The Center for Innovative Clinical Trials at the University of North Carolina at Chapel Hill has been established to look for better ways to design and conduct clinical trials.

The following is excerpted from an interview with center director, Dr. Joseph G. Ibrahim, Alumni Distinguished Professor of biostatistics in the UNC Gillings School of Global Public Health, and Director of the Biostatistics Core at the Lineberger Comprehensive Cancer Center.

Q: What do you hope to accomplish with the Center?

Dr. Ibrahim: We want to find ways to speed up and improve the clinical trial process, and make it a more efficient, precise assessment of treatments, whether they are drugs, medicines, devices or behavior modification programs. We want to find ways to get results faster, but we want them to be more precise and useful in determining if the treatment is effective, and if so, for which patients.

The ultimate goal is better healthcare.

Q: What can you do to improve the clinical trial process? Most of you are statisticians. You deal with math, not molecules. Can you find a mathematical cure for cancer or any other disease? 

Dr. Ibrahim: Well, we don't discover treatments, but statisticians play an enormously important role in assessing whether treatments work and for whom. The statistics - or the mathematics - are a critical part of designing a trial and assessing the results to determine if they are meaningful.

Q: What's different about your approach?

Dr. Ibrahim: Most trials right now are designed to answer two questions. Does the treatment work? And is it toxic? We believe the questions are more complex, and there are statistical approaches that can help us evaluate a broader spectrum of data in the same amount of time.

For years, statisticians have provided technical and practical results for clinical trials, but we believe we can make the trials more precise and definitive if we work more closely  with the investigators as part of a scientific team when designing the trials, deciding what to look for and how to formulate and assess endpoints and interpret results. We can help determine what the most meaningful analyses of the data would be, including toxicity as well as overall benefit. If one is scientifically and statistically sound with their design and analysis plan, then the trials could potentially be shorter, involving fewer patients. That would make them less expensive, which should be reflected in the cost of the treatment to patients. It also should help get effective new treatments to patients sooner.

Q: At what phase in a new treatment's development do you see the Center getting involved? 

Dr. Ibrahim: I believe we could start as early as the Phase I level (early clinical trials to assess appropriate dose levels, toxicity, and side effects) or even at the conceptual stage. In the Phase I setting, we would develop better statistical methods to determine the most tolerable dosage (MTD) and optimal dosing schedules, as well as a better assessment of side effects.  This would make a difference when moving to Phase II and III trials. 

Q: Do you think you could come up with methods that might prevent FDA recalls that we've seen with some other marketed medicines due to side effects? 

Dr. Ibrahim: Yes! I think this is where we could make a real difference in clinical trials. Right now, trials are typically designed to pay special attention to one primary endpoint.

Phase III clinical trials require a multi-dimensional approach where we look at several endpoints simultanously. We believe this will make adverse events and toxicity results more apparent earlier on in the trials. We also think that there is a need for clearer "stopping rules" - points at which you stop a trial either because of toxicity - the drug is hurting patients - or because the drug is so ineffective that it's unethical to continue the treatment,  or potentially for reasons that the drug is so effective that it's unfair to withhold the treatment from other patients in the trial.

Q: You are an expert in Bayesian methods - explain what that is and how you think it will help make safer treatments available faster.

Dr. Ibrahim: There are two main paradigms for analyzing data, namely the frequentist and Bayesian paradigms. In the frequentist paradigm, you, in a certain sense,  "start over" each time in the design and analysis of a new trial without using preconceived notions. That is, the frequentist paradigm does not offer a natural or well accepted mechanism for incorporating preconceived notions or previous data into the design and analysis of a new trial.  The frequentist paradigm is the one that is most often used now, especially for designing studies and analyzing clinical trial data in the Food and Drug Administration's review process.

Among many of its features, Bayesian methods lend themselves to easier understanding interpretation of results, they involve the incorporation of information you know from past experience, they allow for direct probability (chance) statements, such as the chance that one treatment is more effective than another, and they are a more natural paradigm for making predictions about future outcomes with certain treatments or therapies.

There's room for both types of paradigms, but one of the most important features for me in my collaborations with non-statisticians is that Bayesian designs and analyses lend themselves to much easier understanding and interpretation of results.  For example, with frequentist analyses, results are typically reported in terms of a number - a p-value - which really can only be well understood by statisticians. Bayesian analysis methods lend themselves better to graphics and visual descriptions of results that are more easily understood by general audiences

Q: Does the Food and Drug Administration and/or National Institutes of Health recognize Bayesian methods? 

Dr. Ibrahim: It's an issue both the FDA and the NIH are grappling with. In general, the biggest issue that FDA is trying to deal with regarding Bayesian methods is how to incorporate  preconceived notions, called prior information in statistical terms, in the design and analysis of a trial. For example, they're asking themselves if historical data should be considered in the analysis of a new study, or if it would bias the results of a new study? If you do use prior information, how would you use it? Should it be discounted in some way so that it does not have too much influence over the new results?  Right now, there's no conventional or gold standard use of prior information or  historical data in a Bayesian analysis because of a concern that the patient populations in two different studies might not be the same, or there were different ways of administering the study that would lead to non-comparable results.

What we are hoping to do at the Center is get a consensus among experts - find methods everyone can be comfortable with. We believe there is a better way to do the statistical design and analysis of clinical trials, and we believe we can be the catalyst for change.

Q: Will the Center coordinate studies? 

Dr. Ibrahim: No. We are not a coordinating center, but we will work closely with the excellent Collaborative Studies Coordinating Center that is housed in the UNC Gillings School of Global Public Health's biostatistics department.

Q: The center is to be interdisciplinary - what fields do you anticipate being involved? 

Dr. Ibrahim: Right now, center members include faculty from the School of Public Health's departments of biostatistics, epidemiology and maternal and child health, as well as from the UNC School of Medicine and Lineberger Comprehensive Cancer Center.

We anticipate having faculty from many other disciplines, too, including nutrition, environmental health, health policy, disease prevention, economics, business, general medicine, oncology, dentistry, psychology, psychiatry, phychometrics, biomedical engineering, nursing, biology, chemistry, and others.