Conference Schedule

8:15 – 9:00
Registration and Continental Breakfast

9:00 – 9:10
Introductory Remarks
Boris Iglewicz, Temple University

9:10 – 9:55
The Analysis of Proteomics Spectra from Serum Samples
Keith Baggerly, M.D. Anderson Cancer Center

9:55 – 10:40
A New Approach for Evaluating IVIVC That Does Not Require Common Measurement Times for Dissolution and Absorption
A. Lawrence Gould, Merck Research Laboratories

10:40 – 11:05

11:05 – 11:50
Hierarchical Functional Data in Colon Carcinogenesis
Raymond J. Carroll, Texas A&M University

11:50 – 12:35
Making the Leap from One Population to Another in Evaluating Effects of Interventions: Study Design and Analysis for Bridging
Michael D. Hughes, Harvard School of Public Health

12:35 – 1:40

1:40 – 2:25
Bayesian Analysis of Cost-Effectiveness in Clinical Trials
Daniel F. Heitjan, University of Pennsylvania

2:25 – 3:10
Informative Noncompliance in Endpoint Trials
Steven Snapinn and Qi Jiang, Merck Research Labs

3:10 – 3:25
General Discussion


The Analysis of Proteomics Spectra from Serum Samples

Keith Baggerly, M.D. Anderson Cancer Center

Mass spectrometry profiles can provide quick summaries of the relative levels of hundreds of proteins. By surveying profiles from a large number of samples, we can hopefully zoom in on proteins that are linked with a difference of interest such as the presence or absence of cancer. Using examples from two case studies, we will address issues of experimental design, data cleaning and processing, discriminating subsets, and protecting against spurious structure.

A New Approach for Evaluating IVIVC That Does Not Require Common
Measurement Times for Dissolution and Absorption

A. Lawrence Gould, Merck Research Laboratories

IVIVC (In Vitro – In Vivo Correlation) is a body of procedures whose purpose is to provide a basis for approving a change in formulation of a drug using only dissolution information without requiring conventional bioequivalence trials in human subjects. IVIVC is justified by the assumption that the absorption of a drug depends only on the time course of its presentation, e.g., in the gastric lumen, and not on how it actually is delivered. Current IVIVC methods relate the (cumulative) percentage of administered dose absorbed by the body to the cumulative dissolution of a dosage unit under controlled conditions, with both percentages measured at common time points. We describe, and illustrate with data concerning a delayed-release formulation, a different approach to evaluating IVIVC that does not require measurements of absorption and dissolution at a common set of time points, and that explicitly accounts for the variation in the dissolution and absorption processes. We express the plasma concentration curve in terms of concentration instead of time, using the dissolution curve to provide a noisy correspondence between time and concentration. All conventional functionals of the concentration curve such as AUC, Cmax and Tmax can be expressed in terms of concentrations, with uncertainties arising from variability in measuring absorption and concentration accounted for explicitly.

Hierarchical Functional Data in Colon Carcinogenesis

Raymond J. Carroll, Texas A&M University

In colon carcinogenesis experiments, animals are given different treatments, e.g. diets, cancinogenic exposures, etc. After this, extremely detailed measurements are obtained from different regions of their colons. In the typical experiment, numerous colonic crypts are isolated, and within each of these crypts either every cell is measured for DNA damage and repair, or the damage is obtained from an image on a pixel-by-pixel basis. The data are hierarchical (random rats, random crypts within rats). At the lowest level of the hierarchy, the data are functional (location of a measurement within the crypt). We will review two functional data problems arising from these experiments that lead to novel estimation problems. In one case, interest focuses on the interrelationship of DNA damage in different regions of the colon as a function of cell depth in the colon. Here we show that fitting undersmoothed kernel methods at the crypt level results in pleasing asymptotic properties. In the second case, the functional data are highly spiky at the crypt level, but interest focuses on functions of the overall rat. Here we use wavelet methods via Bayesian machinery to show the evolution of DNA repair over time.

Note: This work was done jointly with Naisyin Wang (Texas A&M University) and Jeffrey Morris (M. D. Anderson Cancer Center).

Making the Leap from One Population to Another in Evaluating Effects of Interventions: Study Design and Analysis for Bridging

Michael D.Hughes, Harvard School of Public Health

For many reasons, including practical, ethical and economic ones, interventions are often evaluated in highly selected populations. For example, study populations enrolled in clinical trials of a new drug often involve subjects from selected countries with a restricted ethnic/racial composition and commonly will include more men than women, and a restricted age range. Having established that an intervention is effective in such populations, there is then interest in establishing that it is effective in other “new” populations. The ideal is to do this without repeating a whole program of trials in the new population. Indeed, for some new populations (e.g. newborn infants), even a small trial may be difficult to undertake. In drug development and approval by regulatory authorities, Guidelines of the International Conference on Harmonization (ICH) broach the issue of undertaking studies to bridge from one population to another. These Guidelines discuss circumstances in which different types of study may be required, e.g. just pharmacokinetic studies versus smaller short-term studies using surrogate endpoints versus larger standalone confirmatory trials. However, little detail is given concerning how to design these studies. The talk will describe some approaches to study design and analysis for bridging motivated by issues arising in HIV research such as bridging from adults to children or pregnant women, and from one group of countries to another group.

Bayesian Analysis of Cost-Effectiveness in Clinical Trials

Daniel F. Heitjan, University of Pennsylvania

In recent years it has become common in clinical trials to collect data on health care costs in addition to more traditional measures of clinical effectiveness such as treatment response and survival. A “cost-effectiveness analysis” seeks to evaluate the cost per unit of health that one gains by adopting the more expensive treatment, with a view toward determining whether the more expensive treatment is worth its additional cost and so should be included in health insurance coverage. The analysis of such data raises a number of interesting statistical issues, including the choice of estimands and mode of inference, and methods for dealing with potentially nonignorably censored cost measurements. I will review these topics and present a straightforward Bayesian analysis of cost and survival data from a clinical trial in cardiovascular disease.

Informative Noncompliance in Endpoint Trials

Steven Snapinn, Qi Jiang, Merck Research Labs

Noncompliance with study medications as manifested by permanent discontinuation is an important issue in the design of endpoint clinical trials. When noncompliant patients are included in an intention-to-treat analysis they have the potential to seriously decrease power. While various methods exist to account for noncompliance in the calculation of sample size, standard methods all assume that noncompliance is noninformative; that is, that the risk of discontinuation is independent of the risk of experiencing a study endpoint.

In the first part of this presentation we will describe the concept of informative noncompliance, and give examples from several published clinical trials. In addition, in order to illustrate the impact of noncompliance on the risk of endpoint, we propose an extended Kaplan-Meier estimator for use with time-varying covariates. (Steve Snapinn)

In the second part of the presentation we will describe a modified version of the method proposed by Lakatos (1988, Biometrics 44, 229-241) that can be used to calculate sample size under informative noncompliance. This method is based on the concept of two subpopulations, one with high rates of endpoint and discontinuation, and another with low rates. Using this new method, we will show that failure to consider the impact of informative noncompliance can lead to a seriously underpowered study. (Qi Jiang)


Dr. Keith Baggerly is an Assistant Professor in the Bioinformatics Section of the Department of Biostatistics at the MD Anderson Cancer Center, where he has worked since 2000. Prior to joining the group, he served as a faculty member in the Rice University Statistics Department and as a member of the Statistics Group at the Los Alamos National Laboratory. His current research involves modeling of structure in high-throughput biological data, including the Serial Analysis of Gene Exression (SAGE), cDNA and oligonucleotide microarrays, and mass spectrmetric proteomic spectra.

Dr. Baggerly was a member of the winning team for the 2001 Critical Analysis of Microarray Data (CAMDA) competition (the section also won in 2002), and the leader of the winning team for the first Proteomics Data Mining Conference (2002). He won the 2002 First Place contributed paper award of the Biopharmaceutical Section of the ASA for work on modeling microarray.

Dr. A. Lawrence Gould is Senior Director, Scientific Staff, Biostatistics and Research Data Systems, Merck Research Laboratories. He received his Ph.D. in Biometry from Case Western Reserve University. Dr. Gould is a Fellow of the American Statistical Association, has served on a number of grant review panels, and has served in a variety of positions with the Biopharmaceutical Section of the ASA and the Biometric Society, ENAR. He served as Editor of the Journal of Biopharmaceutical Statistics, received the 1991 Donald Francke Award for most outstanding article published in DIA Journal that year, and the 1994 best presentation award from the Biopharmaceutical Section of ASA. His research interests tend to be driven by problems arising in drug development, and include blinded sample size re-estimation, Bayesian methods, meta-analysis, bioequivalence, analysis of safety data, data mining, clinical trial simulation, and management science.

Dr. Raymond Carroll is Distinguished Professor of Statistics, Nutrition and Toxicology at Texas A&M University. He heads the Bioinformatics Training Program at Texas A&M, as well as the Biostatistics research Core of the NIEHS Center for Environmental and Rural Health. He received his Ph.D. in 1974 from Purdue University, was on the faculty at the University of North Carolina from 1974-87, and joined Texas A&M in 1987. He has spent sabbaticals at the National Cancer Institute, the Australian National University, The University of Heidelberg and the Humboldt University in Berlin.

Carroll has been editor of two of the major Statistics research journals: the Journal of the American Statistical Association (Theory and Methods) and Biometrics. Professor Carroll has been awarded most of the major honors in Statistics, including the Committee of Presidents of Statistical Societies (COPSS) Presidents’ Award (awarded annually to a major research statistician under the age of 40), the COPSS Fisher Lecture (awarded annually for seminal work that has influenced the theory and practice of Statistics), the Snedecor Award (for an outstanding publication in Biometry and Biostatistics) and the Wilcoxon Prize (for an outstanding publication in the practice of Statistics). He is a distinguished alumnus of Purdue University, and has been a Humboldt Fellow. He is an elected Fellow of the American Statistical Association and the Institute of Mathematical Statistics, and an elected member of the International Statistical Institute. Professor Carroll is the chair of the NIH Study Section on Biostatistics and Methods (SNEM-5). He is generally considered to have developed the field of variance function estimation in regression, wrote an initial and highly cited paper on data that are informatively missing, and developed the field of measurement error modeling in nonlinear and generalized linear models. Professor Carroll’s current research interests focus on developing statistical methods in general, with special emphasis on nutritional epidemiology, basic biological processes in nutrition and colon carcinogenesis and in Bioinformatics.

Dr. Michael Hughes is Professor of Biostatistics at Harvard School of Public Health, and is an elected Fellow of the American Statistical Association. He is also Director of the Statistical and Data Management Center (SDMC) for the Adult AIDS Clinical Trials Group (AACTG) and was formerly Director of the SDMC for the Pediatric ACTG. He received his Bachelor’s degree from the University of Cambridge in England and completed his PhD in Medical Statistics at the University of London. His research focuses on issues related to the design and analysis of clinical trials, including sequential methods, the validation of surrogate endpoints, Phase I/II trial design and the analysis of longitudinal data including repeated censored biomarker measurements. Dr. Hughes has over a hundred publications in the statistical and clinical literature. He is a member of the Editorial Board for Statistics in Medicine and is the Statistical Editor for the Journal of Infectious Disease

Dr. Daniel Heitjan earned his PhD in Statistics in 1985 from the University of Chicago. He has served on the faculties of UCLA (1985-1988), Penn State University (1988-1995), Columbia University (1995-2002), and is now at the the University of Pennsylvania, where he is Professor of Biostatistics and Statistics, and Director of Biostatistics in the Abramson Cancer Center. Dr. Heitjan is an associate editor of Statistics in Medicine, a statistical editor of Journal of the National Cancer Institute, and a regular statistical reviewer for the Annals of Internal Medicine. His research interests include clinical trial design, the analysis of longitudinal studies, Bayesian methods, statistical methods in health economics, and the theory of statistical inference with incomplete data. Among the prominent medical studies he has worked on is the REMATCH trial, a landmark randomized study of a mechanical support device in the treatment of advanced heart failure. He was Program Chair of the Biometrics Section of the American Statistical Association (ASA) for the 2002 Joint Statistical Meetings (JSM) in New York, and was recently named overall Program Chair for the 2005 JSM in Minneapolis. He was elected a Fellow of the ASA in 1997.

Dr. Steve Snapinn is Senior Director, Scientific Staff, in the biostatistics department of Merck Research Labs, where he’s been for the past 20 years. Dr. Snapinn has a Ph.D. in biostatistics from the University of North Carolina at Chapel Hill, is an adjunct associate professor in the biostatistics department at UNC, and is a Fellow of the American Statistical Association. He has over 60 publications in the statistical and medical literature; his research interests include statistical methods for the analysis of combination drug trials, the analysis of sequential clinical trials, and the calculation of sample size for survival trials.

Dr. Qi Jiang is Biometrician in the biostatistics department of Merck Research Labs. Prior to joining Merck, she worked as a biostatistician at Harvard School of Public Heath. Dr. Jiang received her Ph.D. in statistics from Temple University earlier this year. She has presented her dissertation work at national and international conferences, and has received three student paper competition awards in 2003: WNAR Student Paper Competition Award, Society for Clinical Trials and International Society for Clinical Biostatistics Student Scholarship Award, and International Chinese Statistical Association Student Award and Travel Fellowship.


This event has passed.


640 W. Germantown Pike, Plymouth Meeting PA 19462
(610) 834-8300

From Airport: Take 95 South to 476 North to the last exit #20(Germantown Pike-West). Merge with Germantown Pike and follow for 3 lights. Make a right onto Hickory Rd. at the 3rd light. The hotel is the 3rd building on the left.

New York/ New Jersey Turnpike: Take the New Jersey Turnpike to exit #6, which is PA turnpike. Go west to exit #333- Norristown. Follow signs to Plymouth Rd. Go to the 1st light and make a left. Go to the next light and make a right onto Germantown Pike. Go to the second light and make a right on Hickory Rd. The hotel is the second driveway on the left.

Washington D.C., Wilmington, and Delaware: Take I-95 North to Route 476 North. Take Route 476 to the Germantown Pike West exit #20. Go to the third light, Hickory Rd., and make a right. The hotel is the 2nd driveway on the left.

Route 476: Take 476 to the Germantown Pike West exit #20. Go to the third light, Hickory Rd., and make a right. The hotel is the 2nd driveway on the left.

From downtown Philadelphia: I-76 west Plymouth Meeting exit #331B (Route 476). Take Route 476 north to Germantown Pike exit. Go to the third light, Hickory Rd., and make a right. The hotel is the 2nd driveway on the left.