J Clin Oncol 25: by American Society of Clinical Oncology INTRODUCTION

Transcription

1 VOLUME 5 NUMBER 8 OCTOBER 7 JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T Global Histone Modifications Predict Prognosis of Resected Non Small-Cell Lung Cancer Fabrice Barlési, Giuseppe Giaccone, Marielle I. Gallegos-Ruiz, Anderson Loundou, Simone W. Span, Pierre Lefesvre, Frank A.E. Kruyt, and Jose Antonio Rodriguez From the Departments of Medical Oncology and Pathology, Vrije Universiteit Medical Center, Amsterdam, the Netherlands; and Statistics Department, and Department of Thoracic Oncology, Université de la Méditerranée, Assistance Publique Hôpitaux de Marseille, Marseille, France. Submitted February, 7; accepted July, 7. Authors disclosures of potential conflicts of interest and author contributions are found at the end of this article. Address reprint requests to Giuseppe Giaccone, MD, PhD, Medical Oncology Branch, CCR, National Cancer Institute, Center Dr, Bldg, Room N6, Bethesda MD 89-9; giacconeg@mail.nih.gov. 7 by American Society of Clinical Oncology 7-8X/7/58-58/$. DOI:./JCO A B S T R A C T Purpose Epigenetic modifications may contribute to the development and progression of cancer. We investigated whether epigenetic changes involving multiple histones influence prognosis of non small-cell lung cancer (NSCLC) patients. Patients and Methods We used immunohistochemistry to assess histone lysine dimethylation (HKdiMe), and acetylation of histone A lysine 5 (HAK5Ac), histone B lysine, histone lysine 9 (HK9Ac), and histone lysine 8 in resected tumor samples of 8 NSCLC patients. Data were analyzed using a recursive partitioning analysis (RPA). Results The RPA classified the patients into seven distinct prognostic groups based on TNM stage (first node), histology, and histone modifications: HKdiMe ( or 85% tumor cells), HK9Ac ( or 68% tumor cells), and HAK5Ac ( or 5% tumor cells). The seven groups were associated with significantly different disease-free (P.) and overall survival (P.). Interestingly, the four groups determined by stage I patients (below the first node) displayed dramatic differences in survival (median, months in adenocarcinoma patients with HK9Ac 68% v 7 months in nonadenocarcinoma patients with HKdiMe 85%). A Cox model retained age and RPA groups as the sole independent factors significantly influencing overall survival. Conclusion The prognostic influence of epigenetic changes involving multiple histones, in particular HA and H, is greater in early NSCLC, and evaluation of these changes may help in selecting early-stage NSCLC patients for adjuvant treatment. Our observations provide a rationale for the use of a combination of standard chemotherapy with drugs interacting with histone modifications, such as histone deacetylase inhibitors. J Clin Oncol 5: by American Society of Clinical Oncology INTRODUCTION Lung cancer is the leading cause of cancer-related death worldwide, with more than. million deaths each year. Non small-cell lung cancer (NSCLC) represents approximately 8% of all lung cancers. The poor prognosis of NSCLC is due mainly to late diagnosis, inasmuch as only approximately % to % of patients are eligible for tumor resection. Despite surgery, recurrences are common even in early NSCLC, occurring in up to two thirds of patients. Although pathologic staging is the most important prognostic factor for patients who undergo resection, there are broad differences within stages, 5 which are due to the different biologic behavior of the tumor. Thus, the identification of biologic markers might help to assess more precisely the prognosis and to address more clearly the use of adjuvant therapy. Several markers have been identified that have prognostic implication in resected NSCLC. 6,7 However, none is used in routine practice because of conflicting results reported (reviewed by Zhu et al 8 ), technological challenges, or both. In addition to genetic alterations, epigenetic changes may contribute to the development and progression of cancer and leukemia. 9 The term epigenetics refers to a number of molecular mechanisms that regulate gene expression without changing the DNA sequence. These mechanisms include alterations in the methylation status of DNA, covalent modification of histone tails, chromatin remodeling, and micrornas. - Epigenetic deregulation may affect several aspects of tumor cell biology, including cell growth, differentiation, and cell death, and thereby it is likely to 58

2 Global Histone Modifications in Predicting Survival in NSCLC be linked to patient prognosis. Most studies on the role of epigenetic alterations in cancer have focused on the aberrant pattern of DNA methylation; in fact, the potential targeting of DNA methylation as a therapeutic approach for lung cancer treatment is being evaluated. In comparison to DNA methylation, other mechanisms of epigenetic regulation, such as the methylation and acetylation of histones, are less well characterized. However, histone modifications together with DNA methylation seem to have a vital role in organizing nuclear architecture, which in turn is involved in regulating transcription and other nuclear processes (reviewed by Esteller 5 ). A recent report identified the global pattern of histone modification as predictor of the risk of recurrence in prostate cancer patients. 6 By applying an unsupervised clustering algorithm, patients were grouped according to the histone H and H modification patterns determined by immunohistochemistry in a tumor tissue microarray. Using this approach, the authors showed that the groups identified were clinically relevant, and provided prognostic information independently of classical prognostic factors, such as tumor grade, preoperative prostate-specific antigen levels, or capsule invasion. In terms of clinical applicability, patients with a lower risk of recurrence could also be identified by examining a single modification, the dimethylation of histone H lysine (HK). We investigated the pattern of global histone modification as a potential prognostic marker in resected NSCLC. We used immunohistochemistry to evaluate the level of HK dimethylation (HKdiMe), 6 and the acetylation of four additional histone residues, including histone A lysine 5 (HAK5Ac), histone B lysine (HBKAc), histone lysine 9 (HK9Ac), and histone lysine 8 (HK8Ac). We analyzed the relationship between epigenetic changes involving multiple histones and patient prognosis using a recursive partitioning analysis (RPA), with overall survival as the primary end point. PATIENTS AND METHODS Patients Tumor samples of 8 patients with early-stage NSCLC (stage I to IIIA) were included in the present study. Patients underwent radical thoracic surgery of primary tumor (lobectomy or pneumonectomy) and regional lymph nodes at the Vrije Universiteit Medical Center (Amsterdam, the Netherlands), between January 988 and December 995. Patients did not receive any preoperative or postoperative treatment. The histopathologic features of the tumor specimens were classified according to the WHO criteria. 7 The TNM staging was determined according to the 997 International Union Against Cancer classification (Table ). Institutional review board approval was obtained to use archived material for research purposes. Immunohistochemistry We used specific rabbit polyclonal antibodies to detect five different histone modifications, including HKdiMe (Abcam, Cambridge, United Kingdom) and HAK5Ac, HBKAc, HK9Ac, and HK8Ac (all from Cell Signaling Technology, Danvers, MA). The use of antibodies was optimized and, once an appropriate dilution and incubation time had been determined, a whole-tissue section of each tumor was stained with each antibody, using standard two-step indirect immunohistochemistry. After deparaffinization in xylenes, the sections were rehydrated in graded alcohols. Endogenous peroxidase was quenched with.% hydrogen peroxide in methanol at room temperature ( C). The sections were placed in a C solution of. M sodium citrate buffer (ph 6.) for antigen retrieval. Primary antibodies were applied as follows: HKdiMe, : for minutes at room temperature; HAK5Ac, :5 overnight; HBKAc, :5 overnight; HK9Ac, : overnight; HK8Ac, :5 overnight). Detection was accomplished with the Dako Characteristic Table. Patient Characteristics (N ) % Age, years Sex Male 9 8 Female 7 Histology Squamous cell carcinoma 66 5 Adenocarcinoma 8 9 Large-cell carcinoma 8 Differentiation grade Well differentiated Moderately differentiated Poorly differentiated Tumor status T 6 T 9 68 T 8 6 Nodal status N 98 7 N 7 N 7 5 Stage IA 5 9 IB 66 5 IIA 6 IIB 7 IIIA 8 6 (Carpinteria, CA) Envision System, followed by chromogen detection with diaminobenzidine. The sections were counterstained with hematoxylin, dehydrated, and mounted. Negative controls were whole-tissue sections stained in the absence of the primary antibody. A semiquantitative assessment of antibodies staining on whole-tissue sections was performed independently by two observers (F.B. and P.L.), who were blinded to all clinicopathologic variables. In case of discordance, both observers defined the final score of the sample together. The score of each whole-tissue section was based on the frequency, regardless of intensity, of cells with positive nuclear staining (range, % to %) within the tumor. Statistical Analysis Follow-up data were updated in February 5. The overall survival time was defined as the time from the date of surgery to date of death as a result of any cause. Patients who were alive at the date of the last follow-up were censored on that date plus one day. Probability of survival was estimated using the Kaplan-Meier method. Differences in survival were tested by means of the log-rank test. To test whether variables differed across groups, the test or Fisher s exact test was used according to test condition. A Cox proportional hazards model was performed to establish independent factor(s) for survival. Statistical significance was defined as P.5. All of the tests are two sided. Statistical analysis was performed using the SPSS software package, version. (SPSS Inc, Chicago, IL). An RPA 8 was performed to test how histone modifications might influence prognosis. In RPA, a classification tree is constructed that provides decision rules for assigning a sample to a category based on a series of sequential decisions. At each stage, a single predictor is used, and depending on whether the value of the predictor is above or below a selected cutoff value, the sample is assigned to a left or right node. Each of the resulting nodes is then analyzed using the same procedure, although different predictor variables and cutoff values may be used. The cutoff value is selected 59

3 Barlési et al A F > 9 B G C D H I > > 9 Fig. Immunohistochemical analysis of NSCLC tumor samples. Representative examples of positively stained sections (original magnification ) for HKdiMe (A), HAK5Ac (B), HBKAc (C), HK9Ac (D), and HK8Ac (E), from different patients; distribution of staining for HKdiMe (F), HAK5Ac (G), HBKAc (H), HK9Ac (I), and HK8Ac (J) antibodies across all the tumor samples. The y-axis represents the number of samples showing positive staining for the indicated percentage of cells stained (x-axis), scored under highpowered field. 5 E J > > 9 6 JOURNAL OF CLINICAL ONCOLOGY

4 Global Histone Modifications in Predicting Survival in NSCLC to maximize the likelihood-ratio statistic for the test that the probability of a particular sample belonging to a given group is independent of whether the predictor of that sample is above or below the cutoff value. Thus, the cutoff value is chosen to make samples above and below the cutoff value as different as possible with respect to classification. This procedure continues until the data in each node are sufficiently well discriminated or until there are too few data in any node to support additional analysis. RESULTS Samples from six patients were not assessable because of insufficient tumor tissue, leaving samples available for analysis. Major clinical and biologic characteristics of these patients are summarized in Table. Immunohistochemical Study of HKdiMe, HAK5Ac, HBKAc, HK9Ac, and HK8Ac The expression of HKdiMe, HAK5Ac, HBKAc, HK9Ac, and HK8Ac was nuclear rather than diffuse, sometimes with a granular staining pattern (Fig A to E). Some slides showed a low level of heterogeneity in the tumor cells staining. The number of tumor cells stained positive by the anti-hkdime, -HAK5Ac, -HBKAc, -HK9Ac, and -HK8Ac ranged from % to % (mean, 57%, 68%, 8%, %, and 6%, respectively; Figs E and F). The overall median expression of HKdiMe, HAK5Ac, HBKAc, HK9Ac, and HK8Ac was 75%, %, %, 5%, and 8%, respectively. The HKdiMe, HAK5Ac, HBKAc, HK9Ac, and HK8Ac histone modifications were rarely associated within the same tumor sample, with the exception of HK9Ac and HK8Ac (Pearson s correlation coefficient, r.9; P.5). HKdiMe expression was correlated with International Union Against Cancer stage (Pearson s correlation coefficient, r.8; P.). No additional relationship was shown between HKdiMe, HAK5Ac, HBKAc, HK9Ac, and HK8Ac stainings and patient clinical characteristics. Global Histone Modification Patterns and Survival The median follow-up was 8 months. One hundred two patients were dead at the time of this analysis, including three lost to follow-up, and were alive, including two patients alive with a local relapse of their disease. The median overall survival and disease-free survival of the whole population was 8 months (95% CI, to months) and months (95% CI, to months), respectively. The RPA classified patients into seven distinct prognostic groups (Fig and Table ). The first prognostic node was based on pathologic tumor stage, with those with stage II or higher having a worse survival than those with stage I (Fig ). Patients with a pathologic tumor stage II or higher were divided into two additional prognostic groups on the basis of their pathologic stage. No other variable could segregate patients with a pathologic tumor stage III. In contrast, pathologic tumor stage II patients were classified into two separate groups by a third node based on acetylation of HAK5. Patients whose tumors expressed HAK5Ac in less than 5% of tumor cells had a worse survival compared with patients whose tumors had more than 5% of cells expressing HAK5Ac (Fig ). Within the stage I patients, the RPA also provided a second node based on histology. Stage I patients with an adenocarcinoma had a worse prognosis than those with a large-cell or squamous cell carcinomas. RPA incorporated histone modifications as a third node for stage I patients. Dimethylation of HK significantly influenced survival of stage I patients with large-cell or squamous cell carcinomas, with a greater survival for patients whose tumor expressed high levels ( 85% of tumor cells) of HKdiMe (group v group ). Moreover, acetylation of HK9 significantly influenced survival of stage I patients with adenocarcinomas, with a greater survival for patients whose tumors expressed lower levels ( 68% of tumor cells) of HK9Ac (group v group ). The seven groups defined the terminal classification of the patients available and were associated with different disease-free (P.) and overall survival (P.; Table and Fig ). Stage I scc/lcc HKdi Me Histology adc HK9Ac HAK5Ac 85% < 85% < 68% 68% 5% < 5% Stage II Group Group Group Group Group 5 Group 6 Group 7 n = 6 n = 6 n = n = 7 n = 9 n = n = 8 Yes No Yes No Fig. Recursive partitioning analysis (RPA) results individualizing seven different prognostic groups among the samples from non small-cell lung cancer patients who underwent resection included in our study. Each node, where the branches of the RPA tree bifurcate, divides patients according to whether the value of a specific feature (predictor) is above or below a selected cutoff value. The first node is represented by the pathologic stage. In stage I patients, histologic subtype provides the second node, and histone modifications (eg, the percentage of cells stained positively for histone H lysine dimethylation [HKdiMe] or for histone lysine 9 acetylation [HK9Ac]) provide the third node. HAK5Ac, histone A lysine 5 acetylation. 6

5 Barlési et al Table. Median Disease-Free and Overall Survival of the Different Groups of NSCLC Patients Who Underwent Resection Established by the Recursive Partitioning Analysis found to influence overall survival significantly and independently (Table ). Group No. of Patients Disease-Free Survival Overall Survival Median (months) 95% CI Median (months) 95% CI DISCUSSION to 75 7 to to 6 to to to to8 to 5 9 6to7 to to 5 to to8 to8 Abbreviation: NSCLC, non small-cell lung cancer. P.. A Cox proportional hazards model including age, sex, pathologic stage, histology, and RPA groups was analyzed to assess independent factors for overall survival. Sex, pathologic stage, and histology were not retained by the model. Conversely, age and RPA groups were In addition to genetic alterations, epigenetic modifications have been implicated in carcinogenesis and cancer development in various types of solid cancer and hematologic malignancies. 9- Epigenetic modifications seem to have different importance in the development of different tumor types. During prostate carcinogenesis, for example, epigenetic changes are so frequent that they allow discrimination between normal and neoplastic tissue with a specificity of up to %. 9, Furthermore, the prognostic value of certain epigenetic alterations in prostate cancer, breast cancer, and leukemia have been established. Conversely, the contribution of epigenetic changes to lung carcinogenesis seems to be less clear. In this study, the pattern of expression and biologic relevance of some histone modifications were analyzed for the first time in NSCLC patients who had undergone resection. Relevant features of the histone modification patterns A B C D Fig. (A) Overall survival and (B) disease-free survival of the different groups of non small-cell lung cancer patients who underwent resection established by the recursive partitioning analysis (RPA). (C) Overall survival and (D) disease-free survival of the different groups of stage I only patients established by the RPA. 6 JOURNAL OF CLINICAL ONCOLOGY

6 Global Histone Modifications in Predicting Survival in NSCLC Table. Multivariate Analysis Using a Cox Proportional Hazards Model Including the Groups Determined by the RPA, Pathologic Stage, Histology, Sex, and Age Variable HR 95% CI P Group Reference.99. to to to to to to 9.9. Sex: male v female.76. to..86 Histology LCC v SCC.86.5 to.5.6 LCC v ADC..5 to Age 7 years.. to.6.7 NOTE. Pathologic stage was not retained by the model because of colinearity with RPA group variable (linearly dependent covariables). Abbreviations: RPA, recursive partitioning analysis; LCC, large-cell carcinoma; SCC, squamous cell carcinoma; ADC, adenocarcinoma. observed in our series include the predominant but not exclusive detection in neoplastic cells in relation to normal lung epithelium, and a characteristic granular staining pattern. The nontumor exclusive staining as well as the low level of heterogeneity noted in some samples is not surprising, giving the role of histone modifications in cell growth, differentiation, and cell death. We show that histone HKdiMe as well as histone HAK5Ac and HK9Ac influenced overall and disease-free survival of NSCLC patients who had undergone resection, suggesting that the role of epigenetic modifications in lung carcinogenesis should be explored further. The postoperative management of stage I NSCLC is still a matter of debate. Although surgery alone represents the standard of care for stage IA patients, conflicting results have been presented regarding the use of adjuvant chemotherapy for stage IB patients., Despite the relatively early diagnosis, a substantial percentage of stage IB patients will experience relapse and succumb to their disease; 5-year survival of pathologic stage IB treated by surgery alone is 57%. Stage I represents a heterogeneous group as demonstrated by Potti et al. 5 On the basis of a gene-expression profile termed lung metagene, these authors were able to predict recurrence for individual stage I patients significantly better than did clinical prognostic factors, and they identified a subgroup of stage IA patients with a high risk of recurrence. Whether these data provide sufficiently robust and reproducible information regarding the use of adjuvant chemotherapy in early-stage NSCLC remains to be established. On the basis of our results, histone modifications, especially methylation and acetylation of histone H, should also be considered in the attempt to better define prognostic subgroups of stage I NSCLC patients. Interestingly, several drugs interacting with histone modifications have been developed in recent years, such as the histone deacetylase inhibitors. Several structurally distinct classes of histone deacetylase inhibitors have been developed 5 and are being tested in clinical trials. A significant antitumor activity was reported in several tumor types, including NSCLC, for some of these agents, such as pivaloyloxymethyl butyrate and CI-99. 6,7 Numerous additional studies are still ongoing and results are awaited. In our study we used the tool of RPA to assess the prognostic value of several histone modifications in relation to known prognostic features in resected NSCLC. There are several methodologies to investigate the potential prognostic value of clinical and biologic features. Multidimensional scaling, as used by Seligson et al, 6 is a set of data analysis techniques that displays the structure of distance-like data as a geometric picture, on which multiple subsequent analyses usually are performed. 8 Conversely, the central result of RPA is a tree, in which the data are organized (partitioned) into nodes (leaves) along branches; data that are more similar according to specified criteria (eg, overall survival) tend to be localized into the same nodes, whereas more dissimilar data tend to occupy different nodes. 9 Therefore, no additional analysis is needed. Furthermore, RPA is particularly well suited when the connection between the variables is complex, and even when there is missing information. For these reasons we used RPA in this study instead of multidimensional scaling. The selection of ptnm stage, a well-known prognostic factor in NSCLC, as the first node supports the validity of the method. Therefore, the histone modifications retained by the RPA should be regarded as valuable new prognostic factors in resected NSCLC patients; in particular, histone HKdiMe and HK9Ac, which are related to prognosis of those patients with resected stage I NSCLC. However, how these specific epigenetic modifications or combination of modifications affect the expression of particular genes or tumor behavior remains to be fully understood. One possibility is that expression of critical genes may be modulated differently by different epigenetic changes, and this hypothesis should be evaluated further by comparative transcriptional analysis of tumors characterized by different patterns of histone modifications. In conclusion, our results highlight the role of epigenetic changes involving multiple histones in lung cancer, identifying possible useful markers for selection of early-stage (especially stage I) NSCLC patients for adjuvant therapy. Our observations provide a rationale for studies of standard chemotherapy, drugs interacting with histone modifications, or combinations of both for the management of these patients. AUTHORS DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The author(s) indicated no potential conflicts of interest. AUTHOR CONTRIBUTIONS Conception and design: Giuseppe Giaccone, Anderson Loundou, Simone W. Span, Frank A.E. Kruyt, Jose Antonio Rodriguez Financial support: Giuseppe Giaccone Administrative support: Giuseppe Giaccone, Jose Antonio Rodriguez Provision of study materials or patients: Giuseppe Giaccone, Simone W. Span, Jose Antonio Rodriguez Collection and assembly of data: Fabrice Barlési, Giuseppe Giaccone, Marielle I. Gellegos-Ruiz, Simone W. Span, Pierre Lefesvre, Frank A.E. Kruyt, Jose Antonio Rodriguez Data analysis and interpretation: Fabrice Barlési, Giuseppe Giaccone, Anderson Loundou, Frank A.E. Kruyt, Jose Antonio Rodriguez Manuscript writing: Fabrice Barlési, Giuseppe Giaccone, Anderson Loundou, Frank A.E. Kruyt, Jose Antonio Rodriguez Final approval of manuscript: Fabrice Barlési, Giuseppe Giaccone, Marielle I. Gellegos-Ruiz, Anderson Loundou, Simone W. Span, Pierre Lefesvre, Frank A.E. Kruyt, Jose Antonio Rodriguez 6

7 Barlési et al REFERENCES. American Cancer Society: Cancer Facts and Figures 6. Atlanta, GA, American Cancer Society, 6. Travis W, Travis LB, Devesa SS: Lung cancer incidence and survival by histologic type. Cancer 75:9-, 995 (suppl ). Fry W, Phillips JL, Menck HR: Ten-year survey of lung cancer treatment and survival in hospitals in the United States. Cancer 86: , 999. Mountain CF: Revisions in the International System for Staging Lung Cancer. Chest :7-77, Potti A, Mukherjee S, Petersen S, et al: A genomic strategy to refine prognosis in early-stage non-small-cell lung cancer. N Engl J Med 55:57-58, 6 6. Thomas P, Doddoli C, Thirion X, et al: Stage I non-small cell lung cancer: A pragmatic approach to prognosis after complete resection. Ann Thorac Surg 7:65-7, 7. Lu C, Soria JC, Tang X, et al: Prognostic factors in resected stage I non-small-cell lung cancer: A multivariate analysis of six molecular markers. J Clin Oncol :575-58, 8. 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 59:79-8, 6 9. Lund AH, van Lohuizen M: Epigenetics and cancer. Genes Dev 8:5-5,. Gibbons RJ: Histone modifying and chromatin remodeling enzymes in cancer and dysplastic syndromes. Hum Mol Genet :R85-R9, 5. Bhalla KN: Epigenetic and chromatin modifiers as targeted therapy of hematologic malignancies. J Clin Oncol :97-99, 5. Esteller M: The necessity of a human epigenome project. Carcinogenesis 7:-5, 6. Jones PA, Baylin SB: The fundamental role of epigenetic events in cancer. Nat Rev Genet :5-8,. Digel W, Lubbert M: DNA methylation disturbances as novel therapeutic target in lung cancer: Preclinical and clinical results. Crit Rev Oncol Hematol 55:-, 5 5. Esteller M: Cancer epigenomics: DNA methylome and histone-modification maps. Nat Rev Genet 8:86-98, 7 6. Seligson DB, Horvath S, Shi T, et al: Global histone modification patterns predict risk of prostate cancer recurrence. Nature 5:6-66, 5 7. Travis WD, Colby TV, Corrin B, et al: In collaboration with pathologists from countries: World Health Organization International Histological Classification of Tumours Histological typing of lung and pleural tumours (ed ). Berlin, Germany, Springer- Verlag, Zhang HP, Singer B: Recursive Partitioning in the Health Sciences. New York, NY, Springer, Jerónimo C, Henrique R, Hoque MO, et al: Quantitative RARbeta hypermethylation: A promising prostate cancer marker. Clin Cancer Res : -,. Tokumaru Y, Harden SV, Sun DI, et al: Optimal use of a panel of methylation markers with GSTP hypermethylation in the diagnosis of prostate adenocarcinoma. Clin Cancer Res :558-55,. Kim H, Kwon YM, Kim JS, et al: Tumorspecific methylation in bronchial lavage for the early detection of non small-cell lung cancer. J Clin Oncol :6-7,. Strauss GM, Herdon 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 96. J Clin Oncol :6s, (suppl; abstr 79). Strauss GM, Herndon JE, Maddaus MA, et al: Adjuvant chemotherapy in stage IB non-small cell lung cancer (NSCLC): Update of Cancer and Leukemia Group B (CALGB) protocol 96. J Clin Oncol :65s, 6 (suppl; abstr 77). Gilbert J, Gore S, Herman J, et al: The clinical application of targeting cancer through histone acetylation and hypomethylation. Clin Cancer Res : , 5. Lindemann RK, Gabrielli B, Johnstone RW: Histone-deacetylase inhibitors for the treatment of cancer. Cell Cycle : , 6. Reid T, Valone F, Libera W, et al: Phase II trial of the histone deacetylase inhibitor pivaloyloxymethyl butyrate (Pivanex, AN-9) in advanced non-small cell lung cancer. Lung Cancer 5:8-86, 7. Pauer L, Olivares J, Cunningham C, et al: Phase I study of oral CI-99 in combination with carboplatin and paclitaxel in the treatment of patients with advanced solid tumors. Cancer Invest : , 8. Young FW, Hamer RM: Theory and Applications of Multidimensional Scaling. Hillsdale, NJ, Erlbaum Associates, Young SS, Ge N: Recursive partitioning analysis of complex disease pharmacogenetic studies: I Motivation and overview. Pharmacogenomics 6:65-75, 5. Zaykin DV, Young SS: Large recursive partitioning analysis of complex disease pharmacogenetic studies: II Statistical considerations. Pharmacogenomics 6:77-89, 5 6 JOURNAL OF CLINICAL ONCOLOGY