|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 35 (December 10), 2007: pp. 5562-5569 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2007.12.0352 Three-Gene Prognostic Classifier for Early-Stage Non–Small-Cell Lung Cancer
From the University Health Network, Ontario Cancer Institute, Princess Margaret Hospital and Toronto General Hospital; Departments of Medical Biophysics, Thoracic Surgery, Computer Science, Medicine, and Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada; and Department of Medical Oncology, Christie Hospital National Health Service Trust, Manchester, United Kingdom Address reprint requests to Ming-Sound Tsao, MD, Princess Margaret Hospital, 610 University Ave, Toronto, Ontario, Canada M5G 2M9; e-mail: Ming.Tsao{at}uhn.on.ca
Purpose Several microarray studies have reported gene expression signatures that classify non–small-cell lung carcinoma (NSCLC) patients into different prognostic groups. However, the prognostic gene lists reported to date overlap poorly across studies, and few have been validated independently using more quantitative assay methods. Patients and Methods The expression of 158 putative prognostic genes identified in previous microarray studies was analyzed by reverse transcription quantitative polymerase chain reaction in the tumors of 147 NSCLC patients. Concordance indices and risk scores were used to identify a stage-independent set of genes that could classify patients with significantly different prognoses. Results We have identified a three-gene classifier (STX1A, HIF1A, and CCR7) for overall survival (hazard ratio = 3.8; 95% CI, 1.7 to 8.2; P < .001). The classifier was also able to stratify stage I and II patients and further improved the predictive ability of clinical factors such as histology and tumor stage. The predictive value of this three-gene classifier was validated in two large independent microarray data sets from Harvard and Duke Universities. Conclusion We have identified a new three-gene classifier that is independent of and improves on stage to stratify early-stage NSCLC patients with significantly different prognoses. This classifier may be tested further for its potential value to improve the selection of resected NSCLC patients in adjuvant therapy.
Non–small-cell lung carcinoma (NSCLC) represents approximately 80% of lung cancers and has a 5-year survival rate of only 16%.1 Tumor stage remains the strongest predictor of survival.2 Early-stage (stage I and II) patients are treated primarily by surgical resection. However, 30% to 55% of these patients develop recurrence and die of the disease,2,3 implying that biologic heterogeneity exists among patients and their tumors. Recent phase III trials have established that adjuvant chemotherapy significantly improves the survival of stage II to IIIA patients.4-8 Therefore, identifying markers that accurately classify early-stage NSCLC patients into different prognostic groups may be used to select patients who should receive adjuvant therapy. Neither tumor histologic features nor the 50 or more associated proteins that have been evaluated to date can represent markers with significant clinical utility.9,10 Several NSCLC mRNA expression microarray studies have identified expression signatures that partition patients into prognostic groups.11-16 However, these putative prognostic gene lists have minimal overlap (Fig 1A). Furthermore, cross-study analyses of the data sets using different statistical approaches17,18 have generated additional prognostic gene sets. This discordance has been attributed to insufficiently powered studies19 and variability in patient cohorts, expression profiling platforms, or statistical methodologies and emphasizes the need to validate candidate prognostic gene lists in independent cohorts of patients and with different assays.
On the basis of the hypothesis that prognostic genes identified in previous microarray studies represent the best candidates for the validation or identification of new prognostic classifiers using an independent assay platform, we evaluated the significance of 158 genes in 147 early-stage completely resected NSCLC patients using reverse transcription (RT) quantitative polymerase chain reaction (qPCR). The concordance index and risk scores identified a three-gene classifier that was validated in microarray data sets from the Dana-Farber Cancer Institute (herein referred to as the Harvard data set)11 and Duke University (Duke).20
Patients and Tissue Samples This study was performed using 165 snap-frozen tumor samples (Toronto) from patients treated by lobectomy or pneumonectomy at the Toronto General Hospital (1996 to 2000) and Mount Sinai Hospital (1995 to 1998). Tissues were harvested within 30 minutes of complete resection, and tumor tissue quality and pathology were confirmed by the study pathologist (M.-S.T.). Tumor cell content ranged from 20% to 100%, with a median cellularity of 80%. All tissues were banked after written informed patient consent, and the University Health Network Research Ethics Board approved the study protocol. Patient characteristics are listed in Appendix Table A1. Total RNA was isolated and prepared for RT-qPCR as described previously.21
Assembly of Prognostic Gene Candidates
Expression Analysis by RT-qPCR
Statistical Analysis
The concordance index was used to select a prognostic classifier from the 158 genes. Similar to C-statistics for binary outcome, the concordance index is a measure of the predictive ability of a covariate when the outcome is the time to event. A concordance index value of 0.5 indicates no predictive ability, whereas a value of 1 represents perfect predictive ability. Genes with concordance index values The sum of the coefficients from the Cox proportional hazards model for each gene was used to calculate a risk score for each patient. The risk scores were median dichotomized to assign patients to one of the prognostic arms. A Cox proportional hazards model was used to test the classifier when adjusting for stage and histology. Concordance indices were calculated for models containing stage and histology information alone (clinical model) and when the classifier was included with the clinical model (full model). The 95% CIs for each of the models and the difference between the concordances indices of the models were calculated using the bootstrap method. A CI for the difference between the two indices that excludes zero suggests that the classifier enhances the predictive ability of clinical information. All reported survival differences were tested using the Wald test within the Cox proportional hazards model. All hazard ratios (HRs) reported have been adjusted for the effects of stage (I v II v III/IV) and histology (adenocarcinoma v squamous cell carcinoma), except those displayed in the Kaplan-Meier plots. The 3-year survival percentages were calculated using the Kaplan-Meier method.
Independent Validation
Protein-Protein Interactions
Univariate survival analysis of the RT-qPCR expression data from 147 Toronto patients demonstrated enrichment for prognostic markers, with 24 (15%) of the 158 genes significant at P  .05 (RT-qPCR data available at http://www.cs.toronto.edu/ juris/data/JCO07). Four genes remained significant after a false discovery rate adjustment for multiple testing (q 0.05)31 (Appendix Table A2). Enrichment of genes involved in cellular homeostasis and stress response was observed (Appendix Table A3). To identify a multigene prognostic classifier, we used a method based on the concordance index and risk scores.
Five genes were above a concordance index threshold of 0.65 (Table 1): syntaxin 1A (STX1A), chemokine (C-C motif) receptor 7 (CCR7), hypoxia-inducible factor 1 alpha (HIF1A), platelet-activating factor acetylhydrolase, isoform Ib, gamma subunit (PAFAH1B3), and chaperonin containing TCP1, subunit 3 (gamma) (CCT3). Neither PAFAH1B3 nor CCT3 was found to contribute to the risk score model (Appendix Table A4); a three-gene classifier comprised of STX1A, CCR7, and HIF1A was used for the remaining analyses. On the basis of median dichotomization of the composite risk scores, patients scoring
Prognostic Significance for Early-Stage NSCLC The three-gene classifier was tested for its ability to group stage I patients into significantly different prognostic arms. Compared with pathologic stage alone, the three-gene set demonstrated significant improvement in its classification ability (HR = 6.0; 95% CI, 2.0 to 19.0; P = .002; Table 2, Fig 1C). In addition, there is a difference of 24% in 3-year survival rate for stage I patients classified as poor prognosis compared with those classified as good prognosis (Appendix Table A6). When adjusted for stage and histology, the classifier also showed a trend to significant separation in stage II patients (Table 2, Fig 1D).
Independent Validation The prognostic ability of the three-gene classifier was validated in both independent data sets (Harvard: HR = 1.7; 95% CI, 1.1 to 2.7; P = .03; Duke: HR = 2.4; 95% CI, 1.3 to 4.4; P = .007; Table 2, Figs 2A and 2B; Appendix Fig A2, online only). The 95% CIs for the difference between the concordance indices showed an improvement in prediction ability when the three-gene classifier was added to the clinical model for each of the validation data sets (Appendix Table A5). The classifier also demonstrated the ability to subclassify stage I patients (Table 2, Figs 2C and 2D), with patients having nearly 30% greater risk of dying at 3 years (Appendix Table A6). Stage II patients could be classified in the Harvard (n = 24) and Duke (n = 18) data sets, although neither was statistically significant (Table 2).
Other lung microarray data sets such as the American College of Surgeons Oncology Group Z0030 study presented by Potti et al32 and the Mayo and Washington University data sets from Lu et al33 are available; the patient cohorts are quite small (n = 25, 36, and 18, respectively), which would severely limit the analysis.
Despite wide discordance among the signature gene sets identified in various lung cancer microarray studies (Fig 1A), we have demonstrated by RT-qPCR in an independent patient cohort that such studies have enriched for genes with prognostic value in NSCLC. We have identified a three-gene, mRNA expression–based classifier that can partition early-stage NSCLC patients into subgroups with significantly different prognoses. We validated the reproducibility and strength of this classifier in two publicly available microarray data sets whose prognostic gene lists did not contain any of our classifier genes. We measured gene expression by RT-qPCR and normalized the data using the geometric mean of four housekeeping genes.34 We then applied concordance indices to select genes and risk scores to assess their ability to predict survival outcome. Risk scores are a well-established method of separating patients into prognostic groups.35,36 The increase in concordance index scores between a model that uses only clinical data and one where the expression data for our three-gene classifier were added to clinical data (Appendix Table A5) shows that gene expression patterns are both independent and additive to the predictive ability of clinical parameters such as stage and histology. The prognostic value of our three-gene classifier was validated in two large publicly available microarray expression data sets, although these data sets were generated from two different generations of microarray chips (Affymetrix HG-U95Av2 and HG-U133 Plus2.0; Affymetrix, Santa Clara, CA). This is even more remarkable considering the heterogeneity of the two cohorts. The Harvard data set included only patients with adenocarcinoma, whereas the Duke data set included both adenocarcinoma and squamous cell carcinoma patients. Of the genes in our classifier, STX1A is deregulated in small-cell lung cancer and has been associated with more aggressive forms of colon and rectal carcinomas,37,38 whereas syntaxin 2, with more than 70% sequence similarity to STX1A, has a transforming role in mouse mammary tissue.39 The chemokine receptor CCR7 is highly expressed in a number of tumors including gastric, hepatic, and colon carcinomas, where it has been linked to increased invasion, lymph node infiltration, metastasis, and poor prognosis.40-42 In NSCLC, high levels of CCR7 have been linked to increased lymphatic invasion at both the mRNA and protein levels, although its association with patient prognosis was not investigated in these studies.43 Most studies have focused on HIF1A protein rather than mRNA, where it has been found to correlate with metastasis, poor prognosis, and resistance to therapy.44,45 Chi et al46 recently reported that a hypoxia response gene expression signature is prognostic in breast and ovarian cancers, where a heightened response is associated with high basal levels of HIF1A mRNA. Our study identified HIF1A mRNA expression as prognostic in NSCLC. We further found that HIF1A was differentially expressed between matched normal and tumor NSCLC samples, and this was also examined in two public microarray data sets (Appendix Table A7). Together, these three genes represent important targets for future biologic and mechanistic studies. Although several microarray studies in NSCLC have identified gene expression signatures that cluster patients into groups with significantly different prognoses, only a few profiles have been validated rigorously in independent patient cohorts using the RT-qPCR technique that is considered the gold standard for mRNA expression analysis (Appendix Table A8).47,48 Potti et al32 identified groups of metagenes that can stratify stage I to III NSCLC patients based on their risk of recurrence using a decision tree model that incorporates clinical data. The model seemed to be predictive in stage IA patients and was validated reasonably well in two independent patient cohorts. However, the precise components of the metagenes were not provided, hampering a direct comparison of their metagenes to signatures from other methods and studies. The Michigan group identified distinct 50-gene signatures for adenocarcinoma13 and squamous cell carcinoma49 that were validated in tumor-specific data sets, as well as combined in a cohort of NSCLC patients. In an in silico meta-analysis of five independent lung microarray data sets of stage I NSCLC patients, Lu et al33 identified a 64-gene prognostic signature. Although this signature validated well in the Duke data set, it did not perform as robustly in an unpublished independent data set from the Memorial Sloan-Kettering Cancer Center. Endoh et al28 used RT-qPCR to assess the prognostic value of 48 candidate genes identified from the Harvard11 and Stanford12 microarray studies in 84 adenocarcinoma patients. An eight-gene prognostic model was identified that could separate 21 patients with different prognoses. Using a custom microarray composed of 672 genes that were associated with invasive activity of NSCLC cell lines, Chen et al29 identified 16 genes that were prognostic in 125 NSCLC samples through univariate analyses. Only five genes were verified using RT-qPCR and were found to be capable of separating patients into two distinct prognostic groups. This classifier was subsequently validated in an independent cohort of 60 patients and the Michigan data set. However, the eight- and five-gene classifiers did not overlap with each other or with our three-gene classifier. The reasons for nonoverlapping prognostic gene lists remain unclear but could be influenced by unique patient cohorts, different probe sets on arrays, and the diverse range of normalization and gene selection algorithms.19,50,51 Several recent reviews and articles52-54 have highlighted key challenges for prognostic classifier identification via microarray analysis, which include heterogeneous patient cohorts, the identification of an excessive number of false-positive results, and perhaps most importantly, the lack of reproducibility in other data sets. Therefore, it remains to be systematically determined which algorithms are best suited for selecting and validating stable prognostic gene lists. On the basis of the reports to date and including ours, it is possible that multiple small NSCLC gene classifiers provide similar prognostic capabilities, especially when they include genes that belong to the commonly deregulated pathways in lung carcinogenesis. For example, prognostic genes from two RT-qPCR–based studies28,29 and ours interact with proteins that are often implicated in cancer such as TP53, ERBB2, and EGFR (Fig 3). Further studies are warranted to identify all possible small gene classifiers or pathways that are critical for the biology and prognosis of NSCLC because these may lead to the development of novel, more robust classifiers that can be clinically used to select patients for adjuvant therapies.
Although some recent publications have emphasized the need for classifiers to identify stage I patients in need of adjuvant chemotherapy, it is equally important that a classifier should identify stage II patients with excellent prognosis who may not require further treatment after complete resection. Our three-gene prognostic classifier was robust for stage I. In stage II patients, only a trend to significant classification ability was observed in all data sets, possibly because of the small size of these cohorts. A number of previous studies discussed have generated models that rely on microarray platforms to assay the expression profiles of dozens or hundreds of genes. RT-qPCR has emerged as a preferred method for independent validation of microarray-based results because it has equivalent or superior technical characteristics. Our group reported previously that snap-frozen surgical samples that are harvested within 30 to 60 minutes of resection demonstrated stable gene expression profiles, although pooling samples from multiple areas of a tumor was recommended to overcome the potential confounding effect of tumor heterogeneity.55 Importantly, the overall robustness and flexibility of the RT-qPCR platform has resulted in its widespread use in clinical diagnostic laboratories and has also been recently applied as part of a large clinical trial of a breast cancer prognostic signature.56 Once further validated in other independent patient cohorts and with standardized protocols in hand, a significant advantage of our robust three-gene classifier for NSCLC is that it may be easily implemented in the clinic using cost-effective multiplex RT-qPCR assays.
The author(s) indicated no potential conflicts of interest.
Conception and design: Suzanne K. Lau, Fiona H. Blackhall, Frances A. Shepherd, Sandy D. Der, Ming-Sound Tsao Financial support: Linda Z. Penn, Igor Jurisica, Sandy D. Der, Ming-Sound Tsao Administrative support: Ni Liu, Davina Lau, Frances A. Shepherd Provision of study materials or patients: Michael R. Johnston, Gail Darling, Shaf Keshavjee, Thomas K. Waddell, Frances A. Shepherd, Ming-Sound Tsao Collection and assembly of data: Suzanne K. Lau, Fiona H. Blackhall, Ni Liu, Davina Lau Data analysis and interpretation: Suzanne K. Lau, Paul C. Boutros, Melania Pintilie, Chang-Qi Zhu, Dan Strumpf, Igor Jurisica, Sandy D. Der, Ming-Sound Tsao Manuscript writing: Suzanne K. Lau, Paul C. Boutros, Melania Pintilie, Frances A. Shepherd, Sandy D. Der, Ming-Sound Tsao Final approval of manuscript: Suzanne K. Lau, Paul C. Boutros, Melania Pintilie, Fiona H. Blackhall, Chang-Qi Zhu, Dan Strumpf, Michael R. Johnston, Gail Darling, Shaf Keshavjee, Thomas K. Waddell, Ni Liu, Davina Lau, Linda Z. Penn, Frances A. Shepherd, Igor Jurisica, Sandy D. Der, Ming-Sound Tsao
Calculation of Risk Scores The concordance index (Harrell FE Jr, Califf RM, Pryor DB, et al: JAMA 247:2543-2546, 1982) was used as a measure of predictive ability. Five genes with a concordance index value exceeding 0.65 were selected as candidate prognostic genes. Although the genes had been originally standardized to be centrally distributed around zero, their variances were slightly different. Thus the gene expressions were further statistically standardized follow a standard normal distribution. The coefficients obtained in the Cox proportional hazards model containing the 5 standardized gene expressions were used to calculate the risk score of each patient. For easier calculation of these scores each gene was dichotomized as positive if the expression value was positive and negative otherwise. The coefficients from the Cox proportional hazards model were multiplied by 10 and rounded to obtain an integer representing the gene score. The risk score for a patient was calculated by adding the gene scores for those genes for which the patient had a positive expression. In mathematical terms this can be expressed as:
2 were classified as having "good prognosis," while those with a risk score more than 2 were classified as having "poor prognosis."
Evaluating the Impact of Clinical Data: Model Assessment by Concordance Index Analysis
Implementation Details
Independent Validation Harvard: http://research.dfci.harvard.edu/meyersonlab/lungca/ Duke: http://data.cgt.duke.edu/oncogene.php Michigan: http://www.camda.duke.edu/camda03/datasets/. Data were preprocessed with the robust multi-array (RMA) algorithm (Irizarry RA, Hobbs B, Collin F, et al: Biostatistics 4:249-264, 2003) using the affy package (v1.6.7; Gautier L, Cope L, Bolstad BM, et al: Bioinformatics 20:307-315, 2004) of the BioConductor open-source library (Gentleman RC, Carey VJ, Bates DM, et al: Genome Biol 5:R80, 2004) for R (v2.1.1). The Harvard data set included technical replicates for some samples, which were collapsed by averaging to yield a unique expression vector for each patient sample. For each gene in our classifiers, all ProbeSets were identified. Inter-ProbeSet Pearson correlations were calculated for each gene represented by multiple ProbeSets. Well-correlated ProbeSets (R > 0.7) were collapsed by averaging. Poorly correlated ProbeSets were collapsed by selecting the sequence most specific to the target transcript as identified via bl2seq against the RefSeq mRNA (Tatusova TA, Madden TL: FEMS Microbiol Lett 174:247-250, 1999). Array data was subjected to normalization and median-scaling (but not pseudocount addition, as RMA-processed data is already log2-transformed) as described previously (Barsyte-Lovejoy D, Lau SK, Boutros PC, et al: Cancer Res 66:5330-5337, 2006). The results from the confirmation study in the Michigan data set are in Appendix Figure A1. In the text, the performance of the three-gene classifier for all patients, or stage I or II patients, was discussed (Fig 2). In Appendix Figure A2, overall survival curves are presented for each study grouped by pathological stage alone.
New Patient Classification For each new patient a composite risk score can be calculated by adding the risk score (Table 1) of each for the three genes in the classifier every time its expression is positive. A composite score of 2 or less would classify that patient in the "good prognosis" group, while a score greater than 2 would classify the patient in the "poor prognosis" group. A simple scheme for patient classification is found in Appendix Table A9.
Differential Expression Analysis To further confirm this pattern of differential expression, the data from the Harvard (Bhattacharjee A, Richards WG, Staunton J, et al: Proc Natl Acad Sci U S A 98:13790-13795, 2001) and Michigan (Beer DG, Kardia SL, Huang CC, et al: Nat Med 8:816-824, 2002) microarray studies were exploited. For each gene in the prognostic subsets, normalized expression values for all adenocarcinomas and normal lung samples were extracted from these two studies. Normalization was performed as described above with initial array preprocessing via RMA followed by house-keeping gene normalization and median centering. The relative expression (Tumor/Normal) in log2 space (the M columns in Appendix Table A7) and the P value for the assessment of whether these two populations have different expression levels were calculated. P values were calculated using a two-tailed t test with the Welch correction for unequal variances and then controlled for multiple-testing with the conservative Bonferroni adjustment for multiple-testing across each array. Although the P values were widely disparate between the Toronto, Harvard, and Michigan data sets, significant correlation was observed between fold-change values in all three data sets.
Protein-Protein Interactions
GO Analysis
Results for the Risk Score Model
2 and "poor" prognosis if the total risk score was more than 2, based on median dichotomization. On careful consideration, it was evident that PAFAH1B3 did not contribute to categorization of a patient into the "good" or "poor" prognosis group. To demonstrate this, we considered all the possible of combinations of STX1A, CCR7, and HIF1A expression patterns. In each case, positive PAFAH1B3 expression did not change a patient's prognostic group (Appendix Table A4). Thus, all further analysis was conducted using the three-gene classifier model composed of STX1A, CCR7, and HIF1A. Appendix Table A5 shows the concordance indices for the Toronto cohort used as training set and for both the validation the confirmation data sets. In all instances, the full model produces larger concordance indices than the clinical model. The differences are always positive, and the CIs do not include zero in the Toronto and Duke data sets.
We thank Geoffrey Liu, MD, for critical reading and advice during the preparation of this manuscript. T.K.W. holds the R. Fraser Elliott Chair in Transplantation Research, F.A.S. holds the Scott Taylor Chair in Lung Cancer Research, and M.-S.T. holds the M Qasim Choksi Chair in Lung Cancer Translational Research at the University Health Network.
Supported by Grant No. 015184 from the National Cancer Institute of Canada (M.-S.T.); Grant No. RGPIN 203833-02 from the National Science and Engineering Research Council; grants from IBM, the Institute for Robotics and Intelligence Systems, and the Fireman and Younger Foundation (I.J.); Genome Canada (I.J., S.D.D.); and Pre-Competitive Applied Research Network, Natural Sciences and Engineering Research Council, and Excellence in Radiation Research for the 21st Century Fellowships (P.C.B.). S.K.L., P.C.B., and M.P. contributed equally to this work. Presented in part at the 11th World Conference of the International Association for the Study of Lung Cancer, July 3-6, 2005, Barcelona, Spain; and 97th Annual Meeting of the American Association of Cancer Research, April 1-5, 2006, Washington, DC. Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
1. Jemal A, Siegel R, Ward E, et al: Cancer statistics, 2007. CA Cancer J Clin 57:43-66, 2007 2. Mountain CF: Staging classification of lung cancer: A critical evaluation. Clin Chest Med 23:103-121, 2002[CrossRef][Medline] 3. Naruke T, Goya T, Tsuchiya R, et al: Prognosis and survival in resected lung carcinoma based on the new international staging system. J Thorac Cardiovasc Surg 96:440-447, 1988[Abstract] 4. Wada H, Miyahara R, Tanaka F, et al: Postoperative adjuvant chemotherapy with PVM (cisplatin + vindesine + mitomycin C) and UFT (uracil + tegaful) in resected stage I-II NSCLC (non-small cell lung cancer): A randomized clinical trial—West Japan Study Group for lung cancer surgery (WJSG). Eur J Cardiothorac Surg 15:438-443, 1999 5. Dunant A, Pignon JP, Le Chevalier T: Adjuvant chemotherapy for non-small cell lung cancer: Contribution of the International Adjuvant Lung Trial. Clin Cancer Res 11:5017S-5021S, 2005[CrossRef][Medline] 6. Winton T, Livingston R, Johnson D, et al: Vinorelbine plus cisplatin vs. observation in resected non-small-cell lung cancer. N Engl J Med 352:2589-2597, 2005 7. Strauss GM, Herndon J, Maddaus MA, et al: Randomized clinical trial of adjuvant chemotherapy with paclitaxel and carboplatin following resection in stage IB non-small cell lung cancer (NSCLC): Report of Cancer and Leukemia Group B (CALGB) protocol 9633. J Clin Oncol 22:621s, 2004 (suppl; abstr 7019) 8. Rosell R, de Lena M, Carpagnano F, et al: ANITA: Phase III adjuvant vinorelbine (N) and cisplatin (P) versus observation in completely resected (stage I-III) non small cell lung cancer (NSCLC) patients (pts). Lung Cancer 49:S3-S4, 2005 (suppl)[Medline] 9. Singhal S, Vachani A, Antin-Ozerkis D, et al: Prognostic implications of cell cycle, apoptosis, and angiogenesis biomarkers in non-small cell lung cancer: A review. Clin Cancer Res 11:3974-3986, 2005 10. Zhu CQ, Shih W, Ling CH, et al: Immunohistochemical markers of prognosis in non-small cell lung cancer: A review and proposal for a multiphase approach to marker evaluation. J Clin Pathol 59:790-800, 2006 11. Bhattacharjee A, Richards WG, Staunton J, et al: Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses. Proc Natl Acad Sci U S A 98:13790-13795, 2001 12. Garber ME, Troyanskaya OG, Schluens K, et al: Diversity of gene expression in adenocarcinoma of the lung. Proc Natl Acad Sci U S A 98:13784-13789, 2001 13. Beer DG, Kardia SL, Huang CC, et al: Gene-expression profiles predict survival of patients with lung adenocarcinoma. Nat Med 8:816-824, 2002[Medline] 14. Wigle DA, Jurisica I, Radulovich N, et al: Molecular profiling of non-small cell lung cancer and correlation with disease-free survival. Cancer Res 62:3005-3008, 2002 15. Ramaswamy S, Ross KN, Lander ES, et al: A molecular signature of metastasis in primary solid tumors. Nat Genet 33:49-54, 2003[CrossRef][Medline] 16. Tomida S, Koshikawa K, Yatabe Y, et al: Gene expression-based, individualized outcome prediction for surgically treated lung cancer patients. Oncogene 23:5360-5370, 2004[CrossRef][Medline] 17. Parmigiani G, Garrett-Mayer ES, Anbazhagan R, et al: A cross-study comparison of gene expression studies for the molecular classification of lung cancer. Clin Cancer Res 10:2922-2927, 2004 18. Jiang H, Deng Y, Chen HS, et al: Joint analysis of two microarray gene-expression data sets to select lung adenocarcinoma marker genes. BMC Bioinformatics 5:81, 2004[CrossRef][Medline] 19. Ein-Dor L, Zuk O, Domany E: Thousands of samples are needed to generate a robust gene list for predicting outcome in cancer. Proc Natl Acad Sci U S A 103:5923-5928, 2006 20. Bild AH, Yao G, Chang JT, et al: Oncogenic pathway signatures in human cancers as a guide to targeted therapies. Nature 439:353-357, 2006[CrossRef][Medline] 21. Barsyte-Lovejoy D, Lau SK, Boutros PC, et al: The c-Myc oncogene directly induces the H19 noncoding RNA by allele-specific binding to potentiate tumorigenesis. Cancer Res 66:5330-5337, 2006 22. Blackhall FH, Wigle DA, Jurisica I, et al: Validating the prognostic value of marker genes derived from a non-small cell lung cancer microarray study. Lung Cancer 46:197-204, 2004[CrossRef][Medline] 23. Tusher VG, Tibshirani R, Chu G: Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98:5116-5121, 2001 24. Sultan M, Wigle DA, Cumbaa CA, et al: Binary tree-structured vector quantization approach to clustering and visualizing microarray data. Bioinformatics 18:S111-S119, 2002 (suppl 1)[Abstract] 25. Liu J, Blackhall F, Seiden-Long I, et al: Modeling of lung cancer by an orthotopically growing H460SM variant cell line reveals novel candidate genes for systemic metastasis. Oncogene 23:6316-6324, 2004[CrossRef][Medline] 26. Yun JJ, Heisler LE, Hwang II, et al: Genomic DNA functions as a universal external standard in quantitative real-time PCR. Nucleic Acids Res 34:e85, 2006 27. Irizarry RA, Bolstad BM, Collin F, et al: Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res 31:e15, 2003 28. 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 29. Chen HY, Yu SL, Chen CH, et al: A five-gene signature and clinical outcome in non-small-cell lung cancer. N Engl J Med 356:11-20, 2007 30. Brown KR, Jurisica I: Unequal evolutionary conservation of human protein interactions in interologous networks. Genome Biol 8:R95, 2007[CrossRef][Medline] 31. Benjamini Y, Hochberg Y: Controlling the false discovery rate: A practical and powerful approach to multiple testing. J R Stat Soc B 57:289-300, 1995 32. Potti A, Mukherjee S, Petersen R, et al: A genomic strategy to refine prognosis in early-stage non-small-cell lung cancer. N Engl J Med 355:570-580, 2006 33. Lu Y, Lemon W, Liu PY, et al: A gene expression signature predicts survival of patients with stage I non-small cell lung cancer. PLoS Med 3:e467, 2006[CrossRef][Medline] 34. Vandesompele J, De Preter K, Pattyn F, et al: Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol 3:RESEARCH0034, 2002[Medline] 35. 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 36. 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 37. Grabowski P, Schonfelder J, Ahnert-Hilger G, et al: Expression of neuroendocrine markers: A signature of human undifferentiated carcinoma of the colon and rectum. Virchows Arch 441:256-263, 2002[CrossRef][Medline] 38. Graff L, Castrop F, Bauer M, et al: Expression of vesicular monoamine transporters, synaptosomal-associated protein 25 and syntaxin1: A signature of human small cell lung carcinoma. Cancer Res 61:2138-2144, 2001 39. Bascom JL, Fata JE, Hirai Y, et al: Epimorphin overexpression in the mouse mammary gland promotes alveolar hyperplasia and mammary adenocarcinoma. Cancer Res 65:8617-8621, 2005 40. Mashino K, Sadanaga N, Yamaguchi H, et al: Expression of chemokine receptor CCR7 is associated with lymph node metastasis of gastric carcinoma. Cancer Res 62:2937-2941, 2002 41. Schimanski CC, Schwald S, Simiantonaki N, et al: Effect of chemokine receptors CXCR4 and CCR7 on the metastatic behavior of human colorectal cancer. Clin Cancer Res 11:1743-1750, 2005 42. Schimanski CC, Bahre R, Gockel I, et al: Chemokine receptor CCR7 enhances intrahepatic and lymphatic dissemination of human hepatocellular cancer. Oncol Rep 16:109-113, 2006[Medline] 43. Takanami I: Overexpression of CCR7 mRNA in nonsmall cell lung cancer: Correlation with lymph node metastasis. Int J Cancer 105:186-189, 2003[CrossRef][Medline] 44. Pugh CW, Ratcliffe PJ: Regulation of angiogenesis by hypoxia: Role of the HIF system. Nat Med 9:677-684, 2003[CrossRef][Medline] 45. Pouyssegur J, Dayan F, Mazure NM: Hypoxia signalling in cancer and approaches to enforce tumour regression. Nature 441:437-443, 2006[CrossRef][Medline] 46. Chi JT, Wang Z, Nuyten DS, et al: Gene expression programs in response to hypoxia: Cell type specificity and prognostic significance in human cancers. PLoS Med 3:e47, 2006[CrossRef][Medline] 47. Ludwig JA, Weinstein JN: Biomarkers in cancer staging, prognosis and treatment selection. Nat Rev Cancer 5:845-856, 2005[CrossRef][Medline] 48. Canales RD, Luo Y, Willey JC, et al: Evaluation of DNA microarray results with quantitative gene expression platforms. Nat Biotechnol 24:1115-1122, 2006[CrossRef][Medline] 49. Raponi M, Zhang Y, Yu J, et al: Gene expression signatures for predicting prognosis of squamous cell and adenocarcinomas of the lung. Cancer Res 66:7466-7472, 2006 50. Verhaak RG, Staal FJ, Valk PJ, et al: The effect of oligonucleotide microarray data pre-processing on the analysis of patient-cohort studies. BMC Bioinformatics 7:105, 2006[CrossRef][Medline] 51. Seo J, Hoffman EP: Probe set algorithms: Is there a rational best bet? BMC Bioinformatics 7:395, 2006[CrossRef][Medline] 52. Simon R, Radmacher MD, Dobbin K, et al: Pitfalls in the use of DNA microarray data for diagnostic and prognostic classification. J Natl Cancer Inst 95:14-18, 2003 53. Simon R: Roadmap for developing and validating therapeutically relevant genomic classifiers. J Clin Oncol 23:7332-7341, 2005 54. Dupuy A, Simon RM: Critical review of published microarray studies for cancer outcome and guidelines on statistical analysis and reporting. J Natl Cancer Inst 99:147-157, 2007 55. Blackhall FH, Pintilie M, Wigle DA, et al: Stability and heterogeneity of expression profiles in lung cancer specimens harvested following surgical resection. Neoplasia 6:761-767, 2004[CrossRef][Medline] 56. Cronin M, Sangli C, Liu M-L, et al: Analytical validation of the Oncotype/Dx genomic diagnostic test for recurrence prognosis and therapeutic response prediction in node-negative, estrogen receptor-positive breast cancer. Clin Chem 53:1084-1091, 2007 Submitted April 4, 2007; accepted August 20, 2007.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2007 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
0.65 were carried forward for risk score analysis. Ten of the 147 patients had expression values less than the dynamic range of the assay for one or more of these five genes and were excluded from subsequent analysis. 
