At a recent Bioinformatics Strategy Meeting in Boston, I chaired an interesting roundtable discussion on the potential convergence of translational research and clinical development.
The usual culprits within the data life cycle were discussed such as effective data integration, large enough (but optimal) datasets, data harmonization, deep learning versus machine learning, but also how the ultimate clinical utility of a biomarker would depend on how “mechanistically understandable” a marker was from a clinician’s point of view.
One of the most interesting thing that kept on cropping up was how can we apply the learnings from Clinical data management to Translational data management.
Of course clinical data managers have seen it all, receiving data from different CROs for different clinical domains in entirely different formats, harmonizing it into data standards and integrating cross-trial data is a challenge they are acutely aware of. Throwing Translational data into the mix to improve Clinical trials whether to identify a drug stratifying biomarker or a drug response biomarker adds further complexity to the types of data being collected for clinical trials.
Certainly, according to ICH-GCP guidelines, sponsors are the people who ultimately have to take the responsibility of harmonizing and integrating these complex data, but what sort of additional harmonizing standards should the CROs be subjected to when producing these data (if at all!) ? Of course, there are data-focused CROs who offer extensive services in clinical data management- but are these CROs really ready for dealing with the complexity that Translational data may bring to the clinical trial process?
The CDISC organization has been working tirelessly to provide standards that help integrate pharmacogenomics with clinical data, and this will allow some structure to be put on these datasets. However, with the technologies evolving at such a fast rate - and producing larger and more complex data- we might need turn to other strategies such as machine learning or other AI approaches to allow for faster data harmonization across various datasets.
Ultimately we want to integrate translational data with clinical data to help us:
Improve trial design by selecting the right subjects
- Identify novel applications and therapeutic areas for approved compounds
However, our classical challenges related to a clinical trial data still remain the same:
- How do we avoid missing a safety signal?
- How can we speed our time to submission?
We live in the fastest moving technological innovations of our time, and they bring a lot of solutions and a whole lot of challenges. The roundtable discussion ultimately came up with more questions than answers, but there was consensus on if we are to make the best use of novel innovations to address data challenges than previous learnings on clinical data management are essential to help us navigate these highly interesting but unequivocally complex ‘data-dependent’ times.
Want to bring together all translational and clinical data into a single platform, then click here
Published on Linkedin on December 12, 2018