is not the intent of the Panel to dictate preferred methods of assays and analysis, but rather to evaluate the results for established assays for specific markers. For each guideline statement, brief descriptions of the most widely used assays and the reagents incorporated into these assays include discussions regarding whether variations in assays would be expected to alter conclusions. These sections also include a discussion regarding collection techniques, if they are important for clinical correlation (for example, whether the assay must be performed in rapidly frozen tissue, or if it can be performed using paraffin-embedded, fixed tissue).
..
This section also includes a discussion of interpretative techniques, and whether confounding factors have been considered in marker analysis. These discussions include definitions of cutoffs that distinguish normal from abnormal, or nonelevated from elevated levels.
..
Literature Review

Relevant literature addressing specific utilities of the marker for the disease in question (colorectal or breast cancer) was identified by use of a comprehensive literature search, as described above. Following each guideline statement, a full description of the number of articles reviewed, the number of articles considered informative, and reasons for excluding articles from the analysis are presented.
..
Analysis

For each potential use, the Panel determined a utility score of the marker for one or more potential outcomes. The Panel did not comment on uses for which few if any studies were performed, for example, the use of HER-2/neu in screening for breast cancer. Although the guideline comments may address correlation of a marker with a biologic process, guidelines recommending the marker were only made if the marker can be used to reliably make clinical practice decisions that result in favorable changes in one of the five clinical outcomes for any given use: survival, disease-free survival, quality of life, toxicity, or cost.
..
Areas for Future Research

For each guideline, if appropriate, areas of research, including both basic and clinical investigations, are proposed and discussed.


..
Carcinoembryonic Antigen as a Marker for Colorectal Cancer

Guidelines

1a. Carcinoembryonic antigen (CEA) is not recommended to be used as a screening test for colorectal cancer.
..
1b. CEA may be ordered preoperatively in patients with colorectal carcinoma if it would assist in staging and surgical treatment planning. Although elevated preoperative CEA (> 5 mg/mL) may correlate with poorer prognosis, data are insufficient to support the use of CEA to determine whether to treat a patient with adjuvant therapy.
..
1c. If resection of liver metastases would be clinically indicated, it is recommended the postoperative serum CEA testing may be performed every 2 to 3 months in patients with stage II or III disease for 3 2 years after diagnosis. An elevated CEA level, if confirmed by retesting, warrants further evaluation for metastatic disease but does not justify the institution of adjuvant therapy or systemic therapy for presumed metastatic disease.
..
1d. Present data are insufficient to recommend routine use of the serum CEA alone for monitoring response to treatment. If no other simple test is available to indicate a response, CEA should be measured at the start of treatment for metastatic disease, and every 2 to 3 months during active treatment. Two values above baseline are adequate to document progressive disease even in the absence of corroborating radiographs. CEA is regarded as the marker of choice for monitoring colorectal cancer.
..
Marker Definition

CEA is a serum glycoprotein with a molecular weight of 180 kd that is one of at least 19 related molecules that are members of the immunoglobulin gene superfamily. As such, CEA functions as a homotypic intercellular adhesion molecule that promotes the aggregation of human colorectal carcinoma cells.4 CEA may facilitate metastasis of colorectal cancer cells to liver and lung.
..
CEA is a normal cell product that is overexpressed by adenocarcinomas, primarily of the colon, rectum, breast, and lung. Smokers have a higher circulating CEA concentration than nonsmokers, but there are no significant effects of age, sex, or race on the normal range. The liver is the major site for clearance of CEA. Moderate to significant elevations of serum CEA can be observed in a variety of chronic and acute inflammatory diseases, including alcoholic cirrhosis, cholelithiasis, obstructive jaundice, cholangitis, liver abscess, emphysema, bronchitis, gastric ulcer, gastritis, diverticulitis, diabetes, and collagen vascular diseases. CEA level elevations are not unique to colorectal cancer, and are observed in carcinomas of the breast, lung, stomach, pancreas, cervix, bladder, kidney, thyroid, liver, lymphoma, and melanoma.
..
Methodology

The current CEA test is an immunoradiometric assay that may use a double monoclonal or monoclonal-polyclonal antibody format. In general, all of the CEA test methods approved by the United States Food and Drug Administration have comparable assay characteristics, with the interassay test precision (measured as the coefficient of variation) ranging from 10% to 12% for mean values ranging from 3.0 to 5.0 ng/mL and 6% to 10% for CEA concentrations of 10 to 100 ng/mL. Similarly, all CEA test methods give a normal range of less than 3.0 ng/mL for nonsmokers and less than 5.0 ng/mL for smokers.5 It is advisable to use the same CEA method consistently for a given patient because different CEA test methods do not give equivalent CEA test values for individual samples.
..
Poorly differentiated tumors may not produce CEA.6 In one study, 14 of 17 patients with poorly differentiated tumors had a CEA value less than 2.5 ng/mL.7 Tumors that are moderately differentiated secrete more CEA than well-differentiated tumors.
..
Literature Review and Analysis

Screening.Several studies have evaluated the potential of serum CEA to detect colorectal cancer at an early stage. Each study provides level III/IV evidence. In each study, a CEA value greater than 5 ng/mL prompted a diagnostic evaluation. The colon cancer detection rate ranged from 0.1% to 4%.8,9 At a colon cancer prevalence of 1/1,000, with a test sensitivity of 40% and specificity of 90% (common for stage A and B cancers), there would be 250 false-positive tests for every one person with cancer.5
..
The Panel considered the following evidence in formulating a recommendation: (1) while screening may occasionally detect individuals with resectable disease, the false-positive rate is unacceptable and costs of a screening program are prohibitive. (2) There are no data that CEA screening provides a better survival, quality of life, or lower costs in a population compared with no screening. Therefore, the ASCO Panel does not recommend CEA testing for colorectal cancer screening.
..
Preoperative prognostic indicator.The ASCO Tumor Marker Panel recognizes there is no literature proving an impact of preoperative CEA testing on the primary outcomes, ie, survival, quality of life, or cost-effectiveness. Nevertheless, the Panel used the following rationale in making its recommendation: (1) a preoperative CEA value predicts a higher rate of recurrence and colon cancer mortality compared with patients with normal levels (level III); and (2) the expert Panel considered the prognostic value of CEA important in this circumstance, and the patient or physician may be efit from the extra vigilance resulting from an elevated value. Some surgeons may change their operative approach based on the CEA value. Therefore, a CEA may be ordered preoperatively in patients with colorectal carcinoma if it will assist in staging and surgical treatment planning. There is a strong association between serum CEA elevation and stage. In one study, the CEA was elevated in 4%, 25%, 44%, and 65% of patients with Dukes' A, B, C, and D disease, respectively.10
..
Preoperative CEA is a prognostic factor independent of Dukes' stage in patients with colorectal carcinoma. Nineteen articles investigated whether CEA was a useful preoperative prognostic indicator (all lev l III/IV).10-28 The majority of the preoperative CEA studies showed the marker was a useful prognostic indicator that remained significant even when Dukes' stage and histologic grade were factored into a multivariate analysis. Some of the results were conflicting, especially on the issue of the prognostic significance of an elevated preoperative CEA for Dukes B versus C, or colon versus rectal cancer.12,15,17,29
..
In the National Surgical Adjuvant Breast Bowel Project (NSABP) study of 945 patients, a preoperative CEA greater than 2.5 ng/mL increased the risk of recurrence by 1.62-fold in Dukes' B patients; if the preoperative CEA value was greater than 10 ng/mL, the risk was increased by 3.25-fold.17 Wanebo et al10 reported the time to recurrence was shorter if the CEA was elevated, 13 versus 23 months for a CEA level greater than 5 ng/mL or less than 5 ng/mL, respectively; stage-specific survival was higher in patients with CEA level less than 5 ng/mL.10 Because 35% of Dukes' D patients will have a CEA value less than 5 ng/mL, and 25% to 45% of Dukes' B and C patients will have a CEA value greater than 5 ng/mL, a preoperative CEA value cannot be reliably used to order or avoid further staging procedures.5 A normal preoperative CEA level does not assure that CEA will remain normal. In a study by Zeng et al12 of 140 patients with normal preoperative CEA (< 5 ng/mL), 32 developed a recurrence and 44% of these patients had a CEA level greater than 5 ng/mL at recurrence.
..
No studies have shown patients benefit from adjuvant therapy based solely on an elevated preoperative CEA value.
..
Postoperative prognostic indicator.Following complete surgical resection of colon carcinoma, elevated plasma CEA levels should return to normal within 4 to 6 weeks.5,30 In one report of 90 patients with complete tumor resection, all but six had a decrease in postoperative CEA (level IV). These six were later found to have tumor relapse.25 Patients with a persistently elevated CEA level postoperatively may indicate an incomplete surgical resection or metastatic disease. Boey et al22 reported that 23 of 33 patients (70%) with an early CEA elevation (> 2 weeks after operation) had a recurrence, compared with 23 of 113 patients (21%) without an elevation (level IV).
..
An elevated postoperative CEA is an adverse prognostic indicator. Yet knowledge of an elevated postoperative CEA does not alter survival, quality of life, or costs, and does not support the use of any intervention. Because none of the major outcomes varies as a result of this information, the test should not be drawn in the immediate postoperative period.
..
Indicator of asymptomatic recurrence.A persistent increase in the serum CEA concentration frequently heralds clinical evidence of recurrent disease. Several studies have examined the usefulness of CEA to detect asymptomatic recurrence.
..
When an elevated CEA is detected in a patient in whom it was previously normal, the study should be repeated before a more extensive investigation. In one study in which 45 patients had false-positive values, 27 had an elevation in only one CEA value, which then returned to normal in the next of several CEA determinations.31 If there continues to be an elevation or if there is an increase in the slope of CEA elevation, liver function tests and renal function must be checked because abnormalities in either can elevate CEA level. No studies show a benefit to providing chemotherapy for an elevated CEA alone in the absence of measurable disease.
..
CEA may become elevated in patients undergoing adjuvant chemotherapy without evidence of cancer recurrence.
..
More than 50 articles about postoperative CEA monitoring were reviewed.14,18,19,21,22,24,25,29-75 Most studies were level III/IV evidence. One study was a randomized, controlled study (level II)75; another was a meta-analysis of seven nonrandomized studies.67 One of the best review articles stated the sensitivity for recurrence disease was approximately 80% (range, 17% to 89%); the false-positive rate was approximately 30% (range, 9% to 66%).5
..
The value of resecting recurrent cancer is controversial. The discussion is complicated by the fact that most articles assume early detection leads to a better patient outcome; few studies specifically test this relationship. The potential implications of finding early metastatic disease is as follows: of 100 patients presenting with a new case of colorectal carcinoma, 70 can be resected with curative intent. Twenty-five of the 70 will have recurrent disease, and three of them can be cured by a second-look procedure, partial hepatectomy, or other surgery to resect an isolated metastasis.76 From a population perspective, aggressive second-look and hepatic surgery for 3% of 150,000 new colorectal cases a year amounts to 4,500 cases potentially cured by resection if detected at an appropriate time point. The critical question is, "Does the information provided by serial CEA monitoring contribute significantly to this resection rate and survival or do other unknown factors play a role?"


..
Lipid-Associated Sialic Acid as a Marker for Colorectal Cancer

Guideline

2. Present data are insufficient to recommend lipid-associated sialic acid (LASA) for screening, diagnosis, staging, surveillance, or monitoring treatment of patients with colorectal cancer.
..
Marker Definition
..
LASA is a complex marker that measures the amount of sialic acid in serum, including sialic acid from gangliosides that are bound to serum proteins as well as sialic acid contained in circulating glycoproteins, especially acute phase reactants like a1-acid glycoprotein. LASA can be elevated in serum from patients with any number of different neoplasms.
..
Methodology

Sialic acid is extracted from serum. The value of the assay is limited because of the different methods across studies for extracting sialic acid residues. In addition, Tautu et al82,83 suggest that the ratio of LASA to total protein is more important than the absolute value of LASA, implying that the amount of lipid needs to be normalized by the amount of total protein in the serum. This results in differences in measured values among investigators and leads to problems with cutoffs. As reviewed by Schutter et al84 glycoprotein-bound sialic acid may be the predominant contributor to LASA rather than lipid-bound sialic acid. LASA is first isolated from serum by chloroform: methanol extraction and the lipids precipitated by phosphotungstic acid or tripotassium citrate or concentrated by evaporation with measurement by resorcinol.84 Because more than 50% of a1-acid glycoprotein may be extracted with the lipid fraction, the sensitivity for malignancy may be good, but the specificity is compromised in many malignancies, where acute-phase proteins are increased as a result of nonspecific inflammation. This also decreases the utility of LASA as a marker for monitoring disease progression.
..
Literature Review and Analysis

There is no practical information concerning outcome and the use of LASA. There are several articles that describe the use of LASA in the diagnosis of colorectal cancer and its association with tumor-node-metastasis (TNM) stage. There is one small study that demonstrates a decrease in LASA with resection of primary colorectal cancers.85 However, Shahangian et al86 demonstrated that patients with colorectal polyps and colorectal carcinoma both had elevated LASA levels. The levels returned to baseline after removal of either polyps or carcinomas. However, these studies are all small, single-institution trials involving fewer than 100 patients. Furthermore, there are no attempts to study prognosis or other measures of outcome. There are no studies that address the use of LASA in monitoring response to therapy. Finally, LASA appears to perform similarly to CEA in level IV utility studies.87,88
..
LASA is a serum cancer marker that has not been widely accepted for use in the detection or prognosis of colorectal carcinoma. When compared with CEA, LASA did not provide information to improve screening, differential diagnosis, prognosis, or monitoring for relapse of either primary metastatic disease. Because LASA represents a combination of glycoprotein and glycolipid markers, it is possible that measurement of the specific markers may be of greater utility.
..
Areas for Future Research

Future research in this area should be directed to improved test methodology with characterization of the substances that are measured. Rather than measure crude levels of sialic acid, specific substances such as sialylated or glycolyl sialic acid derivatives may provide more useful markers


..
CA 19-9 as a Marker for Colon Cancer

Guideline

3. Present data are insufficient to recommend CA 19-9 for screening, diagnosis, staging, surveillance, or monitoring treatment of patients with colorectal cancer.
..
Marker Definition

CA 19-9 assay measures a tumor related mucin that contains the sialylated Lewis-a pentasaccharide epitope, lacto-N-fucopentaose II.89 CA 19-9 is produced by adenocarcinomas of the pancreas, stomach, gall bladder, colon, ovary, and lung, and it is shed into the circulation.
..
Methodology

CA 19-9 is measured with a double monoclonal immunoradiometric assay, using monoclonal antibodies raised against the SW1116 cell line.90 Various commercial assays use the same monoclonal antibodies. Human anti-mouse and heterophile antibodies can interfere with the assay. The upper limit of normal for healthy subjects has been defined by the cutoff value of 37.0 U/mL.90,91 CA 19-9 has become an established marker for pancreatic cancer,91-93 but it must still be regarded as a research test for colorectal cancer.
..
Literature Review and Analysis

Screening. Numerous studies have addressed the potential utility of CA 19-9 in adenocarcinoma of the colon and rectum. The reported incidence of elevated serum CA 19-9 in colorectal cancer ranges from 20% to 40%.94-96 The incidence of elevated CA 19-9 is stage- related, with the highest sensitivity occurring in patients with metastases.97-100 However, the sensitivity of CA 19-9 was always less than that of the CEA test for all stages of disease.94-99 The false-positive rate (> 37.0 U/mL) is 15% to 30% in patients with nonneoplastic diseases of the pancreas, liver, and biliary tract.100,101 Consequently, CA 19-9 cannot be used for screening asymptomatic populations.
..
Indicator of asymptomatic recurrence. Only a few studies have addressed the use of CA 19-9 in monitoring patients post-primary therapy.102-104 Significant postsurgical decreases are observed for CA 19-9, but these decreases have not been correlated with survival or disease- free interval.102 Filella et al103 have reported their experience with CA 19-9 in monitoring disease recurrence. In 370 patients with colorectal cancer, progressive increases of serum CA 19-9 were observed in 48% of 96 patients with relapse. However, the CA 19-9 abnormality preceded clinical manifestation of the disease in only 25% of the cases and provided a median lead time of only 3 months. CEA was abnormal in 84% of the relapsed cases and preceded clinical detection of recurrence in 75%. CA 19-9 and CEA in combination did not improve the performance of the CEA test used alone.103
..
Monitoring response to treatment. Kouri et al104 compared CEA and CA 19-9 for predicting response to chemotherapy in 85 patients. Decreases in CEA more accurately reflected response to therapy than did decreases of CA 19-9. The pretreatment CA 19-9 value was, however, an important prognostic factor. Median survival was 30 months for patients with normal CA 19-9 values and 10.3 months for patients with elevated CA 19-9 values.

Serum CA 19-9 elevations may be observed in as many as 20% to 40% of patients with late- stage colorectal cancer, but cannot be regarded as a diagnostic aid nor can it be used to detect early stage disease. Progressive increases of the marker may signal disease progression in 25% of the patients who express the CA 19-9 marker, but this monitoring provides only a minimal lead time of 1 to 3 months. Monitoring with CA 19-9 has not been shown to improve the management of patients with colorectal cancer. The serum CA 19-9 level does not add significant information to that provided by CEA, which is currently regarded as the marker of choice for this neoplasm.
..
Areas for Future Research

A recent report by Shimono et al105 suggested a prognostic utility of tissue CA 19-9 based on an assessment of the immunochemical staining pattern. The 5-year survival rate for patients with no CA 19-9 staining or a cytoplasmic staining pattern ranged from 68% to 76%, whereas it was only 26% for patients when tissue showed a stromal staining pattern. The potential prognostic utility requires validation. Other than this possible application, the CA 19-9 test does not appear to provide any data relevant to clinical decision making in the management of patients with colorectal cancer.


..
DNA Ploidy or Flow Cytometric Proliferation Analysis as a Marker for Colon Cancer

Guideline

4a. Present data are insufficient to recommend DNA flow cytometrically derived ploidy (DNA index) for the management of colorectal cancer.
..
4b. Present data are insufficient to recommend the use of DNA flow cytometric proliferation analysis (% S-phase and related analyses) for the management of colorectal cancer.
..
Marker Definition

DNA diploid tumors are those in which a single peak containing an amount of DNA similar to normal control cells is generated by flow cytometry. DNA aneuploid tumors have additional peaks on DNA histogram, presumably representing cells containing more or less nucleic acid than is found in 46 normal chromosomes. A more quantitative method of expression is the DNA index (DI), which is the ratio of the mean tumor sample G0/G1 DNA content divided by the mean G0/G1 DNA content of normal diploid reference cells. The greater the deviation of the DI from 1, the more "aneuploid" the tumor.
..
Tumor cell proliferation is defined as the percent of cells in proliferation phase(s) of the cell cycle (S phase or S + G2M phases). In DNA diploid tumors, the presence of only one population of cells should lead to a single proliferation determination. In DNA aneuploid tumors, the multiple peaks each represent a different cellular population; however, proliferation results are generally reported from only one peak, complicating the interpretation of the results. Another problem of definition is that of high versus low S phase; these are frequently defined retrospectively or arbitrarily as being above or below the median for all samples assayed. Such post hoc definitions may introduce bias106 and should in the future be replaced by laboratory- specific, prospectively determined, biologically relevant cutoffs. Reporting the data in tertiles (low, intermediate, or high S phase) may be preferable.107,108
..
Methodology

Flow cytometric DNA and cell cycle analysis can be performed on fresh/frozen tissue or on tissue from paraffin blocks. For fresh tissues, cells or nuclei are obtained via mechanical and or enzymatic disruption of the specimen, which are then washed and resuspended. To obtain material from archival material, the paraffin blocks are cut into sections, dewaxed, and then rehydrated. The individual cells are separated mechanically/enzymatically and the resulting individual cells or naked nuclei are suspended. Care must be taken to differentiate cells obtained from tumor from those obtained from adjacent nonmalignant tissue. After treatment with nucleases to digest single-strand nucleic acids, the cells or nuclei are stained with a fluorescent dye that binds stoichiometrically to the remaining double-stranded DNA. The stained cells or nuclei are passed through an excitation laser with the resultant fluorescence emitted by the DNA-bound dye detected and plotted as a histogram. The intensity of the measured fluorescence should be directly proportional to the amount of DNA present in the cell. DNA content is measured in thousands of individual cells. If a single sharp peak of DNA content is present, the tumor is DNA diploid; multiple discrete peaks are seen in DNA aneuploid tissues. If a large enough number of cells is analyzed, smaller peaks corresponding to DNA content in S and G2M phases will appear adjacent to the predominant peaks that reflect content in G0/G1. For each cell population/predominant peak, these smaller peaks can be used to calculate %S or %S + G2M.
..
Literature Review and Analysis

Postoperative prognostic indicator. The review of DNA ploidy in colorectal cancer included publications in English that appeared between January 1986 and April 1994 that studied 3 50 patients with survival data. Several series were reported more than once.109-114 Only the most recent or inclusive article was considered. Series that evaluated particular subsets of patients, such as colorectal cancer in inflammatory bowel disease or familial cases, were not included.115,116 There were few articles using alternative techniques evaluating DNA ploidy, such as image analysis, that met the criteria for inclusion, and no conclusions can be made as to their utility.
..
Thirty-one articles were included in the DNA ploidy analysis.109-111,114,117-143 Three of the 31 were limited to rectal cancer patients only. Nine studies were prospective using frozen tissue; 20 derived their material from paraffin blocks; two studies contained a mixture of frozen and paraffin-derived specimens. Most of the articles analyzed selected cases from a defined time period; six articles reported on all (consecutive) patients in a time period; three articles looked at selected patients on randomized clinical trials of adjuvant therapy.

Of the 31 studies, 24 found that patients with DNA aneuploid tumors have a statistically significantly (P < .05) worse survival than those with DNA diploid tumors; seven articles found no correlation between DNA ploidy and survival. In 13 of 24 positive studies, it was an independent prognostic feature in multivariate analysis; in seven, it had no independent predictive power; in four, it was not evaluated. Seven studies looked at the subset of patients with Dukes' B (or A plus B) stage colorectal cancer; three of seven found DNA ploidy to be prognostic within the stage; three found no such association; and in one, it was not determined.
..
Seven papers were included in the DNA proliferation analysis (four, %S phase; two, %S + G2M; one, both %S and %S + G2M).112,119,121,125,127,130,132 Four of the seven studies found patients with a high proliferative index to have a significantly worse survival than those with a low index of proliferation112,127,130,132; three found no difference. In three of the four positive studies, proliferative index had independent significance in a multivariate analysis. Two of three studies looking at the subset of patients with early-stage disease found it to be an independent prognostic feature.
..
Although the evidence is almost entirely from retrospective studies, DNA ploidy is of prognostic significance in patients with newly diagnosed colorectal carcinoma. Unfortunately, the data are contradictory as to whether the prognostic information is independent of other recognized predictors, particularly stage. For this reason, the use of DNA ploidy to decide whether to administer adjuvant therapy after curative surgery cannot be recommended.

The situation is more complicated for flow cytometry-derived proliferation data in colorectal carcinoma. The literature is split as to its prognostic value; however, the articles from randomized clinical trials that also contain the largest sample sizes (694 and 194 patients), found %S or S + G2M phase to be prognostically significant in both univariate and multivariate analysis.112,127 Although proliferation analysis should not be used at this time for adjuvant therapy decisions, further studies of the issue are warranted.
..
Areas for Future Research

Retrospective analysis of subgroups of patients with colorectal cancer on cooperative group surgical adjuvant trials suggest that DNA ploidy and %S/%S + G2M have prognostic significance.127 These findings have led to an analysis of paraffin blocks from additional patients in the completed trials. Hopefully, when the reanalysis is completed, there will be sufficient data to determine if either or both may be used to determine which patients with stage II (Dukes' B2) large bowel cancer should be offered adjuvant therapy. Furthermore, it is possible the new data may have sufficient statistical power to conclude whether adjuvant therapy changes the natural history of the disease in patients predicted by flow cytometry to do poorly. Unfortunately, even if positive, the conclusions will be questioned because of the retrospective and incomplete/selective nature of the patient data analyzed. A prospective study should be performed in the setting of a randomized adjuvant trial in colorectal cancer. Such is the case in the new Intergroup colon cancer trial of adjuvant chemotherapy with and without postoperative infusion of fluorouracil. The design of the trial will insure the collection and study of tumor blocks of patients with all stages of disease. The study has the potential to answer definitively whether ploidy or proliferation analysis as determined by flow cytometry has independent prognostic significance.


..
p53 as a Marker for Colorectal Cancer

Guideline

5. Present data are insufficient to recommend the use of p53 expression or mutation for screening, diagnosis, staging, surveillance, or monitoring treatment of patients with colorectal cancer.
..
Marker Definition

p53 is a tumor suppressor gene on the short arm of chromosome 17 that encodes a protein that is important in the regulation of cell division. The p53 gene product appears to regulate transcription of several other genes. The full role of p53 in the normal and neoplastic cell is unknown. There is evidence that the gene product is important in preventing the division of cells containing damaged DNA. p53 gene deletion or mutation is a frequent event along with other molecular abnormalities in colorectal carcinogenesis.
..
Methodology

There are two major techniques used for p53 analysis in colorectal cancer. These include DNA analysis to detect a variety of single base mutations in the p53 gene and immunohistochemical techniques to detect abnormal cytoplasmic and/or nuclear accumulation of p53 protein. The normal p53 protein is not accumulated in the cytoplasm and nuclei of cells. However, when the gene is mutated, accumulation of protein will occur and immunohistochemical techniques using a variety of antibodies may detect this accumulation.
..
p53 analysis has been conducted both on fresh tissue and on paraffin-embedded tissue. Several issues are of concern in the interpretation of the results of p53 analysis in colorectal cancer. First, when mutations are assessed, it is not clear which of the various single base mutations impact on p53 function. Therefore, the possibility exists of interpreting a mutation as important when that mutation has no functional impact on carcinogenesis or prognosis. Secondly, when immunohistochemical techniques assessing p53 protein accumulation in tissue are performed, a variety of antibodies have been used without standardization among antibodies, and the methods used for interpretation of positivity vary among various studies.
..
Literature Review and Analysis

Postoperative prognostic indicator. There are 10 articles140,144-152 correlating p53 abnormality with prognosis in cases of resected large bowel cancer. All articles were retrospective in design. Five of these studies used multivariate analysis techniques. Seven of the studies showed statistically decreased overall survival and increased relapse rate in patients in whom tumors exhibited abnormal p53 expression. In a well performed retrospective analysis150 that used immunohistochemical techniques to measure p53 protein product in tumors from 91 patients with colorectal cancers, 68% of tumors were demonstrated to be positive for p53 protein. However, there was no impact of cellular p53 protein accumulation on survival (P = .4). Cytoplasmic p53 was not differentiated from nuclear p53. Zeng et al152 from the Memorial Sloan-Kettering Cancer Center have reported p53 protein accumulation in nuclei as an independent negative prognostic factor in a retrospective series of patients with stage III colon cancer with normal preoperative CEA levels. Cytoplasmic p53 expression was not prognostic.
..
The literature on p53 abnormality and prognosis in colorectal cancer suffers from a paucity of reported data and the use of a variety of techniques in assay and statistical analysis in the small numbers of cases analyzed. For these reasons, p53 analysis cannot be recommended as a routine approach to assisting in the management of patients with colorectal cancer.
..
Areas for Future Research

Standardization of the immunohistochemical assay for p53 is required. The standardization of antibodies to be used and standardization of measurement to define positivity and negativity is needed. In the genetic assessment of p53, there needs to be a definition of the importance to the function of p53 protein of the various mutations that have been defined.
..
Finally, prospective studies assessing the prognostic value of p53 abnormalities and other abnormalities such as deleted in colon cancer (DCC) allelic deletion will be required before the prognostic value of these genetic abnormalities can be firmly established. Appropriately designed studies using multivariate analysis techniques will be critical to establishing the value of these techniques.


..
ras as a Marker for Colorectal Cancer

Guideline

6. Present data are insufficient to recommend the use of the ras oncogene for screening, diagnosis, staging, surveillance, or monitoring treatment of patients with colorectal cancer.
..
Marker Definition

The ras proto-oncogenes are normal cellular components. These genes encode 21,000-d proteins that bind to the inner surface of plasma membranes. The ras proteins are thought important for transduction of signals required for proliferation and differentiation. Mutation or overexpression of ras results in activated p21. Abnormalities of p21 have been associated with a variety of human cancers including a majority of colorectal cancers.
..
Methodology

ras oncogene activity is assayed in human colon cancer tissue by two methods. First, molecular genetic assays use polymerase chain reaction (PCR) techniques to detect point mutations at codons 12, 13, 31, 61 . A second means of assessing abnormality in ras oncogene is to assay expression of the ras p21 protein. This is done by immunohistochemical techniques and a variety of antibodies, including ras 11, DWP, R256, and E184, have been used in immunohistochemical assays appropriately fixed fresh or paraffin-embedded tissue.
..
PCR assays are complex and expensive. Immunohistochemical techniques suffer from the fact that a number of antibodies are used in such assays, and these antibodies have different binding affinities for the p21 protein. Also, in the studies that have been reported, there are different approaches to defining positivity in the immunohistochemical assays. There has been no attempt to standardize such assays.
..
Literature Review and Analysis

Eight articles144,153-159 evaluating assessment of the ras oncogene and prognosis in resected colon cancer were reviewed. All studies were retrospective, three of the studies used molecular techniques and the others used immunohistochemical staining. Among the five using immunohistochemical staining, four different antibodies (ras-11, Y13-259, RAP-5, and 142- 24EO5) were used. Seven of eight studies reported better disease-free survival or overall survival in tumors with normal ras expression. However, only three studies used multivariate analysis, and, as noted previously, none of the studies was prospective.
..
Measurement of ras oncogene abnormalities has not been demonstrated to be an independent predictor for either survival, quality of life, or disease-free survival in patients with large bowel carcinoma.

Abnormalities of ras in colorectal tissue may correlate with increased relapse rate and decreased survival. However, currently available data do not demonstrate that ras oncogene analysis provides independent prognostic information in colorectal cancer.
..
Areas for Future Research

There is a need for standardization of assays used for ras oncogene overexpression and/or mutation. The plethora of antibodies used for immunohistochemical assay of p21 protein expression have not allowed adequate standardization of this approach. There also needs to be a clear definition of the functional importance of the various point mutations that have been demonstrated in the ras gene. Finally, prospective studies should be performed with standardized assays to assess with appropriately designed multivariate analysis technique the independent predictive value of ras oncogene abnormalities in patients with colorectal cancer.


..
CA 15-3 as a Marker for Breast Cancer

On March 29, 1996, the Food and Drug Administration approved the Tru-Quant BR RIA test (Biomira Diagnostics, Toronto, Canada) for the early detection of recurrent disease in breast cancer patients who have stage II and stage III disease. The Tru-Quant assay uses the monoclonal antibody CA 27.29 to measure a CA 15-3-like antigen. Epitope mapping studies have shown that the epitopes on the MUC-1 gene product recognized by monoclonal antibodies 27.29 and 115-D8 (CA 15-3 test) are similar but not equivalent. The clinical significance of any test difference is not yet clear, as much of the relevant data have not yet been published. The clinical role of this new test and its application in routine practice will be addressed in a future update of these guidelines.
..
Guideline

1a. Present data are insufficient to recommend CA 15-3 for screening, diagnosis, staging, or surveillance following primary treatment. Although an increasing CA 15-3 level can detect recurrence following primary treatment, the clinical benefit is not established; therefore, it cannot be recommended.
..
1b. Present data are insufficient to recommend routine use of CA 15-3 alone for monitoring response to treatment. But, in the absence of readily measurable disease, an increasing CA 15- 3 level may be used to suggest treatment failure.
..
Marker Definition

The CA 15-3 test measures the serum level of a mucin-like membrane glycoprotein, which is shed from tumor cells into the bloodstream. The CA 15-3 epitope is recognized by two monoclonal antibodies in a double-determinant or sandwich radioimmunoassay.
..
Methodology

The CA 15-3 test from all sources uses both DF3 and 115-D8 antibodies. Serum is initially incubated with a polystyrene bead to which 115-D8 antibody has been attached. This antibody binds to antigenic sites on the glycoprotein, pulling it out of solution. The beads are then washed to remove unbound material and incubated with the radioiodine (125I)-labeled DF3 antibody. The radiolabeled DF3 antibody binds its antigenic sites and then the amount of radioactivity is quantitated. Controlled trials comparing these tests are not currently available.
..
Literature Review and Analysis

CA 15-3 is a serum cancer antigen that has been used in the management of patients with breast cancer. Its low detection rate in early stage disease indicates that CA 15-3 cannot be used to screen or diagnose patients with breast cancer. It has been widely used to monitor the effectiveness of treatment for metastatic cancer. Large, prospective, controlled clinical trials were not conducted before widespread marketing of the test. Thus, the available data do not convincingly demonstrate that CA 15-3 can be used alone to determine the response of a patient to treatment. CA 15-3 may be used to provide additional information suggesting that a patient is responding to or failing treatment.
..
Screening. CA 15-3 has been evaluated for its ability to determine diagnosis, prognosis, monitor therapy, and predict recurrence of breast cancer following curative surgery and radiotherapy. Multiple studies have shown that the incidence of CA 15-3 elevation increases with an increasing stage of the disease.160-171 Nine percent of women with stage I and 19% of women with stage II breast cancer have elevated CA 15-3 levels. The incidence of abnormal values increases to 38% and 75% for patients with stage III and IV, respectively. Low CA 15- 3 levels do not exclude metastases, and a given CA 15-3 level cannot be used to determine the stage of disease.172 When CA 15-3 is evaluated before surgery in patients with primary breast cancer, levels have not correlated with prognosis.173,174
..
Very high CA 15-3 levels tend to indicate advanced disease, and a value 5 to 10 times normal could alert a physician to the presence of metastatic disease. CA 15-3 levels are highest in patients with liver or bony metastases,168,175-177 and increasing numbers of metastatic sites correlate with increasing CA 15-3 levels.177,178 Several investigators have suggested that CA 15-3 is a particularly sensitive method for detecting osseous metastases.168,175
..
For a marker to be valuable in screening for cancer, it would have to detect early-stage disease in an asymptomatic population and be infrequently elevated in patients without cancer. Elevations have been observed occasionally in healthy subjects (5% to 6%) and more often in individuals with benign diseases, especially those of hepatic origin, in which false-positive elevations have been observed in 30% of patients.161,179
..
Typically, such CA 15-3 levels do not increase above 100 U/mL, as documented in more than 1,220 patients with benign diseases.179 This level of nonspecific elevation will give false-positive results when attempting to use CA 15-3 in screening or in diagnosis. The low incidence of CA 15-3 elevation in early-stage cancer indicates that it cannot be used in screening or in diagnosis.
..
One can use Bayes' theorem to calculate the predictive value of a positive test in an asymptomatic woman. The prevalence of breast cancer increases with age, but remains fewer than 400 new cases per 100,000 women per year. Incorporating this worst-case prevalence (0.004), the specificity (5% to 6% elevation in healthy women), and sensitivity (9% for stage I and 19% for stage II), the predictive value of a positive test is 0.7% for stage I and 1.5% for stage II cancer. Among individuals with benign disorders (who may have CA 15-3 levels up to 100 U/mL), the incidence of false-positive elevations will be even greater. Thus, in the general population, far more false-positive than true-positive results will be observed in using CA 15-3 in screening. Similar reasoning limits the use of CA 15-3 in diagnosis, where the prevalence of cancer is higher, but the incidence of false-positive elevation is increased as well. Although the use of CA 15-3 in screening has never been tested in large prospective trials, such trials with CEA in colon cancer and CA 125 in ovarian cancer, have confirmed the inability of these serum markers to detect cancer in a reliable way.180-183
..
Indicator of asymptomatic recurrence. A number of studies have addressed the question of whether CA 15-3 can monitor patients for recurrence following breast cancer surgery.164,165,168,179,184-191 Twelve studies were reviewed: one trial was level of evidence III; six were level of evidence IV; and five were level of evidence V. In seven trials, the data were reported in sufficient detail to allow summation of the results.164,179,184,185,187,190,191 Among 1,672 patients, 352 recurrences were noted. Among these, 235 (67%) had CA 15-3 elevation before or at the time of recurrence. Among the 1320 patients without recurrence, 1,218 (92%) had normal CA 15-3 values. The one level III trial by Safi et al164 reported CA 15-3 elevations in 149 of 205 patients (73%) with recurrence, and a 6% false-positive rate among 466 patients without recurrence. Thus, this study parallels the findings in the larger group: 67% of patients with recurrent disease, and 8% of patients without a recurrence will have an elevated CA 15-3 level. The mean lead time from marker elevation to clinical diagnosis of relapse ranges from 2 to 9 months.
..
In comparing CA 15-3 with CEA in patients with breast cancer, most reports indicate a greater sensitivity for CA 15-3 than for CEA.166,192 In fact, the data suggest that CA 15-3 is more sensitive only in advanced disease, with a 10% higher incidence of elevation in stage IV disease. In early-stage disease, in local recurrences, or in single metastases, the incidence of marker elevation is low for both CEA and CA 15-3.160-162,164,167,170,175,188 Summing all of the studies in which comparable data were presented, the incidence of elevation for CA 15-3 and CEA in stage I disease is 9% and 10%, respectively, while in stage II disease, it is 19% for both markers. Comparison of the markers by primary tumor size rather than overall stage resulted in similar findings.176
..
Thus, one would predict that the sensitivity in detecting recurrent, small volume metastases would be low for both CA 15-3 and CEA, and that neither can serve as an indicator of a true early relapse.
..
There are three major obstacles to using CA 15-3 in monitoring postoperative patients. The first, and most important, is that the low incidence of CA 15-3 detection in early-stage disease suggests that the marker is not sensitive enough to detect micrometastatic disease, and that an increase in the marker level during the postoperative period will indicate only a large tumor burden. Thus, "early" detection of recurrence will not be found in most patients. The second impediment relates to the available treatment options for those recurrences detected. The mean lead time from marker elevation to clinical diagnosis of relapse ranges from 2 to 9 months. The absence of curative salvage therapy for breast cancer makes this few-months lead time of questionable value.
..
Third, the efficiency can be questioned. Thirty percent of patients with a documented recurrence do not have elevated CA 15-3 levels.
..
Two of three patients are followed without a recurrence. Eight percent of those without a recurrence have an elevated CA 15-3 (false-positive) at some point. If patients were followed, as in the report by Safi et al,164 with values every 3 months and a median of 51 months until recurrence, 17 tests would be obtained for each patient before recurrence is documented by an increasing CA 15-3 level. Thus, for every two recurrences detected, one will be missed, 153 negative samples will be obtained, and 12 false-positive values will require evaluation along the way. For such a test to be worthwhile, it would have to be very inexpensive (including the cost of evaluating the false- positives reported). These three obstacles suggest that, at the present time, CA 15-3 should not be used in the routine monitoring of patients following surgery for breast cancer.
..
Monitoring response to treatment. Evaluation of CA 15-3 in assessing response to metastatic cancer therapy would be a much-needed clinical tool. Eleven studies reported data that could be summarized; nine presenting level of evidence V studies, and two level of evidence IV.161,162,165,167,169,188,193-197 Summing the numbers from these trials suggests that 66% of patients show decreases in marker levels in the presence of responding disease, 73% of patients show stable levels in the presence of stable disease, and 80% show increasing levels in the presence of progressive disease. Several investigators required a 25% percent change in the level of CA 15-3 to indicate a significant change. The data suggest that there is a good general correlation between changing CA 15-3 levels and response to therapy in metastatic cancer, but that it cannot be relied on in the absence of confirming clinical data.
..
A number of studies have advocated the use of CA 15-3 in monitoring the response to treatment. There is a strong correlation between the response to treatment and the change in CA 15-3 level. However, for all three response categories (response, stabilization, and progression of disease), a sizeable proportion of discordant results have been observed. Indeed, the greatest error will occur in the most important clinical setting: removing a patient from effective therapy because of incorrect information from a tumor marker test, which could occur in 34% of cases.
..
Some of these false-negative results may be caused by a transient increase in marker, which is often observed shortly after the initiation of effective treatment. This increase has been observed in both CA 15-3 and CEA literature, and could be avoided by waiting a number of weeks to draw marker levels after the initiation of chemotherapy.198,199 If false-negative or false-positive results are not considered, one is still left with the open question of whether a correct marker result adds anything to clinical management of breast cancer. Since confirmatory clinical information must be awaited, the lead time (which has not been reliably determined in this setting) offered by the increasing marker level is lost.


..
These results suggest that routine use of the CA 15-3 to monitor the course of therapy of breast cancer cannot be recommended. In exceptional circumstances, such as in the presence of osseous metastases, which are difficult to evaluate clinically, the marker level may be able to support the clinical estimates of disease status. However, the marker cannot, in any situation, stand alone to define response to treatment.


..
Areas for Future Research

Future research in this area will be most benefited by studies that address the two areas above and that report consistent, interpretable data in large numbers of patients. Studies should also report data in a uniform manner, with raw data presented along with statistical calculations; define significant changes in markers; obtain marker levels at defined intervals; present data for patients in separate clinical groups separately (eg, patients who have no evidence of disease (NED) following surgery for stage I or II breast cancer should not have data combined with patients who are NED following chemotherapy for stage IV breast cancer); and study large numbers in each patient group, rather than increasing numbers through increasing the variety of clinical settings. If marker combinations are studied, data for each marker should be presented within the study, rather than derived combined data. In combined studies, statistical analyses need to be performed to determine whether independent contributions of each marker are found. Careful evaluation of the lead time of marker elevation is needed to assess the potential value of a marker elevation. Carefully designed, prospective trials are needed to determine whether there are clinical situations in which marker assessment could replace other diagnostic methods.


..
CEA as a Marker for Breast Cancer

Guideline
..
2a. CEA is not recommended for screening, diagnosis, staging, or routine surveillance of breast cancer patients following primary therapy.
..
2b. Routine use of CEA for monitoring response of metastatic disease to treatment is not recommended. But, in the absence of readily measurable disease, an increasing CEA level may be used to suggest treatment failure.

Marker Definition
..
CEA belongs to a family of cell-surface glycoproteins with increased expression found in a variety of malignancies, including breast cancer. The characteristics of CEA have been described above.

Because CEA has been for two decades the standard to which new breast markers are compared (in the past 5 years replaced in some studies by CA 15-3), publication bias results in reports in which new markers are always better than the old CEA numbers.200 Thus, studies in which a new marker was compared with CEA were not included in the analysis, unless several markers were under evaluation, or both CEA and CA 15-3 were considered.201-203
..
Methodology

Methods to assay CEA have been described in a previous section.

Literature Review and Analysis CEA is elevated in 60% of patients with advanced breast cancer, and the level and frequency of elevation correlates with the tumor burden. Because of its low sensitivity in early-stage disease, CEA cannot be used in screening or diagnosis of breast cancer. The two roles that have been most frequently investigated are in monitoring patients following surgery for primary breast cancer, and in monitoring patients for the response to treatment for metastatic disease. Monitoring patients following breast cancer surgery cannot be recommended. Half of patients with recurrences will not be detected, and there is a 12% false-positive rate. CEA monitoring of treatment of metastatic disease may correlate with other markers of disease progression in patients known to have elevated levels. However, with a high error rate, supporting clinical information must be awaited before modifying therapy. Care should be taken to avoid misinterpreting the paradoxical CEA elevation that can occur shortly after institution of therapy for metastatic disease. CEA determination can only be recommended in patients with elevated levels in whom there is no other ready means of following the disease.
..
The sensitivity and specificity determine whether a marker will be useful in the screening and diagnosis of cancer. Multiple studies have shown that the incidence of CEA level increases with an increasing stage of the disease.160-162,164,166,167,170 Both the incidence and the level are lower in early-stage disease. Merging data from seven studies, 10% of women with stage I and 19% with stage II breast cancer have elevated CEA levels; the incidence increases to 31% and 64% for patients with stages III and IV disease, respectively.

The ability of CEA to predict prognosis, when measured at the time of diagnosis, is not clear. Reports have been issued that suggest a better,204,205 worse,206,207 or unaffected208-210 prognosis when CEA elevation is present at the time of diagnosis.
..
Indicator of asymptomatic response. Multiple studies using CEA to detect relapse following surgery for primary breast cancer were evaluated.179,189,190,197,205,208,211-222 Of these studies, three were level of evidence III, 15 were level of evidence IV, and three were level of evidence V. Summing the data from the patients in the 15 studies in which sufficient data were given, CEA elevations were noted in 563 (50%) of 1,141 patients who had a recurrence of their breast cancer.179,190,197,205,208,211-218,220,222 A normal CEA was present in 1,922 (88%) of the 2,179 patients whose cancer had not recurred, while 257 (12%) false- positive results were recorded. Similar results were reported in one large study in which both CA 15-3 and CEA levels were monitored in 671 patients following breast cancer surgery, and 50% of patients with recurrence had an elevated CEA, while 96% of patients without a recurrence had normal levels.223 Seven studies reported mean lead times ranging from 2.5 to 8 months from CEA elevation to the clinical diagnosis of recurrence.
..
Multiple studies have shown that 80% to 90% of patients present with signs or symptoms of relapse and that various screening tools including chest films, bone scans, and blood chemistries have failed to result in earlier detection of relapse.224 While the hope has been that a significantly earlier diagnosis could be attained by using a serum marker, most studies suggest a lead time of only a few months can be gained by monitoring serum markers following primary therapy. This is almost predictable in view of the low sensitivity in occult disease. Mansi et al213 followed 141 patients after treatment for primary breast cancer and found that 61 relapsed with clinical signs or symptoms indicating the relapse in 58 (95%). In only 13 patients was the relapse predicted by markers (lead time, 7 months), which included CEA.
..
The data suggest that CEA is able to detect recurrence of breast cancer before clinical recurrence. However, its routine use in this setting is not recommended because half of all recurrences will be missed; and because the impact of the small percentage of false-positive levels is amplified by the larger number of women who have not recurred and by the repeated testing required in monitoring therapy. Secondly, the short lead time from marker elevation to clinical recurrence suggests that there will be little biological impact of the earlier diagnosis of recurrence.
..
Monitoring response to treatment. Eighteen studies were reviewed; nine were level of evidence IV and nine were level of evidence V. Six studies reported results only in patients with elevated levels.197,215,225-227 Among these patients, 82% were noted to have decreasing levels in the presence of responding disease. With progressive disease, 74% were noted to have increasing CEA levels. Twelve studies reported data in patients with metastatic disease, whether or not they had an elevated CEA level.185,188,193,195,209,228-235 Among these patients, 61% had a decrease in CEA when tumor response was observed, and 65% had an increase in CEA when tumor progression was observed. Thus, CEA levels, especially when elevated initially, correlated with response to therapy for metastatic breast cancer.197,215,225- 227,229,235
..
Many investigators require a 25% to 30% change in the CEA level before considering the change to be predictive of the clinical course. However, variations greater than these are regularly observed in stable disease, for which CEA is the least able to predict clinical course. In the 14 studies noted above, 30% of 63 patients with stable disease had stable CEA levels.

Both CA 15-3 and CEA have been noted to have marker elevation in the few weeks following initiation of successful chemotherapy, resulting in misleading or paradoxical marker levels in responding patients.199,236 An increased CEA level during the initial 4 to 8 weeks of chemotherapy was seen in 13 patients who subsequently responded to chemotherapy.227 These levels later declined in accordance with the clinical response. Correlations between CEA levels and responses to hormonal therapy have also been recorded, but with a poorer correlation. This may be because of the slower and less complete responses that often occur with hormonal therapy.237 One study noted that, among patients receiving fluorouracil, doxorubicin, and cyclophosphamide (FAC) chemotherapy for metastatic breast cancer, normalization of an elevated CEA level was associated with a longer duration of response (22 v 9 months for patients whose CEA never normalized).227
..
Increasing CEA levels are more likely to be associated with disease progression than with persistently or transiently elevated ones.220,232 Fluctuating abnormal levels were more likely to be false-positive.219 Gion et al238 examined the level of fluctuation in a postoperative population of patients clinically free of disease and noted variations of low levels of nearly 30% for CEA and 20% for CA 15-3. These investigators concluded that the "intrapatient variability of CEA and CA 15-3 is high in a significant number of cases" such that "the potential usefulness of tumor markers as 'fine' detectors of relapse seems to be restricted . . ." The highest proportion and level of CEA expression are found in patients with hepatic metastases.176,235,239 The physician must wait for supporting clinical evidence before modifying treatment in a patient with advanced breast cancer.
..
Because metastatic breast cancer is currently incurable and salvage therapy only partially effective, CEA levels generally should not be drawn. In exceptional cases in which determining disease response is very difficult, CEA may be of limited use. In such a case, levels should be monitored only in patients with documented elevation, a 20% or 30% change in level must be used to indicate a significant response, and supporting clinical evidence must be obtained before discontinuing therapy. Such principles apply to CA 15-3 as well. Both markers need not be used; and when both are elevated, the least expensive alternative (CEA, at the present time) can be chosen.
..
Estrogen Receptors and Progesterone Receptors as Markers for Breast Cancer Guidelines 3a. Estrogen and progesterone receptors are recommended to be measured on every primary breast cancer, and may be measured on metastatic lesions if the results would influence treatment planning.

3b. In both premenopausal and postmenopausal patients, steroid hormone receptor status may be used to identify patients most likely to benefit from endocrine forms of adjuvant therapy and therapy for recurrent or metastatic disease.
..
3c. Estrogen and progesterone receptors are relatively weak predictors of long-term relapse and breast cancer related mortality rates, and are not recommended to be used alone to assign a patient to prognostic groupings.

Marker Definition

The estrogen receptor and progesterone receptor are intracellular receptors that are measured directly in tumor tissue. These receptors are polypeptides that bind their respective hormones, translocate to the nucleus,240,241 and induce specific gene expression. There are three domains on these polypeptides: a C-terminal hormone binding domain, a central DNA binding domain, and an N-terminal domain that is important for transcription.242 When the hormone binds to its respective receptor, the DNA binding domain is modified so it binds to DNA quite avidly and initiates transcription. This process clearly affects the growth of the cell. However, the intricacies of how the hormone and receptor complex affect cell growth are only partially understood.243,244
..
Methodology

The estrogen receptor and progesterone receptor are labile proteins. The accuracy of their measurement is dependent on a timely and efficient collection of the tissue. Activity of these proteins decays at room temperature. Therefore, the tissue should be stored at 4¦ centigrade and, if the assay is not to be performed immediately, the tissue should be frozen in liquid nitrogen until it can be processed. The receptors are also most stable at a pH of 7.4; therefore, this pH should be maintained.
..
Once the specimen is properly collected, it must be separated and the receptors quantified. The two methods of separation have included sucrose density-gradient ultracentrifugation and a dextran-coated charcoal assay. The sucrose density gradient ultracentrifugation was the original method used for the determination of the estrogen receptor. The most commonly used method, because of its ease of performance and its accuracy, is the dextran-coated charcoal assay. This relies on the absorptive properties of dextran-coated charcoal to separate small and large molecular weight compounds.245,246 The assay uses increasing concentrations of radio labeled hormone and unlabeled hormone competitor to determine the quantity of the receptor. The total number of binding sites can actually be determined and expressed as femtomoles per milligram of protein. The dissociation constant (Kd) can be calculated and is a measure of the affinity of the receptor for the estrogen or progesterone.247
..
The estrogen receptors and progesterone receptors of a tumor can also be measured on small samples, including paraffin-fixed tissue. With the advent of monoclonal antibodies, the estrogen receptor immunocytochemical assay (ERICA) and the progesterone receptor immunocytochemical assay (PgRICA) have been developed. These stains are directly applied to a microscope slide containing a section of the tumor and allow the direct counting of the percentage of tumor cells that are positive for estrogen and progesterone receptors.

One of the problems with the dextran-coated charcoal assay is that a reasonably large amount of tumor is necessary to perform this test. Because many of the tumors that are discovered are less than 1 cm, estrogen and progesterone receptors cannot always be measured on these tumors using the technique described above. The dextran-coated charcoal method and sucrose density gradient ultracentrifugation have also been criticized because they measure estrogen and progesterone receptors in normal tissue in the biopsy, and in noninvasive (intraductal) carcinomas within the biopsy, complicating the interpretation of results. The advantages of the ERICA and PgRICA are that they can be performed on very small amounts of tumor, and they can differentiate staining or the presence of the receptor in invasive versus noninvasive carcinomas, or in normal tissue versus the tumor.248 The disadvantage is that no threshold value has been universally accepted for what constitutes a positive ERICA or PgRICA in the tumor.249,250 When the ERICA and PgRICA techniques were compared with the standard dextran-coated charcoal method, the concordance of results in the same tissue has ranged between 70% to 95%. It also appears that the concordance is lower in paraffin-embedded tissue than in frozen tissue. However, this technique does allow for measurement of the receptors in small tumors.
..
Literature Review and Analysis

Most oncologists have used the estrogen receptor and also the progesterone receptor not only to predict the probability of response to hormonal therapy at the time of metastatic disease, but also to predict the likelihood of recurrent disease, and to predict the need for adjuvant hormonal therapy or chemotherapy. Although these latter uses for estrogen and progesterone receptors are commonly accepted by most oncologists, the data on which these conclusions are based are quite controversial. Because estrogen and progesterone receptors are determined from the tumor itself, the estrogen receptor and progesterone receptor might be able to: (1) predict a response to specific therapies for metastatic disease; (2) predict the success of adjuvant hormonal therapy; and (3) assess the prognosis of a specific tumor to metastasize.
..
Predicting response to therapy of metastatic disease.The least controversial area concerning the use of estrogen receptors in breast cancer is their ability to predict a response of metastatic disease to hormonal therapy. Many studies have documented that the estrogen receptor is clearly the most important variable in predicting whether a woman will respond to hormonal therapy when she has developed metastases.251-257 It has been shown that the higher the estrogen receptor level, the more likely that metastatic disease will respond to hormonal therapy.254 Two articles258,259 have, however, questioned this association. The progesterone receptor is also felt to be an independent predictor of hormonal response.252 Women whose tumors are both estrogen receptor-positive and progesterone receptor-positive appear to have the best response to hormonal therapy.
..
The question of whether estrogen receptors and progesterone receptors predict a response of metastatic disease to hormonal therapy was addressed in multiple studies.251-257,259 Five of the studies showed that the estrogen and/or the progesterone receptors were strong predictors that a metastatic tumor would respond to hormonal therapy.252,254-257 Evidence was generally level IV. However, two of the studies that showed a positive relationship between a positive estrogen receptor and a positive progesterone receptor and a response of metastatic disease to hormonal therapy had level II252 and level III254 evidence. Therefore, this Panel recommends the measurement of estrogen receptor and progesterone receptor on every primary breast cancer and every metastatic breast cancer if a positive estrogen or positive progesterone receptor would influence treatment planning (guideline 3a).
..
It is generally accepted at this time that the estrogen and progesterone receptor cannot predict the response of metastatic disease to chemotherapy. A number of investigators have postulated that the estrogen receptor and progesterone receptor should be a predictor for metastatic disease to respond to chemotherapy. The logic is that because there is an association between a low estrogen receptor and a high S phase, and because chemotherapy seems to work better in tumors with a high proliferative index, estrogen receptor-negative metastatic disease should have a higher response to chemotherapy. There is a study that supports this concept.254 However, there have also been studies that have shown just the opposite.260,261 Four articles directly addressed the question of whether the estrogen and progesterone receptors could predict which patients with metastatic breast cancer would respond to cytotoxic chemotherapy.254,256,258,260 One article254 showed a good biologic correlation with a low estrogen receptor in response to chemotherapy. The other three articles did not show a significant correlation between a low estrogen receptor and a positive response to chemotherapy. One article had level III evidence, three had level IV evidence,256,258,260 and the final article had level V evidence.260
..
Predicting response to adjuvant therapy.It is common practice to treat older postmenopausal women with positive estrogen and progesterone receptors with adjuvant hormonal therapy. Although there are some articles262,263 that dispute this recommendation, there is extremely strong evidence264-268 that postmenopausal women with positive estrogen receptor have a better disease-free survival or overall survival with adjuvant tamoxifen. The advantage appears to be greatest with the higher estrogen receptor levels.264 However, two separate studies have shown that there is a significantly increased survival in postmenopausal women treated adjuvantly with tamoxifen even if their estrogen receptor were negative or unknown.262-265
..
The use of adjuvant hormonal therapy for premenopausal women with positive estrogen receptors and progesterone receptors is much more controversial than its use in the postmenopausal woman. Most trials262,264,265 show that there is an improvement in disease- free survival with adjuvant tamoxifen in the premenopausal group, but its significance is clearly less than in the postmenopausal group. The NSABP B-09 trial266 actually showed that the addition of tamoxifen to melphalan and fluorouracil had a poorer survival time than melphalan and fluorouracil alone in the premenopausal women with high estrogen receptors.
..
Seven articles directly address the question of whether the estrogen receptor could predict for the success of adjuvant hormonal therapy. There was an improvement in overall survival and disease-free survival in all of these studies. Three of these studies had a level of evidence that was either I or II.262,264,266 The Early Breast Cancer Trialist Collaborative Group did a meta-analysis of 75,000 patients. They showed a clear relationship between the estrogen receptor and an improvement in overall survival in postmenopausal women who were treated with tamoxifen. Because of the strong evidence in these studies,262,264,266 this Panel recommends that in premenopausal and postmenopausal women, estrogen receptor should be examined and the information may be used to identify patients most likely to benefit from endocrine therapy (guideline 3a).
..
There are insufficient data to use the estrogen and progesterone receptors to predict the success of adjuvant cytotoxic chemotherapy. There were only two studies that directly address the question whether the estrogen receptor or progesterone receptor could predict which patients would be benefited by adjuvant cytotoxic chemotherapy.267,268 Each was only level IV evidence.

Predicting probability of recurrence. Since the late 1970s, there have been numerous studies to determine whether estrogen and progesterone receptors can predict a primary tumor's likelihood of metastasis. Although most oncologists think that women with high estrogen and progesterone receptors have a better chance to remain disease-free, this is still a debatable issue. It is clear, as more data have been obtained, that the estrogen and progesterone receptors are only one prognostic factor of many that predict future behavior of a primary breast cancer tumor. It is also clear that the estrogen and progesterone receptors should be interpreted in the light of other possible prognostic indicators such as tumor size, nodal status, nuclear grade, etc. It is also clear that we do not as yet know the exact usefulness of each of these prognostic factors. Many studies have shown that the estrogen and progesterone receptors are useful in predicting the prognosis of a patient with a primary breast cancer.269-272 However, other studies have shown that the estrogen and progesterone receptors do not predict the outcome of a woman with primary breast cancer.273-275 The NSABP B-06 study of 1,157 women with node-negative breast cancer272 showed a highly significant statistical difference, based on the estrogen receptor, in the disease-free survival of these women. The estrogen receptor-positive patients clearly had an improvement in survival. However, at 5 years, the difference had lessened to only 8%. Even though this was still statistically significant, it raised the question as to whether the estrogen receptor should be used to predict which women should be treated adjuvantly with chemotherapy. In this same study, there was no survival advantage predicted by the progesterone receptor. The data and conclusions from the NSABP B-06 study were similar to those from a large retrospective study.276 In this retrospective study of 2,028 patients, the San Antonio group determined that women with positive estrogen receptors had a better 5-year disease-free survival than women with negative estrogen receptors. They also determined that progesterone receptors in node-negative patients did not seem to predict the behavior of the tumor.
..
In summary, the estrogen receptor in both node-negative and node-positive disease is a predictor for the biologic activity of the tumor and for its potential ability to metastasize, based on two studies,266,272 (level II). The progesterone receptor in node-negative patients does not seem to predict ultimate survival and disease-free survival as it does in node-positive patients.266,272,276 Thirty-eight articles directly address the question of whether the estrogen receptor and/or progesterone receptor could predict the development of metastatic disease when measured in the primary tumor (level III or IV).256,266,269,275,277-304
..
Areas for Future Research

The usefulness of ERICA and PgRICA techniques must be further studied. All of the studies that have led to the guidelines for the use of estrogen receptors and progesterone receptors have used the dextran-coated charcoal method or the sucrose density gradient ultracentrifugation technique. In clinical practice, it has been assumed but never proven that the ERICA is interchangeable with the standard estrogen receptor assays, and the PgRICA is interchangeable with the progesterone receptor assays. Because aggressive screening is detecting smaller tumors that would require the measurement of the estrogen receptor and progesterone receptor by the immunocytochemical techniques, prospective studies are needed to determine whether ERICA and PgRICA have the same predictive value for response to hormonal therapy and metastatic potential of the primary breast cancer as the biochemical estrogen receptor and progesterone receptor measurement.
..
Clinical research is still needed to define the role of the estrogen receptor and progesterone receptor as predictors of the primary tumor's ability to metastasize. Because chemotherapy and hormonal therapy have been shown to increase the cure rate over surgery alone or surgery in combination with radiation therapy for primary breast cancer, it is imperative that we define the group of women who are at risk for a recurrence and who could benefit from adjuvant therapy. It is clear that no single variable is an absolute predictor for recurrence. It is probable that there is a relationship between the known and the as yet undiscovered predictors for recurrence. Research directed at how these prognosticators modulate each other is needed through large prospective studies.


..
DNA Flow Cytometrically Derived Parameters as Markers for Breast Cancer

Guidelines

4a. Present data are insufficient to recommend routinely obtaining DNA flow cytometry-derived estimates of DNA content or S phase in breast tissue.
..
4b. DNA flow cytometry-derived ploidy are not recommended to be used to assign a patient to prognostic groupings. There is insufficient evidence to recommend the use of S-phase determination for assigning patients to prognostic groupings.

4c. Present data are insufficient to recommend the use of DNA flow cytometry-derived ploidy (DI) or flow cytometric measures of proliferation (% S phase and related analysis) for selection of the type of adjuvant therapy to be given.
..
4d. Present data are insufficient to recommend the use of DNA flow cytometry-derived information to select among different treatment options of metastatic disease.

Marker Definition and Methodology

The methods for DNA flow cytometrically derived ploidy (DI) and proliferation analysis (S phase) have been described in the section for colorectal cancer above.
..
Literature Review and Analysis

DNA flow cytometry-derived estimates of DNA content (ploidy or DI) and proliferation (S phase and related estimates) have been correlated with patient outcome in many studies. These studies do not support the use of ploidy as a prognostic variable. The case for the use of S phase appears much stronger with the majority of studies showing S phase to have independent value in some patient subsets. It must be noted, however, that many of these studies are exploratory in nature with multiple cutpoints for defining high- and low-risk populations being used. There are not prospective studies using previously defined techniques and prospective definitions of high- and low-risk groups. With the nature of many of these studies, and the highly variable independent relative risk (1.5- to 6-fold) conferred by "high" S phase in different studies, the Panel cannot endorse the use of DNA flow cytometry-designed S phase as a test to routinely use in making prognostic estimates for breast cancer patients.
..
The most important potential use of DNA flow cytometry-based information is in the refinement of estimates of prognosis of axillary node-negative patients. There are numerous publications288,305-321 (with > 50 patients and > 3 years of follow-up) that address the potential utility of DNA flow cytometrically-based information for defining prognostically distinct populations of node-negative patients. These studies do not strongly support the use of ploidy as a prognostic variable, with only three studies308,316,318 finding ploidy to be of independent predictive value. The case for the use of S phase appears to be much stronger, with the majority of the studies showing S phase to have independent predictive value at least in some patient subsets.307,309,311-314,316,317,319,321 It must be noted, however, that many of these studies are exploratory in nature with multiple cutpoints for defining high- and low-risk populations, and cutpoints were selected retrospectively that gave the highest levels of statistical significance. There are no prospective studies confirming the validity using a given technique and a specific definition of high and low S phase to predict a given level of relative risk.
..
Although some studies have suggested that high S phase predicted a more chemotherapy- responsive state for breast tumors,4 other studies have failed to confirm this effect.5,6 It is therefore not appropriate to use S-phase determinations for selecting the type of adjuvant therapy or therapy for recurrent disease.

Areas for Future Research

Clearly, DNA flow cytometry-based information has the potential to be very useful in the prognostic assessment of breast cancer patients. However, the lack of standardization of techniques and true confirmatory studies are major limitations of current studies. What is needed are a few studies in well-defined populations (T1N0), to prospectively confirm the results of prior studies.


..
c-erbB-2 as a Marker for Breast Cancer

Guideline

5. Present data are insufficient to recommend the use of c-erbB-2 (HER-2/neu) gene amplification or overexpression for management of patients with breast cancer.
..
Marker Definition

The c-erbB-2 gene encodes a transmembrane tyrosine kinase that is the receptor for a family of peptide hormones.

Methodology

Various methods have been used to measure c-erbB-2 and its gene product. These include direct measurement of gene amplification, mRNA level, and protein expression. The most widely studied method is immunohistochemical staining. Because of measurement error, and in an attempt to optimize apparent predictive significance of these tests, the measurements in many studies have been dichotomized into high and low ranges. The threshold and criteria for classification of values have varied from study to study, and there is not an accepted standard. A number of different monoclonal and polyclonal antibody preparations have been used for immunohistochemistry.


..
Literature Review and Analysis

Postoperative Prognosis Indicator

Gene amplification. The first study addressing the possible prognostic significance of c-erbB-2 gene amplification was published in 1987 by Slamon et al.322 This study of 187 patients (including both node-positive and node-negative patients) showed c-erbB-2 gene amplification to be strongly correlated in univariate and multivariate analyses with shortened disease-free survival and overall survival in the subset of 86 node-positive patients. Slamon et al323 then replicated the result, as reported in 1989, in a 526-patient series. Interestingly, in the latter article, no overall analysis of the entire patient group was presented; only separate subset analyses were presented for the node-negative and node-positive patients. c-erbB-2 amplification was again found to be predictive of poorer disease-free survival and overall survival only in the node-positive subset (both in univariate and multivariate analyses), but not in the node-negative subset. Work by the same group published in 1994 showed in univariate analysis that c-erbB-2 did correlate with a worsened disease-free survival in 210 node-negative patients.324
..
Attempts by other investigators to replicate these results have had mixed success, with some studies finding that c-erbB-2 gene amplification was an independent predictor in breast cancer patients,325,326 and others failing to find any useful correlations,327-331 even in the node-positive subset.327,329,330 Perhaps most interesting for clinicians is that, of the five studies with more than 100 patients and at least 3 years of median follow-up,323,324,326,329,330 only one326 found c- erbB-2 to have multivariate predictive value in node-negative patients. Thus, overall, these studies correlating c-erbB-2 gene amplification with either disease-free survival or overall survival have produced no clear rationale for the usefulness of this test in evaluating the prognosis of those breast cancer patients most in need of prognostic information.
..
Gene expression.Various studies329,330,332 have shown good correlations between c-erbB-2 gene amplification and gene expression. One could also make a case that gene expression might be more directly related to tumor cell behavior. Thus, the initial amplification studies were extended to studies of gene expression.

Of the two most commonly used methods for measuring c-erbB-2 protein levels, immunoblotting and immunohistochemistry, most of the work has been done with immunohistochemistry, because if appropriate antibodies are used, this technique can often be performed on paraffin-embedded archived material. Immunoblotting is more technically complete and requires fresh or frozen material, and consequently is far less amenable to use in most centers.
..
Immunoblotting. There are only three articles329,333,334 using immunoblotting that have greater than 3 years of follow-up and greater than 100 patients studied, but eight reports found it not to be predictive332,335-341 on work performed at single institutions. The first study in 1989 investigated the prognostic significance of c-erbB-2 protein levels in 728 patients.333 In the node-negative patients (n = 378), no univariate or multivariate prognostic value was found. However, in the node-positive population (n = 350), c-erbB-2 was a strong independent predictor of disease-free survival and overall survival. An attempt to replicate this study in 1991 again showed that c-erbB-2 expression was a predictor in a node-positive population (301 patients), but no multivariate analysis was performed and node-negative patients were not included.334 In the other large study using immunoblotting, with 159 node-negative patients and 120 node-positive patients, no evidence for any predictive value of c-erbB-2 was found.329 Thus, there is no rationale for using immunoblotting of c-erbB-2 as a prognostic test.
..
Immunohistochemistry. Eighteen immunohistochemistry studies meeting criteria for patient numbers (> 100) and follow-up (> 3 years) and including both node-positive and node-negative patients have been performed. There are some studies suggesting that c-erbB-2 might be of some value, and others that do not support this hypothesis. No study reported it to be an independent predictor of disease-free survival in multivariate analysis. Six of the studies reported c-erbB-2 to be a predictor of overall survival in multivariate analysis.331,342-346 Two of the studies that appeared to be positive in univariate analyses were negative in multivariate analyses, perhaps because of the inclusion of other uncommon prognostic variables: in one study, internal mammary lymph node sampling,335 and, in the other, complex histopathologic variables.347 In general, the pattern that emerges is that c-erbB-2 by immunohistochemistry adds little to the prediction of disease-free survival, but may add something to prediction of overall survival.342 This general conclusion is consistent with the one cooperative group study. The analysis of patients from NSABP B-06 showed that c-erbB-2 expression was not at all predictive of disease-free survival but was predictive of overall survival. Still, the majority of studies were negative.
..
Examination of 10 studies in node-negative patients suggests little clinical utility in this group.324,332,343,344,346,349,355-359 In only one of 10 of these studies was there a positive finding in multivariate analysis.345 Even in the univariate analyses the correlations appear weak. Thus, there seems to be little support for the use of c-erbB-2 in node-negative patients, which is the group for whom prognostic factors are most important for making adjuvant treatment decisions.

c-erbB-2 and resistance to therapy. If c-erbB-2 were predictive of resistance to adjuvant therapy, then studies stratifying adjuvant therapy patients between no (or minimal) therapy and more intensive therapy might be expected to show predictive value of c-erbB-2 only or primarily in the groups getting more intensive therapy. For example, in the study by Tatu and Brisson,348 c-erbB-2 was not a significant predictor of either disease-free survival or overall survival for the untreated node-positive patients, while it was a powerful predictor of disease- free survival and overall survival for patients treated either with adjuvant chemotherapy or hormone therapy. Another study that supports this concept is the Ludwig study,349 in which the node-negative population that received no adjuvant therapy showed no predictive value of c- erbB-2, while in the node-negative group that did receive adjuvant therapy (a single perioperative cycle), c-erbB-2 was a significant predictor of overall survival. Similar effects were seen in the node-positive population in that study, where patients were randomized between a less intensive therapy (one cycle of perioperative chemotherapy) and a more conventional multicycle, multidrug adjuvant regimen. A greater predictive value of c-erbB-2 was noted in the more heavily treated group than in those receiving less intensive therapy.349 However, not all studies have agreed that c-erbB-2 is more predictive in the more heavily treated patients. A recent report by Muss et al,350 of a trial randomizing patients between different intensities of adjuvant chemotherapy, can be interpreted as showing that c-erbB-2 was less predictive of outcome in the high-dose arm. Whether this apparent discrepancy between different trials is caused by differences in the types of chemotherapy used, the relative doses, statistical aberrations, or other factors is unclear at this time.
..
There is also more direct preclinical and clinical evidence that c-erbB-2 may be a predictor of treatment resistance. For example, Benz et al351 have shown that, in animal models, transfection of breast tumor cells with c-erbB-2 results in treatment resistance to tamoxifen. Wright et al352 addressed this effect of c-erbB-2 overexpression on resistance to tamoxifen in 65 patients with recurrent metastatic breast cancer. Response rates were 7% in the c-erbB-2- positive patients and 37% in the c-erbB-2-negative patients (P < .05). Although the subsets were small, c-erbB-2 expression appeared predictive of treatment failure in estrogen receptor- positive patients, with response rates of 20% (one of five) in the estrogen receptor-positive/c- erbB-2 patients and 48% (12 of 25) in estrogen receptor-positive/c-erbB-2 patients. The work of Klijn et al353 is consistent with these tamoxifen results, though the conclusion was not bolstered by a multivariate analysis. Paradoxically, Klijn et al353 also found that c-erbB-2 overexpression was actually a predictor of good response to cyclophosphamide, methotrexate, and fluorouracil (CMF) chemotherapy in patients with metastatic disease. However, a study by Wright et al354 in 68 patients with metastatic breast cancer treated with mitoxantrone showed a marginally poorer response rate in patients who were c-erbB-2 overexpressors (50% v 58%), and the survival was significantly but only marginally shorter. It is clear that the ability of c-erbB- 2 to predict treatment responsiveness will have to be evaluated more carefully in studies of specific treatment regimens.
..
Areas For Future Research

It seems unlikely that measurements of c-erbB-2 or its gene product will be useful in predicting the prognosis of untreated stage I breast cancer patients. The most interesting and important avenues of research would seem to be in determining if this marker can be used to predict response to endocrine or cytotoxic chemotherapy. Crucial issues in these studies will be the use of standardized methodologies, and the prospective validation of the results of prior exploratory analyses.


..
p53 as a Marker for Breast Cancer

Guideline

6. Present data are insufficient to recommend use of p53 measurements for management of patients with breast cancer.
..
Marker Definition and Methodology

Definitions and methods for detecting abnormalities in p53 have been described above.

Literature Review and Analysis

Approximately 40 reports of studies of p53 as a prognostic indicator for breast cancer were reviewed, all of level III or IV evidence. Only studies that included more than 100 patients and compared p53 expression to other prognostic indicators were used to develop the guideline. The other studies were not used because they had ioo few patients for a meaningful analysis; did not provide sufficient information about the technical approaches or the patient population; or used multiple analyses of the data without prespecified thresholds or outcomes. All of the studies reviewed were retrospective, using tumor specimens from archival collections. There have been no prospective studies designed specifically to determine whether p53 predicts disease-free survival, overall survival, or response to specific therapies.
..
Correlations between the presence of p53 mutations and the stage of disease have suggested that the mutations are not primary events, that they occur later as part of the progression of the tumor rather than its initiation. However, if the mutation occurs early in the development of a particular tumor, it could result in a more rapidly growing tumor or even in a tumor that is more likely to metastasize. This has formed the basis for the effort to assess the value of measurements of p53 alterations for prognosis. Detection of these mutations in tumors could, in theory, provide information about the expected behavior of the tumors. To be useful in treatment planning, it will be necessary to show that the presence of mutations in a tumor predicts a poor outcome in a patient who would otherwise be expected to have an excellent prognosis based on standard indicators. Therefore, most of the focus has been on patients who have no apparent involvement of the axillary nodes at the time of diagnosis.
..
Thor et al360 examined 295 archived paraffin-embedded specimens (selected from approximately 1,600 eligible for the study), approximately one half of which were from women with node-negative breast cancer. They measured p53 immunoreactivity using the monoclonal antibody PAb 1801 and compared the ability of p53 to predict disease-free and overall survival with other prognostic indicators, including lymph node involvement, T stage, estrogen receptor status, and age. Using a regression tree analysis, p53 staining (positivity defined as any cell staining) proved to be the most important indicator of overall survival in patients who were node-negative or had 1 to 3 positive nodes.
..
Isola et al361 examined 289 archived paraffin-embedded specimens, representing almost half of the node-negative patients diagnosed during a 10-year period ending in 1986. They used a polyclonal antibody preparation, CM-1, and defined a case as positive when more than 20% of the cells were stained. When the multivariate analysis was performed, including p53, c-erbB-2 overexpression, and tumor size, p53 demonstrated independent prognostic value. However, when S-phase fraction was included in the model, only S-phase fraction and tumor size remained as independent prognostic indicators.
..
Two reports from the San Antonio group362, Guideline Marker Definition and Methodology Literature Review and Analysis 319 present data on p53 in node-negative breast cancer patients. Elledge et a l362 used the single-strand conformational polymorphism technique to evaluate mutations in the p53 gene in frozen, pulverized tumor specimens from 200 patients. The multivariate analysis included estrogen and progesterone receptor status, S-phase fraction, DNA ploidy, patient age, tumor size, and nuclear grade. Complete data were only available for 155 patients. Only age and p53 mutational status proved to be statistically significant independent prognostic factors. The second article319 reports on immunohistochemical analysis of p53 on frozen sections from 700 patients, using a cocktail of two monoclonal antibodies, PAb 1801 and PAb 240. Only 316 patients had S-phase fraction data; therefore, only data from these patients could be used in the full multivariate analysis. p53 accumulation and S-phase fraction were the only factors that achieved statistical significance as independent predictors in the multivariate analysis.
..
An unusual patient population was examined by Silvestrini et al363 in that 60% of the 256 women were premenopausal and 46% had tumors less than 2 cm in diameter. Paraffin- embedded specimens were stained using the monoclonal antibody, PAb 1801, and a series of cutoff values in 5% increments were tested as separate variables to establish a suitable cutoff value. The multivariate analysis included p53 (positive defined as > 5%), thymidine labeling index, estrogen receptor status, and tumor size. p53 and thymidine labeling index proved to be independent predictors of relapse-free survival. p53 and estrogen receptor status were independent predictors of overall survival. A more recent report from this group364 added evaluation of the oncogene bcl-2 to the analysis of 283 tumors. It is not clear whether the two reports reflect analyses of the same tumor population. The recent article still shows a strong effect of p53 in predicting poorer outcome. Most of the studies cited did not report the number of outcome events, ie, recurrences or deaths, making it difficult to determine whether there was sufficient power to draw meaningful conclusions from the multivariate analyses. Given the total number of patients in most of the studies, the fact that they had early-stage disease, and the relatively short median and range of follow-up time, it is unlikely that the number of events was sufficient for the number of variables included in the analyses.
..
Areas for Future Research

To determine definitively the usefulness of p53 as a prognostic indicator, it will be necessary to perform large studies to assure sufficient numbers of events for adequate statistical power. The standard prognostic indicators along with other promising indicators must all be measured on the same tumors. Appropriate cutoffs for the markers should be established using a training set, and a separate set of samples should be used for validation. To make interpretation of results meaningful, reproducible methods for measurement will have to be chosen and validated. Once p53 is shown in retrospective studies to aid in the prediction of patient outcome, p53 measurement should be included as a stratification variable in treatment trials to determine whether these values predict outcome prospectively or whether they predict improved response to particular therapies.


..
Cathepsin-D as a Marker for Breast Cancer

Guideline

7. Present data are insufficient to recommend use of cathepsin-D measurements for management of patients with breast cancer.
..
Marker Definition

In 1979, a glycoprotein was discovered in the culture medium of hormone-dependent human breast cancer. It had a dependence on estrogens in that it could be increased by estrogens and inhibited by antiestrogens. It was discovered to be a 52-kd protein, which is a precursor to a lysosomal acidic protease, cathepsin-D. This proteolytic enzyme can react against basement membranes. Cathepsin-D also has mitogenic activity on MCF-7 cells that are estrogen- depleted. Further studies showed that cathepsin-D was relatively low in resting mammary cells but was elevated in malignant and benign proliferative breast diseases. These findings raised the suspicion that the cathepsin-D could both promote abnormal growth of cells as well as contribute to the metastatic potential of malignant cells through its disruption of the basement membrane and therefore might be a marker for a poor prognosis in breast cancer.365
..
Methodology

Cathepsin-D can be measured both on paraffin fixed tissue and in the cytosol of fresh tissues. A number of different techniques can be used to measure cathepsin-D on paraffin embedded tissue. Most of them are a modification of a technique described by Tatu.366 The microscopic slides are incubated with a primary antibody to cathepsin-D, and then processed using a modification of the avidin-biotin-complex (ABC) method of Hsu et al.367 There are a number of different antibodies to cathepsin-D that can be used in this technique. The slides are then reviewed by microscopic examination and compared with controls. A semiquantitative assessment is made by determining the percent of tumor cells that are positive for cathepsin-D.
..
Cathepsin-D can also be measured in the cytosol of tumor cells. There are different techniques for preparing the cytosols for the measurement of cathepsin-D but a common method is described by Maudelonde.368 After they are prepared the cytosols are analyzed for Cathepsin- D levels using either immunohistochemical techniques, enzymatic activity techniques, Western blotting, or double antibody techniques.

Literature Review and Analysis
..
It is difficult to determine the prognostic importance of cathepsin-D.369 There are a number of techniques used to measure cathepsin-D that have different levels of sensitivity and that measure different parts of the molecule. The most widely used assay for cathepsin-D is with the monoclonal antibody M1G8. A review of the studies that evaluate the prognostic properties of cathepsin-D as measured by this monoclonal antibody show that the data are difficult to interpret. Of the seven studies that specifically examine the prognostic significance of cathepsin- D in node-negative patients, three are positive316,370,371 and four are negative. 372-375 In the four studies specifically examining the prognostic significance in node-positive patients, all show that this is a predictor of overall survival and disease-free survival. 366,372,375,376 However, only one of these was confirmed by multivariate analysis. By using a polyclonal cathepsin-D in Western blotting, Tandon et al365 showed cathepsin-D to be the strongest independent predictor of prognosis in primary breast cancers.
..
In reviewing these studies, there are a number of different cut points used. This makes the interpretation of the data very difficult. There are also no studies that look at the concordance between the level of cathepsin-D as measured either from paraffin-embedded tissue or from the cytosol of fresh or frozen tissue.

Multiple studies were reviewed to derive the guidelines for cathepsin-D.316,358-382 The only outcome that was addressed was the ability to predict the overall survival and disease-free survival from the level of the cathepsin-D that was measured in the primary tumor. The level of evidence in these studies was generally a level III or IV. On the basis of these studies, the Panel thinks that the utility of cathepsin-D should not be used at this time in the routine management of a women with breast cancer.
..
Areas for Future Research

Prospective studies are needed in large numbers of patients to determine the role of cathepsin-D as an independent predictor for prognosis in women with breast cancer. Prospective studies are also necessary to determine the relationship of cathepsin-D to other risk factors in breast cancer. It is possible that a woman's ultimate risk for recurrence of breast cancer may not lie in a single variable, but in an interplay of many variables with one another. Because there are a number of different assays to measure cathepsin-D studies, determining the method that gives the most reproducible and useful data is also indicated.


..
Appendix

---

..
www.asco.org

c Copyright 2002 American Society of Clinical Oncology All rights reserved worldwide.