Alliance 2020
The Alliance held a two-day meeting on February 27th & February 28th, 2020, prior to USCAP in Los Angeles, CA.
Hosted by the DPA and FDA, with support from Visiopharm and Matt Leavitt (LUMEA).
“The Alliance 2020 pre-USCAP meeting had great speakers and productive working groups. I am honored to be an Alliance member. I am looking forward to the next meeting.”
The Alliance 2020 pre-USCAP meeting, hosted by the DPA and FDA with support from Visiopharm and Matt Leavitt (LUMEA), at the J.W. Marriott Los Angeles L.A. Live.
THE AIMS OF OUR INAUGURAL TWO-DAY MEETING
Day 1: Thursday, February 27, 2020
Provide a status update of the Alliance
Provide additional perspectives from thought-leaders in the field
Enable networking among members
Day 2: Friday, February 28, 2020
Host the FDA's high-throughput truthing (HTT) session for tumor infiltrating lymphocytes (TILs)
Provide the unique perspective of two leaders Mike Leavitt and Abdul Hamid Halabi
Take content of prior breakout sessions and continue development
Please find the presentations (Thursday) and the summaries of the talks and breakout sessions (Friday) below.
Thursday, February 27, 2020 - Presentations
Welcome to Alliance 2020 – and Overview and Status Update
Joe Lennerz
Massachusetts General Hospital
Joe provided an introduction and overview of the aims of the Alliance. The slide-deck includes a brief status update regarding current developments. The collaborative spirit of the Alliance includes being open to other vantage points.
Regulatory Science through NGS and beyond
Keynote address by Lakshman Ramamurthy
GSK Sr. Dir. of Global Regulatory for Oncology, former Lead Reviewer at FDA/CDRH, former Global Regulatory Lead of Foundation Medicine
Lakshman started by telling his personal journey, provided an overview of the organization structure of the FDA including a brief overview of device classification. He also outlined the co-development of drug-diagnostics. He outlined that he will use the NGS approval of Foundation Medicine’s NGS panel as an analogy to novel regulatory approvals. He outlined a regulatory strategy for a WSI/AI/ML algorithm by applying existing systems and the In Vitro Diagnostic Multivariate Index Assays (IVDMIA) framework. He also outlined that most current ML tools are in the clinical decision support realm (typically class II).
Digitizing Pathology Slides for ML Applications: Opportunities and Lessons Learned
Craig Mermel
Product Manager at Google
Craig outlined his talk composed of three distinct topics: digitization of archival slides, ML applications in pathology, and the impact of image quality on ML applications. He first outlined Google’s scanning operations. He distinguished strongly vs. weakly supervised vs. unsupervised learning. He shared second, several of the applications that Google has developed and published including performance comparison between pathologists and ML tools, their contributions in the CAMELYON challenges, and a Gleason grading tool. He finally shared their approaches to apply ML tools for image quality and focus-related issues.
The WHO International Standards for Tumour Classification and Diagnosis
Ian Cree
Head of WHO Tumor Classification
Ian introduced: IARC and WHO as a complementary partnership. He outlined that his responsibilities include the IARC monographs and the blue books (among many other issues). He has introduced the 5th edition of the blue books (GI and breast) and outlined the introduced changes including 2-column layout, larger images, standardized language and essential vs. desirable diagnostic criteria. He also showed the new website including whole-slide imaging capabilities. He emphasized the importance of numerous image-based tools from planimetry to AI/ML tools. Finally, he outlined the IC3R initiative, a large scale effort that we will hear more about in an upcoming webinar.
Introducing 2nd part of Alliance 2020 Thursday Sessions
Brandon Gallas, FDA/CDRH/OSEL/DIDSR
Brandon described his role at OSEL where he’s been since 2001. Half his time is devoted to the regulatory review of studies of imaging devices and related algorithms. The other half is devoted to research on least burdensome imaging study designs and protocols and the statistical analysis of such studies. Brandon pivoted from radiology to pathology starting with a presentation at the first FDA panel meeting on WSI in 2009. He also outlined the scope of the data-collection “truthing” event (February 28, 2020).
Standardization and Quality Gap in Tissue Pathology? Challenges and potential solutions
Michael Grunkin
CEO of Visiopharm
Michael started by providing a brief history of Visiopharm and his background. He emphasized the challenges of interpretive accuracy and comparing biomarkers across laboratories and pathologists. Their studies on quality studies led them to creating an infrastructure and data model (“qualitomics”) used to measure and report staining quality. He also mentioned monitoring and calibration as key components to monitoring image-based analytics.
Criteria for reimbursement decisions
Frank Dookie
Consultant, former contracting professional at Humana
After a brief introduction, Frank outlined that digital pathology is part of laboratory medicine and “you don’t have to be sick to use it”. He outlined the intricate relationship between diagnosing a condition and how this change the healthcare delivery moving forward with changing treatment and diagnostic regimen. He outlined that digital pathology might be an opportunity to provide added value including faster and/or more reliable diagnosis or in/excluding conditions (e.g., excluding the diagnosis of cancer). He outlined what payors are looking for when it comes to coverage determination including FDA approvals, CMS (national and local coverage policies), CPT/PLA codes. He summed up by outlining models for the discussion in the breakout session (February 28, 2020).
Value of Digital Pathology
Charles Mathews
Principal at ClearView Healthcare Partners
Charles talks about how to get the added value recognized in digital pathology. Charles outlining how to provide supportive evidence – including cost-effectiveness analyses. He distinguished three value-propositions: (1) efficiency gains vs. (2) improved quality vs. (3) novel information related to treatment and patient managements. Coverage adoption, healthcare systems and payors become more and more explicit. Charles mentioned the design and importance of the TAILORx study design and relevance to demonstrating outcome changes.
The ORIEN Alliance. Accelerating cancer research and delivering hope through collaborative learning and partnerships.
Ming Poi
Senior Director of Scientific Affairs, M2Gen
Ming outlined the scope and aims of ORIEN. The concept entails a comprehensive data sharing platform initially developed by MOFFIT Cancer Center that is now growing into a multi-center network. Ming reported on various data sharing projects including data sharing agreements, the infrastructure, an AVATAR project (whole exome and RNA sequencing), and the ORIEN clinical trials network (OCTN). ORIEN also entails creating real world evidence (RWE) for FDA submissions. ORIEN is interested in contributing to the Alliance aims: truthing and payor initiatives.
Friday, February 28, 2020 - Summaries
MIKE LEAVITT
We were excited and very thankful that Mike Leavitt joined us and delivered an impactful speech on February 28, 2020. In his talk entitled "The New Frontier of Innovation" he introduced his work on Value Alliances and the key elements of successful Value Alliances.
The Alliance for Digital Pathology wants to thank Mike Leavitt for taking the time and his support.
Mike is a prominent business leader, and the former 14th Governor of Utah, Secretary of Health and Human Services (HHS), and Administrator of the Environmental Protection Agency (EPA).
ABDUL HAMID HALABI
We were also excited to learn from Abdul Hamid Halabi and his amazing talk "What Pathology Can Learn From Self-Driving Cars”.
We collaboratively developed this topic because the regulatory considerations for self-driving cars also have to take into account numerous rare and unexpected situations, which is quite similar to AI/ML in digital pathology. Abdul presented this highly complex topic in a fascinating and entertaining way. The Alliance very much appreciates the support and seeing behind the curtain of some of the technologies we use everyday.
Abdul is NVIDIA's Global Business Development Lead for Healthcare & Life Sciences and leads the product team for NVIDIA’s health offerings and has been at the forefront of advanced technologies in AI and high performance computing for several decades.
Deliverables Breakout Session
Think about it as a hackathon where the output is a regulatory science deliverable (e.g. a white paper).
We have the following topics and session leads – for other topics or projects, please contact the steering committee members (e-mails provided below).
Pre-analytics (Amanda Lowe) alo[[visiopharm.com
Join our working session and become a voice and contributor to this very important topic! Tissue samples support critical decisions in diagnostics, research, drug development, business, and the regulatory approvals of drugs and medical devices (including SaMD). It is essential to understand the error sources of tissue samples, the resulting images and metadata, the impact of those error sources and how to mitigate them. Although broadly relevant, the goal of the pre-analytics committee is to generate guidelines for pre-analytical standardization that will promote pathways for regulatory clearance based on high quality images of tissue samples. During our working session we’ll be conducting searches for supporting literature, reviewing existing guidelines, and prioritize areas for the committee to focus on. Our goal is a framework and/or Table-of-Contents by the end of the working session.
Slide Scanning (Scott Blakely) sblakely[[hamamatsu.com
The overall goals of the this group within the Digital Pathology Alliance is to seek to find ways to standardize the characterization of slide scanners, define these characterizations and how they can be measured and (hopefully) be able to describe a non-tissue based phantom that can be used in conjunction with one or more FDA qualified MDDTs such that the review and approval of new scanners, or the updating of existing scanners can be standardized and expedited to the benefit of digital pathology as a whole.
Currently, I envision this as being a phased process with one or more white papers being published at the end of each phase. Some of the topics this committee will discuss will also touch on pre-analytic topics and I envision some cross talk between the two committees.
The phases I am suggesting are:
Phase I: The standardization of the characterization of slide scanners.
This would include topics like:
Physical characterizations for scanners, regulatory requirements .
Evaluation of scanner image formation/optical and detector evaluation.
Data characterizations – What information needs to be delivered to the scanner and what must the scanner output.
Phase 2: Proposals for Validation tests to become MDDTs that can be used to accelerate FDA review and approval for Technical Performance Assessments (TPAs).
What needs to be tested?
What is the testing/reporting format?
Stretch goal – can we envision a way to see full validation based upon MDDTs and TPA information without the need for full clinical trials?
Phase 3: Development of phantom or set of phantoms that can be used to objectively characterize all scanners.
One of the goals of the upcoming workshop is to review this phased approach and to identify specific projects for Phase 1, identify a leader for each identified project and begin constructing a timeline with milestones.
I encourage anyone who has an interest in these topics to join either as a full participant or as a reviewer of content as breadth and diversity of viewpoints and expertise will be a key factor in the success and acceptance of the deliverables from this work.
Truthing (Brandon Gallas, Sarah Dudgeon, Hetal Marble) brandon.gallas[[fda.hhs.gov or sarah.dudgeon[[fda.hhs.gov or hmarble[[mgh.harvard.edu
Tumor infiltrating lymphocytes (TILs) in breast cancer establishing ground truth dataset (HTT group).
We are looking for board-certified anatomic pathologists and residents to contribute 30 minutes each to this effort on-site.
Please see here for full data collection details: HTTatUSCAP.pdf.
Share IRB and RCA materials and enlist partners to source slides at other sites and coordinate data-collection events. Discuss stat plan and agency feedback on MDDT submission.
AI/ML/Continuous Learning (Richard Huang) rhuang6[[mgh.harvard.edu
What are the relevant FDA publications (guidances, discussion papers) that discusses regulatory approaches to adaptive algorithms?
Digital Health Software Precertification (Pre-Cert) Program: https://www.fda.gov/media/119722/download
Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD): https://www.fda.gov/media/122535/download
Use of Real-World Evidence to Support Regulatory Decision-Making for Medical Devices: https://www.fda.gov/media/99447/download
Software as a Medical Device (SAMD) Clinical Evaluation: https://www.fda.gov/media/100714/download
Draft whitepaper outlining the regulatory approaches
Prepare mock submission to test our regulatory approaches
Standards (Markus Herrmann and Mike Isaacs) isaacs[[wustl.edu
Standard image database: collection of digital whole slide microscopy images in DICOM format
DICOM software tools for digital and computational pathology
Payor strategies (Esther Abels and Joe Lennerz) jlennerz[[partners.org or esther.abels[[pathai.com
Best practices
Approaching the cost-saving aspect against reduced reimbursement
Policy development
During our workshop we have two goals: 1) develop an overview of all entities and stakeholders involved in reimbursement for Digital Pathology (DP) and 2) kick off the development of a whitepaper.
To get reimbursed for DP one would have to consider the DP workflow and all related phases, from pre-analytics - the technical component - to the pathologist signing out the case - the professional component. Many entities are involved in or working on different aspects of these phases in this entire workflow. To come to a sustainable and scalable approach to properly set up reimbursement for DP, we have to better understand who is involved in what and when, who is an influencer, decision maker, champion or a stakeholder. For example, zooming in on pre-analytics, one could think about proficiency testing and which entities are involved, such as CLIA, the health care provider, industry associations, and maybe even the international color consortium. All these entities touch upon reimbursement in some way.
This overview of stakeholders, will be one of the key inputs for authoring the whitepaper about reimbursement for DP. The whitepaper will outline the current landscape of reimbursement, concerns from users, the value proposition and what to focus on for getting DP reimbursed.