Originally published as JCO Early Release 10.1200/JCO.2006.05.6861 on August 8 2006

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 Rights & Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Hess, K. R. Articles by Pusztai, L. Search for Related Content PubMed PubMed Citation Articles by Hess, K. R. Articles by Pusztai, L. Social Bookmarking                            What's this?

Pharmacogenomic Predictor of Sensitivity to Preoperative Chemotherapy With Paclitaxel and Fluorouracil, Doxorubicin, and Cyclophosphamide in Breast Cancer

Kenneth R. Hess, Keith Anderson, W. Fraser Symmans, Vicente Valero, Nuhad Ibrahim, Jaime A. Mejia, Daniel Booser, Richard L. Theriault, Aman U. Buzdar, Peter J. Dempsey, Roman Rouzier, Nour Sneige, Jeffrey S. Ross, Tatiana Vidaurre, Henry L. Gómez, Gabriel N. Hortobagyi, Lajos Pusztai

From the Departments of Biostatistics and Applied Mathematics, Pathology, Breast Medical Oncology and Radiology, The University of Texas M.D. Anderson Cancer Center, Houston, TX; Breast Cancer Unit and Unité Propre de l'Enseignement Supérieur; Equipe d'Accueil 3535 of the Institut Gustave Roussy, Villejuif, France; Albany Medical College, Albany NY; Departamento de Medicina Instituto Nacional de Enfermedades Neoplásicas, Lima, Perú

Address reprint requests to Lajos Pusztai, MD, DPhil, Department of Breast Medical Oncology, The University of Texas M.D. Anderson Cancer Center, Unit 1354, PO Box 301439, Houston, TX 77230-1439; e-mail: lpusztai{at}mdanderson.org

	ABSTRACT

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
Purpose We developed a multigene predictor of pathologic complete response (pCR) to preoperative weekly paclitaxel and fluorouracil-doxorubicin-cyclophosphamide (T/FAC) chemotherapy and assessed its predictive accuracy on independent cases.

Patients and Methods One hundred thirty-three patients with stage I-III breast cancer were included. Pretreatment gene expression profiling was performed with oligonecleotide microarrays on fine-needle aspiration specimens. We developed predictors of pCR from 82 cases and assessed accuracy on 51 independent cases.

Results Overall pCR rate was 26% in both cohorts. In the training set, 56 probes were identified as differentially expressed between pCR versus residual disease, at a false discovery rate of 1%. We examined the performance of 780 distinct classifiers (set of genes + prediction algorithm) in full cross-validation. Many predictors performed equally well. A nominally best 30-probe set Diagonal Linear Discriminant Analysis classifier was selected for independent validation. It showed significantly higher sensitivity (92% v 61%) than a clinical predictor including age, grade, and estrogen receptor status. The negative predictive value (96% v 86%) and area under the curve (0.877 v 0.811) were nominally better but not statistically significant. The combination of genomic and clinical information yielded a predictor not significantly different from the genomic predictor alone. In 31 samples, RNA was hybridized in replicate with resulting predictions that were 97% concordant.

Conclusion A 30-probe set pharmacogenomic predictor predicted pCR to T/FAC chemotherapy with high sensitivity and negative predictive value. This test correctly identified all but one of the patients who achieved pCR (12 of 13 patients) and all but one of those who were predicted to have residual disease had residual cancer (27 of 28 patients).

	INTRODUCTION

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
Despite the critical importance of selecting the most effective adjuvant/neoadjuvant chemotherapy for an individual, diagnostic tests to guide selection of the optimal regimen for a particular patient are lacking.^1-4 Estrogen receptor (ER) –negative status, high grade, and high proliferative activity are histologic characteristics that tend to indicate more chemotherapy-sensitive cancer.^5-7 However, these clinicopathologic variables predict general chemotherapy sensitivity and therefore, have little potential to guide selection of a specific regimen. Neoadjuvant (preoperative) chemotherapy provides an opportunity to directly assess tumor response to therapy. Furthermore, complete eradication of all invasive cancer from the breast and regional lymph nodes, pathologic complete response (pCR), is associated with excellent long-term cancer-free survival.^8,9 Our goal was to evaluate gene expression profiling as a potential tool to predict who may achieve pCR to sequential anthracycline paclitaxel preoperative chemotherapy. We selected a complex multidrug regimen for study because combination chemotherapy represents the current clinical standard for patients who require systemic cytotoxic treatment. Also, gene signatures that are predictive of response to individual drugs may not fully capture sensitivity to combination chemotherapy. In an earlier small study (n = 42) we reported that it is possible to perform gene expression profiling on fine-needle biopsies of newly diagnosed breast cancer and that a multigene predictor of pCR has been developed promising predictive performance.¹⁰ The current study represents an extension of the earlier work. We used a larger sample size for predictor discovery (n = 82) and used a commercially available standard gene expression profiling technology. We systematically examined the performance of a large number of potential predictors in cross validation in the training data and selected one final predictor for independent validation in 51 new cases. In the current study, we also examined the reproducibility of prediction results in 31 replicate experiments.

	PATIENTS AND METHODS

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
Patients and Samples
This biomarker discovery trial was conducted at the Nellie B. Connally Breast Center of the University of Texas M.D. Anderson Cancer Center (MDACC) in Houston, TX, and at the Instituto Nacional de Enfermedades Neoplásicas (INEN) in Lima, Peru. During this research, patients were asked to undergo pretreatment fine-needle aspiration (FNA) of the primary breast tumor or ipsilateral axillary metastasis before starting chemotherapy as part of an ongoing pharmacogenomic marker discovery program.¹¹ The aspiration was performed using a 23- or 25-gauge needle. Cells from two to three passes were collected in vials containing 1 mL RNA later solution (Ambion, Austin, TX) and stored at –80°C. FNA samples on average contain 80% neoplastic cells and contain little or no stromal cells or normal breast epithelium.¹² Gene expression data generated from FNAs capture the molecular characteristics of the invasive cancer, including the molecular class.¹³ Approximately 70% of all aspirations yielded at least 1 µg total RNA required for the gene expression profiling. The main reason for failure to obtain sufficient RNA was acellular aspirations. One hundred thirty-one consecutively accrued patients with at least 1 µg RNA were included in this analysis. All patients received 24 weeks of sequential paclitaxel and fluorouracil-doxorubicin-cyclophosphamide preoperative chemotherapy. Metallic markers were placed under radiologic guidance in the shrinking tumor bed for any patient interested in breast-conserving surgery, whose tumor became less than 2 cm measured by ultrasonogram or mammogram during the course of treatment. At the completion of neoadjuvant chemotherapy, all patients had modified radical mastectomy or lumpectomy and sentinel lymph node biopsy or axillary node dissection as determined appropriate by the surgeon. Grossly visible residual cancer was measured and representative sections of the cross sectional area were submitted for histopathologic study. When there was not grossly visible residual cancer, the slices of the specimen were radiographed, and all areas of radiologically and/or architecturally abnormal tissue were entirely submitted for histopathologic study. pCR was defined as no residual invasive cancer in the breast or lymph nodes. Residual in situ carcinoma without invasive component was also considered a pCR. This study was approved by the institutional review boards of MDACC and INEN, and all patients signed an informed consent for voluntary participation. Clinical characteristics of the patients are presented in Table 1.

Data Analysis
dCHIP V1.3 (http://www.dchip.org) software was used to generate probe level intensities and quality measures including median intensity, percent of probe set outliers, and percent of single probe outliers for each chip. This program normalizes all arrays to one standard array that represents a chip with median overall intensity. This reference chip and the normalization procedure is available online (http://www.bioinformatics.mdanderson.org/pubdata.html). Normalized gene expression values were transformed to the log₁₀ scale for analysis. To identify differentially expressed genes between cases with pCR and those with residual disease (RD) genes, two-sample, unequal-variance t tests were performed and genes rank ordered by P values. Beta uniform mixture analysis (BUM) of the P values showed a nonuniform distribution and was used to estimate false discovery rates (FDRs).¹⁵ We constructed multigene classifiers using combinations of the most informative genes and several different class prediction algorithms including support vector machines with linear, radial, and polynomial kernels (SVM), Diagonal Linear Discriminant Analysis (DLDA), and K-nearest neighbor (KNN) using Euclidean distance.¹⁶ Monte Carlo cross validation (CV) was performed by repeated iteration (n = 100) of stratified random sampling to estimate the prediction performance of the different classifiers in the training data and to facilitate selection of a single classifier for independent validation. Stratification was performed to insure that the relative proportion of outcomes sampled in both cross-validation training and test sets was similar to the original proportions for the full training data. We performed complete CV including gene selection in each iteration to avoid selection bias.¹⁷

To assess whether the performance of a chosen predictor differed significantly from what chance alone could produce, a random-label permutation test was performed.¹⁸ During this process, the outcome label of each case (ie, pCR v RD) was randomly reassigned. One thousand such data sets with randomly permutated labels were created, and predictors were generated with repeat selection of the top probes with each iteration. The observed prediction error rate for the data set is compared with the distribution of the error rates observed with the randomly permutated data sets to calculate a permutation P value.

Classifier performance on the validation data were assessed by using the area under the receiver operating characteristic (ROC) curve (AUC) and its compliment, the area above the curve (AAC; AAC = 1–AUC). The ROC curve is a graphical display of the false-positive rate and the true-positive rate from multiple classification rules.¹⁹ The ROC curve arises when a continuous predictor value is calculated for each subject for a broad range of thresholds. A case is called test-positive (eg, predicted to have pCR) if the threshold is above a defined value. The total area under the ROC curve is a summary measure of the test's ability to correctly classify those with and without the outcome of interest. An AUC of 1 represents a perfect test; an AUC of 0.5 represents a test no better than random prediction.

We also evaluated the performance of a multivariate clinical predictor. The predictor utilized clinical variables only to predict pCR and was constructed similarly to the genomic predictor. Informative variables were identified through logistic regression performed on the training set, and a multivariate predictor was constructed using a DLDA machine learning algorithm identical to that used for the pharmacogenomic predictor. A three-variable–based (nuclear grade, age, and ER status) DLDA prediction model was tested on the independent validation set and its performance compared to the multigene predictor using ROC analysis. A combined clinical plus genomic model was also assessed that included age, dichotomized versions of ER status determined by routine immunohistochemistry ( 10% of cells with positive nuclear staining versus < 10%) and grade (grade 3 v grades 1 or 2) as variables in addition to the genes already included in the pharmacogenomic prediction models.

Predictor learning was also evaluated for selected pharmacogenomic models. One hundred and twenty cases from the training and validation sets were included in this analysis. Nine different training sample sizes ranging from 20 to100 by increments of 10 were created. The test set size was kept constant at 20. Stratified sampling was used to preserve the ratio of pCR and RD cases in each training and test sets. For each training set, feature selection was repeated to identify the top 30 informative genes. The area above the ROC curve and the misclassification error rate were calculated for 50 random sample sets generated for each nine training set sizes (n = 20, 30, 40, ...100). The following learning curve model was fit to the resulting AAC and misclassification error rate (MER) values:

Gene expression data are available at http://bioinformatics.mdanderson.org/pubdata.html.

	RESULTS

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
Pathologic Response Rate and Selection of Informative Genes
The overall pCR rate in the 133 patients was 26% (n = 34), which is consistent with results from a larger randomized study using the same preoperative therapy.²⁰ The first 82 cases were used as a training set (including 21 cases of pCR) to develop pharmacogenomic and clinical predictors. This training set size was determined by fitting learning curves to gene expression data (see below). The next 51 consecutive cases, including 13 cases of pCR, were used as independent test set to assess accuracy. In univariate analysis including clinical variables, age, nuclear grade, and ER status were significantly associated with pCR in the training set. In a logistic regression model including age, pretreatment T stage, N stage, nuclear grade, ER status, and HER2 status as predictors, only ER status (P = .0037) and age (P = .012) remained significant. To select informative genes for response prediction, we compared gene expression data from cases with pCR to those with RD in the training set. Setting the FDR to 5% resulted in 395 probe sets, 1% in 56 probe sets (corresponding to 49 genes) and 0.5% in 31 probe sets (27 genes). Table 2 presents the top 31 probe sets that were differentially expressed between the 2 response groups at an FDR of 0.5%. When similar analyses were performed separately for ER-positive (n = 35) and ER-negative cases (n = 47), BUM analysis of P values indicated the possibility of high false discovery rates even with small P values. Small sample size and relatively few events, particularly in the ER-positive group (three pCR), may have prevented the identification of differentially expressed genes with confidence within these subsets. Indeed, when predictors were constructed from genes separately associated with pCR in ER-negative and ER-positive cancers, these predictors were inferior to predictors that used genes selected from all cases.

View this table: [in this window] [in a new window]   Table 2. Top 31 Differentially Expressed Probe Sets by Unequal-Variance t Test (n = 82, false discovery rate 0.5%)

View larger version (30K): [in this window] [in a new window] [PowerPoint Slide for Teaching]   Fig 1. Mean area above the receiver operating characteristic (ROC) curves plotted against the number of top genes included in the classifiers. Complete 5-fold cross validation results (means over the 100 iterations) for 20 classifier algorithms including different numbers of probe sets (39 gene sets) are shown. Green and black horizontal dotted lines indicate the mean +/– 2SD for the nominally best Diagonal Linear Discriminant Analysis (DLDA) classifier with 30 probe sets that was selected for independent validation. polynomial kernels (SVM), and K-nearest neighbor

To estimate the training set size that is necessary to develop a pharmacogenomic predictor that operates near to its plateau, we examined how the performance characteristics of DLDA-30 changed as the training set size increased. One would expect some improvement in prediction accuracy and narrowing of confidence intervals as the predictor is developed and trained on larger and larger sample sets. The steepness of this "learning curve" may help determine a reasonable sample size for predictor discovery.²¹ Figure 2 shows the change in mean AAC as the classifier was developed from 20 to 100 cases in increments of 10. The results indicate a steady but modest improvement in performance as the training set size increases. Based on these observations a DLDA-30 predictor developed from 80 cases was projected to be only marginally inferior to a predictor developed from 200 cases. Essentially similar learning curves were obtained for KNN and SVM methods that all exhibited minimal improvement in projected performance beyond 60 to 100 training cases.

View larger version (13K): [in this window] [in a new window] [PowerPoint Slide for Teaching]   Fig 2. Learning curve for Diagonal Linear Discriminant Analysis–30 classifier. Mean area above the receiver operating characteristic (ROC) curves (AAC) developed from 20 to 100 cases in increments of 10 are shown. Fifty individual point estimates of AAC (small dots) and their means (large dots) are plotted for each of the training set sizes. The projected AAC is 0.12 if 200 cases.

View larger version (21K): [in this window] [in a new window] [PowerPoint Slide for Teaching]   Fig 3. Receiver operating characteristic curves of three distinct pathologic complete response prediction models. The performance of the Diagonal Linear Discriminant Analysis–30 predictor and a predictor based on clinical variables and a combined clinical + pharmacogenomic prediction model are shown in the validation set (n = 51). ER, estrogen receptor; AUC, area under the curve.

View larger version (13K): [in this window] [in a new window] [PowerPoint Slide for Teaching]   Fig 4. (A) Diagonal Linear Discriminant Analysis–30 (DLDA-30) probe set model prediction outcome (black dots indicate misclassified cases). (B) Clinical (estrogen receptor status, grade, age) model prediction (black dots indicate misclassified cases). RD, residual disease; pCR, pathologic complete response.

View larger version (14K): [in this window] [in a new window] [PowerPoint Slide for Teaching]   Fig 5. Prediction agreement in replicate experiments. Thirty-one RNA specimens were profiled twice to assess reproducibility of prediction results. All but one of the 31 replicates yielded identical prediction outcome for the Diagonal Linear Discriminant Analysis–30 predictor. Prediction scores < 0 predict for residual disease (RD) and > 0 predict for pathologic complete response (pCR).

	DISCUSSION

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

We selected the nominally best predictor with the least number of genes for independent validation in 51 cases. This predictor, DLDA-30, includes 30 probe sets and uses DLDA to formulate outcome prediction rule. The predictor showed substantially higher sensitivity (92% v 61%) and slightly better NPV (96% v 86%) than a clinical variable (ER, grade, and age) –based model. The high sensitivity indicates that the predictor correctly identified almost all of the patients (92%) who actually achieved pCR. The positive predictive value (PPV) of the pharmacogenomic predictor was 52% (95% CI, 31% to 73%). Importantly, the lower bound of the 95% CI did not overlap with the 26% pCR rate observed with this regimen in unselected patients.¹⁷ This indicates that the predictor can define a patient population who is more likely to achieve pCR than the general patient population. However, the 52% PPV also indicates that many who were predicted to have pCR had a lesser response. This type of error may be considered acceptable in the adjuvant treatment setting. The NPV of the test was also high (96%; 95% CI, 82% to 100%), which indicates that less than 5% of test-negative patients (ie, predicted to have residual disease) achieved pCR. These performance statistics are similar, with regards to the NPV, and better with regards to PPV, than those seen with ER immunohistochemistry or HER2 gene amplification as predictive markers to endocrine or trastuzumab therapies, respectively. We also examined the reproducibility of the pharmacogenomic prediction results in 31 replicate experiments and observed a 97% concordance in prediction outcome for the DLDA-30 model and 100% concordance for the combined clinical and pharmacogenomic model. This indicates a very high level of technical reproducibility of pharmacogenomic predictions when the same RNA is used.

How do these results fit in with other recently reported molecular predictors of pCR? There is more than one method to predict probability of pCR. Patients with high recurrence score determined by the Oncotype Dx assay have greater probability to achieve pCR (to paclitaxel/FAC chemotherapy) than those with low recurrence score.²² Patients with basal-like breast cancer determined by hierarchical clustering using the "intrinsic gene list" also have higher probability of pCR than other molecular classes of breast cancer.¹³ Reassuringly, all of these methods tend to identify the same group of patients, those with ER-negative, high grade and highly proliferative tumors. However, they also seem to add some incremental predictive value to the known clinical characteristics. Appendix Tables 1 and 2 (online only) shows pCR rates in the training and validation sets as a function of ER-status and Appendix Figure 1 (online only) shows the correlation between Ki67 mRNA expression (probe sets 212022_s_at and 212023_s_at) and pCR. The discriminating value of these single gene variables is less than that of the combined model.

Predicting extreme chemotherapy sensitivity to paclitaxel/anthracycline therapy can be useful in some clinical situations. However, the greatest contribution will come from the molecular test that can discriminate between likelihoods of response to different chemotherapy regimens and can therefore guide selection of one treatment over another. To develop such predictor gene expression data from cohorts of patients who received different chemotherapy regimens will be needed. Pharmacogenomic research also has the potential to open new insights into the biology of breast cancer and treatment response. Indeed, some of the genes from our predictive signature seem to contribute to the mechanism of sensitivity to paclitaxel.²⁴

In summary, we developed a 30-probe set DLDA classifier that predicts pathologic response to preoperative paclitaxel/FAC chemotherapy with higher sensitivity (92% v 61%) than a clinical variable–based predictor. In an independent validation set of 51 patients, this test correctly identified all but one of the patients who achieved pCR, and all but one of those who were predicted to have residual disease had residual cancer.

	Appendix

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 

View larger version (21K): [in this window] [in a new window] [PowerPoint Slide for Teaching]   Appendix Fig 1. (A) Response versus Ki-67 expression (212022_s_at; n = 82). (B) Response versus Ki-67 expression (212022_s_at; n = 51). (C) Response versus Ki-67 expression (212023_s_at; n = 82). (D) Response versus Ki-67 expression (212023_s_at; n = 51). pCR, pathologic complete response; RD, residual disease.

	Authors' Disclosures of Potential Conflicts of Interest

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

Dollar Amount Codes (A) < $10,000 (B) $10,000-$99,900 (C) $100,000 (N/R) Not Required

	Author Contributions

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 

Conception and design: Kenneth R. Hess, W. Fraser Symmans, Gabriel N. Hortobagyi, Lajos Pusztai Financial support: Gabriel N. Hortobagyi, Lajos Pusztai Administrative support: Gabriel N. Hortobagyi, Lajos Pusztai Provision of study materials or patients: W. Fraser Symmans, Vicente Valero, Nuhad Ibrahim, Daniel Booser, Richard L. Theriault, Aman U. Buzdar, Peter J. Dempsey, Nour Sneige, Tatiana Vidaurre, Henry L. Gomez, Gabriel N. Hortobagyi, Lajos Pusztai Collection and assembly of data: W. Fraser Symmans, Jaime A. Mejia, Peter J. Dempsey, Roman Rouzier, Nour Sneige, Lajos Pusztai Data analysis and interpretation: Kenneth R. Hess, Keith Anderson, W. Fraser Symmans, Roman Rouzier, Jeffrey S. Ross, Lajos Pusztai Manuscript writing: Kenneth R. Hess, Lajos Pusztai Final approval of manuscript: Kenneth R. Hess, Keith Anderson, W. Fraser Symmans, Vicente Valero, Nuhad Ibrahim, Jaime A. Mejia, Daniel Booser, Richard L. Theriault, Aman U. Buzdar, Peter J. Dempsey, Roman Rouzier, Nour Sneige, Jeffrey S. Ross, Tatiana Vidaurre, Henry L. Gomez, Gabriel N. Hortobagyi, Lajos Pusztai

 

	GLOSSARY

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 

ROC (receiver operating characteristic) curves:

ROC curves plot the true positive rate (sensitivity) against the false-positive rate (1-specificity) for different cut-off levels of a test. The area under the curve is a measure of the accuracy of the test. An area of 1.0 represents a perfect test (all true positives), whereas an area of 0.5 represents a worthless test.

FDR (false discovery rate):

FDR is a statistical method that is used to correct for multiple comparisons. FDR is the expected proportion of false positives (as opposed to the more traditional Type II error rate, which is the probability of any false positives).

Monte-Carlo cross validation:

Cross validation (CV) is a process using a dataset to build a prediction algorithm and to estimate how well it will perform on new but similar data. A portion of the data is used to build an algorithm while the remainder is used to estimate the performance. Classical k-fold CV randomly divides the data into k portions using each portion in turn as a test set and the remaining data as training sets. One random partitioning of the data is used. The performance on new data is estimated as the average performance over the k test sets. In Monte-Carlo CV, the partitions are randomly selected hundreds or thousands of times.

BUM (beta-uniform mixture analysis):

This model is used to quickly compute the FDR estimates for the analysis of microarray data by taking the distribution of individual p-values to follow a mixture of a Beta density function from the genes that differentially expressed and a uniform density function from the genes that are not differentially expressed.

Pharmacogenomic:

The study of how a person's genome can affect their reaction to medications.

Class prediction algorithms:

Computer programs used to combine known data on samples to predict unknown characteristics.

	ACKNOWLEDGMENTS

 
We would like to acknowledge the contributions of Stephen Tirrell, James Stec, Mark Ayers and Edwin Clark who contributed gene expression data and several useful comments to this project while employed at Millennium Pharmaceuticals Inc (Cambridge, MA).

	NOTES

 
published online ahead of print at www.jco.org on August 7, 2006.

Supported by grants from the National Cancer Institute (RO1-CA106290), the Breast Cancer Research Foundation, the Gilder Foundation, the Dee Simmons Fund, and the Nellie B. Connally Breast Cancer Research Fund.

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

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

	REFERENCES

TOP ABSTRACT INTRODUCTION PATIENTS AND METHODS RESULTS DISCUSSION Appendix Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
1. Early Breast Cancer Trialists' Collaborative Group: Polychemotherapy for early breast cancer: An overview of the randomized trials. Lancet 352:930-942, 1999[CrossRef]

2. Carlson RW, Anderson BO, Bensinger W, et al: NCCN Practice guidelines for breast cancer. Oncology (Huntington) 14:33-49, 2000

3. National Institutes of Health Consensus Development Conference. http://consensus.nih.gov/2000/2000AdjuvantTherapyBreastCancer114html.htm

4. Goldhirsch A, Wood WC, Gelber RD, et al: Meeting Highlights: Updated International Expert Consensus Panel on the Primary Therapy of Early Breast Cancer. J Clin Oncol 21:3357-3366, 2003[Abstract/Free Full Text]

5. Rouzier R, Pusztai L, Delaloge S, et al: Nomograms to predict pathologic complete response and metastasis-free survival after preoperative chemotherapy for breast cancer. J Clin Oncol 23:8331-8339, 2005[Abstract/Free Full Text]

6. Ross JS, Linette GP, Stec J, et al: Breast cancer biomarkers and molecular medicine, Part I. Expert Rev Mol Diagn 3:573-585, 2003[CrossRef][Medline]

7. Bast Jr RC, Ravdin P, Hayes D, et al: 2000 Update of recommendations for the use of tumor markers in breast and colorectal cancer: Clinical practice guidelines of the American Society of Clinical Oncology. J Clin Oncol 19:1865-1878, 2001[Abstract/Free Full Text]

8. Fisher B, Bryant J, Wolmark N, et al: Effect of preoperative chemotherapy on the outcome of women with operable breast cancer. J Clin Oncol 16:2672-2685, 1998[Abstract]

9. Kuerer HM, Newman LA, Smith TL, et al: Clinical course of breast cancer patients with complete pathologic primary tumor and axillary lymph node response to doxorubicin-based neoadjuvant chemotherapy. J Clin Oncol 17:460-469, 1999[Abstract/Free Full Text]

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

11. Pusztai L, Symmans WF, Hortobagyi GN: Development of pharmacogenomic markers to select preoperative chemotherapy for breast cancer. Breast Cancer 12:73-85, 2005[CrossRef][Medline]

12. Symmans WF, Ayers M, Clark EA, et al: Total RNA yield and microarray gene expression profiles from fine needle aspiration and core needle biopsy samples of breast cancer. Cancer 97:2960-2971, 2003[CrossRef][Medline]

13. Rouzier R, Perou CM, Symmans WF, et al: Different molecular subtypes of breast cancer respond differently to preoperative chemotherapy. Clin Cancer Res 11:5678-5685, 2005[Abstract/Free Full Text]

14. Stec J, Wang J, Coombes K, Ayers M, et al: Comparison of the predictive accuracy of DNA array based multigene classifiers across cDNA arrays and Affymetrix GeneChips. J Mol Diagn 7:357-367, 2005[Abstract/Free Full Text]

15. Pounds S, Morris SW: Estimating the occurrence of false positive and false negatives in microarray studies by approximating and partitioning the empirical distribution of p-values. Bioinformatics 19:1236-1242, 2003[Abstract/Free Full Text]

16. Hastie T, Tibshirani R, Friedman J: The Elements of Statistical Learning. Springer-Verlag, New York, 2001

17. Simon R: Roadmap for developing and validating therapeutically relevant genomic classifiers. J Clin Oncol 23:7332-7341, 2005[Abstract/Free Full Text]

18. Hsing T, Attoor S, Dougherty E: Relation between permutation-test p-values and classifier error estimates. Machine Learning 52:11-30, 2003[CrossRef]

19. Baker SG: The central role of receiver operating characteristic (ROC) curves in evaluating tests for the early detection of cancer. J Natl Cancer Inst 95:511-515, 2003[Free Full Text]

20. Green MC, Buzdar AU, Smith T, et al: Weekly paclitaxel improves pathologic complete remission in operable breast cancer when compared with paclitaxel once every 3 weeks. J Clin Oncol 23:5983-5992, 2005[Abstract/Free Full Text]

21. Mukherjee S, Tamayo P, Rogers S, et al: Estimating dataset size requirements for classifying DNA microarray data. J. Comput Biol 10:119-142, 2003[CrossRef]

22. Gianni L, Zambetti M, Clark K, et al: Gene expression profiles in paraffin-embedded core biopsy tissue predict response to chemotherapy in women with locally advanced breast cancer. J Clin Oncol 23:7265-7277, 2005[Abstract/Free Full Text]

23. Ein-Dor L, Kela I, Getz G, et al: Outcome signature genes in breast cancer: Is there a unique set? Bioinformatics 21:171-178, 2005[Abstract/Free Full Text]

24. Rouzier R, Rajan R, Hess KR, et al: Microtubule associated protein tau is a predictive marker and modulator of response to paclitaxel-containing preoperative chemotherapy in breast cancer. Proc Natl Acad Sci U S A 102:8315-8320, 2005[Abstract/Free Full Text]

Submitted January 11, 2006; accepted May 1, 2006.

CiteULike    Complore    Connotea    Del.icio.us    Digg    Facebook    Reddit    Technorati    Twitter    What's this?

This article has been cited by other articles:

				P. Roepman, H. M. Horlings, O. Krijgsman, M. Kok, J. M. Bueno-de-Mesquita, R. Bender, S. C. Linn, A. M. Glas, and M. J. van de Vijver Microarray-Based Determination of Estrogen Receptor, Progesterone Receptor, and HER2 Receptor Status in Breast Cancer Clin. Cancer Res., November 15, 2009; 15(22): 7003 - 7011. [Abstract] [Full Text] [PDF]

				P. D. Williams, S. Cheon, D. M. Havaleshko, H. Jeong, F. Cheng, D. Theodorescu, and J. K. Lee Concordant Gene Expression Signatures Predict Clinical Outcomes of Cancer Patients Undergoing Systemic Therapy Cancer Res., November 1, 2009; 69(21): 8302 - 8309. [Abstract] [Full Text] [PDF]

				C. Speers, A. Tsimelzon, K. Sexton, A. M. Herrick, C. Gutierrez, A. Culhane, J. Quackenbush, S. Hilsenbeck, J. Chang, and P. Brown Identification of Novel Kinase Targets for the Treatment of Estrogen Receptor-Negative Breast Cancer Clin. Cancer Res., October 15, 2009; 15(20): 6327 - 6340. [Abstract] [Full Text] [PDF]

				L. Pusztai, J.-H. Jeong, Y. Gong, J. S. Ross, C. Kim, S. Paik, R. Rouzier, F. Andre, G. N. Hortobagyi, N. Wolmark, et al. Evaluation of Microtubule-Associated Protein-Tau Expression As a Prognostic and Predictive Marker in the NSABP-B 28 Randomized Clinical Trial J. Clin. Oncol., September 10, 2009; 27(26): 4287 - 4292. [Abstract] [Full Text] [PDF]

				C. Min, Z. Yu, K. H. Kirsch, Y. Zhao, S. R. Vora, P. C. Trackman, D. B. Spicer, L. Rosenberg, J. R. Palmer, and G. E. Sonenshein A Loss-of-Function Polymorphism in the Propeptide Domain of the LOX Gene and Breast Cancer Cancer Res., August 15, 2009; 69(16): 6685 - 6693. [Abstract] [Full Text] [PDF]

				A. Grigoriadis, O. L. Caballero, K. S. Hoek, L. da Silva, Y.-T. Chen, S. J. Shin, A. A. Jungbluth, L. D. Miller, D. Clouston, J. Cebon, et al. CT-X antigen expression in human breast cancer PNAS, August 11, 2009; 106(32): 13493 - 13498. [Abstract] [Full Text] [PDF]

				F. Correa Geyer and J. S. Reis-Filho Microarray-based Gene Expression Profiling as a Clinical Tool for Breast Cancer Management: Are We There Yet? International Journal of Surgical Pathology, August 1, 2009; 17(4): 285 - 302. [Abstract] [PDF]

				B. Fayard, F. Bianchi, J. Dey, E. Moreno, S. Djaffer, N. E. Hynes, and D. Monard The Serine Protease Inhibitor Protease Nexin-1 Controls Mammary Cancer Metastasis through LRP-1-Mediated MMP-9 Expression Cancer Res., July 15, 2009; 69(14): 5690 - 5698. [Abstract] [Full Text] [PDF]

				C. Liedtke, C. Hatzis, W. F. Symmans, C. Desmedt, B. Haibe-Kains, V. Valero, H. Kuerer, G. N. Hortobagyi, M. Piccart-Gebhart, C. Sotiriou, et al. Genomic Grade Index Is Associated With Response to Chemotherapy in Patients With Breast Cancer J. Clin. Oncol., July 1, 2009; 27(19): 3185 - 3191. [Abstract] [Full Text] [PDF]

				T. A. DiMeo, K. Anderson, P. Phadke, C. Feng, C. M. Perou, S. Naber, and C. Kuperwasser A Novel Lung Metastasis Signature Links Wnt Signaling with Cancer Cell Self-Renewal and Epithelial-Mesenchymal Transition in Basal-like Breast Cancer Cancer Res., July 1, 2009; 69(13): 5364 - 5373. [Abstract] [Full Text] [PDF]

				V. Guarneri and P. Conte Metastatic Breast Cancer: Therapeutic Options According to Molecular Subtypes and Prior Adjuvant Therapy Oncologist, July 1, 2009; 14(7): 645 - 656. [Abstract] [Full Text] [PDF]

				D. R. Rhodes, B. Ateeq, Q. Cao, S. A. Tomlins, R. Mehra, B. Laxman, S. Kalyana-Sundaram, R. J. Lonigro, B. E. Helgeson, M. S. Bhojani, et al. AGTR1 overexpression defines a subset of breast cancer and confers sensitivity to losartan, an AGTR1 antagonist PNAS, June 23, 2009; 106(25): 10284 - 10289. [Abstract] [Full Text] [PDF]

				A. M. Sieuwerts, J. Kraan, J. Bolt-De Vries, P. Van Der Spoel, F. Elstrodt, M. Smid, M. Timmermans, B. Mostert, M. Schutte, J. W. M. Martens, et al. Response: Re: Anti-Epithelial Cell Adhesion Molecule Antibodies and the Detection of Circulating Normal-Like Breast Tumor Cells J Natl Cancer Inst, June 16, 2009; 101(12): 896 - 897. [Full Text] [PDF]

				M. Lupien, J. Eeckhoute, C. A. Meyer, S. A. Krum, D. R. Rhodes, X. S. Liu, and M. Brown Coactivator Function Defines the Active Estrogen Receptor Alpha Cistrome Mol. Cell. Biol., June 15, 2009; 29(12): 3413 - 3423. [Abstract] [Full Text] [PDF]

				Y. Zhang, J. W.M. Martens, J. X. Yu, J. Jiang, A. M. Sieuwerts, M. Smid, J. G.M. Klijn, Y. Wang, and J. A. Foekens Copy Number Alterations that Predict Metastatic Capability of Human Breast Cancer Cancer Res., May 1, 2009; 69(9): 3795 - 3801. [Abstract] [Full Text] [PDF]

				J. S. Parker, M. Mullins, M. C.U. Cheang, S. Leung, D. Voduc, T. Vickery, S. Davies, C. Fauron, X. He, Z. Hu, et al. Supervised Risk Predictor of Breast Cancer Based on Intrinsic Subtypes J. Clin. Oncol., March 10, 2009; 27(8): 1160 - 1167. [Abstract] [Full Text] [PDF]

				T. Sorlie Introducing Molecular Subtyping of Breast Cancer Into the Clinic? J. Clin. Oncol., March 10, 2009; 27(8): 1153 - 1154. [Full Text] [PDF]

				C. Sotiriou and L. Pusztai Gene-Expression Signatures in Breast Cancer N. Engl. J. Med., February 19, 2009; 360(8): 790 - 800. [Full Text] [PDF]

				J. Baselga, M. Zambetti, A. Llombart-Cussac, G. Manikhas, E. Kubista, G. G. Steger, A. Makhson, S. Tjulandin, H. Ludwig, M. Verrill, et al. Phase II Genomics Study of Ixabepilone as Neoadjuvant Treatment for Breast Cancer J. Clin. Oncol., February 1, 2009; 27(4): 526 - 534. [Abstract] [Full Text] [PDF]

				J. Chiquet, A. Smith, G. Grasseau, C. Matias, and C. Ambroise SIMoNe: Statistical Inference for MOdular NEtworks Bioinformatics, February 1, 2009; 25(3): 417 - 418. [Abstract] [Full Text] [PDF]

				F. Andre, B. Job, P. Dessen, A. Tordai, S. Michiels, C. Liedtke, C. Richon, K. Yan, B. Wang, G. Vassal, et al. Molecular Characterization of Breast Cancer with High-Resolution Oligonucleotide Comparative Genomic Hybridization Array Clin. Cancer Res., January 15, 2009; 15(2): 441 - 451. [Abstract] [Full Text] [PDF]

				F. Huang, A. Greer, W. Hurlburt, X. Han, R. Hafezi, G. M. Wittenberg, K. Reeves, J. Chen, D. Robinson, A. Li, et al. The Mechanisms of Differential Sensitivity to an Insulin-like Growth Factor-1 Receptor Inhibitor (BMS-536924) and Rationale for Combining with EGFR/HER2 Inhibitors Cancer Res., January 1, 2009; 69(1): 161 - 170. [Abstract] [Full Text] [PDF]

				O. I. Olopade, T. A. Grushko, R. Nanda, and D. Huo Advances in Breast Cancer: Pathways to Personalized Medicine Clin. Cancer Res., December 15, 2008; 14(24): 7988 - 7999. [Abstract] [Full Text] [PDF]

				M. J. Duffy and J. Crown A Personalized Approach to Cancer Treatment: How Biomarkers Can Help Clin. Chem., November 1, 2008; 54(11): 1770 - 1779. [Abstract] [Full Text] [PDF]

				J. K. Miller, D. L. Shattuck, E. Q. Ingalla, L. Yen, A. D. Borowsky, L. J.T. Young, R. D. Cardiff, K. L. Carraway III, and C. Sweeney Suppression of the Negative Regulator LRIG1 Contributes to ErbB2 Overexpression in Breast Cancer Cancer Res., October 15, 2008; 68(20): 8286 - 8294. [Abstract] [Full Text] [PDF]

				J. S. Ross, C. Hatzis, W. F. Symmans, L. Pusztai, and G. N. Hortobagyi Commercialized Multigene Predictors of Clinical Outcome for Breast Cancer Oncologist, May 1, 2008; 13(5): 477 - 493. [Abstract] [Full Text] [PDF]

				L. Pusztai Current Status of Prognostic Profiling in Breast Cancer Oncologist, April 1, 2008; 13(4): 350 - 360. [Abstract] [Full Text] [PDF]

				E. Azambuja, V. Durbecq, D. D. Rosa, M. Colozza, D. Larsimont, M. Piccart-Gebhart, and F. Cardoso HER-2 overexpression/amplification and its interaction with taxane-based therapy in breast cancer Ann. Onc., February 1, 2008; 19(2): 223 - 232. [Abstract] [Full Text] [PDF]

				A. Dhasarathy, M. Kajita, and P. A. Wade The Transcription Factor Snail Mediates Epithelial to Mesenchymal Transitions by Repression of Estrogen Receptor-{alpha} Mol. Endocrinol., December 1, 2007; 21(12): 2907 - 2918. [Abstract] [Full Text] [PDF]

				M. Kaufmann, G. von Minckwitz, H. D. Bear, A. Buzdar, P. McGale, H. Bonnefoi, M. Colleoni, C. Denkert, W. Eiermann, R. Jackesz, et al. Recommendations from an international expert panel on the use of neoadjuvant (primary) systemic treatment of operable breast cancer: new perspectives 2006 Ann. Onc., December 1, 2007; 18(12): 1927 - 1934. [Abstract] [Full Text] [PDF]

				K. S. Garman, J. R. Nevins, and A. Potti Genomic strategies for personalized cancer therapy Hum. Mol. Genet., October 15, 2007; 16(R2): R226 - R232. [Abstract] [Full Text] [PDF]

				L. Pusztai, K. Anderson, and K. R. Hess Pharmacogenomic Predictor Discovery in Phase II Clinical Trials for Breast Cancer Clin. Cancer Res., October 15, 2007; 13(20): 6080 - 6086. [Abstract] [Full Text] [PDF]

				A. A.N. Rose, F. Pepin, C. Russo, J. E. Abou Khalil, M. Hallett, and P. M. Siegel Osteoactivin Promotes Breast Cancer Metastasis to Bone Mol. Cancer Res., October 1, 2007; 5(10): 1001 - 1014. [Abstract] [Full Text] [PDF]

				L. Gianni Ixabepilone and the Narrow Path to Developing New Cytotoxic Drugs J. Clin. Oncol., August 10, 2007; 25(23): 3389 - 3391. [Full Text] [PDF]

				F. Peintinger, K. Anderson, C. Mazouni, H. M. Kuerer, C. Hatzis, F. Lin, G. N. Hortobagyi, W. F. Symmans, and L. Pusztai Thirty-Gene Pharmacogenomic Test Correlates with Residual Cancer Burden after Preoperative Chemotherapy for Breast Cancer Clin. Cancer Res., July 15, 2007; 13(14): 4078 - 4082. [Abstract] [Full Text] [PDF]

				F. Andre, C. Hatzis, K. Anderson, C. Sotiriou, C. Mazouni, J. Mejia, B. Wang, G. N. Hortobagyi, W. F. Symmans, and L. Pusztai Microtubule-Associated Protein-tau is a Bifunctional Predictor of Endocrine Sensitivity and Chemotherapy Resistance in Estrogen Receptor-Positive Breast Cancer Clin. Cancer Res., April 1, 2007; 13(7): 2061 - 2067. [Abstract] [Full Text] [PDF]

				F. Huang, K. Reeves, X. Han, C. Fairchild, S. Platero, T. W. Wong, F. Lee, P. Shaw, and E. Clark Identification of Candidate Molecular Markers Predicting Sensitivity in Solid Tumors to Dasatinib: Rationale for Patient Selection Cancer Res., March 1, 2007; 67(5): 2226 - 2238. [Abstract] [Full Text] [PDF]

				L. Pusztai Chips to Bedside: Incorporation of Microarray Data into Clinical Practice Clin. Cancer Res., December 15, 2006; 12(24): 7209 - 7214. [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 Rights & Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Hess, K. R. Articles by Pusztai, L. Search for Related Content PubMed PubMed Citation Articles by Hess, K. R. Articles by Pusztai, L. Social Bookmarking                            What's this?