Originally published as JCO Early Release 10.1200/JCO.2005.01.4746 on August 29 2005

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Espinosa, E. Articles by Barón, M. G. Search for Related Content PubMed PubMed Citation Articles by Espinosa, E. Articles by Barón, M. G.

Breast Cancer Prognosis Determined by Gene Expression Profiling: A Quantitative Reverse Transcriptase Polymerase Chain Reaction Study

E. Espinosa, J.A. Fresno Vara, A. Redondo, J.J. Sánchez, D. Hardisson, P. Zamora, F. Gómez Pastrana, P. Cejas, B. Martínez, A. Suárez, F. Calero, M. González Barón

From the Service of Medical Oncology, Hospital La Paz; and Translational Oncology Unit and Department of Preventive Medicine, School of Medicine, Universidad Autónoma; and the Services of Pathology and Gynecology, Hospital La Paz, Madrid, Spain

Address reprint requests to E. Espinosa, MD, Servicio de Oncología Médica, Hospital La Paz, P° de la Castellana, 261—28046 Madrid, Spain; e-mail: eespinosa00{at}terra.es.

	ABSTRACT

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

 
PURPOSE: We sought to reproduce with quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR) the results obtained with a 70-gene expression profile that has been described previously in breast cancer.

PATIENTS AND METHODS: Frozen breast cancer samples from patients who were operated on were used to isolate tumor RNA. Ninety-six patients with stage I to II disease were included. Median age was 57 years (range, 27 to 80 years). Forty-eight patients had lymph node–negative and 48 lymph node–positive disease. qRT-PCR amplifications were performed and the results were correlated with clinical data.

RESULTS: After a minimum follow-up of 5 years, 25 patients had a relapse. The gene profile divided patients into two groups with poor and good prognosis. Significant differences with regard to grade of differentiation, size and hormone receptors were seen between the two groups. The gene profile was significantly associated with relapse-free survival and overall survival in the whole group of 96 patients. Multivariate analysis showed that only lymph node status and gene profile were significantly correlated to overall survival.

CONCLUSION: qRT-PCR reproduced the results obtained with microarrays for a prognostic gene profile in women with early-stage breast cancer.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

 
Breast cancer is a prevalent disease and one of the main causes of death from cancer. The main prognostic factor is the lymph node status, followed by size of the tumor and other parameters (such hormone-receptor status, HER-2 expression, grade of differentiation, or patient's age).^1-5 Theoretically, these factors should guide therapeutic decisions, but in practice they are not powerful enough and offer little help to identify people who benefit from adjuvant treatments. As a consequence, most women with early-stage disease undergo adjuvant chemotherapy, radiotherapy and, if appropriate, hormonal therapy, even when only a minority will experience a benefit.^6-8

New high-throughput technologies have opened the possibility to study the gene expression profile of tumors. In breast cancer, investigators have used DNA microarrays to obtain information about the prognosis^9-13 and the response to some cytotoxic agents.^14,15 Gene expression profiles may determine outcome even more accurately than the lymph node status.¹⁶ This was the result obtained with a 70-gene expression signature in 295 patients: the hazard ratio for distant metastases was 5.1 in the poor-prognosis group as compared with the good-prognosis group, and the ratio remained significant when groups were analyzed according to lymph node status. The genomic information obtained through this kind of study, however, is not still being used in the clinic for a number of reasons. First, results have been obtained from relatively small and selected groups of patients, which require confirmation in independent series. Even in the case of big series, experts agree that profiles must be validated in independent series before their widespread use.^17,18 Second, microarray technology is not available in many centers. Finally, the predictive value of these profiles should be further improved before they are widely accepted. Microarrays usually include a high number of genes, many of which may not be very relevant from a clinical point of view.

Quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR), also known as real-time PCR, consists of the detection of PCR products as they accumulate.¹⁹ It allows the absolute and relative quantification of specific RNA. On the basis of the 70-gene expression signature previously reported to indicate good or poor prognosis in breast cancer,¹⁶ we used qRT-PCR to measure the expression of these genes, as well as four additional genes related to prognosis, in breast cancers biopsies. Our objective was to reproduce the results obtained with the profile through an alternative method that could have some advantages over DNA-microarrays.

	PATIENTS AND METHODS

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

 
Patients and Clinical Features
Samples from patients who had an operation for early-stage invasive breast carcinoma between 1991 and 1997 were snap-frozen in liquid nitrogen and kept at –70°. Early-stage disease was defined as stage I or II, according to the American Joint Committee on Cancer (AJCC) 1997 staging system (in the 2002 version of the AJCC, patients with four or more involved lymph nodes have at least stage IIIA disease, but our study began in 2001). We retrospectively selected 96 patients who had available both full follow-up information and frozen tumor samples from our institution. No further selection criteria were applied. Patients were routinely followed every 4 months the first year, every 6 months from the second until the fifth year and once a year thereafter. The following data were taken from the clinical record: age at diagnosis, size of the primary tumor, grade of differentiation, number of positive lymph nodes, hormonal status, adjuvant treatment, date of relapse, sites of relapse, relapse-free survival (RFS), and overall survival (OS). Relapses were divided into low-risk if the patient survived 2 years or longer, and high-risk otherwise, regardless of the site of first relapse. Samples were processed only if full clinical data were available.

Median age was 57 years (range, 27 to 80 years). The tumor was up to 2 cm in size in 36 patients (37%) and larger in the other cases. Half of the patients were positive for lymph nodes and 75% were positive for hormone receptors (either estrogen, progesterone, or both). The histology was ductal in 80 cases and lobular in 16. Table 1 summarizes patient characteristics.

Institutional approval from our ethical committee was obtained for the conduct of the study. Patients had agreed that samples from their tumors could be used for future investigation, although they did not provide written permission for this particular study, which was performed years after the initial diagnosis.

Sample Processing
Frozen sections were stained with hematoxylin and eosin and analyzed by an experienced breast pathologist. Eligible samples had to include at least 90% of tumor cells. Total RNA was isolated from frozen sections with TRIzol Reagent (Invitrogen, Carlsbad, CA) and cleaned-up with use of the Qiagen RNeasy spin columns (Qiagen, Venlo, the Netherlands). Total RNA isolated was quantified using spectrophotometer OD₂₆₀ measurements. First-strand cDNA was synthesized from 1 µg of total RNA using the High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA). From the published 70-gene prognostic signature, 60 underwent TaqMan Gene Expression Assays (Applied Biosystems) products available by the time our study was begun. We could not find assay products for the 10 remaining genes at that time. Expression of mRNA for these previously described genes was measured in each specimen by qRT-PCR. The list of TaqMan Gene Expression Assays is shown in Appendix A (online only). We also decided to assess the possible predictive value for relapse of four genes whose expression has been previously related to prognosis: HER-2,²⁰ EGFR,²¹ PLAT,^22,23 and MUC-1.²⁴ qRT-PCR amplifications were performed with TaqMan Gene Expression Assays products in an ABI PRISM 7900 HT Sequence Detection System (Applied Biosystems). The reactions were carried out using the new Applied Biosystems TaqMan Low Density Arrays containing 50 µL TaqMan Universal PCR Master Mix (Applied Biosystems) and 50 µL of a cDNA template corresponding to 50 ng total RNA per channel of the microfluidic card.

The relative changes in gene expression were calculated by the Ct method using the Sequence Detection System (SDS) 2.1 software (Applied Biosystems). Derivations of the Ct equation, including assumptions, experimental design, and validation tests, have been described previously.²⁵ The Ct method gives the amount of target normalized to an endogenous reference and relative to a calibrator. We selected this method to allow comparison with the previously published microarray study, because most microarrays provide relative intensities of genes, whereas both absolute and relative quantification can be used to analyze data obtained through qRT-PCR.

The expression of each of the genes in all specimens was related to its expression in a reference RNA pool (made by pooling equal amounts of total RNA from each of the specimens) used as a calibrator. The ribosomal 18S RNA was used as the endogenous control.

Statistical Analysis
We applied a method for classifying breast tumor into good or poor prognostics categories based on the gene expression profile previously published.¹⁶ For this reason, good- and poor-prognosis groups in our study are defined exclusively by this profile. The method consisted of four steps: (1) for each gene, data sets from qRT-PCR arrays were normalized by measuring distance from the mean in standard deviation units; (2) using the cutoff value of 0, each gene was categorized as overexpressed or underexpressed; (3) the distance/similarity coefficient²⁶ for each patient was calculated with regard to the poor-prognosis gene profile previously described¹⁶ (the Russel & Rao similarity coefficient yields a proportion, "1" indicating full coincidence of a patient's profile with the poor-prognosis signature); and (4) different cutoff values for the similarity coefficient were performed to find the highest significance level in log-rank test for the disease-free survival (DFS) curve. This uses the so-called minimum P value method, where the cutpoint is taken such that the P value for the comparison of observations below and above the cutpoint is a minimum.²⁷ In the case of a positive result (ie, if the signature significantly predicted relapse), stepwise discriminant analysis^28,29 would be performed (both forwards and backwards) to try to reduce the number of genes.

Data on patients were analyzed from the date of surgery to the time of the relapse or death or the date on which data were censored. Both OS and RFS curves were obtained with the Kaplan-Meier product limit method for each prognostic factor and risk group. RFS was referred to either local or distant relapses from the primary tumor and did not include the occurrence of second tumors. Date of surgery was used to calculate both OS and RFS. Comparisons were made with the bilateral log-rank test. In order to compare patients' features in the two risk groups, the Mann-Whitney test was used because of lack of normality of continuous variables. ² and Fisher's exact tests were used to compare categoric variables. Cox proportional regression³⁰ was used for multivariate models, adjusting the hazard ratio of the risk group by the number of positives nodules, both for OS and DFS. The multivariate analysis included all variables having statistical value at the univariate analysis. Normalized data for all genes were included in the discriminant analysis, where the dependent variable was the inclusion in the group of low or high risk. All statistical analyses were carried out at 5% level of significance and a power of 80%. All analyses were performed with the Statistical Package for the Social Sciences, version 11.5 (SPSS Inc, Chicago, IL).

	RESULTS

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

 
Twenty-five patients (26%) had a relapse. There were eleven high-risk relapses (survival < 2 years from first relapse) and fourteen low-risk relapses. First relapse was local in eight patients and distant in 17. After a median follow-up of 70 months (range, 18 to 157 months), the median OS has not been reached. Patients were divided into two groups—good and poor prognosis—according to the gene expression in their tumors. Similarity coefficient values higher of 0.47 defined patients in the poor-prognosis group because this cutoff value provided a minimum P value for the log-rank test related to relapse (ie, highest prediction of relapse). For this reason, if the expression of > 47% of the genes of a patient coincided with that of the poor-prognosis signature in van't Veer et al,¹⁶ this patient was included in the poor-prognosis group and vice versa. These results were obtained with the 60 genes from the original signature. In the univariate analysis, the expression of HER-2, EGFR, PLAT and MUC-1 did not correlate with relapse, so we decided not to include them in the model. Table 2 shows the clinical features of patients in both groups. For raw data, see Appendix B (online only).

RFS and OS differed significantly between these two groups, as shown in Figure 1A and 1B. Estimated RFS at 70 months was 85% in the good-prognosis group and 62% for the poor-prognosis group (P = .03). The poor-prognosis profile correctly predicted 68% of relapses. OS at 70 months was 97% and 72%, respectively (P = .002). When divided by nodal status, the gene profile did not delineate significant differences in node-negative patients (P = .17 for RFS and P = .44 for OS; Fig 2). However, differences appeared in node-positive patients (P = .07 and P = .003, respectively; Fig 2). Of note, OS for patients with positive lymph nodes and a favorable gene profile was 100%. The reason for that lies in that relapses in these patients were associated with a prolonged survival (ie, they were low-risk relapses). We also limited the analysis to women older than 52, who accounted for two thirds of our sample: OS (P = .15) and RFS (P = .34) did not differ between good and poor-prognosis groups.

View larger version (19K): [in this window] [in a new window]   Fig 1. (A) Overall survival and (B) relapse-free survival in groups with a good- or poor-prognosis profile.

View larger version (26K): [in this window] [in a new window]   Fig 2. (A and C) Overall survival and (B and D) relapse-free survival in patients with N0 (A and B) or N+ (C and D) disease according to the gene profile.

The multivariate analysis included age, hormone status, lymph node status, number of positive lymph nodes ( 3 versus > 3) and gene profile. Only the lymph node status (hazard ratio, 1.2; 95% CI, 1.09 to 1.36) and the gene profile (hazard ratio, 6.3; 95% CI, 1.28 to 31.07) had a significant influence on OS (ie, they were independent prognostic variables for survival). The number of positive lymph nodes ( 3 versus >3) (hazard ratio, 1.13; 95% CI, 1.05 to 1.25) and again the gene profile (hazard ratio, 2.74, 95% CI, 1.13 to 6.61) were independent prognostic variables for RFS. We tried to reduce the number of genes with predicting relapse, so the sample was randomly split into two groups, one with 50 patients for the discriminant analysis, and the other with 46 patients for the validation, if indicated. Discriminant analysis could not reduce significantly the number of genes with predictive capacity for relapse: all the combinations with fewer genes obtained a Pvalue higher than that of the 60-gene signature. Principal component (factorial) analysis was also performed, with a negative result.

	DISCUSSION

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

 
In this study, we used qRT-PCR to reproduce the results obtained with a 70-gene prognostic signature in patients with early breast cancer. Both RFS and OS were significantly longer in patients harboring a good-prognosis gene profile. The gene signature was first described by using microarray technology.¹⁶ Although microarrays are an excellent tool for initial target discovery, there is a broadly recognized variability in microarray results depending on the user and platform.³¹ We used qRT-PCR, which has some advantages over microarrays. qRT-PCR requires smaller quantities of valuable tumor tissue and provides accurate, reproducible, and quantitative results.^32,33 Moreover, recent reports suggest the possibility to quantify gene expression with the use of sections of routinely prepared blocks of fixed, paraffin-embedded tumor tissue, which represent the most abundant source of tissue specimens associated with clinical records.^34,35 For these reasons, if a profile containing a limited number of genes were accepted for use in the clinical practice, qRT-PCR would be the technique of choice.

The original gene signature was validated in women younger than 52 years,¹⁶ and younger age is an unfavorable prognostic factor.^36,37 Our patients' median age coincide with that of the general population of patients with breast cancer. Also of note, the original signature was validated for distant relapses, whereas we included both local and distant relapses. As a local relapse usually heralds distant spread of the disease, this may add value to the profile. This indicates that the profile predicts both local and distant relapses in the general population of women with breast cancer. In the poor-prognosis group, most patients survived less than 2 years after relapse, regardless of the site of first relapse. As opposite, patients in the good prognosis group usually had low-risk relapses and survived longer than 2 years after relapse (Table 2). It is striking that patients were equally distributed in good-/poor-prognosis groups, as well as positive/negative lymph nodes: this happened by chance, because we did not performed further selection of patients if they fulfilled the inclusion criteria.

Even when our patients presented with some adverse prognostic factors (50% positive lymph nodes, 63% tumors > 2 cm), there were very few relapses. After a median follow-up of 70 months, RFS and OS were excellent. These favorable results should have decreased the chances of validating a prognostic marker, which suggests that the gene profile is a powerful predictor.

Subgroup analysis in our study should be considered with caution because of the limited number of patients. In patients with N0 disease, the profile was not useful, maybe because most patients received CMF adjuvant chemotherapy (unlike those in the previous study¹⁶). This could suggest that, at least in patients with N0 disease, adjuvant therapy could partially abrogate the poor outcome defined by the gene profile. Other investigators have found that, in the node-negative setting, classical clinical markers have a similar power in breast cancer prognosis as microarray gene expression profilers.³⁸ On the other hand, we cannot rule out the possibility that this gene profile is less useful in women older than 52 years. For this reason, we think that further studies should be performed in postmenopausal patients.

Microarrays and RT-PCR studies of previously nonvalidated profiles usually divide the sample into two groups, one to build and optimize a prediction model and the second to validate it. However, because the profile had been previously validated in the original microarray study,¹⁶ we think that the possibility of a chance association is low. Our main objective was not to find a new signature, but to validate a previous one. Even so, unsupervised clustering was performed (data not shown), but too many subgroups resulted, so results would have been unreliable.

Although it was not our primary objective, we tried to reduce the number of genes, because a profile with fewer genes would be desirable, but discriminant analysis did not allow this reduction. A different profile with only 16 genes has yielded conflicting results.³⁹ The reason for that may lie in the heterogeneity of breast cancer. Microarray studies have shown that what we call breast cancer is in fact a group of diseases with different outcomes.¹¹ This could hinder the prediction of prognosis with few genes, unlike the case of lymphoma³³ or even lung adenocarcinoma.³² Another reason to explain why we could not reduce the list of useful genes in our study might be the limited number of patients. A cutoff value of 0.47 for the coefficient suggests that the inclusion of further patients could have allowed the identification of a smaller set of genes predicting relapse. Moreover, two recent studies using expression levels of the 70-gene signature strongly suggest that the same information can be obtained with fewer genes.^40,41 Quantification of gene expression could be of some help in this regard but, although we used a quantitative technique, we categorized gene expression levels instead of using pure quantitative values, because our objective was to confirm previous results in an independent series. Considering the small number of patients in our study and their heterogeneity, cutoff points on the similarity coefficient and odds ratio cannot be regarded as definitive.

The incorporation of new genes into this profile might improve its accuracy, but this remains hypothetical and will certainly be the matter of future studies. We studied whether HER-2, EGFR, PLAT, and MUC-1 would be of some help, but even when they have been shown to have prognostic value individually, the information provided by the profile as a whole was more consistent. We cannot discard a sampling error that would have impeded validation of these four genes as prognostic markers, although an alternative explanation is that, given the limited number of patients, only the most powerful marker stood out (ie, the gene signature).

This gene profile could help deciding about the use of adjuvant treatment. Patients with very good prognosis—negative lymph nodes plus favorable gene profile—may not need chemotherapy, but this should be demonstrated in prospective studies. Studies like ours do not allow distinction between prognostic and predictive markers because the majority of patients receive some kind of adjuvant treatment; hence the need for phase III trials where a subgroup of patients receives no treatment. The European Organisation for Research and Treatment of Cancer is planning to initiate a trial in which the risk for women with node-negative disease will be determined either with St Gallen's index or the 70-gene profile previously published: women at low risk will not receive adjuvant chemotherapy. We do not think that this profile is now ready for clinical use outside a clinical trial, because it does not vastly improve the information provided by the nodal status, even considering that it was an independent prognostic factor. For the future, we need to refine our molecular tools. Microarrays and PCR studies could allow the incorporation of further genes that enhance the accuracy of the profile, as well as genes predicting resistance to cytotoxic agents or response to targeted molecules. Also, a quantitative technique such as qRT-PCR may be of help if further studies demonstrate that gene quantification offers additional prognostic information.

In summary, we used qRT-PCR to reproduce the results obtained with microarrays for a prognostic gene profile in patients with breast cancer. Further studies are needed to incorporate this profile into the clinical practice.

	Appendix A

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

 

	Appendix B

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

	Authors' Disclosures of Potential Conflicts of Interest

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

	Acknowledgment

 
We thank Verónica Torres, Rocío López, Esther Díaz and Paloma Nebreda for their assistance in the preparation of the RNA.

	NOTES

 
Supported by Grant No. FIS 020568, a grant from Red de Centros de Epidemiología y Salud Pública (Universidad Autónoma of Madrid), Grant No. RTICC C03/10 from Instituto de Salud Carlos III, and unrestricted grants from Amgen, AstraZeneca, Bristol-Myers Squibb, Lilly, Novartis, Roche, Sanofi-Aventis, and Schering-Plough.

Presented at the 27th Annual San Antonio Breast Cancer Symposium, San Antonio, TX, December 7-11, 2004.

E.E. and J.A.F.V. contributed equally to this study.

Terms in blue are defined in the glossary, found at the end of this issue and online at www.jco.org.

Authors' disclosures of potential conflicts of interest are found at the end of this article.

	REFERENCES

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix A Appendix B Authors' Disclosures of... REFERENCES

 
1. Carter CL, Allen C, Henson DE: Relation of tumor size, lymph node status, and survival in 24,740 breast cancer cases. Cancer 63:181-187, 1989[CrossRef][Medline]

2. Galea MH, Blamey RW, Elston CE, et al: The Nottingham Prognostic Index in primary breast cancer. Breast Cancer Res Treat 22:207-219, 1992[CrossRef][Medline]

3. Mirza AN, Mirza NQ, Vlastos G, et al: Prognostic factors in node-negative breast cancer: A review of studies with sample size more than 200 and follow-up more than 5 years. Ann Surg 235:10-26, 2002[CrossRef][Medline]

4. Press MF, Bernstein L, Thomas PA, et al: HER-2/neu gene amplification characterized by fluorescence in situ hybridization: Poor prognosis in node-negative breast carcinomas. J Clin Oncol 15:2894-2904, 1997[Abstract]

5. Weiss RB, Woolf SH, Demakos E, et al: Natural history of more than 20 years of node-positive primary breast carcinoma treated with cyclophosphamide, methotrexate, and fluorouracil-based adjuvant chemotherapy: A study by the Cancer and Leukemia Group B. J Clin Oncol 21:1825-1835, 2003[Abstract/Free Full Text]

6. Coleman RE: Current and future status of adjuvant therapy for breast cancer. Cancer 97:880-886, 2003[CrossRef][Medline]

7. Mincey BA, Palmieri FM, Perez EA: Adjuvant therapy for breast cancer: Recommendations for management based on consensus review and recent clinical trials. Oncologist 7:246-250, 2002[Abstract/Free Full Text]

8. Winer EP, Hudis C, Burstein HJ, et al: American Society of Clinical Oncology technology assessment working group update: Use of aromatase inhibitors in the adjuvant setting. J Clin Oncol 21:2597-2599, 2003[Free Full Text]

9. Bertucci F, Nasser V, Granjeaud S, et al: Gene expression profiles of poor-prognosis primary breast cancer correlate with survival. Hum Mol Genet 11:863-872, 2002[Abstract/Free Full Text]

10. Huang E, Cheng SH, Dressman H, et al: Gene expression predictors of breast cancer outcomes. Lancet 361:1590-1596, 2003[CrossRef][Medline]

11. Sorlie T, Perou CM, Tibshirani R, et al: Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A 98:10869-10874, 2001[Abstract/Free Full Text]

12. Sotiriou C, Neo SY, McShane LM, et al: Breast cancer classification and prognosis based on gene expression profiles from a population-based study. Proc Natl Acad Sci U S A 100:10393-10398, 2003[Abstract/Free Full Text]

13. West M, Blanchette C, Dressman H, et al: Predicting the clinical status of human breast cancer by using gene expression profiles. Proc Natl Acad Sci U S A 98:11462-11467, 2001[Abstract/Free Full Text]

14. Ayers M, Symmans WF, Stec J, et al: Gene expression profiles predict complete pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 22:2284-2293, 2004[Abstract/Free Full Text]

15. Chang JC, Wooten EC, Tsimelzon A, et al: Gene expression profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer. Lancet 362:362-369, 2003[CrossRef][Medline]

16. van de Vijver MJ, He YD, van't Veer LJ, et al: A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 347:1999-2009, 2002[Abstract/Free Full Text]

17. Dobbin KK, Beer DG, Meyerson M, et al: Interlaboratory comparability study of cancer gene expression analysis using oligonucleotide microarrays. Clin Cancer Res 11:565-572, 2005[Abstract/Free Full Text]

18. Michiels S, Koscielny S, Hill C: Prediction of cancer outcome with microarrays: A multiple random validation strategy. Lancet 365:488-492, 2005[CrossRef][Medline]

19. Mocellin S, Rossi CR, Pilati P, et al: Quantitative real-time PCR: A powerful ally in cancer research. Trends Mol Med 9:189-195, 2003[CrossRef][Medline]

20. Ross JS, Fletcher JA, Linette GP, et al: The Her-2/neu gene and protein in breast cancer 2003: Biomarker and target of therapy. Oncologist 8:307-325, 2003[Abstract/Free Full Text]

21. Nicholson RI, McClelland RA, Finlay P, et al: Relationship between EGF-R, c-erbB-2 protein expression and Ki67 immunostaining in breast cancer and hormone sensitivity. Eur J Cancer 29A:1018-1023, 1993

22. Bertucci F, Viens P, Hingamp P, et al: Breast cancer revisited using DNA array-based gene expression profiling. Int J Cancer 103:565-571, 2003[CrossRef][Medline]

23. Corte MD, Verez P, Rodriguez JC, et al: Tissue-type plasminogen activator (tPA) in breast cancer: Relationship with clinicopathological parameters and prognostic significance. Breast Cancer Res Treat 90:33-40, 2005[CrossRef][Medline]

24. Resetkova E, Gonzalez-Angulo AM, Sneige N, et al: Prognostic value of P53, MDM-2, and MUC-1 for patients with inflammatory breast carcinoma. Cancer 101:913-917, 2004[CrossRef][Medline]

25. Livak K, Schmittgen T: Analysis of relative gene expression data using real-time quantitative PCR and the 2-CT method. Methods 25:402-408, 2001[CrossRef][Medline]

26. Gower JC: A general coefficient of similarity and some of its properties. Biometrics 27:857-872, 1971[CrossRef]

27. Schumacher M, Holländer N, Schwarzer G, et al: Prognostic factor studies, in Crowley J (ed): Handbook of Statistics in Clinical Oncology. New York, NY, Marcel Dekker, 2001, pp 321-378

28. Anderson T: An Introduction to Multivariate Statistical Analysis (2nd ed.). New York, NY, Wiley, 1984

29. Barnett V: Interpreting Multivariate Data. New York, NY, Wiley, 1981

30. Cox DR: Regression models and life-tables. J. R. Stat Soc [B] 34:187-200, 1972

31. Miller LD, Long PM, Wong L, et al: Optimal gene expression analysis by microarrays. Cancer Cell 2:353-361, 2002[CrossRef][Medline]

32. Endoh H, Tomida S, Yatabe Y, et al: Prognostic model of pulmonary adenocarcinoma by expression profiling of eight genes as determined by quantitative real-time reverse transcriptase polymerase chain reaction. J Clin Oncol 22:811-819, 2004[Abstract/Free Full Text]

33. Lossos IS, Czerwinski DK, Alizadeh AA, et al: Prediction of survival in diffuse large-B-cell lymphoma based on the expression of six genes. N Engl J Med 350:1828-1837, 2004[Abstract/Free Full Text]

34. Paik S, Shak S, Tang G, et al: A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 351:2817-2826, 2004[Abstract/Free Full Text]

35. Ma XJ, Wang Z, Ryan PD, et al: A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell 5:607-616, 2004[CrossRef][Medline]

36. Albain KS, Allred DC, Clark GM: Breast cancer outcome and predictors of outcome: Are there age differentials? J Natl Cancer Inst Monogr, 35-42, 1994

37. Nixon AJ, Neuberg D, Hayes DF, et al: Relationship of patient age to pathologic features of the tumor and prognosis for patients with stage I or II breast cancer. J Clin Oncol 12:888-894, 1994[Abstract/Free Full Text]

38. Eden P, Ritz C, Rose C, et al: "Good Old" clinical markers have similar power in breast cancer prognosis as microarray gene expression profilers. Eur J Cancer 40:1837-1841, 2004

39. Fiets WE, Nortier JW: A multigene RT-PCR assay used to predict recurrence in early breast cancer: Two presentations with contradictory results. Breast Cancer Res 6:185-187, 2004[CrossRef][Medline]

40. Glinsky GV, Higashiyama T, Glinskii AB: Classification of human breast cancer using gene expression profiling as a component of the survival predictor algorithm. Clin Cancer Res 10:2272-2283, 2004[Abstract/Free Full Text]

41. Bair E, Tibshirani R: Semi-supervised methods to predict patient survival from gene expression data. PLoS Biol 2:511-522, 2004

Submitted February 2, 2005; accepted July 7, 2005.

This article has been cited by other articles:

				C.-C. Huang and M. Bredel Use of Gene Signatures to Improve Risk Estimation in Cancer JAMA, April 2, 2008; 299(13): 1605 - 1606. [Full Text] [PDF]

				I. G. Yulug and B. Gur-Dedeoglu Functional genomics in translational cancer research: focus on breast cancer Brief Funct Genomic Proteomic, March 7, 2008; (2008) eln009v1. [Abstract] [Full Text] [PDF]

				L. Harris, H. Fritsche, R. Mennel, L. Norton, P. Ravdin, S. Taube, M. R. Somerfield, D. F. Hayes, and R. C. Bast Jr American Society of Clinical Oncology 2007 Update of Recommendations for the Use of Tumor Markers in Breast Cancer J. Clin. Oncol., November 20, 2007; 25(33): 5287 - 5312. [Abstract] [Full Text] [PDF]

				S. K Gruvberger-Saal, H. E Cunliffe, K. M Carr, and I. A Hedenfalk Microarrays in breast cancer research and clinical practice - the future lies ahead Endocr. Relat. Cancer, December 1, 2006; 13(4): 1017 - 1031. [Abstract] [Full Text] [PDF]

				S. Loi, C. Sotiriou, M. Buyse, E. Rutgers, L. V. Veer, M. Piccart, and F. Cardoso Molecular Forecasting of Breast Cancer: Time to Move Forward With Clinical Testing J. Clin. Oncol., February 1, 2006; 24(4): 721 - 722. [Full Text] [PDF]

				J. D. Brenton, A. A. Ahmed, and C. Caldas In Reply: J. Clin. Oncol., February 1, 2006; 24(4): 722 - 723. [Full Text] [PDF]

				S. Koscielny, S. Michiels, V. Boige, and C. Hill Validation of Microarray Data by Quantitative Reverse-Transcriptase Polymerase Chain Reaction J. Clin. Oncol., December 20, 2005; 23(36): 9439 - 9440. [Full Text] [PDF]

				P. Workman and P. G. Johnston Genomic Profiling of Cancer: What Next? J. Clin. Oncol., October 10, 2005; 23(29): 7253 - 7256. [Full Text] [PDF]

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Espinosa, E. Articles by Barón, M. G. Search for Related Content PubMed PubMed Citation Articles by Espinosa, E. Articles by Barón, M. G.