National Cancer Policy Summit 2

New Paradigms

Razelle Kurzrock, MD, Murray Professor of Medicine, UC San Diego Moores Cancer Center, asked one of the most disconcerting questions of the day: What if every patient with metastatic disease is genomically unique?

If this were true, trial design would have to be turned on its head. “In the model we use now, patients are not selected by molecular characteristics, and this results in a poor likelihood of efficacy because many of them do not have the molecular defect in question.”

If most (or all) patients require a specially tailored treatment regimen, the current clinical trial endeavor is almost impossible. Given an estimate of at least 300 current oncology drugs (that are either approved or in advanced testing), said Dr. Kurzrock, there are 45,000 two-drug combinations and 4,455,100 three-drug combinations. This means 1,000 years of trial and error to figure out what would work for a particular patient.

A new model uses smaller trials where all patients have the relevant mutation or genetic defect, leading to a greater chance of success. “In phase II, for instance, we could use multiplex markers to diagnose and classify cancers, validate a specific strategy, and understand convergence pathways.”

There are problems with this model, though. Drugs may not be easily acquired or repurposed, and there are greater regulatory burdens. “Despite this, we need new paradigms: N=1 strategies, drug trials that take place earlier in the course of a disease, and more customized drug combinations.”

Drug Development Strategies

Mace Rothenberg, MD, Senior Vice President for Clinical Development and Medical Affairs, Pfizer Oncology, New York, addressed four provocative questions. First, how can we integrate genomics and other platform diagnostics into drug development strategies that will meet regulatory requirements for companion diagnostics? His response was that the era of “one-off” diagnostics is over for certain diseases where multiple driver or actionable mutations are identified. Moreover, there is a limited amount of tissue for stand-alone diagnostics. But even if genomics and similar platforms provide sufficient accuracy, consistency, reliability, and reproducibility, regulatory agencies may not accept single-analyte results.

Second, can the results of minimally invasive diagnostic tests be used to expedite drug development and inform therapeutic decisions? Dr. Rothenberg said that technology is improving rapidly for detection and analysis of circulating tumor cells, and we know that enumeration of and change in circulating tumor cells are useful prognostic and predictive tools. But can we take the next step and evaluate the relationship between molecular characteristics of circulating tumor cells and development of clinical resistance?

Third, how can we make the greatest use of N=1 studies? This question led to others: What threshold of evidence should qualify a biomarker for patient selection? When are single genetic abnormalities sufficient? What threshold of activity should we expect in biomarker-selected cohorts? And how can we learn from failed N=1 studies? Dr. Rothenberg urged the group to consider these questions as it settled on priorities.

Fourth, is it possible to define a preferred sequence or hierarchy of therapies for a specific disease? For instance, a drug with a longer progression-free survival in first-line therapy but which confers cross-resistance to other drugs may not be preferable over a drug with a shorter progression-free survival, which can then be followed by second- and third-line drugs that, taken together, result in a longer progression-free survival, overall.

Keynote Speech

In his keynote presentation, Warren A. Kibbe, PhD, Director, NCI Center for Biomedical Informatics and Information Technology, Rockville, Maryland, offered three policy issues:

Does informed consent enable research? Is there a role for standardized consent, and can translational research be undertaken without it?
Identification of specimens and data—any tissue will be rapidly identifiable, and sequence is like a fingerprint. What is privacy? If patients would like to see their specimens and data available for research and discovery, how do we enable that?
Open access to data—how can we promote the desire of patients to enable the open access to specimens and data to make truly transformative inferences and observations? Consent is a process, and how can we give patients better control over their consent decision on an ongoing basis? What does that mean for data and specimens that have already been released. How do we incorporate this into standardized consent and open science discussions?

He noted that social media could measure and explain attitudes and behavior in a population, perhaps enabling better risk estimates. “Social media are a big data opportunity, but what are the ethics of using them for research? People say they want more data sharing, but using data in a way that translates into health statistics and personal information changes the paradigm. To put that in cancer terms, we need to find an ethical way to combine systems biology and social data with clinical care and outcomes from providers.”

The future, said Dr. Kibbe, may reside in elastic computing clouds, that is, social networks and big data analytics to produce semantic and synoptic data that can result in precision medicine, accurate health measurement, and protective medicine—intervention before health is compromised.