|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 5 (February 10), 2007: pp. 517-525 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2006.06.3743 An Integrated Genomic-Based Approach to Individualized Treatment of Patients With Advanced-Stage Ovarian Cancer
From the Divisions of Gynecologic Surgical Oncology and Cancer Prevention and Control, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL; Institute for Genome Sciences and Policy, Department of Molecular Genetics and Microbiology, Department of Obstetrics and Gynecology/Division of Gynecologic Oncology, and Departments of Surgery and Medicine, Duke University Medical Center; Institute of Statistics and Decision Sciences, Duke University, Durham, NC; and Institute of Medical Genetics, University Hospital of Wales, Cardiff, United Kingdom Address reprint requests to Johnathan M. Lancaster, MD, PhD, H. Lee Moffitt Cancer Center and Research Institute, 12902 Magnolia Dr, Tampa, FL 33612; e-mail: lancasjm{at}moffitt.usf.edu
Purpose: The purpose of this study was to develop an integrated genomic-based approach to personalized treatment of patients with advanced-stage ovarian cancer. We have used gene expression profiles to identify patients likely to be resistant to primary platinum-based chemotherapy and also to identify alternate targeted therapeutic options for patients with de novo platinum-resistant disease. Patients and Methods: A gene expression model that predicts response to platinum-based therapy was developed using a training set of 83 advanced-stage serous ovarian cancers and tested on a 36-sample external validation set. In parallel, expression signatures that define the status of oncogenic signaling pathways were evaluated in 119 primary ovarian cancers and 12 ovarian cancer cell lines. In an effort to increase chemotherapy sensitivity, pathways shown to be activated in platinum-resistant cancers were subject to targeted therapy in ovarian cancer cell lines. Results: Gene expression profiles identified patients with ovarian cancer likely to be resistant to primary platinum-based chemotherapy with greater than 80% accuracy. In patients with platinum-resistant disease, we identified expression signatures consistent with activation of Src and Rb/E2F pathways, components of which were successfully targeted to increase response in ovarian cancer cell lines. Conclusion: We have defined a strategy for treatment of patients with advanced-stage ovarian cancer that uses therapeutic stratification based on predictions of response to chemotherapy, coupled with prediction of oncogenic pathway deregulation, as a method to direct the use of targeted agents.
Ovarian cancer is a leading cause of cancer death among women in the United States and Western Europe and has the highest mortality rate of all gynecologic cancers. Currently, platinum drugs are the most active agents in epithelial ovarian cancer therapy.1-3 Consequently, the standard treatment protocol used in the initial management of advanced-stage ovarian cancer is cytoreductive surgery, followed by primary chemotherapy with a platinum-based regimen that usually includes a taxane.4 Approximately 70% of patients will have a complete clinical response to this initial therapy, with absence of clinical or radiographic detectable residual disease and normalization of serum CA-125 levels.5,6 The remaining 30% of patients will demonstrate residual or progressive platinum-resistant disease. The inability to predict response to specific therapies is a major impediment to improving outcome for women with ovarian cancer. Empiric-based treatment strategies are used and result in many patients with chemotherapy-resistant disease receiving multiple cycles of often toxic therapy without success before the lack of efficacy is identified. In the course of these empiric treatments, patients may experience significant toxicities, compromise to bone marrow reserves, detriment to quality of life, and delay in the initiation of therapy with active agents. Moreover, the lack of active therapeutic agents for patients with platinum-resistant disease limits treatment options. As such, many patients receive chemotherapy with little or no benefit. The clinical heterogeneity of ovarian cancer, resulting from the acquisition of multiple genetic alterations that contribute to the development of the tumor, underlies the heterogeneity of response to chemotherapy.7 Although a variety of gene alterations have been identified, no single gene marker can reliably predict response to therapy and outcome.8-12 Recent advances in the use of DNA microarrays, which allow global assessment of gene expression in a single sample, have shown that expression profiles can provide molecular phenotyping that identifies distinct classifications not evident by traditional histopathologic methods.13-20 Our group and others have applied this approach to describe gene expression profiles associated with ovarian cancer development, surgical debulking, response to therapy, and survival.21-27 We have now applied genomic methodologies to identify gene expression patterns within primary tumors that predict response to primary platinum-based chemotherapy. We have coupled this analysis with gene expression signatures that reflect the deregulation of various oncogenic signaling pathways to identify unique characteristics of the platinum-resistant cancers that can guide the use of these drugs in patients with platinum-resistant disease. We propose integrating gene expression profiles that predict platinum response and oncogenic pathway status as a strategy for developing personalized treatment plans for individual patients.
Patients and Tissue Samples Clinicopathologic characteristics of the 119 patients who contributed the ovarian cancer samples included in this study are listed in Table 1. All ovarian cancer samples were obtained at initial cytoreductive surgery from patients treated at Duke University Medical Center and H. Lee Moffitt Cancer Center and Research Institute, who then received platinum-based primary chemotherapy. The samples were divided (70:30 ratio) into training and validation sets. As a result, 83 (70%) of 119 samples were randomly selected for the training set, and 36 (30%) of 119 samples were selected for the validation set. In the training set, a total of 59 (71%) of 83 patients demonstrated a complete response (CR) and 24 (29%) of 83 patients demonstrated an incomplete response (IR) to primary platinum-based therapy after surgery. In the validation set, a total of 26 (72%) of 36 patients demonstrated a CR and 10 (28%) of 36 patients demonstrated an IR to primary platinum-based therapy. The distribution of CR and IR in both the training and validation sets was selected to reflect clinical CR rates of approximately 70%. The distribution of debulking status within the training and validation sets was equally balanced. All tissues were collected under the auspices of respective institutional review board–approved protocols with written informed consent.
Measurement of Clinical Response Response to therapy in ovarian cancer patients was evaluated from the medical record using standard WHO criteria for patients with measurable disease.28 CA-125 was used to classify responses only in the absence of a measurable lesion; CA-125 response criteria were based on established guidelines.29,30 A CR was defined as a complete disappearance of all measurable and assessable disease or, in the absence of measurable lesions, a normalization of the CA-125 level after adjuvant therapy. Patients were considered to have an IR if they demonstrated only a partial response, had stable disease, or demonstrated progressive disease during primary therapy. A partial response was considered a 50% or greater reduction in the product obtained from measurement of each bidimensional lesion for at least 4 weeks or a decrease in the CA-125 level by at least 50% for at least 4 weeks. Disease progression was defined as a 50% or greater increase in the product from any lesion documented within 8 weeks of initiation of therapy, the appearance of any new lesion within 8 weeks of initiation of therapy, or any increase in the CA-125 from baseline at initiation of therapy. Stable disease was defined as disease not meeting any of the above criteria.
RNA and Microarray Analysis
Statistical Analysis
Cell Lines and RNA Extraction
Cell Proliferation Assays
Gene Expression Profiles That Predict Platinum Response With the ultimate objective of developing a strategy for determining the most appropriate therapy for an individual patient with ovarian cancer, we developed a predictive tool that identifies patients with platinum-resistant disease at the time of initial diagnosis. The 83-sample training set was used to identify a gene expression pattern that could predict clinical outcome. Using a cutoff of 0.47 predicted probability of response, as determined by receiver operating characteristic curve analysis (Fig 1A), platinum response in patients was predicted accurately in 70 of 83 samples, achieving an overall accuracy of 84.3% (specificity 85%; sensitivity 83%; Fig 1B). Applying a Mann-Whitney U test for statistical significance (P < .001) demonstrates the capacity of the predictor to distinguish nonresponder patients from responder patients.
A validation of the predictive performance of the gene expression model was performed on a randomly generated set of 36 samples to evaluate the ability of the model to predict platinum response. Both the training and validation sets were balanced with respect to platinum response rates seen in the clinic (ie, approximately 70% complete responders). On the basis of the cutoff of 0.47 as defined in the training set, it is evident that the predicted platinum response in the training set performs well to predict the response within the separate validation set (78% accuracy; Fig 1C). When other clinical variables, such as debulking status and CA-125, were included in the SSS to determine platinum response predictions, there was no effect on the predicted accuracy or gene content of the models, suggesting that the signature of platinum response is independent of other clinical variables. Given these results, we conclude that it is possible to develop gene expression profiles that have the capacity to predict response to platinum-based chemotherapy and thus serve as a mechanism to stratify patients with respect to treatment. Although the ability to identify responsive patients is not likely a primary goal, a capacity to identify the patients resistant to platinum therapy would be a significant benefit in guiding more effective treatment for these patients. In this context, an emphasis on the specificity of predicting resistance might be the most appropriate goal. A total of 1,727 genes were included in the averaged predictive model, and the 100 genes most weighted in achieving the prediction are listed in Table 2. Analysis of gene ontology categories represented by these genes is depicted in Appendix Table A1 (online only). The analysis reveals an enrichment for genes reflecting cell proliferation and cell growth, certainly consistent with a mechanism of action of cytotoxic chemotherapeutic agents such as cisplatin and paclitaxel that generally are directed at the proliferative capacity of the cancer cell.
Identifying Therapeutic Options for Patients With De Novo Platinum-Resistant Ovarian Cancer The development of a predictor that can identify patients likely to be resistant to primary platinum therapy provides an opportunity to effectively identify the population most likely to benefit from additional therapeutic intervention. The challenge is determining what other therapies might benefit these patients. Although in principle it might be possible to use the gene expression data to deduce the critical biologic distinction(s) that predicts platinum response, in practice this is difficult because of our limited knowledge of the integration of biologic pathways and systems. We believe an alternative strategy is one that makes use of an ability to profile the status of various oncogenic signaling pathways within the tumor. We have recently described the development of gene expression signatures that reflect the activation status of several oncogenic pathways and have shown that these signatures can evaluate the status of the pathways in a series of tumor samples, providing a prediction of relative probability of pathway deregulation of each tumor.34 To explore the potential for using this as an approach to identify new therapeutic options, we made use of the previously developed signatures to predict the status of these pathways in the tumors. In each case, the probability of pathway activation in a given tumor is predicted from the signature developed by expression of the activating oncogene in quiescent epithelial cell cultures. Evidence for high probability of pathway activation is indicated by red, and evidence for low probability is indicated by blue (Fig 2A). Initial analyses revealed that a substantial number of the tumors exhibit Src pathway deregulation. In Figure 2A, the tumor samples are sorted based on the predicted level of Src activity. The Kaplan-Meier survival analysis in Figure 2B illustrates further that those patients with deregulated Src pathway also exhibit the worst prognosis. However, in complete responders, there was no evident relationship between Src and E2F3 pathway deregulation and survival (Fig 2D and 2E). An examination of other pathways in the context of the Src pathway deregulation revealed Myc and E2F3 to be frequently deregulated in the tumors lacking Src activity. Although Myc pathway deregulation does not link with available therapeutics, E2F3 deregulation does suggest an opportunity for use of a cyclin-dependent kinase (CDK) inhibitor. We further explored the potential of these two pathway signatures (Src and E2F3) to direct the use of inhibitors that target these pathways.
In parallel with the determination of pathway status in the tumors, we characterized the status of the pathways in a series of ovarian cancer cell lines (Fig 3A). This analysis provides a baseline measure of the status of these pathways that can be compared with the sensitivity of the cells to therapeutic drugs known to target specific activities within given oncogenic pathways. The goal is to determine whether a cell line is sensitive to a drug based on the knowledge of the pathway deregulation within that cell. For the Src pathway, we made use of an Src-specific inhibitor (SU6656), and for the E2F3 pathway, we made use of a CDK inhibitor (roscovitine). The ability of these agents to inhibit growth of the ovarian cancer cell lines was assessed using assays of cell proliferation. In Figures 3B and 3C, a clear and statistically significant relationship can be seen between prediction of either Src or E2F3 pathway deregulation and sensitivity to the respective therapeutic of that pathway. As such, it is evident from these results that predicted pathway deregulation predicts sensitivity to the pathway-specific therapeutic agent.
Although the goal of the use of pathway predictions is to identify options for patients with platinum-resistant ovarian cancer, it is nevertheless true that most of the patients with platinum-resistant disease will show some evidence of response to platinum therapy. The utilization of targeted therapeutics, such as the Src or CDK inhibitor, likely would be in conjunction with standard cytotoxic chemotherapies such as carboplatin and paclitaxel. We have further investigated the extent to which there may be an additive effect of combined therapies. A collection of ovarian cancer cell lines was assayed for sensitivity to cisplatin either with or without SU6656 or roscovitine. In Figure 4, the response was plotted as a function of pathway prediction (either Src or E2F3), and as seen previously, there is a relationship between pathway deregulation and SU6656 or roscovitine drug sensitivity. In contrast, there was no evident relationship between pathway deregulation and cisplatin sensitivity. Nevertheless, there was evidence for a greater sensitivity to the combination of cisplatin and SU6656 compared with either agent alone, whereas there was no evident added benefit of cisplatin combined with roscovitine versus roscovitine alone.
Taken together, these results demonstrate a capacity of a pathway signature to not only predict deregulation of the pathway but also to predict sensitivity to therapeutic agents that target the corresponding pathways. We suggest that this is a viable approach for directing the use of various therapeutic agents.
Treatment of patients with advanced-stage ovarian cancer is empiric, and almost all patients receive a platinum drug, usually with a taxane. Although many patients achieve a clinical CR to platinum-based primary therapy, a significant fraction of patients either have an IR or develop progression of disease during primary therapy. Recently, several groups have used genomic approaches to delineate genes that may impact ovarian cancer platinum responsiveness.24-27 Although we can identify some commonality of gene family/function (ie, zinc finger proteins, ubiquitin-specific proteases, protein phosphatases, and DNA mismatch repair genes) between our platinum predictor and those of others,24-27 common genes do not seem to be represented, which could be a result of the use of cDNA-based microarrays by other groups. Strategies for the treatment of patients determined to be resistant to platinum-based chemotherapy involve the use of various empiric-based salvage chemotherapy agents that often have only marginal benefit. Although it is possible that, based on knowledge that the patient is unlikely to benefit from platinum therapy, initiation of salvage agents as first-line therapy would achieve a greater benefit, we believe a more effective strategy may be the use of agents that target components of pathways that are seen to be deregulated in individual cancers. Thus, the therapeutic strategy is tailored to the individual patient based on knowledge of the unique molecular alterations in their tumor. Individualizing treatments by identifying those patients unlikely to respond fully to the primary platinum-based therapy coupled with an ability to identify characteristics unique to this group of patients can direct the use of novel therapeutic strategies. This truly represents a move toward the goal of personalized treatment. An outline of the approach afforded by these developments is summarized in Figure 5. The capacity to predict likely response to platinum chemotherapy based on gene expression data obtained from the primary tumor can identify those patients most appropriate for additional therapies. The purpose of this assessment is not to direct the use of primary platinum-based chemotherapy but rather to identify that subset of patients who most likely will benefit from additional therapies. The use of pathway predictions provides a basis for utilization of drugs specific to the deregulated pathway in patients predicted to have platinum-resistant disease. In Figure 5, this might involve a choice of either an Src inhibitor or a CDK inhibitor based on the observation that these two pathways dominate ovarian cancers and the results that demonstrate a capacity of these pathway predictors to also predict sensitivity to these agents. Given the fact that most patients demonstrate some (if not complete) response to platinum therapy, we would expect that, for now, all patients would still receive standard platinum therapy, but patients predicted to have an IR to platinum would also receive a targeted therapeutic.
We believe that the approach described here, using gene expression profiles that predict primary chemotherapy response coupled with expression data that identify oncogenic pathway deregulation to stratify patients to the most appropriate treatment regimen, represents an important step toward the goal of personalized cancer treatment. We further suggest that a major benefit of this approach (and, in particular, the use of pathway information to guide the use of targeted therapeutics) is the capacity to ultimately direct the formulation of combinations of therapies (multiple drugs that target multiple pathways) based on information that details the state of activity of the pathways.
The authors indicated no potential conflicts of interest.
Conception and design: Holly K. Dressman, Andrew Berchuck, Andrea Bild, Jonathan Gray, Joseph R. Nevins, Johnathan M. Lancaster Financial support: Joseph R. Nevins, Johnathan M. Lancaster Administrative support: Joseph R. Nevins, Johnathan M. Lancaster Provision of study materials or patients: Andrew Berchuck, Robyn Sayer, Janiel Cragun, Johnathan M. Lancaster Collection and assembly of data: Holly K. Dressman, Andrew Berchuck, Gina Chan, Andrea Bild, Robyn Sayer, Janiel Cragun, Jennifer Clarke, Regina S. Whitaker, LiHua Li, Joseph R. Nevins, Johnathan M. Lancaster Data analysis and interpretation: Holly K. Dressman, Andrew Berchuck, Gina Chan, Jun Zhai, Andrea Bild, Jennifer Clarke, LiHua Li, Jeffrey Marks, Geoffrey S. Ginsburg, Anil Potti, Mike West, Joseph R. Nevins, Johnathan M. Lancaster Manuscript writing: Holly K. Dressman, Andrew Berchuck, Jun Zhai, Andrea Bild, Jennifer Clarke, LiHua Li, Jonathan Gray, Geoffrey S. Ginsburg, Anil Potti, Mike West, Joseph R. Nevins, Johnathan M. Lancaster Final approval of manuscript: Holly K. Dressman, Andrew Berchuck, Gina Chan, Jun Zhai, Andrea Bild, Robyn Sayer, Janiel Cragun, Jennifer Clarke, Regina S. Whitaker, LiHua Li, Jonathan Gray, Jeffrey Marks, Geoffrey S. Ginsburg, Anil Potti, Mike West, Joseph R. Nevins, Johnathan M. Lancaster
RNA isolation. Approximately 30 mg of tissue was added to a chilled BioPulverizer H tube (Bio101, Solon, OH). Lysis buffer from the Qiagen RNeasy Mini kit (Qiagen, Valencia, CA) was added, and the tissue was homogenized for 20 seconds in a Mini-Beadbeater (BioSpec Products, Bartlesville, OK). Tubes were spun briefly to pellet the garnet mixture and reduce foam. The lysate was passaged through a 21-gauge needle 10 times to shear genomic DNA. Total RNA was extracted using the Qiagen RNeasy Mini kit. Quality of the RNA was measured using an Agilent 2100 Bioanalzyer (Agilent, Santa Clara, CA). Statistical analysis. The expression intensities for all genes across the samples were normalized using the robust multiarray average (RMA) (Irizarry RA, Hobbs B, Collin F, et al. Biostatistics 4:249-263, 2003), including probe-level quantile normalization and background correction, as implemented in the Bioconductor software suite (Bolstad BM, Irizarry RA, Astrand M, et al: Bioinformatics 19:185-193, 2003). RMA data were prescreened to remove genes/probes with trivial variation across the sample and low median expression levels; thus, 6,088 genes/probes were used in the analysis. The remaining RMA data were further processed by applying sparse regression model methods (Lucus J, Carvalho C, Wang Q, et al: Sparse Statistical Modeling in gene Expression Genomics. Cambridge, United Kingdom, Cambridge University Press, 2006) to correct for assay artifacts; the resulting expression files are available at http://data.cgt.duke.edu/platinum.php. A binary logistic regression model analysis and a stochastic regression model search, called shotgun stochastic search, was used to determine platinum response prediction models in the training set of 83 samples. The predictive analysis evaluated regression models linking log values of observed expression levels of small numbers of genes to platinum response and debulking status. As mentioned in previous publications (Rich J, Jones B, Hans C, et al: Cancer Res 65:4051-4058, 2005; Hans C, Dobra A, West M. J Am Stat Assoc [in press]), the challenge of statistical analysis is to search for subsets of genes that together define significant predictive regressions, that is, to select both the number k of genes, or variables (platinum response and debulking status), and then the specific set of genes {x1, ..., xk} by searching over subsets. This includes the possibility of no association with any genes (ie, k = 0). Technically, with many genes available, this requires some form of stochastic search (ie, shotgun stochastic search) that, in a distributed computer environment, allows the rapid evaluation of many such models so long as the search is constrained to values of k that are reasonably small, a precept consistent with both the small sample size constraint of many gene expression studies and also scientific parsimony and the need to penalize models on larger numbers of predictors to avoid overfitting. With several thousand genes as possible predictors (subsets of the 6,088 genes/probes), there is a large number of candidate regressions to explore even when restricting the number of genes in any one model to be no more than eight genes. The parallel computational strategies implemented are very efficient, and the search over models generally focuses quickly on subsets of relevant models with higher probability (if such exist). In the analysis here with the training set of 83 samples, the average of 5,000 small models (total number of genes = 1,727) confirms that a number of models containing one to five genes are of some interest. The Bayesian analysis heavily penalizes more complex models, initially very strongly favoring the null hypothesis of no significant predictors in this model context among the thousands of genes in a manner that naturally counters the false discovery propensity of purely likelihood-based model search analyses. In addition, routine calculations confirm that the false-positive rate for discovery of single variable regressions as significant as those identified among the top candidates here is small. From the 5,000 regression models that identify a total of 1,727 genes, Table 2 lists the 100 genes that contribute the most weight in the prediction and that appeared most often within the models. The full list of 1,727 genes is posted on the Web site mentioned earlier. The overall practical relevance of the set of regressions identified (as opposed to nominal statistical significance of any one model) is evaluated by cross-validation prediction. Predictions are based on standard Bayesian model averaging–weighted model averaging; the models identified are evaluated according to their relative data-based probabilities of model fit, and these probabilities provide weights to use in averaging predictions for the hold-out (or future) tumor samples. Analysis of sensitivity and specificity in the prediction of platinum response in the training set was performed by using the receiver operating characteristic curve to define estimated sensitivity and specificity with respect to each prediction of platinum response. The percent accuracy of the models for the validation set (n = 36) was determined by the predicted probability of sensitivity and specificity determined by the receiver operating characteristic curve (probability = 0.47) for the training set. The analysis approach for the prediction of oncogenic pathway deregulation has been previously described.36 Cell proliferation assays. Briefly, growth curves for the ovarian cancer cell lines were carried out by plating 300 to 4,000 cells per well of a 96-well plate. The growth of cells at 12-hour time points (from t = 12 hours) was determined using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay Kit (Promega, Madison, WI), which is a colorimetric method for determining the number of growing cells. Sensitivity to an Src inhibitor (SU6656), cyclin-dependent kinase/E2F inhibitor (roscovitine), and cisplatin was determined by quantifying the percent reduction in growth (v dimethylsulfoxide controls) at 120 hours using a standard MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium] colorimetric assay (Promega). Concentrations used for individual and combination treatments were from 0 to 50 µM for SU6656, roscovitine, and cisplatin. The degree of proliferation inhibition was plotted as a function of probability of Src pathway activation or E2F3 pathway activation. A linear regression analysis demonstrates statistically significant relationships between percent response and probability of Src activity. Significant relationships included P < .001 between cisplatin plus SU6656 versus cisplatin alone, P = .0003 between cisplatin plus SU6656 versus SU6656 alone, and P = .01 for cisplatin versus SU6656 in relationship to probability of Src activity. A linear regression analysis of inhibition of proliferation plotted as a function of E2F3 pathway activity demonstrates statistically significant (P = .02) relationship only between roscovitine and probability of E2F3 activity.
We thank Kaye Culler for her help in the preparation of the article. This research was supported by the NIH 1R21CA110499-01A2 and Department of Defense, National Functional Genomics Center project, under Award No. DAMD17-02-2-0051. Views and opinions of, and endorsements by, the author(s) do not reflect those of the US Army or the Department of Defense.
Supported by the Ovarian Cancer Research Fund, Gynecologic Cancer Foundation's Molly Cade Ovarian Cancer Award, Hearing the Ovarian Cancer Whisper Organization, and the Jacquie Liggett Fellowship in Ovarian Cancer Research. Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
1. Levin L, Simon R, Hryniuk W: Importance of multiagent chemotherapy regimens in ovarian carcinoma: Dose intensity analysis. J Natl Cancer Inst 85:1732-1742, 1993 2. McGuire WP, Hoskins WJ, Brady MF, et al: Assessment of dose-intensive therapy in suboptimally debulked ovarian cancer: A Gynecologic Oncology Group study. J Clin Oncol 13:1589-1599, 1995[Medline] 3. Jodrell DI, Egorin MJ, Canetta RM, et al: Relationships between carboplatin exposure and tumor response and toxicity in patients with ovarian cancer. J Clin Oncol 10:520-528, 1992 4. McGuire WP, Hoskins WJ, Brady MF, et al: Cyclophosphamide and cisplatin compared with paclitaxel and cisplatin in patients with stage III and stage IV ovarian cancer. N Engl J Med 334:1-6, 1996 5. McGuire WP, Brady MF, Ozols RF: The Gynecologic Oncology Group experience in ovarian cancer. Ann Oncol 10:29-34, 1999 6. Piccart MJ, Bertelsen K, Stuart G, et al: Long-term follow-up confirms a survival advantage of the paclitaxel-cisplatin regimen over the cyclophosphamide-cisplatin combination in advanced ovarian cancer. Int J Gynecol Cancer 13:144-148, 2003[CrossRef][Medline] 7. Wenham RM, Lancaster JM, Berchuck A: Molecular aspects of ovarian cancer. Best Pract Res Clin Obstet Gynaecol 16:483-497, 2002[CrossRef][Medline] 8. Berchuck A, Kohler MF, Marks JR, et al: The p53 tumor suppressor gene frequently is altered in gynecologic cancers. Am J Obstet Gynecol 170:246-252, 1994[Medline] 9. Kohler MF, Marks JR, Wiseman RW, et al: Spectrum of mutation and frequency of allelic deletion of the p53 gene in ovarian cancer. J Natl Cancer Inst 85:1513-1519, 1993 10. Havrilesky L, Alvarez AA, Whitaker RS, et al: Loss of expression of the p16 tumor suppressor gene is more frequent in advanced ovarian cancers lacking p53 mutations. Gynecol Oncol 83:491-500, 2001[CrossRef][Medline] 11. Reles A, Wen WH, Schmider A, et al: Correlation of p53 mutations with resistance to platinum-based chemotherapy and shortened survival in ovarian cancer. Clin Cancer Res 7:2984-2997, 2001 12. Schmider A, Gee C, Friedmann W, et al: p21 (WAF1/CIP1) protein expression is associated with prolonged survival but not with p53 expression in epithelial ovarian carcinoma. Gynecol Oncol 77:237-242, 2000[CrossRef][Medline] 13. Wong KK, Cheng RS, Mok SC: Identification of differentially expressed genes from ovarian cancer cells by MICROMAX cDNA microarray system. Biotechniques 30:670-675, 2001[Medline] 14. Welsh JB, Zarrinkar PP, Sapinoso LM, et al: Analysis of gene expression profiles in normal and neoplastic ovarian tissue samples identifies candidate molecular markers of epithelial ovarian cancer. Proc Natl Acad Sci U S A 98:1176-1181, 2001 15. Shridhar V, Lee J-S, Pandita A, et al: Genetic analysis of early- versus late-state ovarian tumors. Cancer Res 61:5895-5904, 2001 16. Schummer M, Ng WW, Bumgarner RE, et al: Comparative hybridization of an array of 21,500 ovarian cDNAs for the discovery of genes overexpressed in ovarian carcinomas. Gene 238:375-385, 1999[CrossRef][Medline] 17. Ono K, Tanaka T, Tsunoda T, et al: Identification by cDNA microarray of genes involved in ovarian carcinogenesis. Cancer Res 60:5007-5011, 2000 18. Sawiris GP, Sherman-Baust CA, Becker KG, et al: Development of a highly specialized cDNA array for the study and diagnosis of epithelial ovarian cancer. Cancer Res 62:2923-2928, 2002 19. Jazaeri AA, Yee CJ, Sotiriou C, et al: Gene expression profiles of BRCA1-linked, BRCA2-linked, and sporadic ovarian cancers. J Natl Cancer Inst 94:990-1000, 2002 20. Schaner ME, Ross DT, Ciaravino G, et al: Gene expression patterns in ovarian carcinomas. Mol Biol Cell 14:4376-4386, 2003 21. Lancaster JM, Dressman H, Whitaker RS, et al: Gene expression patterns that characterize advanced stage serous ovarian cancers. J Soc Gynecol Investig 11:51-59, 2004 22. Berchuck A, Iversen ES, Lancaster JM, et al: Patterns of gene expression that characterize long term survival in advanced serous ovarian cancers. Clin Cancer Res 11:3686-3696, 2005 23. Berchuck A, Iversen E, Lancaster JM, et al: Prediction of optimal versus suboptimal cytoreduction of advanced stage serous ovarian cancer using microarrays. Am J Obstet Gynecol 190:910-925, 2004[CrossRef][Medline] 24. Jazaeri AA, Awtrey CS, Chandramouli GV, et al: Gene expression profiles associated with response to chemotherapy in epithelial ovarian cancers. Clin Cancer Res 11:6300-6310, 2005 25. Helleman J, Jansen MP, Span PN, et al: Molecular profiling of platinum resistant ovarian cancer. Int J Cancer 118:1963-1971, 2005[CrossRef] 26. Spentzos D, Levine DA, Kolia S, et al: Unique gene expression profile based on pathologic response in epithelial ovarian cancer. J Clin Oncol 23:7911-7918, 2005 27. Spentzos D, Levine DA, Ramoni MF, et al: Gene expression signature with independent prognostic significance in epithelial ovarian cancer. J Clin Oncol 22:4700-4710, 2004 28. Miller AB, Hoogstraten B, Staquet M, et al: Reporting results of cancer treatment. Cancer 47:207-214, 1981[CrossRef][Medline] 29. Rustin GJ, Nelstrop AE, Bentzen SM, et al: Use of tumor markers in monitoring the course of ovarian cancer. Ann Oncol 10:21-27, 1999 30. Rustin GJ, Nelstrop AE, McClean P, et al: Defining response of ovarian carcinoma to initial chemotherapy according to serum CA 125. J Clin Oncol 14:1545-1551, 1996 31. Irizarry RA, Hobbs B, Collin F, et al: Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4:249-263, 2003[Abstract] 32. Bolstad BM, Irizarry RA, Astrand M, et al: A comparison of normalizaton methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19:185-193, 2003 33. Lucus J, Carvalho C, Wang Q, et al: Sparse Statistical Modeling in Gene Expression Genomics. Cambridge, United Kingdom, Cambridge University Press, 2006 34. Bild A, 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] Submitted February 27, 2006; accepted November 17, 2006. Related Correspondence
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
INTRODUCTION
