Research Interests
Dr. Ibrahim's general areas of research are Bayesian inference, missing data problems, and genomics. He has an established track record in methodological and collaborative research, and I have four NIH supported R0-1 grants for developing statistical methodology related to cancer and genomics research. His work in Bayesian statistics includes prior elicitation, model selection, survival analysis, generalized linear models, and computational methods. His missing data work has focused on maximum likelihood methods for parameter estimation in the presence of missing response and/or covariate data in generalized linear models, models for longitudinal data, and survival models. His genomics work has focused on Bayesian model-based methods for the analysis of DNA microarray data. He has also directed 7 doctoral dissertations to date, one of which was titled Bayesian Models for Gene Expression Analysis". He is also the Principal Investigator (PI) of a recently funded Training Grant titled ``Biostatistics for Genomics and Cancer." This prestigious grant was awarded to the Department of Biostatistics at UNC in May, 2004, and provides funding for 5 pre-doctoral and 2 post-doctoral students in Biostatistics. He has published over 120 research papers, most of which are in the top statistical journals. He has written two books at the advanced graduate level on Bayesian survival analysis and Monte Carlo methods in Bayesian computation.He is a fellow of the American Statistical Association (ASA) and the Institute of Mathematical Statistics (IMS), and an elected member of the International Statistical Institute (ISI). On the collaborative front, Dr. Ibrahim has over ten years of experience working in cancer clinical trials at the Dana-Farber Cancer Institute at Harvard and the Lineberger Cancer Center at UNC. He is currently the biostatistical core leader of a recently funded GI SPORE grant at UNC, and is currently a senior faculty advisor on the Lineberger Cancer Center Core Grant. Most of Dr. Ibrahim's statistical papers have been published in leading statistical journals, including Biometrika, the Journal of the American Statistical Association, the Journal of the Royal Statistical Society - Series B, and Biometrics. In addition, he is the author/co-author of thirteen book chapters.

In his cancer genomics research, Dr. Ibrahim has examined Bayesian parametric models for examining differential gene expression between two or more groups of subjects. He has studied three classes of models. These are log-normal models allowing for truncation in the gene expression levels; hierarchical analysis of variance (ANOVA) models allowing for left censored data; and a new class of continuous mixture models that separately model the expressed and unexpressed genes by continuous distributions, and therefore, do not censor or truncate the data in any way. For each class of models, Dr. Ibrahim and his colleagues have proposed methods for characterizing the relationships between tissue types (groups) and genes, investigated Bayesian gene selection
Algorithms for identifying genes that are differentially expressed between the tissue types, examined novel prior distributions that allow the mean gene expression levels to be correlated a priori, and studied the inclusion of subject specific covariates. Theoretical and computational properties of the proposed models have been examined in detail, and their robustness properties have been investigated and their performance evaluated. Dr. Ibrahim has examined this methodology using datasets from endometrial cancer in collaboration with Dr. George Mutter, a pathologist at Brigham and Women's Hospital in the Harvard Medical School. Some of this work has been published in Ibrahim, Chen, and Gray (2002), and some of this is ongoing research work. Dr. Ibrahim has also studied formal Bayesian methodologies for controlling the False Discovery Rate (FDR) in the analysis of DNA microarray data. Dr. Ibrahim and colleagues have proposed a Bayesian criterion to controlling the FDR, investigated its theoretical properties, and compared its performance to non-Bayesian approaches for controlling the FDR. He has also investigated the Bayesian FDR procedure for non-normal models as well as develop and examine this procedure with correlated response data. The Bayesian FDR procedure is currently being explored for the three classes of models mentioned above. Some of this work appears in Tadesse, Ibrahim, et al. (2005). Dr. Ibrahim has also developed Bayesian methods for joint models of outcome and survival data. Specifically, he has developed Bayesian semiparametric measurement error survival models for examining relationships between a time-to-event, such as survival time or time to remission, and gene expression level. He has studied a classical measurement error model for the gene expression level, and links this with a piecewise constant hazard model for the survival time. The resulting model provides a new and novel class of models for jointly modeling gene expression and time-to-event data. Prior distributions and gene selection algorithms for determining which genes are most highly associated with survival time have been be studied, and computational algorithms are being investigated and implemented. Theoretical properties of the proposed joint models are currently being examined, their robustness properties are being investigated, and their performance is being evaluated. Dr. Ibrahim and his colleagues have applied this methodology to several cancer datasets including a recent study in Adult Lymphoblastic Leukemia (ALL). Some of this work has appeared in his paper with Tadesse, Ibrahim, et al. (2005).

In addition to his work methodological work on genomics, he has made significant contributions to prior elicitation, Bayesian model selection, and Bayesian computation. In his work on missing data, he has made significant contributions in the development of maximum likelihood methods for estimating parameters with ignorable or nonignorable missing data for a wide variety of models. Dr. Ibrahim has developed a novel class of informative prior distributions constructed from historical data called the power priors (Ibrahim and Chen, 2000) which are quite useful for prior elicitation. Dr. Ibrahim has published an extensive number of papers on the theoretical development, properties, and implementation of the power prior for proportional hazards models, generalized linear models, random effects models, generalized linear mixed models for longitudinal data, time series models, and cure rate models. Some select papers on the power prior include Ibrahim and Laud (1994), Laud and Ibrahim (1995), Ibrahim, Ryan, and Chen (1998), Ibrahim and Chen (1998), Ibrahim, Chen, and MacEachern (1999), Chen, Ibrahim, and Sinha (1999), Chen, Ibrahim, and Yiannoutsos (1999), Chen, Ibrahim, and Shao (2000), Ibrahim, Chen, and Ryan (2000), Chen and Ibrahim (2000), Ibrahim, Chen, and Sinha (2001), Chen, Harrington, and Ibrahim (2002), Chen, Ibrahim, and Lipsitz (2002), and Ibrahim, Chen, and Sinha (2003b). A recent review paper on the power prior is given by Ibrahim and Chen (2000) in Statistical Science. The power prior has received international recognition and is currently widely used in the Bayesian community. The software package BUGS, for example, has incorporated the power prior into its suite of priors for its next release. Dr. Ibrahim has also done significant work in Bayesian model selection. He has proposed criterion based methods as well as fully Bayesian procedures for variable selection and model selection involving non-nested models. Some select papers include Ibrahim and Laud (1991, 1994), Laud and Ibrahim (1995), Laud and Ibrahim (1996), Ibrahim, Chen, Ryan (2000), and Ibrahim, Chen, and Sinha (2001). Also, Dr. Ibrahim also has made significant contributions in Markov chain Monte Carlo methods for Bayesian model selection and computation. Dr. Ibrahim and his colleagues derived several novel methods for posterior sampling and computation of posterior model probabilities (Ibrahim, Chen, and MacEachern, 1999, Chen, Ibrahim, and Yiannoutsos, 1999, Chen, Ibrahim, and Sinha, 1999). One of Dr. Ibrahim's most significant contributions in model selection is the development of Bayesian criterion-based methods for model assessment. Dr. Ibrahim and his colleagues have developed the now well-known measure as a general Bayesian model assessment statistic. This model assessment statistic is quite general, and can be used for univariate or multivariate data, continuous or discrete data, univariate or multivariate censored survival data, and longitudinal data.

In his work on missing data, Dr. Ibrahim (1990) has developed a very novel and general implementation of the EM algorithm for missing data problems, called the EM by the Method of Weights. This very general result shows that the E-step of the EM algorithm for any regression problem can be expressed as a weighted complete data likelihood. The EM by the method of weights can be used for missing nonignorable (or ignorable) covariate and/or missing response data. In addition, the response data can be correlated and/or right censored, as in longitudinal and survival outcomes. Dr. Ibrahim has demonstrated the EM by the method of weights in a variety of regression models with missing covariate and/or missing response data. These include generalized linear models (Ibrahim, Chen, and Lipsitz, 1999, Ibrahim, Lipsitz, and Chen, 1999, Lipsitz and Ibrahim, 1996), proportional hazards models (Lipsitz and Ibrahim 1998, Herring and Ibrahim, 2001), cure rate models (Chen and Ibrahim, 2000), random effects models (Herring and Ibrahim, 2001, Herring and Ibrahim, 2001), generalized linear mixed models (Ibrahim, Chen, and Lipsitz, 2001), and models for clustered survival data (Lipsitz and Ibrahim, 2000). The EM by the method of weights has become one of the leading methodologies for dealing with missing data, and has been used by a number of other researchers. It is becoming one of the most highly referenced methods for dealing with missing data. The novelty and generality of the EM by the method of weights has led Dr. Ibrahim to consult with Dr. Cyrus Mehta of Cytel Software corporation to incorporate this methodology into several commercial statistical software packages, including SAS, S-PLUS, and EGRET. Dr.'s Ibrahim and Mehta have recently been awarded a Phase II SBIR grant to develop commercial software for implementing the EM by the method of weights for various types of regression models. The software development will be completed within the next two years. Dr. Ibrahim has two published two books at the advanced graduate level. The first book, titled Monte Carlo Methods in Bayesian Computation, Springer-Verlag, was published in 2000, and the second book, titled Bayesian Survival Analysis, Springer-Verlag, was published in 2001. These two areas are at the cutting edge of current research in Bayesian statistics and are motivated from real data problems. Dr. Ibrahim has presented day long short courses based on his two books at the JSM Annual meeting, the ENAR meeting, and the WNAR meeting.

Dr. Ibrahim's collaborative statistical activities are mainly in cancer research. Before coming to UNC, he collaborated are with oncologists at the Dana-Farber Cancer Institute (DFCI) and with the Melanoma and Gastrointestinal Committees of the Eastern Cooperative Oncology Group (ECOG) for 8 years. Currently, he is in close collaboration with members of the Lineberger Cancer Center, including Dr. Robert Sandler, Dr. Charles Perou, and Dr. David Threadgill on cancer genomics projects. The Melanoma Committee of ECOG has established itself as one of the world's leading groups for initiating and carrying out therapeutic trials for the treatment of melanoma. In addition to the planning and analysis of many melanoma clinical trials, Dr. Ibrahim has developed a cure rate model which attempts to estimate the proportion of patients cured by therapy. The application of the cure rate model to melanoma clinical trials has provided some new insights into the potential for cure. With his introduction of the cure rate model for melanoma, Dr. Ibrahim has changed the direction of the statistical activities of the melanoma committee in ECOG. The cure rate model is now becoming the standard in the design and analysis of melanoma clinical trials in ECOG. Dr. Ibrahim has followed up on this melanoma collaboration with methodological papers on his approach to cure rate modeling. Some papers include Chen, Ibrahim, and Sinha (1999), Chen and Ibrahim (2000), Ibrahim, Chen, and Sinha (2001) Chen, Harrington, and Ibrahim (2001), and Herring and Ibrahim (2002).

Dr. Ibrahim is also currently a PI on two R0-1 statistical grants from the National Cancer Institute. These grants provide support for conducting statistical methodological research for Bayesian methods for the analysis of DNA microarray data (CA 70101), and for research in missing data problems (CA 74015). He is also the Co-PI for two other RO-1 grants studying methodology for semiparametric survival analysis and methods for analyzing repeated categorical data.

Recent Accomplishments and Honors
1999- Fellow, American Statistical Association
2000- Fellow, Institute of Mathematical Statistics
2000- Elected member, International Statistical Institute

Training
University of Minnesota , Ph.D., 1988, Statistics

Publications
Collaborative Publications

1. Ayash, LJ, Elias, AE, Schwartz, G, Wheeler, CW, Ibrahim, JG, Teicher, B, Warren, D, Lynch, C, Richardson, P, Schnipper, L, Frei, E, III., and Antman, K, "Double Dose-Intensive Chemotherapy with Autologous Stem Cell Support for Metastatic Breast Cancer: No Improvement in PFS by the Sequence of High-Dose Melphalan Followed by CTCb", Journal of Clinical Oncology, 1996; 14:2984-2992.
2. Wheeler C, Eickhoff C, Elias AE, Ibrahim JG, Ayash L, McCauley M, Mauch P, Schwartz G, Eder JP, Mazanet R, Ferrara J, Rimm IJ, Bierer B, Gilliland G, Churchill HW, Ault K, Parsons S, Antman K, Schnipper L, Tepler I, Gaynes L, Frei E III, Kadin M, Antin JH, "High Dose Cyclophosphamide, Carmustine and Etoposide with Autologous Transplantation in Hodgkin's Disease: A Prognostic Model for Treatment Outcomes," Biology of Blood and Bone Marrow Transplantation, 1997; 3:98-106.

3. Ayash, LJ, Elias AE, Ibrahim JG, Schwartz G, Wheeler C, Reich E, Lynch C, Warren D, Shapiro C, Richardson P, Hurd D, Schnipper L, Frei E, Antman K, "High-Dose Multimodality Therapy with Autologous Stem Cell Support for Stage IIIB Breast Carcinoma," Journal of Clinical Oncology, 1998; 16:1000-1007.
4. Falkson, CI, Ibrahim, JG, Kirkwood, JM, Coates, AS, Atkins, MB, Blum, RH, "A Phase III Trial of Dacarbazine versus Dacarbazine with Interferon o:2b versus Dacarbazine with Tamoxifen (TMX) versus Dacarbazine with Interferon o:2b and Tamoxifen in Patients with Metastatic Malignant Melanoma: an Eastern Cooperative Oncology Group Study (E3690)," Journal of Clinical Oncology, 1998; 16:1743-1751.

5. Elias AE, Wheeler C, Ayash LJ, Schwartz G, Ibrahim JG, ills L, McCauley M, Coleman N, Warren D, Schnipper L, Antman KH, Teicher BA, Frei, E, "Dose Escalation of the Hypoxic Cell Sensitizer Etanidazole Combined With Ifosfamide, Carboplatin, Etoposide, and Autologous Hematopoietic Stem Cell Support," Journal of Clinical Cancer Research, 1998; 4:1443-1449.

6. Elias, E, Ibrahim, JG, Skarkin, AT, Wheeler, C, McCauley, M, Ayash, L, Richardson, P, Schnipper, L, Antman, K and Frei, E, "Dose-Intensive Therapy for Limited-Stage Small-Cell Lung Cancer: Long-Term Outcome," Journal of Clinical Oncology, 1999; 17:1175-1184.

7. Hochster, H, Ibrahim, JG, O'Dwyer PJ, Liebes, L, Benson, AB, "A Phase II Study of Topotecan 21-day Infusion in Advanced Colorectal Cancer: An Eastern Cooperative Oncology Group Study (E4293)," Cancer Therapeutics, 1999; 2:37-43.
8. Kirkwood, JM, Ibrahim, JG, Sondak, VK, Richards, J, Flaherty, LE, Ernstoff, MS, Smith, TJ, Rao, U, Steele, M, and Blum, RH, "The Role of High- and Low-Dose Interferon Alfa-2b in High-Risk Melanoma: First Analysis of Intergroup Trial E1690jS9111jC9190," Journal of Clinical Oncology, 2000; 18:2444-2458.

9. Manola, J, Atkins, M, Ibrahim, JG, Borden, E, Blum, R, Cunningham, T, Golumb, F, Kirkwood, JM , "Prognostic Factors in Metastatic Melanoma: A Pooled Analysis of Eastern Cooperative Oncology Group Trials," Journal of Clinical Oncology, 2000; 18:3782-3793.

10. Frank, DA, Meuse, J, Hirsch, D, Ibrahim, JG, and Abbeele, A, "The Treatment and
Outcome of Cancer Patients with Thromboses on Central Venous Catheters," Journal of
Thrombosis and Thrombolysis, 2000; 10:271-275.
11. Legedza, ATR, and Ibrahim, JG "Longitudinal Design for Phase I Clinical Trials Using
the Continual Reassessment Method," Controlled Clinical Trials, 2000; 21:574-588.

12. Legedza, ATR, Ibrahim, JG, "Heterogeneity in Phase I Clinical Trials: Prior Elicitation and Computation Using the Continual Reassessment Method," Statistics in Medicine, 2001; 20:867-882.

13. Elias AD, Richardson P, Avigan D, Ibrahim J, Joyce R, Demetri G, Levine J, Warren D, Arthur T, Hieng S, Reich E, Frei E III, Ayash LA., "Short Course ofInduction Chemotherapy Followed by Two Cycles of High-dose Chemotherapy With Stem Cell Rescue for Chemotherapy-naive Metastatic Breast Cancer," Bone Marrow Transplantation, 2001; 27:269-278.
14. Wheeler C, Khurshid A, Ibrahim J, Elias A, Mauch P, Ault K, Antin J, "Incidence of Post Transplant Myelodysplasiajacute Leukemia in Non-Hodgkin's Lymphoma Patients Compared with Hodgkin's Disease Patients Undergoing Autologous Transplantation Following Cyclophosphamide, Carmustine and Etoposide (CBV)," Leukemia and Lymphoma, 2001; 40:499-509.

15. Elias AD, Richardson P, Avigan D, Ibrahim JG, Joyce R, McDermott D, Levine J, Warren D, McCauley M, Wheeler C, Frei E III, "A short course of induction Chemotherapy Followed by Two Cycles of High-dose Chemotherapy with Stem Cell Rescue for Chemotherapy Naive Metastatic Breast Cancer: Sequential Phase Ijll Studies," Bone Marrow Transplantation, 2001; 28:447-454.
16. Kirkwood, JM, Ibrahim, JG, Lawson, DH, Atkins, MB, Agarwala, SS, Collins, K, Mascari, R, Morrissey, DM, Chapman, PB, "High-dose Interferon alfa-2b Does not Diminish Antibody Response to GM2 Vaccination, in Patients with Resected Melanoma: Results of the Multicenter ECOG Phase II Trial E2696," Journal of Clinical Oncology, 2001; 19:1430-1436.

17. Kirkwood, JM, Ibrahim, JG, Sosman, JA, Sondak, VK, Agarwala, SS, Ernstoff, MS, Rao, U, "High-Dose Interferon Alfa-2b Significantly Prolonged Relapse-Free and Overall Survival Compared with the GM2-KLH/QS-21 Vaccine in Patients with Resected Stage lIB-III Melanoma: Results of Intergroup Trial E1694/S9512/C509801," Journal of Clinical Oncology, 2001; 19:2370_2380.

18. Glover, DG, Ibrahim, JG, Kirkwood, JM, Glick, J, Karp, D, Stewart, J, Ewell, M, and Borden, E, "Phase III Randomized Trial of Cis-Platinum and WR-2721 Versus Cis-Platinum Alone for Metastatic Melanoma: An Eastern Cooperative Oncology Group Study (E1686)," Melanoma Research, 2002; in press.

19. Elias, AD, Ibrahim, JG, Richardson, P, Avigan, D, Joyce, R, Reich, E, McCauley, M, Wheller, C, Frei E III, "The Impact of Induction Duration and the Number of High Dose Cycles on the Long-Term Survival of Women With Metastatic Breast Cancer Treated with High Dose Chemotherapy with Stem Cell Rescue: An Analysis of Sequential Phase I/II Trials From the Dana-Farber/Beth Israel STAMP Program," Biology of Bone Marrow Transplantation, 2002; 28:198-205.

20. Rao, UNM, Ibrahim, JG, Flaherty, LE, Richards, J, Kirkwood, JM, "Implications of Microscopic Satellites of the Primary and Extracapsular Lymph Node Spread in Patients with High Risk Melanoma: Pathologic Corollary of Intergroup Trial E1690," Journal of Clinical Oncology, 2002; 20:2053-2057.

21. Kirkwood, JM, Richards, T, Zarour, HM, Sosman, J, Ernstoff, M, Whiteside, TL, Ibrahim, JG, Blum, R, Wiend, S, Mascari, R, "Immunomodulatory Effects of High- and Low-dose IFN a2b in Patients with High-risk Resected Melanoma: The E2690 Laboratory Corollary ofIntergroup Adjuvant Trial E1690," Cancer, 2002; 95:1101-1112.

E-mail: ibrahim@bios.unc.edu
Telephone: 919-843-2715
FAX: 919-966-3804
Address: McGavran-Greenberg Hall Chapel Hill, NC

Click here to update this profile