Reimagining biopharma R&D and clinical trials in the age of AI

World Clinical Trials Day, May 20, marks the date of the first randomised clinical trial, held aboard a ship in 1747. Today, as part of biopharmacetuical R&D, clinical trials is the area where the spend is the highest and provides the most opportunity for automation and adoption of artificial intelligence (AI). Sowmyanarayan Srinivasan, Managing Director – Life Sciences, Accenture Advanced Technology Centers in India (ATCI) predicts that the future of clinical trials will be fully AI-enabled, starting right from protocol to clinical operations around site and investigator management and eventually pivoting to patients and making it easier for the patients to be part of clinical trials

The global biopharma industry is undergoing significant transformation across the spectrum of drugs, drug delivery and business model. There is an increasing focus on patient at the core. Some of the key drivers of the transformation are –

• End of blockbuster era and beginning of niche buster era. The drugs are getting more personalised and relevant for much smaller populations. This also means that the number of drug launches are increasing
• Focus on new science driving more companion and digital solutions as approved drugs, genomics is driving personalised drugs
• Healthcare consumerisation is being driven by the digital revolution. Patients do not want to be held back by disease and want to be treated more like consumers. They are looking for same experiences as in their daily lives, for example retail
• In North America and Europe, the focus on outcomes has increased and reimbursements are tied to defined and agreed outcomes

According to a recent report (1), while the transformation is driving a change in pricing models and patient engagement, the costs of bringing a drug to the market continues to increase and is projected to be worth $2 billion. This situation has created a perfect storm to reimagine the entire biopharma R&D process which is the main cost contributor in getting the drug to the market.

As biopharma organisations focus on reimagining the R&D process, one of the key enablers of this change is Artificial Intelligence (AI). Though AI has existed for several years (in fact decades), the adoption in the biopharma industry was slow and has increased significantly in the last six to eight years. This adoption can be directly attributed to increased generation and availability of data, ‘big data.’ There has been a tremendous increase in both volume and variety of data – genomics data becoming mainstream and cost effective, imaging data and the ability to store and analyse such data, patient data from digital solutions and patient medical data with higher adoption of EMR and EHR. In this evolution, the bottleneck moved from quantity of data to quality of data and curation, though the fact that data is available for leveraging artificial intelligence techniques is a critical factor for increase in adoption.

Reimagining R&D

If you look at the history (2) of clinical trials, the first clinical trial of a novel therapy was accidentally conducted in 1537 and it took over 200 years (year 1747) to the first controlled clinical trials and another 200 years (year 1946) to the first double-blinded-controlled trials, which has been the norm since that time. It clearly shows us that scientific establishments take a long time to change and modernise processes. However, if you consider technology evolution in the same timeframe, for instance, it has grown by leaps and bounds. Today, we are at a point where there is a need to get medicines to the market faster and there are several technologies that are available to help the overall R&D, specifically clinical trials to leapfrog many challenges.

While clinical trials is an obvious place to start given the disproportionate amount of spends, the future of R&D will not only have a different approach for research, but to safety and regulatory as well. Let us start by looking at potential use cases in R&D and how they can drive transformation in R&D.

Potential use cases in R&D

The first step to reimagine the R&D process is to identify use cases which are ready and mature to be adopted for AI. The primary drivers for this would be quality and quantity of data and predictability of the use case. Some of the most common use cases across research, clinical, safety and regulatory are outlined in the next sections.

Summary view of potential use cases

Research reimagined

Science is evolving rapidly, and it is humanly impossible to track these changes. Research is hypothesis-driven and is amenable to using AI. In fact, use of AI in research has existed for a very long time. The earliest set of databases, be it for compound, target or pathways, used AI and curation to build and deliver at scale.

Some of the use cases in research that are already in play are as follows:
1. Next Generation Sequencing or genetics and genomics data in general – This has been at the heart of big data for biopharma. Requires multiple steps to run the analysis to get to relevant insights around disease target or biomarkers.
2. Pharmacokinetics (PK)/ Pharmacodynamics (PD) Modeling – It has been in existence for very long. There are well-known products in the market that support this and is commonly used and adopted.
3. Target, compound profiling – There are companies that are focussed on building this as a product by leveraging AI and then curating to ensure high-quality data.
4. Pathways – AI plays a key role in extracting entity interactions from publications / other data sources which become the base for pathway building.

There are industry-leading solutions in this area which have been around for long and are well-accepted solutions. The continuing challenge in research has been around bringing data together, drive standardisation and leverage analytics to enable business decisions. The next horizon in research is data platforms that will enable broader AI and analytics to reduce overall cost of lab experiments and support rapid hypothesis testing.

Clinical trials reimagined

Clinical trials and research is the area where the spend is the highest and provides the most opportunity for automation and adoption of AI. Some of the main challenges in clinical trials are around patients, site, investigators and collaboration.
Patient identification and recruitment continues to be a challenge as the number of trials and studies continue to increase. Several of them are in the same therapeutic area with similar disease and target focus. This means that the target population for several companies is the same and success in getting to patients is limited. This is compounded by the fact that less than five per cent of the patients opt to participate in clinical trials.

While recruitment is key, patient engagement is also important. Although the first steps are to identify and recruit patients, the industry also struggles to retain patients for the duration of the trials. Patient drop out is a big ongoing challenge in clinical trials. On the other side, site or investigator selection is a critical part of trials success. As drugs become more specialised and targeted, the expertise for managing trials have also increased. Finally, lack of clinical stakeholder collaboration is also a challenge. It is important to bring together the various stakeholders on a common platform which is a critical success factor for trials completion on time. There are several opportunities that are fit for driving transformation in clinical trials.

The clinical process starts with a protocol document which outlines the approach and intent of the clinical study. The key content of this protocol document then needs to be used to set up an electronic data capture system that will help capture defined data of patients as outlined in the protocol during the various visits to the site. Today, the process of setting up the electronic data capture system from the protocol takes anywhere between 12-16 weeks since the process is completely manual. The data entry and quality-check iteration time take the bulk of this total time. The future of this process will be AI-driven, where AI will learn from past protocols and provide suggestions as new protocols are drafted. It will support data extraction from the protocol to load into the electronic data capture system and increase productivity by 80-90 per cent which will provide significant competitive advantage.

Today, there are some opportunities especially in the clinical operations area that have adopted AI to varying degree while there are several others that are ripe for disruption. Identifying the right patients and retaining them along with right sites and investigators is predominantly AI-driven and will continue to be dominated by AI, given the large pool of data available.

• Patient recruitment helps identify the right set of patients for a given study by using data across past studies, EHR/EMR, other data sources like patient communities based on data access
• Patient adherence and drop out is based on patient behaviours like medication adherence, diet, fitness regimen, ongoing health record etc., predict patients that are at risk of not complying with the requirements of the clinical trial or dropping out of it completely
• Site selection helps profile sites based on therapeutic area across various parameters like ability to recruit patients on time, data quality, training etc.
• Investigator selection/profiling helps profile investigator performance specific to therapeutic area leveraging proprietary data, commercial data and public data like publications/clinical trials disclosure sites etc.

In addition, Automated Clinical Study Report (CSR) generation, AI/ML-enabled CSR generation based on available data, are some of the other areas to benefit from AI. As one can clearly predict, the future of clinical will be fully AI-enabled starting right from protocol to clinical operations around site and investigator management and eventually pivoting to patients and making it easier for the patients to be part of clinical trials.

Some of the start-ups that are seeing success in driving adoption of AI in clinical include:

• Deep 6 AI:- Trial matching software can find eligible patients for complex trials within minutes
• Viz.ai: It helps analyse CT scans to identify potential stroke.
• AiCure: It helps improve medication adherence by visually confirming medication ingestion.
• Trials.ai: In its first trial, AI retained 98 per cent of patients, had one critical deviation throughout the entire trial, and continued to completion without interruptions.
• Several biopharma companies are focussed on setting up control room like centres for monitoring clinical operations and interacting with sites real time.

Patient safety reimagined

As the pharma industry pivots to patients, their safety has become the key differentiator to drive success in the market. An increased focus on safety implies an additional cost burden which adds to the ever-increasing cost of bringing drugs to the market. Pharmacovigilance (PV) is the process of monitoring/managing drug/patient safety.

For instance, focus on drug safety is being driven by many parameters. As patients become more aware, there has been a surge in safety data volumes and the number of reported cases is increasing steadily. The regulatory authorities across the world are becoming more stringent and the regulations continue to evolve, which in turn again drives increase in reported events. Further, with digitisation taking center stage, there is an increased proliferation of social media usage and patients/consumers reporting potential adverse events in these forums, thereby increasing the burden on biopharma companies.

These trends together put an undue pressure on biopharma companies to increase their spends on adverse event processing. However, AI has started to play a crucial role leading to several use cases , where it has showcased  reduction in time and effort required  to process adverse events.

Below are some of the key use cases in safety which are being adopted by biopharma companies globally:

• Literature scanning to support the mandatory requirement to scan literature in regular intervals to look for adverse events. AI makes this search simpler and easier.
• Proactive signal analytics to continuously look for potential signals which can cause significant harm to patients and proactively address by label changes or changing target patient population, etc.
• Case intake and processing leveraging AI for case intake across both structured forms like CIOMS or unstructured sources like publications, patient support programmes etc. This will be the game changer.
• PV social listening to monitor social/digital media, AI can help monitor for adverse events. While the guidelines are broad, this is a best practice that can at least help alert for any real adverse events in the social forums.

The future of PV will have an integrated AI approach – today an adverse event processing can take place anywhere between several hours to days, and this can be reduced dramatically to a few to several hours by leveraging AI for case intake and processing, with humans reviewing the results and approving the submission. Increasing data will mean better signal detection and improved algorithms will help drive better outcomes.

Regulatory reimagined

Regulatory was the last to start adopting technology solutions and start the transformation journey from manual processes, emails and excels to newly evolving focussed platform solutions. As the market becomes hyper competitive and crowded in specific therapeutic areas, it is extremely critical to get the submissions right the first time. The key challenges that need to be addressed to achieve this mission are keeping track of changing regulatory requirements which continues to evolve – as newer forms of drugs hit the market, new science resulting in several companion digital solutions with additional requirements, collaboration and coordination with local markets for region specific requirements and the inability to qualify submissions based on past experiences and interactions with regulatory authorities.

The challenges nicely lead to several opportunities where AI can play a role in solving them. Here are few opportunities that can directly address immediate needs:

• An AI-driven regulatory intelligence engine tracks changes in regulatory requirements. It is designed as an intelligent engine to track changes in submission requirements for regulatory agencies with digital presence. It will alert significant changes to avoid delays during the submission process.
• One of the significant and critical AI solution is Health Authority Interaction Mining/Intelligent Submissions. This intelligent engine will mine past health authority interactions to enable response to new queries. This can also be forward looking to act as a filter to avoid any aspects that may lead to queries from health authorities.
• When it comes to actual submissions, the AI engine that will drive the future will be automated regulatory submission for dossier compilation. This will be an AI/ML enabled data and document collection based on the specific country/geography specific regulations and preparing a dossier ready for submission and in some cases e-submission to select authorities. This will enable automated conversion of unstructured data/documents to a structured format, based on pre-defined configurations per country/geography specific regulations and AI/ML-enabled data extraction and data entry into regulatory portal for e-submission.
• Finally, an automated label creation solution based on AI/ML-enabled label creation based on master specifications pre-configured on brand category per regulations of country/geography.

This holistic regulatory solution suite based on AI will dramatically improve the functioning of regulatory team and help them focus on the launch planning with laser focus. The solution suite will provide biopharma companies with competitive advantage and help them get to market faster.

Getting ready for NOW

Biopharma companies are in the middle of a major change and one of the important drivers for success would be people. The future is bringing AI and people together to derive key insights and drive rapid decision making. The success of biopharma companies will depend on getting their people ready for the present as AI becomes mainstream. While AI can bring the art of the possible, the decision-making will still be driven by humans. Training and upskilling the workforce will be a major focus in the immediate future to ensure we are ready to deliver in the reimagined world of biopharma R&D. Some repetitive tasks will be eliminated, but more complex decisions with new science will be there where the humans will have a major role to play.

How to get started

It is evident that AI has become mainstream and will have a major role to play in reimagining the biopharma R&D process. With increasing pressure on cost, need to establish outcomes-based models, shift from volume to value, and most importantly putting the patient at the core, it is obvious that biopharma companies will need a dramatic change to the current processes to drive innovation and stay competitive. With increasing availability of data, data-driven approaches will take center stage and companies that will succeed are the ones that can get the most out of this data and AI will be at the heart of driving this transformation.

The very first step in getting started is to do an internal assessment on the current state across the R&D functions. The output of the assessment should be a maturity index for each of the R&D function in terms of leading or lagging. A blueprint should then be created for each of the functions on way forward with clearly outlined success criteria. The blueprint should call out specific use cases, technology solutions, data access and veracity and anticipated challenges. The use cases outlined in this point of view are a good starting point to assess. Finally, translating the blueprint into an execution roadmap and following through to closure will be critical in getting ready for the future and positioning for growth and success.

References
1. https://www.scientificamerican.com/article/cost-to-develop-new-pharmaceutical-drug-now-exceeds-2-5b/
2.https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3149409/