Reuters Events | Pharma

Thought leadership and innovation for the Pharmaceutical Industry

 

Medicines only work well in a fraction of their target populations. Many clinical trials perform poorly or fail. 

 

But a grand, cross-industry project to gather and harness data could change all this.

 

The creation of the ‘Digitome’, a data commons aggregating structured and unstructured patient data as well as population, claims and trial data, would be fuel for many potential new applications.

 

Harnessing this wealth of data, mapping it and putting it into a data commons for use in different applications would enable innovators to create a series of applications for disease detection, clinical trials and healthcare management.

 

Such an initiative would be far more ambitious in its scope and conception than existing initiatives, says Jacob LaPorte, Co-Founder, Novartis Biome. “All of us can point to scores of initiatives out there around a single diagnostic where we’re trying to detect disease earlier. The problem, as we all know, is that many of these things are sitting in silos.” 

 

Transforming trials

One silo problem the Digitome could help solve is the lack of accessible trial design information from existing and past studies for trial teams when developing new ones, says Mohammed Ali, Global Head Digital Development, Boeringer Ingelheim. 

 

“Identifying sources from past studies and ensuring that all of this is harmonised and standardised in a location where you can access it will give you the best intelligence and outcome for writing your study protocols and designing the overall clinical trial design.”

 

Digitome-powered data analysis need not just stop at that level either, says Ali. “We should also compare at the industry level what it means to standardise disease trial design across the board. So looking at white papers, looking at lessons learned through other engagements or other forums and making sure that is captured too.”

 

Taking such a strategic approach would help identify which points in a protocol are driving progress and which have prevented it in the past, says Ali. “If you do it right, you should be able to get a protocol that is much more efficacious and acceptable by your external regulatory bodies and key stakeholders, saving time, reducing amendments and reducing costs.

 

“It may not seem like a lot but if you do hundreds of protocols a year and you’re running them altogether concurrently, that time can actually add up significantly and you’re able to accelerate your overall pipeline process.”

 

The mass availability of clinical trial data has many other potential applications. It could, for example, also help form heat map analyses of where patients live, enabling targeted outreach to the best local sites, adds Ali.

 

Accelerating research

The Digitome has obvious applications in rare disease too, where it’s hard to find patients, and where data and time for research is scarce, says Adama Ibrahim, Associate Director, POC, Global Clinical Ops, Biogen. “There is a category of diseases that will benefit from actually accessing existing data repositories that are aggregated into a meaningful context. 

 

“For example, some data repositories contain genomic information or EMR information that help create a full picture for a particular clinical trial. The technology already exists to do this if we are able to work differently, think differently and start capturing patient data in non-traditional ways. These can be reused and repurposed multiple times to accelerate the research activity.

 

“We have been doing pilots and proof of concepts to target some of those data sources and I think we're at the point where it is important to create a global ecosystem because for rare disease we cannot be localised in one region or one country.”

 

There is also the opportunity to use data, says Ibrahim, “in a prospective and retrospective way that creates a better picture for new indications, especially around indications where control is challenging. 

 

“You can start to isolate some of the control data and the control groups and have a dialogue with regulators so that as you progress into the advanced phases of your trials, you can leverage some of the data that is cleaned up and relevant to that particular asset.”

 

Baby steps

It won’t be necessary for the Digitome to be fully formed for pharma to begin realise some of its potential benefits. Taking some steps along the way to pooling data at scale will realise improvements and efficiencies in trials.

 

For example, conventional clinical trials could be digitised to a much greater degree, says Michelle Longmire, CEO and co-founder of Medable. “When you look at the phases of clinical drug development from trial design, to feasibility, to recruitment, enrolment and then study conducts, these exist in discrete processes that are largely done by human-first approaches.

 

“A digital-first approach generates synergies across those processes. For example, implementing a data commons across clinical trials can help identify what worked and what didn’t work, insights that will drive better trial design. 

 

“From there, ideally we’re piping in real data from patients that are recruitable, engaged and interested in clinical trials. We map data sets from these patients and to that digitally optimized trial design, while we’re also connected to [trial] sites and we know the investigator qualifications and what patients are actually being seen in these healthcare ecosystems. 

 

“Clearly we’re very far away from that but I believe that as we advance the use of technology in these processes, we’ll get much closer to realising this Digitome concept because we’ll be utilising data in new ways, creating new synergies across those processes.”

 

There is also scope to blend typical clinical research with real-world activities that are generating health data. In the near future, data from a multitude of sensors in smart devices, will help make comparisons across people’s different genomes to create nuanced phenotypic information, driving deeper understanding of health outcomes over time, says LaPorte.

 

“You can imagine patients generating very important data that’s being captured in a very systematic and structured way. This would give us much better insight on how they're performing across different therapy regimens, the ability to change therapeutic courses and to experiment more rapidly in the real world.

 

“A lot of us are nervous about crossover experimentation but if you think about it, this actually goes on a lot in the real world, where there’s an uncertainty around how treatments are performing. All of this needs to be captured in a meaningful way and used to give more meaningful data back to our research community so that we can make better decisions about treatments.

 

“The problem is right now a lot of the data that’s being generated is either not systematically captured or, if it is captured, it’s not captured in a way that’s analyzable across the broader research community.”

 

Interoperable, reproducible

Opening up access to different types of data would enrich the Digitome’s overall diversity, allowing access to minorities and people who are not currently participating in trials. 

 

To fully realise the potential of the Digitome, then, the healthcare community needs to put much more time and attention into making data interoperable, says LaPorte.

 

“Not only will the data ecosystem need to contain the right, correctly structured data, it also needs the owners of that data to repurpose it for going into new therapeutic areas. We need to create an engaged community that’s really talking about the standards we need to bring to the table and join together what may have already been developed in order to make this work.

 

“This community needs to be much broader than industry. It needs to be healthcare professionals, payers and ultimately we hope it goes way beyond any independent stakeholders biases and motives.”

 

The tech giants may also have a role to play in driving progress, suggests Ali. “The acquisitive moves by Google, Amazon and Apple into healthcare may well help the drive to standardise the data they’re capturing, which could prove useful for overall trial designs and in primary or secondary endpoints.”

 

But the tools and systems that emerge to harness data must be open for use by all. Making machine learning (ML) tools open source is important for establishing a testable framework that different scientists can use to demonstrate reproducibility on the AI and so the validity of the findings from digital data for diagnosis and for disease detection.

 

Without such capabilities, regulators will not be comfortable if they lack clarity about the nature of deep-learning AI driven by a ‘black box’ neural network that supports diagnosis decisions, says Longmire. “Being able to create algorithms across those datasets that other people can really test and validate [is important]. There will continue to be models that don’t have full clarity around some of their underlying logic. Reproducibility is one way to really show the validity of the model.”

 

Harnessing AI

Done right though, the advent of a Digitome will be the means of unlocking the promise of AI and ML. “The way you make AI work well is really good data training sets,” says LaPorte. “The problem with healthcare data is that there are very few data sets that we trust to be very good training sets for the machine learning algorithms. 

 

“That’s why you’re seeing very narrow applications of machine learning right now like diagnosing pathology reports, but that doesn’t necessarily translate into these broader applications that we really want to get to. So by creating a data commons and interoperability, I think it will power a new set of machine learning applications that will have significant impact on healthcare and research.”

 

If such challenges can be overcome, the benefits of the Digitome could go far beyond improving trials and will present whole new opportunities as well as threats for ‘legacy’ pharma. It could advance healthcare, taking it beyond the existing paradigm of intervention at disease onset or progression. It will, to a far greater degree, enable prevention, says Longmire. 

 

“The onset of disease really starts with a deviation in homeostasis and that what we call the symptoms of today probably manifest far earlier with deviations in homeostasis and baseline changes that could be detected through a new era of diagnostics.”

 

Once detected “a multi-faceted intervention with medications and behaviour changes that target deviations in homeostasis prior to the typical phenotypic manifestation of disease” would be possible, says Longmire.

 

“By collecting the data from our everyday lives and by bringing intelligence to this, we can detect disease at a far earlier stage and even open the door to interventions that could reestablish homeostasis precluding the full development of disease.”