|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2005.02.9363 on October 3 2005 © 2005 American Society of Clinical Oncology. Unique Gene Expression Profile Based on Pathologic Response in Epithelial Ovarian CancerFrom the Program of Gynecologic Medical Oncology, Beth Israel Deaconess Medical Center, Genomics Center and Bioinformatics Core, Beth Israel Deaconess Medical Center, Harvard Medical School, and the Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, NY Address reprint requests to Stephen A. Cannistra, Program of Gynecologic Medical Oncology, Beth Israel Deaconess Medical Center, 330 Brookline Ave, Boston, MA 02215; e-mail: scannist{at}bidmc.harvard.edu
PURPOSE: We investigated whether tumor tissue obtained at diagnosis expresses a specific gene profile that is predictive of findings at second-look surgery in patients with epithelial ovarian cancer (EOC). PATIENTS AND METHODS: Tumor tissue obtained at the time of diagnosis was profiled with oligonucleotide microarrays. Class prediction analysis was performed in a training set of 24 patients who had undergone a second-look procedure. The resultant predictive signature was then tested on an independent validation set comprised of 36 patients. RESULTS: A 93-gene signature referred to as the Chemotherapy Response Profile (CRP) was identified through its association with pathologic complete response. When applied to a separate validation set, the CRP distinguished between patients with unfavorable versus favorable overall survival (median 41 months v not yet reached, respectively, log-rank P = .007), with a median follow-up of 52 months. The signature maintained independent prognostic value in multivariate analysis, controlling for other known prognostic factors such as age, stage, grade, and debulking status. There was no genetic overlap between the CRP and our previously described Ovarian Cancer Prognostic Profile (OCPP), which demonstrated similar prognostic value. The combination of the CRP and OCPP yielded better prognostic discrimination then either profile alone. Genes present in the CRP include BAX, a proapoptotic protein previously associated with chemotherapy response in ovarian cancer. CONCLUSION: Identification of a gene expression profile based on pathologic response in EOC provides independent prognostic information and offers potential insights into the mechanism of drug resistance. Efforts to identify a more tailored profile using selected genes from both the CRP and OCPP are underway.
Epithelial ovarian cancer (EOC) is a major cause of gynecologic cancer mortality, with approximately 16,210 deaths expected in 2005.1 Most patients present with advanced disease that is managed with surgical cytoreduction, followed by postoperative chemotherapy.2 First-line chemotherapy with taxane and platinum agents is capable of achieving a complete clinical response (CCR) in the majority of patients with advanced disease, as defined by normal physical examination, CA-125 level, and computed tomography scan.2 Most patients who obtain a CCR will experience relapse due to the presence of subclinical disease that persists after completion of first-line treatment. In this regard, second-look laparotomy or laparoscopy (SLL) have been shown to be more sensitive than clinical assessment for the detection of residual disease in patients with advanced ovarian cancer, who have achieved a CCR. Specifically, approximately 50% to 75% of such patients will have persistent gross or microscopic residual disease at the time of SLL.2 However, there is currently no convincing evidence that additional treatment based on the results of second-look procedures confers a survival advantage. Consequently, this technique has not been extensively performed outside of a clinical trial setting. As a research tool, the SLL provides a more sensitive measure of chemotherapy responsiveness than is possible through clinical means. In the present study, we investigated whether tumor tissue obtained at diagnosis expresses a specific gene profile that is predictive of findings at SLL, and whether such a profile might provide information of clinical or biologic relevance.
Patients The study population consisted of 60 patients with epithelial ovarian cancer from Beth Israel Deaconess Medical Center (BIDMC; n = 30) and Memorial Sloan-Kettering Cancer Center (MSKCC; n = 30). All patients underwent exploratory laparotomy for diagnosis, staging, and debulking, and subsequently received first-line platinum/taxane–based chemotherapy. These patients represent a subset of those previously reported3 and were selected on the basis of having achieved a CCR after first-line chemotherapy (and were therefore candidates for SLL). Patients with stage III or IV disease at MSKCC were considered for second-look laparoscopy (n = 17) or laparotomy (n = 7) to assess pathologic response. During SLL, any nodules or dense adhesions were biopsied and sent for frozen section. If the frozen section revealed disease, or if gross tumor was identified, debulking was attempted. If there were no suspicious lesions or if frozen sections were negative, all adhesions were lysed, and random biopsies were obtained from approximately eight to 10 anatomic areas (anterior and posterior cul-de-sac, right and left pelvic side walls, right and left paracolic gutters, right and left diaphragms, and the anterior abdominal wall). On average, approximately 20 biopsies were obtained per second-look procedure. Postchemotherapy surveillance of patients who achieved a CCR included serial physical examination and serum CA-125 levels, with subsequent computed tomography scanning performed for clinical suspicion of relapse. Follow-up data for this study were extracted from the Ovarian Cancer Relational Database at BIDMC and the Ovarian Cancer Clinical Database at MSKCC.3 The study protocol for collection of tissue and clinical information was approved by the institutional review boards at both institutions. Written informed consent was obtained for the collection and use of tumor tissue.
Clinical Definitions
RNA Isolation, cDNA Synthesis, Microarray Probe Preparation, and Affymetrix GeneChip Hybridization
Bioinformatics and Statistical Analysis
Validation of the Gene Expression Profile
Patient Characteristics Characteristics of the 60 patients in this study are shown in Table 1. The median age at diagnosis was 54 years (range, 36 to 80 years), and the majority had advanced-stage (FIGO stages III/IV, 95%), grade 3 EOC tumors (78%), with serous histology (92%). Sixty-seven percent of patients were optimally cytoreduced after initial surgery (  1 cm residual diameter), and all received postoperative taxane/platinum–based adjuvant chemotherapy. Twenty-four patients from MSKCC underwent second-look procedure, with the finding of RD in 14 patients, and no evidence of disease (pCR) in 10 patients. Within the RD group, 79% had gross residual disease, and 21% had microscopic residual disease. Second-look surgery consisted of a laparoscopy in 17 patients, while a laparotomy was performed in seven patients for adhesiolysis, colostomy reversal, and/or hernia repair. None of the BIDMC patients underwent a second-look procedure.
Development of a Gene Expression Profile Associated With Pathologic Response Pattern-recognition analysis was performed in the training set as previously described.3,7,8,10-13 We derived 134 multigene patterns that included 176 unique genes that could discriminate between pCR and RD at P < .001 (for each pattern), thus suggesting that the two phenotypes expressed distinct molecular signatures. Using the compound covariate predictive algorithm, a 93-gene panel was developed with maximum accuracy in predicting pathologic response status by leave-one-out cross validation (91%, Fisher's P = .001, permutation P = .03). Other predictive algorithms performed similarly, such as weighted voting (86%), k nearest neighbor (91%) and diagonal linear discriminant analysis (91%). The expression pattern of the genes in this panel is shown in the color plot in Figure 1, and is referred to as the chemotherapy response profile (CRP). For the analyses described below, the gene expression signature that correlated with pCR is defined as the "sensitive" profile, and the signature that correlated with RD is defined as the "resistant" profile.
Independent Validation of the CRP We applied the CRP to an independent validation set of patient samples (n = 36, 83% from BIDMC), in order to determine whether the signature was capable of identifying a clinically relevant end point. Because none of these patients underwent SLL, we could not use achievement of pCR as an end point in the validation set, but instead chose DFS and OS as clinically relevant surrogates for this metric. Using hierarchical clustering, two distinct groups were identified in the validation set that were highly correlated with the gene expression pattern shown in Figure 1 (ie, sensitive and resistant groups as defined by the CRP) (Fisher's P = .006). As shown in Figure 2, the CRP was capable of discriminating between groups with short versus prolonged median DFS, with the resistant and sensitive profiles having a median DFS of 9 and 22 months, respectively (log-rank P = .009). Overall survival for the resistant and sensitive CRP groups was 41 months and not yet reached, respectively (P = .007), with a median follow-up of 52 months. The hazard ratio for recurrence (resistant versus sensitive groups) was 2.7 (95% CI, 1.2 to 6.1) and for death 3.9 (95% CI, 1.3 to 11.4) with the Cox Proportional Hazards Model. Excluding the three patients with early stage disease (Table 1) from the survival analysis did not appreciably change the magnitude or statistical significance of the survival split.
Association of the CRP With Clinical Characteristics Table 2 presents the distribution of clinical features as a function of CRP assignment (sensitive or resistant) in the validation set. The two groups defined by the CRP were well balanced for debulking status, grade, and histology (presence or absence of clear-cell elements), although there was a nonsignificant trend toward a higher median age in the sensitive group. In univariate analysis, debulking status showed a prognostic trend for both DFS and OS, while there were too few low grade and clear-cell samples to allow for adequate assessment of their prognostic value. We then performed multivariate analysis including the CRP assignment (ie, sensitive or resistant), age, and debulking status, which demonstrated that the profile maintained an independent association with outcome. The hazard ratio for recurrence (resistant v sensitive groups) was 3.1 (95% CI, 1.3 to 7.6; P = .014), and the hazard ratio for death was 5.0 (95% CI, 1.6 to 15.6; P = .006) by the Cox proportional hazards analysis (Table 3) . Furthermore, in the 27 patients with optimally debulked disease, we found a significant difference in both median DFS (10 v 28 months, P = .01) as well as median OS (36 months v not yet reached, P = .01) for the "resistant" versus the "sensitive" profile, respectively.
Prognostic Assessment by Combining Two Distinct Gene Profiles We recently reported another gene expression profile (Ovarian Cancer Prognostic Profile [OCPP]), which was strongly prognostic of survival in EOC.3 Unlike the gene expression signature identified in the CRP, which reflects findings at SLL, the OCPP was selected for its ability to discriminate between patients with long versus short survival. Interestingly, when comparing the CRP with the OCPP, we found no evidence of overlap between the two gene signatures, despite the fact that both of these profiles provide independent prognostic information. Given this lack of overlap, we considered the possibility that these two profiles are capturing distinct biologic features that dictate tumor behavior, such as chemoresponse on the one hand (eg, the CRP), versus growth rate and metastatic potential on the other (eg, the OCPP). In view of these considerations, we were interested in determining whether the prognostic power of these gene profiles could be further refined by using them in combination. For this purpose, we utilized the validation set from this study and assigned one of 3 categories to each sample as follows: "excellent" (favorable OCPP and sensitive CRP labels); "poor" (unfavorable OCPP and resistant CRP labels); and "intermediate" (profiles that did not fit into the excellent or poor categories). The Kaplan-Meier survival curves as a function of these groups are shown in Figure 3. Because the poor and intermediate groups showed overlapping survival curves, these categories were collapsed into one group for the purpose of this analysis. The median DFS for the intermediate/poor versus excellent groups was 9 v 34 months (P = .001), respectively, and the corresponding median OS was 36 months versus not yet reached (P = .0001), respectively. The hazard ratio for recurrence (intermediate/poor versus excellent groups) was 4.6, (95% CI, 1.6 to 12.8) and for death was 18.0 (95% CI, 2.4 to 137.3), stronger than corresponding hazard ratio estimates observed for the OCPP or the CRP by themselves. Specifically, the hazard ratios for recurrence and death for the OCPP were 3.5 and 4.8, respectively, as reported previously.3 For the CRP, the hazard ratios for recurrence and death were 2.7 and 3.9, respectively, as noted above. Notably, in the combined analysis, the "excellent" group demonstrated a high likelihood of survival despite the continued occurrence of relapse (compare Figs 3A and 3B), suggesting either an indolent natural history and/or prolonged responsiveness to chemotherapy.
Functional Description of the CRP Seventy-seven genes were overexpressed in the resistant group, and 16 genes were overexpressed in the sensitive group (Fig 1). Gene families with potential relevance to chemoresponsiveness were represented in the profile (Table 4). As discussed below, these include genes that regulate apoptosis, cell cycle entry, and DNA repair. For some genes such as BAX or Rb, well-described mechanisms have been previously reported that might explain their role in mediating chemoresponse.21-27 The identities of all genes belonging to the CRP are provided in the online supplement (https://www.bidmcgenomics.org/ovariancancerresponse/index.htm).
The molecular determinants of response to chemotherapy in ovarian cancer remain poorly understood. In this study, we used tumor tissue obtained at the time of initial diagnosis to define a multigene signature referred to as the CRP, selected for its ability to discriminate between patients who achieved a pCR after first line therapy versus those who did not. The fact that this profile strongly correlated with both DFS and OS when applied to an independent validation set of patients suggests that the CRP has clinical relevance. The survival split observed with the CRP is very similar to that previously observed with the OCPP,3 and yet the two profiles share no genetic overlap. This is an interesting observation that undoubtedly relates to how these profiles were identified, and it has important implications for interpreting the biologic significance of microarray data. In the present study, the CRP was identified on the basis of findings at SLL and therefore represents a more proximal metric of chemoresponse. In contrast, the OCPP was identified on the basis of long versus short median survival, without regard to how the patient might have responded to first line chemotherapy, and therefore represents a complex interaction between factors that govern natural history (eg, tumor growth rate and metastatic potential), as well as those that mediate response to treatment. Thus, it is perhaps not surprising that a combination of the CRP and OCPP provides more powerful prognostic information than the use of either profile separately (Fig 3). Comparison of DFS and OS in Figure 3 show that the added prognostic value of the combined CRP/OCPP profiles is partly due to identification of a group with long survival despite experiencing relapse. Whether this is related to a more indolent natural history, and/or continued chemoresponsiveness to agents used in the relapsed setting, will require further investigation. Regardless of the exact mechanism involved, these observations demonstrate how two independently derived gene profiles might be combined to form a powerful prognostic tool, based on their ability to characterize different aspects of tumor biology (eg, chemoresponse and natural history). Furthermore, the fact that the CRP and OCPP share no genetic overlap, and yet together confer complementary prognostic information, suggests that interpretation of microarray results must be taken in the context of the selection process used to define a given profile. Thus, differences in gene expression signatures between two groups of investigators may be due to the end point of interest (eg, chemoresistance v natural history), the algorithms used to define gene expression thresholds, the bioinformatics used to define expression clusters, as well as more subtle differences between the patient populations under study. Our experience suggests that many different gene expression signatures will be identified over the next several years, each containing a divergent set of genes, but each converging on an equally powerful and valid prognostic end point. The gene content of the CRP offers potential insight into mechanisms associated with chemotherapy response in EOC (Table 4). The importance of BAX gene expression in the CRP was of particular interest, in view of past work suggesting a correlation between high BAX protein levels and sensitivity to drugs such as paclitaxel.23,28 BAX is a pro-apoptotic member of the BCL-2 family and has been shown to promote paclitaxel-mediated apoptosis in stable cell line transfectants as well as in preclinical animal models.21-23 In addition, BAX protein expression in ovarian cancer tissue has been shown to correlate with response to paclitaxel-containing chemotherapy and survival in EOC.28 Specifically, high BAX levels are associated with increased paclitaxel responsiveness and prolonged survival, an observation consistent with its known mechanism of action as a pro-apoptotic protein.28 The fact that the BAX gene was identified as a component of the CRP, and that low BAX expression is correlated with the resistant phenotype (Table 4), provides additional support for the clinical importance of this protein. This observation also validates the use of microarray gene expression profiling as a tool to investigate mechanisms of chemotherapy responsiveness in vivo. In contrast to the presence of BAX in the CRP, it is noteworthy that this gene did not appear in our previously published OCPP.3 This is another demonstration of how genes expressed in a given microarray profile must be interpreted in the context of how the profiles were derived, even if the profiles have equivalent prognostic value. Other genes within the CRP may play a role in mediating chemoresponse, although this remains to be proven. For instance, overexpression of the Rb gene in the resistant group is consistent with previous reports demonstrating that high Rb levels are associated with cell cycle arrest and increased DNA repair in response to cytotoxic stimuli.24-27 Likewise, the Ras association domain 1 (RASSF1) gene is also overexpressed in the resistant group and, like Rb, is known to promote cell cycle arrest at the G1 phase.29 Whether RASSF1 overexpression induces a state of relative chemoresistance by promoting DNA repair in G1 is unknown, but represents a testable hypothesis. It is also noteworthy that two receptor tyrosine kinases of the ephrin family (EphB2 and EphB3) are down-regulated in the resistant group (Table 4). Other members of this family (such as EphA2) function as pro-apoptotic molecules, suggesting that decreased expression of EphB2 and EphB3 may promote chemoresistance by impairing the apoptotic response to cell damage.30,31 Other genes of interest such as MKNK1 and ICAM2 (Table 4) have a less well-established role in mediating chemotherapy resistance, although their inclusion in the CRP suggests that they are candidates for further study.323435-36 Although individual genes such as these may prove to have functional relevance, it is likely that the coordinated expression of several distinct gene families will be a more important determinant of chemotherapy responsiveness. Thus, our future efforts will use gene expression profiling to determine how networks of genes may interact to promote a drug resistant phenotype. We and others have thus far defined distinct gene expression profiles that are correlated with clinical outcome in ovarian cancer.3,37,38 We intend to streamline the gene profile by using relevant genes from the CRP and OCPP, in order to develop a tailored gene array that is more amenable to possible clinical use. Although our observations suggest that gene expression profiling is capable of defining prognosis and yielding mechanistic insights into the process of chemoresistance, additional prospective validation is required to determine the ultimate value of this technique in clinical practice and in the conduct of clinical trials.
The authors indicated no potential conflicts of interest.
We acknowledge the efforts of gynecologic oncologists at BIDMC and MSKCC in providing tissue samples used in this analysis.
Supported in part through grants from the Paul Weisman Fund, the Ovarian Cancer SPORE (P50 CA105009, Career Development Award), 1R21CA107352, and the Director's Challenge Grant (U01 CA88175), RO1 CA85467, and U24 DK58739. Authors' disclosures of potential conflicts of interest are found at the end of this article.
1. Jemal A, Murray T, Ward E, et al: Cancer statistics, 2005. CA Cancer J Clin 55:10-30, 2005
2. Cannistra SA: Cancer of the ovary. N Engl J Med 351:2519-2529, 2004
3. 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
4. Li C, Wong WH: Model-based analysis of oligonucleotide arrays: Expression index computation and outlier detection. Proc Natl Acad Sci U S A 98:31-36, 2001
5. Mitsiades N, Mitsiades CS, Poulaki V, et al: Molecular sequelae of proteasome inhibition in human multiple myeloma cells. Proc Natl Acad Sci U S A 99:14374-14379, 2002
6. Mitsiades N, Mitsiades CS, Richardson PG, et al: The proteasome inhibitor PS-341 potentiates sensitivity of multiple myeloma cells to conventional chemotherapeutic agents: Therapeutic applications. Blood 101:2377-2380, 2003 7. Califano A, Stolovitzky G, Tu Y: Analysis of gene expression microarrays for phenotype classification. Proc Int Conf Intell Syst Mol Biol 8:75-85, 2000[Medline]
8. Califano A: SPLASH: Structural pattern localization analysis by sequential histograms. Bioinformatics 16:341-357, 2000 9. Kuppers R, Klein U, Schwering I, et al: Identification of Hodgkin and Reed-Sternberg cell-specific genes by gene expression profiling. J Clin Invest 111:529-537, 2003[CrossRef][Medline] 10. Lepre J, Rice JJ, Tu Y, et al: Genes@Work: An efficient algorithm for pattern discovery and multivariate feature selection in gene expression data. Bioinformatics, 2004
11. Klein U, Tu Y, Stolovitzky GA, et al: Gene expression profiling of B cell chronic lymphocytic leukemia reveals a homogeneous phenotype related to memory B cells. J Exp Med 194:1625-1638, 2001 12. Klein U, Tu Y, Stolovitzky GA, et al: Gene expression dynamics during germinal center transit in B cells. Ann N Y Acad Sci 987:166-172, 2003[Medline]
13. Klein U, Tu Y, Stolovitzky GA, et al: Transcriptional analysis of the B cell germinal center reaction. Proc Natl Acad Sci U S A 100:2639-2644, 2003
14. Hedenfalk I, Duggan D, Chen Y, et al: Gene-expression profiles in hereditary breast cancer. N Engl J Med 344:539-548, 2001 15. Tukey JW: Tightening the clinical trial. Control Clin Trials 14:266-285, 1993[CrossRef][Medline] 16. Armstrong SA, Staunton JE, Silverman LB, et al: MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet 30:41-47, 2002[CrossRef][Medline]
17. Wu B, Abbott T, Fishman D, et al: Comparison of statistical methods for classification of ovarian cancer using mass spectrometry data. Bioinformatics 19:1636-1643, 2003
18. Golub TR, Slonim DK, Tamayo P, et al: Molecular classification of cancer: Class discovery and class prediction by gene expression monitoring. Science 286:531-537, 1999 19. Pomeroy SL, Tamayo P, Gaasenbeek M, et al: Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 415:436-442, 2002[CrossRef][Medline]
20. Eisen MB, Spellman PT, Brown PO, et al: Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 95:14863-14868, 1998
21. Tai YT, Strobel T, Kufe D, et al: In vivo cytotoxicity of ovarian cancer cells through tumor-selective expression of the BAX gene. Cancer Res 59:2121-2126, 1999
22. Strobel T, Kraeft SK, Chen LB, et al: BAX expression is associated with enhanced intracellular accumulation of paclitaxel: A novel role for BAX during chemotherapy-induced cell death. Cancer Res 58:4776-4781, 1998
23. Strobel T, Swanson L, Korsmeyer S, et al: BAX enhances paclitaxel-induced apoptosis through a p53-independent pathway. Proc Natl Acad Sci U S A 93:14094-14099, 1996
24. Bosco EE, Mayhew CN, Hennigan RF, et al: RB signaling prevents replication-dependent DNA double-strand breaks following genotoxic insult. Nucleic Acids Res 32:25-34, 2004
25. Knudsen KE, Booth D, Naderi S, et al: RB-dependent S-phase response to DNA damage. Mol Cell Biol 20:7751-7763, 2000 26. Mayhew CN, Bosco EE, Solomon DA, et al: Analysis of RB action in DNA damage checkpoint response. Methods Mol Biol 281:3-16, 2004[Medline]
27. Plath T, Peters M, Detjen K, et al: Overexpression of pRB in human pancreatic carcinoma cells: Function in chemotherapy-induced apoptosis. J Natl Cancer Inst 94:129-142, 2002 28. Tai YT, Lee S, Niloff E, et al: BAX protein expression and clinical outcome in epithelial ovarian cancer. J Clin Oncol 16:2583-2590, 1998[Abstract]
29. Shivakumar L, Minna J, Sakamaki T, et al: The RASSF1A tumor suppressor blocks cell cycle progression and inhibits cyclin D1 accumulation. Mol Cell Biol 22:4309-4318, 2002 30. Depaepe V, Suarez-Gonzalez N, Dufour A, et al: Ephrin signalling controls brain size by regulating apoptosis of neural progenitors. Nature, 2005 31. Dohn M, Jiang J, Chen X: Receptor tyrosine kinase EphA2 is regulated by p53-family proteins and induces apoptosis. Oncogene 20:6503-6515, 2001[CrossRef][Medline] 32. Waskiewicz AJ, Flynn A, Proud CG, et al: Mitogen-activated protein kinases activate the serine/threonine kinases Mnk1 and Mnk2. Embo J 16:1909-1920, 1997[CrossRef][Medline] 33. Fukunaga R, Hunter T: MNK1, a new MAP kinase-activated protein kinase, isolated by a novel expression screening method for identifying protein kinase substrates. Embo J 16:1921-1933, 1997[CrossRef][Medline]
34. Takata R, Katagiri T, Kanehira M, et al: Predicting response to methotrexate, vinblastine, doxorubicin, and cisplatin neoadjuvant chemotherapy for bladder cancers through genome-wide gene expression profiling. Clin Cancer Res 11:2625-2636, 2005 35. Lee S, Choi EJ, Jin C, et al: Activation of PI3K/Akt pathway by PTEN reduction and PIK3CA mRNA amplification contributes to cisplatin resistance in an ovarian cancer cell line. Gynecol Oncol 97:26-34, 2005[CrossRef][Medline] 36. Perez OD, Kinoshita S, Hitoshi Y, et al: Activation of the PKB/AKT pathway by ICAM-2. Immunity 16:51-65, 2002[CrossRef][Medline]
37. Berchuck A, Iversen ES, Lancaster JM, et al: Patterns of gene expression that characterize long-term survival in advanced stage serous ovarian cancers. Clin Cancer Res 11:3686-3696, 2005
38. Hartmann LC, Lu KH, Linette GP, et al: Gene expression profiles predict early relapse in ovarian cancer after platinum-paclitaxel chemotherapy. Clin Cancer Res 11:2149-2155, 2005 Submitted June 1, 2005; accepted July 21, 2005.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2005 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
