People can be forgiven for thinking that the messages coming out of the American Association for Cancer Research annual meeting in Chicago this week seem to conflict. Finishing up today, the meeting hosted nearly 17,000 scientists, exhibitors and guests and had several talks expounding the dizzying pace of genome technologies being applied to cancer diagnosis and treatment. At the same time, some speakers warned of the challenges inherent in doing cancer ‘omics.’
A plenary talk Sunday evening by Elaine Mardis of the Genome Institute at Washington University in St Louis covered her group’s ongoing work to characterize individual patients’ tumours using what she calls deep digital sequencing, which looks at the whole-genome sequence from a patient and his or her cancer and then resequences and verifies individual mutations in DNA and RNA recovered from multiple biopsies. Her methods can show not only differences between cancer cells and normal cells but also how cancer cells change and evolve over time and in response to treatment. She has published recently on this for acute myeloid leukaemia (AML) and for myelodysplastic syndromes that can progress to AML.
Each progression to AML “has a different story to tell”, she said, meaning that different mutations can drive the evolution of the cancer. Another project, yet to be published, uses the same approach on solid tumours, namely breast-cancer samples from several patients enrolled in a clinical trial to test different aromatase inhibitor drugs. “We took this on because we wanted to be able to determine whether a genomic predictor of aromatase-inhibitor response could be identified,” Mardis said. The data suggested that cancers with lots of breaks, fusions and swapping of DNA in the genome responded less well in general to the drugs, but there were clues that some of the tumours may be responsive to other drugs not normally prescribed for breast cancer. Mardis called this a harbinger of things to come, in which genetic information, rather than classic clinical characterization of tumours alone, may direct treatment. What a tumour looks like genetically may be more important than where in the body it occurs, for example.
Many talks at the meeting focused on how using gene sequence could point to combination therapies tailored to specific cancer characteristics. Complex cancers, such as the ones resistant to the aromatase inhibitors, may require more than one therapy to prevent them from coming back. But it will be difficult to rigorously test such personalized approaches in the clinic, says George Sledge of the Indiana University Cancer Center in Indianapolis. At a session on Sunday morning, he discussed some back-of-the-envelope calculations regarding how a clinical trial might recruit patients. For example, if investigators wanted to evaluate a combination therapy that should work in a subset of tumours, the number of patients they would need to screen to find even a single study subject could quickly escalate to well beyond 100. That’s not feasible and necessitates a new way of looking at trials, Sledge says.
Leading scientists have proposed some ideas for fixing clinical trials. But any solution will have to contend with inertia and the wreckage of recent history. At an afternoon discussion on Tuesday, members of a US Institute of Medicine (IOM) panel presented their recommendations, released in late March, for how to develop genomics-based cancer tests for patients. They had developed the recommendations in response to clinical trials begun prematurely by researchers at Duke University in Durham, North Carolina, that were based on flawed, genome-based prediction schemes. The trials have since been terminated, and much of the research they were based on has been retracted. The committee concluded that a lack of oversight by individual researchers, their institutions, their funding agencies and the journals that published the initial studies allowed the trials to commence. Predominant among their recommendations was that data be shared widely so that others can replicate methods and that the Food and Drug Administration be contacted before the tests being evaluated start affecting patient treatments. “A bad tumour marker is every bit as bad as a bad drug,” said Daniel Hayes, one of the IOM committee members.
But their recommendations mainly dealt with the use of a few genomic markers to predict response to a specific drug, not whole-genome sequences to direct combination therapies. I asked Gilbert Omenn, who chaired the IOM committee, how ever-more-complex genomic tests might be evaluated in the clinic. That, he said, is still an open question.
Report this comment
Thank you for this post. There is an error that should be corrected. No one named Dan Foley was on the IOM committee. Dan Hayes from the University of Michigan was a member of the committee and a speaker at the AACR session, perhaps this is something he said? The IOM committee roster can be viewed at https://iom.edu/Reports/2012/Evolution-of-Translational-Omics.aspx by selecting the “View Full Committee” option on the right-hand side of the screen.
Report this comment
Thank you, Christine.
I’ve made the correction.
Report this comment
Thanks, Brendan! I hope this was the right way to go about suggesting a correction.
Report this comment
Are these findings surprising given the well understood nature of the instability of tumour genetic integrity? Once tumours have sufficiently compromised their DNA repair processes, their rates of mutation are significantly elevated and heterogeneity is bound to increase. We also know that selective pressure (acquisition of drug resistance) can cause watershed shifts in cell populations, allowing new populations to emerge and dominate. Tracking these shifts should yield important prognostic information, unless genomic integrity is shot, at which point using precision tools is likely a wasted effort. Expansion of deep sequencing of genetically unstable tumours will undoubtedly generate reams of data. However, documenting the precise shapes of clouds on one day does not help predict the shapes the following day, or the next.
Report this comment
However, documenting the precise shapes of clouds on one day does not help predict the shapes the following day, or the next.
I’m going to have to borrow that, thanks.