The case for a global clinical oncology digital archive

Author: Sanjay Joshi
Source: Original, reproduced with permission

Image source: “Innovation or Stagnation: Challenge and Opportunity on the Critical Path to New Medical Products”, US FDA, 2004

 

Clinical Trials are costly, complex, compliant and continual processes with well-defined pre-determined endpoints (especially for prospective randomized trials). Cancers are costly, complex, heterogeneous, multi- and epi-omic processes with usually one end-point. Currently, precision oncology affects less than 7% of patients with minimal benefit across large trials. The combination of the clinical trials for oncology drugs and therapies become a maze of multiplicative complexities. This complexity is succinctly illustrated in the figure below:

Image source: Rosenblatt M, “The Large Pharmaceutical Company Perspective” N Engl J Med 2017; 376:52-60. DOI: 10.1056/NEJMra1510069

 

Small molecule and biological drug companies strive to contain this complexity with good design of experiment, well defined processes and multi-variate operational decision trees. Digital Imaging and Pathology are two pivotal technologies that will redefine the way trials are conducted, recorded and analyzed in the future of Precision Imaging.

While the “Cloud” is widely speculated to become the universal flattener of clinical workflows, it would behoove us, looking forward, to keep our collective hubris in our back pockets and consider infrastructures and architectures that provide the best clinical functionality at the right geo-spatial and temporal points within the clinical trial workflow.

In my previous posting, I had discussed the Vendor Neutral Archive (VNA). In this post, I present the case that clinical trials need to scale this critical imaging infrastructure component (the VNA) globally as a value-add and integrate it with clinical trials standards like the Clinical Data Interchange Standards Consortium (CDISC) along with the large ecosystem of applications that manage trials.

Halted oncology clinical trials in 2016

2016 was a tough year for oncology clinical trials. A list of the drugs halted by the company (or the FDA) for not meeting endpoints are listed in the table below:

Source: http://clinicaltrials.gov , http://open.fda.gov, data from respective companies

 

Note: The toughest blow to the neuroscience community in 2016 was the halting of the Eli Lilly Alzheimer’s trial.

Legend:

MM: Multiple Myeloma, NSCLC: Non-small Cell Lung Cancer, GBM: Glioblastoma, CLL: Chronic Lymphocytic Leukemia, SLL: Small Lymphocytic Lymphoma, NHL: Non-Hodgkin’s Lymphoma, ALL: Acute Lymphocytic Leukemia, PaC: Pancreatic Cancer, AML: Acute Myeloid Leukemia, MDS: Myelodysplastic Syndrome, mCRPC: Castration Resistant Prostate Cancer

What is a clinical trial endpoint?

For a drug to be clinically meaningful, it needs to have three results: improved survival, detectable benefit by the patient and decreased chances of adverse effects. A primary endpoint is one of these three results which is important to the patient. A surrogate endpoint is a laboratory measure or physical observation meant to be a substitute for a primary endpoint. A biomarker is a characteristic that is objectively measured and evaluated as an indicator of normal biologic processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention. Sometimes endpoints are patient reported called “patient reported outcomes” or PRO. This is the subject for another discussion. A summary of the clinical and biomarker endpoints are illustrated below:

Image source: Atkinson AJ, et al., “Biomarkers and Surrogate Endpoints: Preferred Definitions and Conceptual Framework”, Clinical Pharmacology and Therapeutics, 69:3 (March 2001), 89-95

 

Summary of imaging biomarkers

Continuing my “RNSA Beyond Imaging” report, two of the tracks that I attended at the RSNA annual meeting in Dec 2016 were The Cancer Imaging Archive (TCIA http://cancerimagingarchive.net) and Precision Imaging. I learned about Krippendorff’s Alpha, The Stanford Quantitative Image Feature Pipeline (QIFP) and the Quantitative Imaging Biomarker Alliance (QIBA, pronounced kee-ba).

Imaging improves the scale and accuracy of ‘omics trials by providing temporal, spatial and a non-invasive functional context to the molecular pathways under study. Computed Tomography (CT), Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) are the current basket of biomarkers for Precision Medicine (PM). In my discussion with several Radiologists who are practicing Precision Imaging, MRI and Ultrasound are widely considered as the best modalities for imaging biomarkers (especially with the error rates up to 20% in cancer histopathology). There is some bias (mostly within the Radiology community) that these modalities may also supersede digital pathology since both MRI and Ultrasound are non-invasive and non-ionizing. I am not quite convinced of the bias toward Precision Radiology and will reiterate my previous statement that it would be beneficial to the various Precision Medicine initiatives if the Radiology and Pathology communities along with their quantitative and protocol standards were combined.

Biomarker endpoints for imaging

Imaging biomarkers are now the most explored of the “surrogate endpoints” described above. The keywords in biomarker endpoints for imaging are “Quantitative Validation.” The rise in imaging biomarkers in clinical trials is due to novel machine and deep learning algorithms that can more completely cover the range of quantitative features that can describe tumor heterogeneity, such as texture, shape, or margin gradients or, importantly, different environments (niches) within the tumors.

The case for a global clinical oncology imaging archive

One of the first Big Data laws way back in 1703 by Jakob Bernoulli was the Law of Large Numbers where the same experiment, repeated many times leads away from random results toward the expected. The core of the p-value hypothesis as I learned it: a low p-value only means the experiment needs to be repeated again. Reproducibility is the scourge of science. How do we scale Imaging results across clinical trials and geographies? In October 2016, Cancer Research UK (CRUK) led a multi-organizational consensus statement on building an image biomarker roadmap for cancer studies. The translational progress across quantitative validation and cost effectiveness are shown in the diagram below:

Image source: O’Connor JPB, et al, “Imaging biomarker roadmap for cancer studies”, Nature Reviews | Clinical Oncology, Oct 2016, doi:10.1038/nrclinonc.2016.162

 

There needs to be a single, standards based effort for imaging biomarkers to cross the following translation gaps.

1. to become robust medical research tools, and

2. to be integrated into routine patient care.

This goal is achieved through the three parallel tracks of

· Assay Validation;

· Biological and Clinical Validation; and

· Cost Effectiveness.

Radiogenomics or Radiomics is defined and identified within the imaging phenotype. From a reproducibility perspective, it is much harder to doctor the imaging phenotype than it is to doctor rectangular, tabular data. The source of truth remains within the 256 shades of gray. The same holds true for Pathogenomics.

At population scale, these globally coherent caches of images from local organizations to regional to national archives stitched together with metadata and IDs become the fulcrum of the future of clinical trials. We need an infrastructure to support standards, security and scale.

Notable clinical trials for 2017

Listed below are the notable trials whose data will be available in 2017:

 

I am seeking individuals and organizations within similar wavelengths to discuss this topic further. Please drop me a note if you are interested. Here is looking forward to 2017 becoming both the catalyst and the accelerator of the Global Clinical Imaging Digital Archive!

About the author:

Sanjay Joshi is the Isilon CTO of Healthcare and Life Sciences at the Dell EMC Emerging Technologies Team. Based in Seattle, Sanjay’s 28+ year career has spanned the entire gamut of life-sciences and healthcare from clinical and biotechnology research to healthcare informatics to medical devices. A “skunkworks” engineer, bioengineer and informaticist, he defines himself as a “non-reductionist” with a “systems view” of the world. His current focus is a systems-view of Genomics, Proteomics and Healthcare infrastructures. Sanjay was the recipient of a National Institutes of Health (NIH) Small Business Innovation Research (SBIR) grant and has been a consultant or co-Principal-Investigator on several NIH grants.