Artificial intelligence (AI) is increasingly being used in the pharmaceutical and medical device industries, where it promises to improve the efficiency of product development and provide innovative solutions for extending patient lives. However, this emerging field will challenge the existing regulatory system. Therefore, stakeholders must work together to ensure the smooth development of the regulatory system to adapt to the changes brought about by AI.
Four opportunities: AI is expected to improve medical product development and life-cycle management
Opportunity 1: Use AI tools to evaluate inclusion/exclusion criteria for clinical trials
In clinical trials, AI can be used to evaluate inclusion and exclusion criteria related to imaging or histopathology. Such applications are to be expected as the first diagnostic tools using artificial intelligence are already on the market. With these tools, the process of assessing existing inclusion/exclusion criteria will become faster and cost less as standardization increases.
AI tools are particularly important for low – to middle-income countries. When biological samples, such as blood or tissue, are needed to diagnose diseases, these countries often lack national experts to assess the biological samples. AI tools can effectively simplify this process, helping researchers evaluate samples locally, without the need for complex and time-consuming cross-country transportation.
Opportunity 2: Using AI to identify clinical activity in a Phase II clinical trial
Using AI to evaluate the clinical efficacy of new drugs can reduce costs, accelerate clinical development, and bring new therapies to patients early, such as evaluating imaging endpoints of CT or MRI scans in phase II trials. Ai-based algorithms can optimize the reading and evaluation of imaging results, reducing variability between and within readers, thus improving the sensitivity and specificity of measurements. If a radiologist is no longer needed for this work, it can effectively speed up the measurement process and reduce costs.
Another application is the development of new clinical trial endpoints, as AI algorithms can help reduce the number of patients in trials. For example, people with Parkinson’s disease could wear accelerometers on their wrists, like fitness trackers. The accelerometer will provide continuous data ON a patient’s movement disorders and their changes over time, which will then be evaluated by an artificial intelligence algorithm to distinguish whether a patient is in the ON or OFF state, thus recording whether the medication is making a difference. This assessment method can significantly reduce variability when compared to patient diaries or the Parkinson’s Disease Comprehensive Score Scale (UPDRS), which cannot measure the exact time of ON and OFF status. If clinical endpoints are identified, reduced variability may help to recruit fewer stage II patients to determine the therapeutic efficacy of a new drug.
The researchers expect this technological advance to have the greatest impact on phase II clinical trials, which require enough patients to accurately evaluate the safety of new products and to validate phase II clinical trial results in a larger sample size. In addition, a substantial validation process is required before any new clinical endpoint can be used as a routine alternative endpoint to demonstrate clinical benefit.
Opportunity 3: Extracting data from unstructured text
We can get valuable information from unstructured text from health boards, healthcare companies, and the Internet. This includes relatively complex information about intelligent regulation, for example, but also simple data that can be easily evaluated by researchers once extracted and transferred to a database.
New tools for text mining using natural language processing (NLP) have opened up new possibilities for extracting information and data from documents and then automatically uploading them to databases for analysis. Ai-based tools are available to identify drug products (IDMP)(such as substance name or strength) by extracting data from unstructured text (such as a summary of product characteristics)(see Figure 2).
Text mining tools enable health authorities and pharmaceutical companies to better produce documentation and guidance on chemical composition production and control (CMC). These tools help health authorities evaluate documentation across different applications and marketing mandates. Such as finding the same chemical impurity in a product during production or finding a specific raw material to use to make a new biological entity. This will help the health department’s reviewers improve their decision-making, while at the same time helping pharmaceutical companies automatically extract information from the health authority’s rules and feed it into smart regulatory systems. Both of these tasks require NLP software that understands CMC documentation. The software requires access to large amounts of data in order to achieve the desired results quickly, efficiently and with high quality that will bring maximum benefit to health authorities as well as industry stakeholders.
Opportunity 4: Automate administrative work
Health authorities and healthcare workers manage a large amount of administrative work, and robotic process automation (RPA) and machine learning (ML) can help reduce their workload. In the EU, for example, about 400 full-time employees are employed by authorities and industry to manage type IA variants, according to a review of the Organisation for Regulatory Excellence (ROG). At the AI Alliance conference, participants discussed how AI/RPA could help automate the treatment of type IA variants, provided that companies could implement them without authorization, but would need to inform health authorities at a specific time.
One application of AI in this area is to intelligentially extract information from scanned documents, such as copies of registration certificates or trade registers, and use the “SPOR” standard to transfer this information to databases, including entity, product, organization, and reference data (see Figure 3). This technique has been used for automatic processing of invoices, where the data on the invoices can be extracted into an ERP system.
The application of artificial intelligence technology in the medical field is both an opportunity and a challenge. No matter the regulators or the industry, countries are not fully prepared to meet this new thing and move forward in the exploration.