Applied Clinical Trials
Benjamin Hughes, SVP of AI and Real-World Data at IQVIA, discusses how AI is changing the clinical trial paradigm, and how the technologies IQVIA is developing are different from other tech companies that specialize in AI/ML
Artificial Intelligence (AI) and Machine Learning (ML) have been an area of focus for many big pharma companies. Pfizer, for instance, focused on leveraging AI for drug discovery. Now, many companies are seeking improvements in drug discovery as well as clinical operations; hence, large CROs are stepping in by offering AI/ML solutions to both emerging biopharmaceutical companies, as well as Big Pharma. In this article, Benjamin Hughes, SVP of AI and Real-World Data at IQVIA, discusses how AI is changing the clinical trial paradigm and provides some guidance on where AI/ML is heading.
Moe Alsumidaie: Can you tell me how AI/ML is changing the drug development landscape?
Benjamin Hughes: AI/MLis a series of techniques that leverage machine learning, natural language processing-NLP-or other techniques that help create new insights, drive operational efficiencies and accelerate decision-making in new ways.
In drug development, AI is likely to be felt in three main areas: discovery and more reliable targets, better clinical trial design and planning, and more efficient execution from recruitment through monitoring.
Pharmaceutical companies need to be looking at all three of these opportunity areas and optimizing for them across the drug development process. More efficient trial design and planning are the most actionable or potentially impactful. It can help identify sites with the most patients to drive faster trials and estimate the outcomes of particular trial designs, which is exciting. It is important to note that this requires widespread access to global data, which means not everyone can operate at this scale.
At the execution phase, there are many things you can do with AI, from automating clinical management and monitoring or automation in eTMF, that help drive quality and don’t require access to vast data.
From a drug discovery perspective, AI/ML capabilities show great promise, which is essential for investment funding and startups. Unlike other areas of development, it doesn't require as large of an upfront investment, so you are seeing hundreds of startups applying AI/ML and deep learning models to things like publicly available data sets, proteomics and metabolomics. The impact on drug development is yet to be fully proven.
MA: How has AI/ML benefited clinical development specifically?BH:At IQVIA, we’ve had remarkable results applying AI techniques to clinical design and execution. The AI engine and methods that we developed are helping to optimize trials, what we call our CORE-enabled trials, using data analysis for critical-step change and improvement. We've done a number of these trials with clients, and the benefits are mostly around accelerating the clinical process. Where we're using these techniques, we see about a 40 percent faster site identification process and 30 percent faster recruitment rate. That has enormous benefits and cost-saving implications for clients. Other benefits can be seen in clinical monitoring with improvements in quality management and data capture.
MA:How are AI/ML services at CROs different from AI/ML technology companies that are focusing on R&D?
BH:There are significant differences between the CROs, AI/ML companies focusing on R&D and the large tech players.
You have some CROs that are much more traditionally positioned and others that are a blend between CROs and tech and data companies. Probably the most significant difference in approaching the opportunity is how much data does a CRO have to develop their AI capabilities. For those that don’t have a considerable amount of data to lever, they tend to focus their AI development on select use cases. At IQVIA, we are applying AI more broadly to many different use cases because we have more comprehensive data access and leverage for more opportunities.
Regarding tech companies specially focused on R&D, there are probably two main types -those in discovery and those trying to lever software to optimize the end-to-end clinical development process. There are now hundreds of startups in the discovery space levering various types of biodata and public literature, as well as techniques in NLP, Graph, and machine and deep learning to optimize target identification. And then there are a smaller number of players looking at trial orchestration and AI across the clinical development process.
Large tech companies can drive AI capabilities that are not specific to healthcare, even if they have some healthcare-specific activities. They are building excellent generic AI capabilities but don't necessarily have the domain expertise to optimize applications across a series of use cases.
AI isn't just about data and technology. It’s about change management and deployment of processes and decision-making to create and change processes and how people work. It is an iterative process, working on the performance of the AI or its position within that business process to create more value. Bottom line, there are tremendous opportunities for healthcare organizations to lead applied AI beyond what we see with some of the large tech companies.
MA:How are the technologies IQVIA is developing different from some of the other tech companies that specialize in AI/ML?BH:Large tech companies are making advances in generic AI frameworks, and these are generally available to organizations like ourselves. We then have the ability to implement, improve, and customize them within specific business processes and decision points to create significant value for our clients. We bring the domain expertise within the clinical development space to make that progress possible.
Compared to companies focused explicitly on R&D, there are three main differences in the IQVIA approach. First, we are levering a portfolio of data assets for our modeling that is an order of magnitude larger than any other actor, which allows us to build things like site ID AI models at a global scale. Second, we build our approaches for leverage across the healthcare sector, supporting R&D but also commercial and healthcare stakeholders, such as providers. Third, we build the solutions in many forms, embedded in services, as stand-alone consultative projects, as software or backed into our overall infrastructure and the IQVIA Core™. Our dedicated AI teams have exposure to all these different contexts and, thus, have greater experience than companies specialized in one area.
MA: Is the demand for AI/ML coming from emerging biopharmaceutical companies (EBPs) or Big Pharma? And if you're targeting both, how can you help from a services portfolio standpoint?BH:We haven’t observed an enormous difference in market appetite for AI between large pharma and EBPs. All of them are exploring it. The nature of the work that we may do for them and how they’re going about sourcing their capabilities does differ. EBPs are working in more targeted disease areas and often need more rapid decision-making. This then lends itself potentially to use of AI in helping to locate patients for challenging trials, such as rare diseases. Big Pharma can apply those techniques as well, and we are seeing them do this across R&D drug development and commercial. We also see Big Pharma build in more systematic capabilities in house, even if they still recognize that there are some AI capabilities that are best sourced externally.
An EBP client is much more likely to commission AI and predictions around clinical trials as a full-service offering and incorporate that within the context of their goals in clinical development. Big Pharma would ask us to work within other initiatives and infrastructures they're trying to develop. In some cases, they may have their own data analytics hubs and their own AI initiatives and then ask us to work with them to apply our techniques to their infrastructure to improve their overall capabilities. The use cases are very similar; however, Big Pharma wants us to train and enable and do a lot more knowledge transfer in AI than EBPs do.
MA: Where do you see the most significant opportunities for disruption and innovation in clinical trials?BH: The biggest area of opportunity may be using AI to model diseases and treatment response. By leveraging extensive historic patient or real-world data, we are already able to get a better sense as to which specific patient populations to target for a protocol and their likely positive response or the potential number of adverse events. This will increase trial success and make participating in clinical research a more positive experience for patients.