Personalizing Clinical Trials
Because of various genetic and/or environmental factors, treatment response may vary across patient populations. As a result, increased focus on targeted therapies has led to great interest in subgroup analysis. Therefore, proactive planning is imperative for subgroup analysis. Here, I focus on trial designs that foster subgroup analysis, and on analysis methods that can be explored further to optimize this type of analysis. These proposed designs can help optimize designs as it relates to evidence based research for precision medicine.
Precision medicine challenges classical approaches. It is often the case that a patient enrolls into a clinical trial where the treatment is randomly allocated, ignoring almost all information about the patient. However, advances in both technology and molecular biology are changing this paradigm. Therefore, other methods that attempt to personalize patients’ treatment such as basket trials, umbrella trials and/or enrichment designs should be explored. Basket trials are designed to test the effect of a single drug, on a single mutation, in a variety of cancer types.
Umbrella studies are designed to test the impact of different drugs, on different mutations, in a single cancer type . Basket trials especially, have generated a great amount of interest, because they implement a hypothesis-driven strategy that incorporates precision medicine into clinical trials, even for rare diseases . Enrichment designs screen patients for the presence or absence of a marker, or a panel of markers, and then only include patients who either have or do not have a certain marker characteristic or profile .This design is based on the assumption that not all patients will benefit from the study treatment under consideration, but rather that the benefit will be restricted to a subgroup of patients who either express or do not express a specific molecular feature .
Since in many cases, there are many subgroups available, it is important to plan for subgroup analysis before a trial begins: this makes subgroup analysis a hypothesis testing problem. Otherwise, there can be a tendency to test for subgroup effects in a large number of un-planned subgroups, resulting in spurious significant findings. Still, even when only pre-planned subgroups are tested, studies may not adjust for multiple testing, and report biased (i.e inflated) p-values for each subgroup. In addition, sample sizes are only adequate for detecting possible overall treatment effects, and true subgroup effects, when present, may not be detected due to insufficient power. This is especially a challenge for statisticians in oncology because phase III trials may be underpowered for hypothesis testing.
To account for these shortcomings, observational data and Bayesian approaches are recommended . Though this post will not detail these methods, it has been shown that Bayesian methods address important goals in a subgroup analysis: multiple testing and controlling the error of choosing a subgroup effect when there is an overall effect .
Statisticians may not agree on the optimal approaches to apply in precision medicine, but it is clear that our role is essential in assisting our clients in this area. At BSSI, we work closely with medical and scientific experts to help identify subgroups of patients with increased benefits. Feature selection and classification through multivariate logistic regression are some key statistical tools that we use in subgroup identification. By developing and evaluating both existing and new approaches, statisticians are charged with applying these techniques in ways that are innovative and that will ultimately, serve the patient.
Illustration by Simona Ceccarelli, copyright F. Hoffmann-La Roche
- Redig, A. J., & Jänne, P. A. (2015). Basket trials and the evolution of clinical trial design in an era of genomic medicine. Journal of Clinical Oncology, 33(9), 975-977.
- Mandrekar, S. J., & Sargent, D. J. (2011). All-comers versus enrichment design strategy in phase II trials. Journal of thoracic oncology: official publication of the International Association for the Study of Lung Cancer,6(4), 658.
- Sivaganesan, S., Laud, P. W., & Müller, P. (2011). A Bayesian subgroup analysis with a zero‐enriched Polya Urn scheme. Statistics in medicine, 30(4), 312-323.