Dr. Richard Simon, Chief

The Biometric Research Branch (BRB) is the statistical and biomathematical component of DCTD. Its members provide statistical leadership for the Division's national research programs in developmental therapeutics, developmental diagnostics, cancer imaging, and clinical trials. Staff functions include reviewing protocols for therapeutic and diagnostic clinical trials sponsored by the Division, serving on data safety monitoring committees, and providing statistical advice to DCTD scientific administrators. Branch members conduct research in biostatistics, bioinformatics, and computational biology on topics ranging from methodology for DNA microarrays and biomarker development to design and analysis of clinical trials for new therapeutic and diagnostic approaches. BRB staff members also collaborate with intramural and extramural scientists on basic and translational cancer research.

Reduced Mortality for Breast Cancer Patients
Citron ML, Berry DA, Cirincione C, et al. Randomized trial of dose-intense versus conventionally scheduled and sequential versus concurrent combination chemotherapy as postoperative adjuvant treatment of node-positive primary breast cancer: First report of intergroup trial C9741. Journal of Clinical Oncology 2003;21(8):1431-9.

Combinatorial Peptide Library Data Analysis Identifies T-cell Epitopes in Autoimmune Diseases and Cancers
The Molecular Statistics and Bioinformatics Section collaborated with Dr. Roland Martin's laboratory at the National Institute of Neurological Disease and Stroke and the Torrey Pines Institute of Molecular Medicine to develop methods for analyzing combinatorial peptide library data to identify T-cell epitopes in autoimmune diseases and cancers. The paper by Sung, Zhao, Martin, and Simon reports an improved method for such analysis that is the first algorithm to consider the multiple register alignments possible for peptides in the major groove of major histocompatibility complex (MHC) molecules. Papers by Rubio-Godoy et al. and Borras et al. report on applying methods developed by the BRB to autoimmune diseases and cancer. The paper by Sung and Simon reports on a new model developed by BRB for predicting whether a peptide will bind to a specified MHC molecule based on the peptide's biophysical properties. The model is more accurate than several other MHC-binding models and the paper reports using the model to identify T-cell epitope candidates in tumors that may be good targets for therapeutic vaccines.

References 1. Sung M, Zhao Y, Martin R, Simon R. T-cell epitope prediction with combinatorial peptide libraries. Journal of Computational Biology 2002;9:527-39. 2. Sung MH, Simon R. Candidate epitope identification using peptide property models: Application to cancer immunotherapy. Methods 2003; (in press). 3. Rubio-Godoy V, Dutoit V, Zhao Y, et al. Positional scanning-synthetic peptide library-based analysis of self and pathogen derived peptide cross-reactivity with tumor reactive melan-A specific CTL. Journal of Immunology 2002;169:5696-707. 4. Rubio-Godoy V, Pinilla C, Dutoit V, et al. Towards PS-SCL based identification of CD8+ tumor reactive T cell ligands: a comparative analysis of PS-SCLs recognition by a single tumor-reactive CD8+ CTL clone. Cancer Research 2002;62:2058-63. 5. Borras E, Martin R, Judkowski V, et al. Findings on T cell specificity revealed by synthetic combinatorial libraries. Journal of Immunological Methods 2002;267:79-97. 6. Rubio-Godoy V, Ayyoub M, Dutoit V, et al. Combinatorial peptide library based identification of peptide ligands for tumor reactive cytolytic T lymphocytes of unknown specificity. European Journal of Immunology 2002;32:2292-9.

Identifying Cancer-Associated Genes Based on Family and Population Studies
These papers offer improved methods for identifying disease genes in family studies or case-control association studies. The first paper proposes an adaptive method for use when the mode of inheritance (e.g., dominant or recessive) is unknown. The second method provides a method of sample-size planning for such studies.

References 1. Zheng G, Freidlin B, Gastwirth JL. Robust TDT-type candidate gene association tests. Annals of Human Genetics 2002;66:145-55. 2. Freidlin B, Zheng G, Li Z, Gastwirth JL. Trend tests for case-control studies of genetic markers: Power, sample size and robustness. Human Heredity 2002;53:146-52.

Evaluating Diagnostic Procedures This paper develops methods for evaluating covariate effects on the accuracy of diagnostic tests. An example of an interesting covariate is lead time in cancer screening. The method is the first to permit the use of continuous covariates.

Reference 1. Dodd LE, Pepe MS. Semi-parametric regression for the area under the receiver operating characteristic curve. Journal of the American Statistical Association 2003; (in press).

New Ways to Analyze Survey and Case-Control Data
Health surveys provide an abundance of information on individuals. Because of their complex sampling designs that involve differential rates of selection of individuals, however, classical methods of estimating variance components do not work. The paper by Korn and Graubard develops new methods to estimate variance components using survey data. The paper by Graubard and Korn develops new methods for analyzing surveys when interest is in superpopulation parameters. The paper by Hunsberger et al. presents and compares two new tests for seasonal trend in monthly incidence data and applies the methods to melanoma incidence data collected by the Surveillance, Epidemiology, and End Results program. The paper by Shih and Chatterjee develops improved methods for assessing familial aggregation of a disease and the relationship between the disease and genetic or environmental risk factors based on age of disease onset in family-based case-control studies.

References 1. Korn EL, Graubard BI. Estimating variance components using survey data. Journal of the Royal Statistical Society B 2003; (in press). 2. Graubard BI, Korn EL. Inference for superpopulation parameters using sample survey. Statistical Science 2002;17:73-96. 3. Hunsberger S, Albert PS, Follmann DA, Suh E. Parametric and semiparametric approaches to testing for seasonal trend in serial count data. Biostatistics 2002;3:289-98. 4. Shih JH, Chatterjee N. Analysis of survival data from case-control family studies. Biometrics 2002;58:502-9.

New Ways to Analyze Serial Biomarker and Quality-of-Life Data
The paper by Albert, McShane, and Korn discusses how the results of a pilot study can be used to design subsequent studies of a binary biomarker when there are interlaboratory differences. Papers by Albert and Follmann develop a random-effects model for binary longitudinal studies with missing data. The paper by Borkowf et al. was motivated by collaborations with the NCI Cancer Studies Prevention Branch and provides methods for removing systematic trend in biomarker level values.

References 1. Albert PS, McShane LM, Korn EL. The design of a binary biomarker study from the results of a pilot study. Biometrics 2002;58:576-85. 2. Albert PS, Follmann DA. A random effects transition model for binary longitudinal data with informative missingness. Statistica Neerlandica 2003; (in press). 3. Albert PS, Follmann DA, Wang SA, Suh EA. A latent autoregressive model for longitudinal binary data subject to informative missingness. Biometrics 2002;58:631-42. 4. Borkowf CB, Albert PS, Abnet CC. Using Lowess to remove systematic trends in predictive variables before logistic regression with quintile-categories. Statistics in Medicine 2003; (in press).

TARGETED INITIATIVES

New Ways to Design and Analyze DNA Microarray Data
DNA microarrays are a breakthrough technology that biologists have rapidly adopted for a wide range of research objectives. The technology can provide data on gene expression levels for tens of thousands of genes in a single hybridization assay. Planning and analyzing experiments that effectively use this technology is a major challenge. Few biologists have the statistical sophistication to effectively use DNA microarrays, and few statisticians have the necessary biological knowledge. Many biologists do not recognize the nature of the challenge and believe all they need is proper software. The literature contains many examples of invalid analyses on important biomedical studies and statistical methods, and recommendations that are inappropriate for most biomedical studies. The Molecular Statistics and Bioinformatics Section has worked for five years to advance the effective use of DNA microarrays. This work includes (i) designing major studies, including several studies funded by Director's Challenge Grants; (ii) analyzing many DNA microarray studies; (iii) developing statistical methodology for the design and analysis of DNA microarray studies; (iv) developing BRB-ArrayTools software that encapsulates many best practices for analyzing DNA microarray data for use by biologists in analyzing their own data; (iv) training biologists in NIH-sponsored courses; and (v) training post-doctoral statisticians, computer scientists, and computational biologists in developing methodologies and software for DNA microarray studies.

References 1. Korn EL, McShane LM, Troendle JF, Rosenwald A, Simon R. Identifying pre-post chemotherapy differences in gene expression in breast tumors: A statistical method appropriate for this aim. British Journal of Cancer 2002;86:1093-6. 2. Radmacher MD, McShane LM, Simon R. A paradigm for class prediction using gene expression profiles. Journal of Computational Biology 2002;9:505-11. 3. McShane LM, Radmacher MD, Freidlin B, Yu R, Li M, Simon R. Methods for assessing reproducibility of clustering patterns observed in analyses of microarray data. Bioinformatics 2002;18:1462-9. 4. Simon R, Radmacher MD, Dobbin K, McShane LM. Pitfalls in the analysis of DNA microarray data for diagnostic and prognostic classification. Journal of the National Cancer Institute 2003;95:14-18. 5. Simon R, Radmacher MD, Dobbin K. Design of studies with DNA microarrays. Genetic Epidemiology 2002;23:21-36. 6. Dobbin K, Simon R. Comparison of microarray designs for class comparison and class discovery. Bioinformatics 2002;18:1462-9. 7. Dobbin K, Shih J, Simon R. Statistical design of reverse dye microarrays. Bioinformatics 2003; (in press). 8. Simon R, Dobbin K. Experimental design of DNA microarray experiments. Biotechniques 2003; (in press). 9. Simon R, Korn EL, McShane LM, Radmacher MD, Wright G, Zhao Y. Design and Analysis of DNA Microarray Investigations. Heidelberg: Springer Verlag; 2003; (in press).

BRB-ArrayTools Software for DNA Microarray Data Analysis
BRB-ArrayTools, developed by Drs. R. Simon and Amy Lam, give biologists state-of-the-art statistical methods for analyzing DNA microarray data, and interactive tutorials on the effective use of DNA microarray technology in biomedical research. This integrated package is provided without charge for non-commercial uses from the BRB-ArrayTools Web site and has been downloaded by more than 1700 investigators at more than 700 institutions worldwide.

New Methods for Designing and Analyzing Clinical Trials
Therapeutic vaccines and molecularly targeted cytostatic agents represent a large proportion of cancer therapeutics research. Although principles for the design and analysis of phase III trials developed for cytotoxics are generally applicable to vaccines and cytostatics, the standard phase I and phase II paradigm is often inappropriate for these newer agents. The BRB is developing more appropriate study designs and development strategies for therapeutic vaccines and cytostatic agents based on experience with intramural and extramural investigators. The paper by Korn et al. describes designs for cytostatic agents and the papers by Simon address efficient approaches for early clinical development of therapeutic vaccines.

The paper by Freidlin and Korn demonstrates serious problems with using statistical methods that are overly aggressive in early termination of phase III clinical trials. The second paper by Freidlin and Korn proposes a method for improving the power to detect survival differences when less than 50% of patients are expected to respond to the experimental treatment.

References 1. Korn EL, Rubinstein LV, Hunsberger SA, Pluda JM, Eisenhauer E and Arbuck SG. Clinical trial design for cytostatic agents and agents directed at novel molecular targets. In: Adjei AA, Buolamwini J, eds. Strategies for Discovery and Clinical Testing of Novel Anticancer Agents. New York: Elsevier (in press). 2. Simon R. Clinical trial designs for therapeutic vaccine studies. In: Morse MA, Lyerly HK, Clay TM, eds. Handbook of Cancer Vaccines. Totawa NJ: Humana Press; 2003; (in press). 3. Simon R. Clinical trial designs for therapeutic cancer vaccines. In: Khleif S, ed. Tumor Immunology and Cancer Vaccines. Norwell MA: Kluwer Academic Publishers; 2003; (in press). 4. Freidlin B, Korn EL. A note on futility monitoring. Controlled Clinical Trials 2002;23:355-356. 5. Freidlin B, Korn EL. A testing procedure for survival data with few responders. Statistics in Medicine 2002;21:65-78.

Collaborative Research
The papers by Rosenwald et al. reflect collaboration with Dr. Lou Staudt's laboratory on an international study of the molecular signatures of lymphomas using DNA microarrays. The first paper reports on a new molecular classification system for large-cell lymphoma that is more predictive of survival than the international prognostic index for patients receiving CHOP chemotherapy. The second paper reports a similar study for mantel cell lymphoma. Both studies represent the largest studies of their kind performed for lymphomas.

The paper by Vasselli et al. reflects collaboration with Dr. Richard Klausner's laboratory and Dr. Marston Linehan on using DNA microarrays to molecularly characterize metastatic kidney cancer.

The paper by Desai et al. results from collaboration with the Mouse Model Consortium and Dr. Jeffrey Green's NCI laboratory to characterize several transgenic mouse models of human breast cancer in terms of DNA microarray-based gene expression profiles. The paper by Muira et al. is based on collaboration with the NCI Laboratory of Clinical Carcinogenesis to characterize expression profiles of human lung cancers based on DNA microarray expression profiles of microdissected specimens. The paper by Assersohn et al. results from BRB collaboration with Dr. Ed Liu and Dr. Trevor Powles of the Royal Marsden Hospital in comparing expression profiles of small breast cancer specimens based on aspiration and biopsy. The paper by Wang et al. results from BRB collaboration with Dr. Franco Marincola's laboratory in the NCI Surgery Branch on expression profiles of melanoma lesions. In this work it was possible to predict which patients would obtain complete remission from IL2-based vaccine therapy based on their DNA microarray expression profiles. The paper by Best et al. results from BRB collaboration with Dr. Emmert-Buck's laboratory in the NCI Laboratory of Pathology to develop DNA microarray-based expression profiles of human prostate cancers. The papers by Tangrea et al., Lee et al., and Velasco et al. result from the collaborations of Dr. Paul Albert of the Biometric Research Branch with the NCI Center for Cancer Research. The paper by Janne et al. results from BRB collaboration with Dr. Bruce Johnson of the Dana Farber Cancer Center on a quantitative review of 25 years of therapeutic research for patients with limited-stage small-cell lung carcinoma, finding meaningful survival improvements.

References 1. Rosenwald A, Wright G, Chan WC, et al. The use of molecular profiling to predict survival after chemotherapy for diffuse large B-cell lymphoma. New England Journal of Medicine 2002;346:1937-47. 2. Rosenwald A, Wright G, Wiestner A, et al. The proliferation gene expression signature is a quantitative integrator of oncogenic events that predicts survival in mantle cell lymphoma. Cancer Cell 2003; (in press). 3. Vasselli J, Shih JH, Iyengar SR, et al. Predicting survival in patients with metastatic kidney cancer by gene expression profiling in the primary tumor. Proceedings of the National Academy of Sciences USA 2003; (in press). 4. Tangrea JA, Albert PS, Lanza E, et al. Non-steroidal anti-inflammatory drug use is associated with reduced risk of colorectal adenomas: a prospective study among participants of the polyp prevention trial. Cancer Causes and Control 2003; (in press). 5. Janne PA, Freidlin B, Saxman S, et al. Twenty-five years of clinical research for patients with limited-stage small cell lung carcinoma in North America-Meaningful improvements in survival. Cancer 2002;7:1528-38. 6. Lee J, Hampl M, Albert PS, Fine HA. Antitumor activity and prolonged expression from a TRAIL-expressing adenoviral vector. Neoplasia 2002;4:3123-23. 7. Velasco A, Hewitt SM, Albert PS, et al. Differential expression of the mismatch repair gene hMSH2 in malignant prostate tissue is associated with cancer recurrence. Cancer 2002;94:690-9. 8. Velasco A, Albert PS, Rosenberg H, et al. Clinicopathologic implications of hMSH2 gene expression and microsatellite instability in prostate cancer. Cancer Biology and Therapy 2002;1:362-7. 9. Desai KV, Xiao N, Wang W, et al. Initiating oncogenic event determines gene-expression patterns of human breast cancer models. Proceedings of the National Academy of Sciences USA 2002;99:6967-72. 10. Miura K, Bowman ED, Simon R, et al. Laser capture microdissection and microarray expression of lung adenocarcinoma reveals tobacco smoking and prognosis-related molecular profiles. Cancer Research 2002;62:3244-50. 11. Assersohn L, Gangi L, Zhao Y, et al. The feasibility of using fine needle aspiration from primary breast cancers for cDNA microarray analyses. Clinical Cancer Research 2002;8:794-801. 12. Wang E, Miller LD, Ohnmacht GA, et al. Evolving molecular portraits of metastatic melanoma. Cancer Research 2002;62:3581-6. 13. Best C, Leiva IM, Chuaqui RF, et al. Comparative gene expression profiling of different grades of human prostate cancer and prostate cell lines. Diagnostic Molecular Pathology 2003; (in press).