Results of randomized clinical trials can be somewhat limited. Currently used statistical methods to test drug candidates only consider a single criterion—the primary endpoint. This can represent an incomplete assessment of trial results, particularly in oncology and chronic diseases, where quality of life and relief of symptoms are of prime importance. 
 
The International Drug Development Institute (IDDI) has contributed to the development of a new statistical method called “generalized pairwise comparisons” (GPC), which allows the analysis to take into account all the endpoints at once, whether they represent treatment benefit or harm, as long as these endpoints can be ranked in an order of priority. The method aims to allow physicians and/or patients to define their own priorities regarding treatment outcomes, thereby leading to “personalized medicine”.
 
IDDI recently initiated collaborative research on a statistical method for individualization of therapy. The goal is to establish proof of the concept that the GPC method can complement traditional methods for the design, analysis and interpretation of clinical trial results. The research will be carried out by a Consortium led by IDDI and comprising Bristol-Myers Squibb, the European Organization for Research and Treatment of Cancer (EORTC), the Université Catholique de Louvain (UCLouvain) and the University Hospital and Cancer Center of Lyon.
 
Contract Pharma spoke with Marc Buyse, founder of IDDI, and Everardo Saad, medical director of IDDI, about the goals of the new collaboration, the shortcomings of current methods in clinical trials and how new statistical methods can be used to improve outcomes. –KB
 
 
Contract Pharma: What is the current standard method used in clinical trials?
 
IDDI: Randomized clinical trials, which ultimately lend support to the approval and clinical use of effective treatments, have served us well and will continue to do so in the foreseeable future. In these trials, what is typically done is to compare groups of patients and consider a single primary outcome to assess treatment benefit, with secondary outcomes analyzed separately and, most of the times, informally. As a result, these trials provide “an average” response to the question “which treatment is better, A or B?”.
 
CP: What are the benefits and shortcomings of this method?
 
IDDI: The main advantage of the methods currently used is that the statistical significance and magnitude of the treatment effect can be assessed across different trials using universally accepted primary outcomes which are both clinically relevant and statistically sensitive. However, there are situations for which there is no agreed magnitude of benefit that is desirable. Also, sometimes there is no well-accepted single primary outcome. A potential solution to this problem is the use of composite measures, which typically focus on the time to occurrence of the first on a list of possible outcomes, but not necessarily the most important one of that list. In this age of personalized therapy we need to allow for responses to the question mentioned above in a way that takes into account personal preference with regard to the importance of different outcomes, not only effectiveness, but also toxicity and quality of life. 
 
CP: What new statistical methods can be used to improve outcomes?
 
IDDI: Improving “average” outcomes requires better treatments, but outcomes can also be improved for individuals if a method allows personalized decisions. The method studied by the BENEFIT Consortium is called “Generalized Pairwise Comparisons”. In this approach, each patient in, say, group A, is compared with each patient in group B, taking into account an order of priorities of the different outcomes following individual preferences. The statistical analysis then comes up with a probability of more benefit from treatment A than treatment B, given that preference. For instance, a given patient may prefer a more aggressive treatment, even if accompanied by serious side-effects, if this improves their chance of responding to treatment, whereas another patient with lower tolerance to serious side-effects may prefer a less aggressive treatment. The method of Generalized Pairwise Comparisons will take into account the order of preference of the outcomes by these two patients, compute different probabilities for each of them, and give a personalized response to the question “which treatment is better, A or B?”.
 
CP: Are there any regulatory hurdles to overcome leveraging these new methods?
 
IDDI: The method of Generalized Pairwise Comparisons has never been used in the regulatory setting. This is why the BENEFIT Consortium will focus on studying the properties of this method in a wide variety of settings, with publication of papers on the method itself as well as its applications.
 
CP: What are the goals of the collaboration?
 
IDDI: The goals of the collaboration are to further develop the method of Generalized Pairwise Comparisons, to study its statistical properties, to come up with ways of making the method intuitively simple to understand and use, and finally to make it accessible to the various stakeholders: drug developers, clinicians, statisticians, regulatory agencies, health care organizations, and ultimately patients.
 
CP: What will each party contribute?
 
IDDI: The collaboration will focus on the requirements of the various stakeholders, in terms of regulatory acceptance for pharmaceutical companies (Bristol-Myers Squibb); in terms of clinical trial design for large cooperative groups (European Organization for Research and Treatment of Cancer); in terms of clinical interpretation for clinical oncologists and cancer patients (Centre Léon Bérard, Lyon); and in terms of statistical properties (Université Catholique de Louvain and Université de Lyon). The International Drug Development Institute (IDDI) will coordinate the research efforts and has contracted an independent software company to develop open-source software as well as a Software as a Service (SaaS) implementation of the method.