Decision Making in Health and Medical Care

IMSI - October 2021
Decision Making in Health and Medical Care Thumbnail Image
As COVID-19 pandemic has shown, decision making in health care is fraught with a multi-dimensional array of connected complexities including scientific discovery, accelerated drug development and approval driven by global urgency, allocation of limited resources, potential efficacy of unproven therapies, and political and economic pressures driven by governments, voters, companies, and patients.  COVID-19 and the subsequent discovery and distribution of effective vaccines put a lens on public health and medicine, magnifying many of the issues that are present on a daily basis for sufferers of other potentially fatal diseases.  Two such diseases are breast cancer and glioblastoma (an aggressive and often incurable form of brain cancer).  

Traditional drug discovery is a long process designed to maximize the safety and efficacy of drug candidates in a linear fashion, driven by the statistics of blinded, controlled experiments on patients to test a tightly defined scientific question about the potential drug (e.g., is there statistically significant evidence that this particular drug formulation, at this dose & frequency of administration, under these set circumstances, shows evidence of effectiveness as predicted while not harming the patient?).  These so-called “single arm trials,” test a single hypothesis and look for evidence of safety and effectiveness that meets a minimum level of measurable impact as compared to having done nothing at all.

The research groups of Laura Esserman and Donald Berry have been experimenting with alternative approaches to phase II and III drug trials for breast cancer (BC) (see Barker et al) and glioblastoma (GBM) (see Alexander et al), respectively. Their innovation is to address multiple hypotheses within the same trial, with the goals of accelerating the discovery of effective therapies and increasing the likelihood that approved drugs are adapted to the distinctive characteristics of patient subgroups. 

Berry and Esserman spoke recently at IMSI’s workshop on “Decision Making in Health and Medical Care,” held virtually May 17-21, 2021, and were joined in their session by Gary Gordon from the Global Coalition for Adaptive Research.  The three of them are leading teams that are demonstrating the effectiveness of multi-armed hypotheses within a single drug trial, using Bayesian statistics and adaptively randomizing patients throughout the process based on adaptively learning from results as the trial progresses.  Esserman, Berry, and their collaborators laid critical groundwork for these new approaches with I-SPY 2 which holds the promise of developing therapies adapted to the patient-specific biomarkers of their breast cancer.  

More recently, Berry has extended his statistical innovations to address the challenge of GBM.  GBM is incurable and the only therapeutic hopes right now are to extend patients’ lives and the quality of the time they have remaining.  They employ a multi-armed trial to address emergent hypotheses as a way to efficiently and more rapidly identify promising clinical improvements to patient treatment.  The figure below illustrates the GMB AGILE process.  As Alexander, et al. note, “Each experimental arm may participate in two stages during the trial: an initial adaptively randomized screening stage and a second confirmatory stage for those experimental arms that graduate.” Patient outcome data are used to update the model and adaptively modify the predictions of future probabilities of success as compared to the control.

GBM_Agile.jpg 132.75 KB
These new models are challenging old norms for how regulatory bodies oversee the drug discovery process and approve much needed new therapies.  Importantly, I-SPY 2 and now GBM AGILE, are prototypes for drug platform trials for COVID-19, Alzheimer’s Disease, Amyotrophic Lateral Sclerosis, and other diseases.  Modern statistical methods, and new computational learning and decision-making tools are contributing to a hopeful future for those stricken with deadly and debilitating diseases.

Barker, AD, Sigman CC, Kelloff GJ, Hylton NM, Berry DA, Esserman LJ. I-SPY 2: an adaptive breast cancer trial design in the setting of neoadjuvant chemotherapy. Clin Pharmacol Ther 2009; 86:97-100.
Alexander BM, Ba S, Berger MS, Berry DA, Cavenee WK, Chang SM, Cloughesy TF, Jiang T, Khasraw M, Li W, Mittman R, Poste GH, Wen PY, Yung WKA, Barker AD; GBM AGILE Network. Adaptive Global Innovative Learning Environment for Glioblastoma: GBM AGILE. Clin Cancer Res. 2018 Feb 15;24(4):737-743.