Session 4: Creating Knowledge Systems to Guide Cancer Treatment and Clinical Practice

Session 4: Creating Knowledge Systems to Guide Cancer Treatment and Clinical Practice

Session 4: Creating Knowledge Systems to Guide Cancer Treatment and Clinical Practice - Data Sharing, Widespread Implementation, and Ethical Considerations

Top Tier Recommendations

What are the top 3 recommendations that would significantly accelerate progress and what must be done to achieve them? Please be as specific as possible.

Recommendations from the presentations and notes from the discussion can be found beginning on page 2.

Key Questions, Panelist Recommendations, and Discussion Notes

Key Questions

  • What are the challenges of providing decision support within a healthcare delivery model for precision cancer therapy?
  • What are the challenges in creating and maintaining knowledge bases regarding the clinical significance of genomic alterations in cancer?
  • What are the challenges related to widespread sharing of linked genomic and clinical data?

Recommendations

Levy

  • Innovation in clinical decision support for precision cancer medicine can exist outside traditional clinical information systems but needs to evolve into more action oriented recommendations (versus interpretations) and should be seamlessly integrated into clinical workflows.
  • Publically available knowledge bases are essential to scalability but evolved models are needed for governance and sustainability.
  • Develop strategies to improve the quality of “BIG Data” derived from clinical care and evaluate its utility for evidence generation and decision support.

Van Allen

  • Consensus on the language of genomics reporting
  • Foster interdisciplinary work between medical groups and UI/UX, technology, and EHR experts
  • Empower patients to access and share their (genomic) data

Staudt

Impediments to Cancer Genomic Data Sharing

  • Sequencing done in the course of clinical care – HIPAA

Develop universal consent for data sharing and offer to patients

  • Patients do not own their own cancer genomic data

Efforts to rectify include Blue Button and the NIH PMI

  • Cost of sequencing – third party coverage not routine

Support coverage of comprehensive cancer genomics by CMS

  • Contractural limitations from pharma

Encourage pharma to consider genomics as largely pre-competitive

Joffe

•Incentivizing collaboration

•Building & maintaining a common infrastructure for ‘omic, phenotype & outcomes data

•Overcoming (real and perceived) regulatory barriers

•Gaining patient engagement, consent & trust

•Demonstrating improved patient outcomes

Discussion Notes

  • Amy Abernethy –
  • Need to make sure that datasets are of highest quality
  • Comes down to the management of unstructured documents
  • Can get physicians to structure 3-5 variables (beyond that it falls apart)
  • There’s not just one way of getting there
  • Core elements
  • Aggregate across EHRs
  • Manage unstructured documents and have human curation of data points in a technical framework – need to understand quality of underlying data points
  • To get high quality datasets, you need to know how to standardize them
  • Mia Levy
  • Would the FDA ever consider a comparator arm from this type of data?
  • Girish Putcha
  • Can Amy go one layer deeper
  • How to incentivize collaboration, cultural and practical issues
  • Amy – datasets are continuously aggregating, we need to consistently analyze them
  • Eliezer Van Allen – incentives
  • Cancer Genome Atlas - probably won’t get promoted based on it, but was a worthwhile experience. Would hope that this kind of thing goes beyond looking for promotion etc.
  • Todd Golub - Days of holding onto data are over. Still in a tenuous position though. Concerned that we’re approaching a bit of a stalemate with all of these databases.$ is also a concern to doing this well.
  • Mia Levy – these are all little experiments – sustainability model is also completely unknown
  • Steven Joffe – Need to get away from institution competition, strings attached to funding, can we learn something from other collaborative fields
  • Lillian Siu – white paper can address collection of data
  • Challenge that the group can help to standardize
  • We typically have rules that we cannot disclose data – how can we get around this to share knowledge
  • Louis Staudt – Are making a significant contribution to make the date available. Will take any good ideas. This will be a sustainable system. There’s a good deal that can be learned from reasonable well curated data
  • George Demetri
  • learning about the disease is a big thing we can do with this
  • If we had some rules….
  • Barbara Conley - exceptional responders, maybe some of these projects can inform but how much can we trust what’s being said in the community
  • Amy Abernethy - once you develop a way to manage data you can craft real world endpoints.
  • Lillian Siu - is time on treatment a good predictor of response
  • Amy Abernethy - other assessments can be better, but sometimes it works
  • Gideon Blumenthal – respond to whether FDA would take real world evidence
  • We have some precedence
  • Boils down to magnitude of effect
  • Need for guidance on what types of endpoints we should be looking at
  • Mia Levy – goes back to challenge of keeping up with what’s on drug labels. Could make big gains on using this type of data
  • Girish Putcha - for an oncology drug, we discuss what’s the right endpoints. Anything that involves duration of response is as much about safety and efficacy as the diagnostic
  • Lillian Siu – Is time on treatment the best surrogate? Don’t know what is the right solution
  • Amy Abernethy – need to work through portfolio of endpoints
  • Jose Baselga – every case is different
  • If you know the natural history of the disease
  • If this is going to work, it’s because we have a Gleevec-type story (have no doubt that it’s real)
  • Steven Joffe – 2 outcomes that matter to patients – living longer with a good quality of life…save money as a 3rd outcome
  • Girish Putcha - talking about criteria for coverage – challenge is that if we have survival endpoints, realize that goes against what others have argued for
  • Susan –
  • Ronald Kline - caution people not to latch on to CED
  • Peter Yu – is CMS allowed to do research as we think of it? There are limitations to CED policy
  • George Demetri – not asking CMS to do research
  • Jose Baselga – we have plenty of data on NGS, mutations have been define.
  • There is evidence for 15-25 therapies that work for specific mutations
  • Question is does implementation of this assay: 1) enable finalization of clinical trial,

2) does approach lead to a better outcome

  • George Demetri – drug should come from pharma sector. Necessary part is the testing part. Understanding of the cancer with precision is the patient element
  • Girish Putcha – need to define the patient that should be tested
  • George Demetri – want testing in the right patients – there’s good evidence that this is the way to find patients with rare mutations
  • David Hyman – currently CMS cannot assess analytic validity
  • Girish Putcha – correct
  • Mia Levy - love that we’re discussing if real world evidence can inform decisions