Journal for Clinical Studies shares AMPLEXOR's view on Clinical Precision: What does AI Offer Life Sciences?

18 September 2017

Where information overload has started to slow innovation and efficiency in medicinal markets, could artificial intelligence and machine learning help – by pinpointing what’s important and suggesting better ways of doing things. AMPLEXOR’s Elvis Paćelat ponders the bigger picture in the Journal for Clinical Studies.

It would be easy to assume that the more adventurous end of technological innovation belongs to markets other than life sciences. To those with more direct contact with the public, and which are less bound by red tape – from retail to hospitality and travel, even banking. Yet ignoring the potential of digital disruption could put companies on the back foot as their own market starts to transform itself around them.

Just look at the impact artificial intelligence (AI) and machine learning are expected to have on healthcare, in accelerating and transforming patient diagnoses, for instance. Investments in healthcare-focused AI start-ups have more than tripled in recent years. In life sciences, the opportunities are not dissimilar: start- ups are already using machine learning algorithms to reduce drug discovery times.

AI promises to make clinical trials cheaper, faster and more targeted. Last November, British AI startup, BenevolentAI, announced that its technology – which offers to speed up late-stage development of drugs and provide richer clinical data – would be tested under exclusive licence for a series of novel clinical stage drug candidates with Janssen Pharmaceutica, part of Johnson & Johnson, this year.

In an age of data overload, AI offers a way to find and keep teams focused on what’s important – from what’s being said in the market, to how drugs are designed and developed.

AI's Role
So what is it that AI does differently, and how broad is its potential in life sciences?

AI takes automation and makes it smart. Where robots in factories excelled at doing repetitive, mundane tasks efficiently and tirelessly, with precision, AI can be programmed to carry out more complex tasks (e.g. robotics taking measurements in hostile environments which would be too risky for humans to go into, and where there are lots of variables).

Machine learning improves upon that, allowing AI-based systems to find better ways of doing things5. Rather than humans having to foresee every possibility and programme a system for every eventuality, AI-based systems can learn and adapt from what they know to create effective and powerful shortcuts – as though they are ‘thinking’ and problem-solving using their own intelligence and reasoning. (What they’re really using are complex algorithms - or clever maths.)

If they’re faced with a deluge and range of data that would tie up a human team for days or months, machine learning systems can perform analyses and distil subtle trends that humans might overlook. In the context of pharmacovigilance, they can help scour the internet for relevant patient feedback about life sciences products, or identify unmet needs or gaps in the market. Operationally, such tools can help companies navigate routine processes more promptly, thoroughly and economically, freeing up teams to use their skills where they will add greater value.

file

 

Cutting Through Compliance Demands

An obvious area where AI and machine learning can help here is in managing matters of regulatory compliance – where requirements are multiplying and changing all the time. Not only does this increase the burden on regulatory affairs and quality teams; it also potentially slows companies’ time to market. Moves towards international standards, and deployment of sophisticated content management systems, go a long way towards alleviating the additional work involved and maintaining data quality. Yet, with each new regulatory initiative or submissions hoop that companies need to jump through, the business agility and creativity they are aiming for appears to become further out of reach.

In everyday life, AI has begun to transform the way people interact with diverse information and achieve end results. Over breakfast, thanks to AI-enabled ‘personal assistants’, they can check their various message sources, get a precis of the news, peruse their diary, search for travel options and make bookings, without touching a keypad or mouse. They simply speak their requests to a voice-enabled user interface (Siri, Alexa, Google Assistant or Cortana), and an AI-enabled ‘assistant’ does the rest – instantly interacting with all the different applications and performing the various analyses and transactions, returning its results before the user has taken a second bite of their toast.

It may feel like a leap now, but there is no reason to suppose regulatory affairs teams couldn’t enjoy a similarly unencumbered user experience when managing health authority submissions. The ideal is that their product lifecycle content systems will make it more intuitive to manage data changes, document authoring and reviews, quality control, and submission. Currently much of this is managed via comprehensive rules, templates and workflow which help to streamline processes and ensure that the right data is used in support of the given requirement. But what if AI and machine learning could promote reliable shortcuts, and issue red flags or suggestions if rogue actions are taken, the wrong master data is used, or someone tries to alter approved ISO IDMP-compliant source content?

René Kasan, a visionary speaker from IT consultancy NNIT, outlined his own personal take on AI’s potential as part of regulatory information management at AMPLEXOR’s recent annual conference. He explored how it might help transform companies’ ability to consume and harness vast volumes of complex data, and unprecedented inter-connections between different data sources and systems – without compromising the integrity of the content, and with significant benefits to speed and reliability. And without time limits, because AI doesn’t get tired or slow down.

As well as freeing up skilled people’s time to do more satisfying and productive work, AI could also reduce the risk of dependence on a single person’s knowledge of how things are done. (If a highly skilled team member moves on, there is usually a productivity gap as their replacement gets up to speed.) It’s easy to see how companies would benefit: it is getting harder to attract experienced skills for important regulatory roles6 as demand increases, yet the pressures of the job take their toll.

Could  Alexa  Automate  Regulatory  Information  Preparation?

Life sciences firms are already harnessing more automation to streamline regulatory information processes: annual surveys by Gens & Associates repeatedly show increasing sophistication in the industry’s approach to regulatory information management. Although exploiting AI in particular is something that is only now appearing on their radar, AI and machine learning have become hot topics, and events and conference sessions on the theme are well attended.

We can speculate about a number of ways AI could transform regulatory information and submissions management transformation – improving the process of planning, structuring, authoring, publishing and archiving content.

Example scenarios might include using AI to monitor and determine which content elements of a submission are routinely included, so that they become a structural component in their architecture. AI could also help ensure referential integrity, so that the correct, approved master content is reliably drawn on every time, and that protected sources (e.g. ISO IDMP data) cannot be tampered with, without a formal change request.

Document and dossier authoring and reviews could be streamlined as AI capabilities learn to spot content that has been changed frequently in the past. Drawing on this knowledge, the system could propose changes as a document is being put together, saving rounds of redrafting. Alternatively, the user could ‘ask’ the system what the implications would be if they made a particular change to content, and have the Alexa-style voice interface list all the ramifications. An intuitive, user-friendly interface combined with smart, machine-based deductions could save a lot of clicks, system navigation and time.

AI could reinforce compliance and content quality along the supply chain, too, helping to restrict what country affiliate representatives are able to do with content. Where there have been quality violations, AI could provide the analysis and insight so teams can act and prevent repeated issues. Related to this, the technology could help identify and avoid common submission queries, to prevent delays in getting products approved. Similarly, it might support strategic decisions about which health authorities/ markets to target first.

file

 

Crossing the Digital Divide

As organisations start to make the connection and understand the role artificial intelligence could have in transforming their operations, new opportunities may begin to present themselves – for example, the potential to improve R&D by sharing ‘learning’ back along the product lifecycle.

In the meantime, there are associated issues  companies will need to consider. For example, how might AI use affect the compliance processes, especially if checks have previously relied on conditions remaining static (rather than continuously tweaked and adjusted, as AI finds scope for improvement)? If the AI capability is becoming an additional reviewer or actor in the workflow associated with creating regulated or regulatory content, how does this change the auditing requirements and where does responsibility lie?

Finally, if knowledge management and retention comes to rely more heavily on machines rather than skilled individuals, what is the migration plan if technology moves on?

None of these details are a reason to dismiss AI’s potential; they merely need to be factored into the design plan. It’s all part of understanding the technology’s potential in the context of the challenges life sciences organisations are trying to overcome – and at this stage nothing should be ruled out.

Returning to the example of BenevolentAI, the British AI startup has bold ambitions for the way drugs are developed and brought to market – and it plans to do this itself, within as short a timeframe as four years7. It envisages a world where, thanks to AI’s ability to cut to the chase, it is possible to provide “first-in-class and best-in-class stratified medicines to help patients with high unmet needs”. At a recent conference in London, the company hosted a session exploring how digital, data and technology developments are disrupting clinical trials and the big leap the life sciences industry now needs to make if it is to prepare itself for a brave new future.

The drivers for change are very clear – from the inefficient management and strategic use of operational data, to the high cost of developing medicines and preparing them for market. Now all companies need to do is clear the way for a new approach. Because if they don’t, someone else will.

About the author

Elvis Paćelat
Executive Vice President, Life Sciences, AMPLEXOR

Elvis is a business and technology executive with more than two decades of international experience in the life sciences market. With detailed technical understanding and expertise in compliance and regulatory content management solutions for Life Sciences, Elvis is a specialist in business impact analysis. At AMPLEXOR, he is responsible for driving the corporate strategy and market success of the AMPLEXOR Life Sciences business. Elvis is committed to delivering benefit for clients, partners and shareholders, whilst supporting client-centric strategies and spearheading groundbreaking innovations.

From our blog