Risks and Opportunities of Generative AI in Clinical Trials

05 Sep 2023 · 10 minutes read

Minseok Kim, Chief Business Officer at JNPMEDI

The frenzy around ChatGPT, a prototype conversational AI chatbot developed by OpenAI, shows no signs of abating. Though it has only been a mere 8 months since the beta service was launched, the changes it has brought during this time have been truly revolutionary.

The AI industry, which experienced a lull after the appearance of AlphaGo, is undergoing a restructuring due to ChatGPT’s emergence. Over the past decade, AI technology has rapidly advanced. Particularly, generative AI technology is expanding its application beyond engineering fields into industries spanning law, culture, arts, presenting new avenues for business innovation.

In the realm of clinical trials, there have been various recent attempts to apply AI technology. Global pharmaceutical companies are using AI models to simulate diverse trial conditions, aiming to find effective conditions through simulations.

AI models trained on vast amounts of clinical trial data support the simulation of optimal countries, trial institutions, and recruitment scales, facilitating efficient clinical trials. Furthermore, they are being used to predict results for new patients or trial conditions.

AI is also being employed to create virtual patient groups. In clinical trials, to prove the efficacy of a drug, a control group receiving a placebo is recruited. However, ethical concerns surrounding administering placebos to patients with rare cancers or chronic diseases, where patients are scarce and their lives are at stake, make recruiting a control group challenging.

In such cases, AI technology is used to create a virtual patient group (Synthetic Control Arm) based on past patient data, applied in clinical trials.

With the integration of generative AI technology, AI’s influence on the field of clinical trials is steadily growing. Generative AI refers to artificial intelligence that produces results tailored to a specific request made to the AI through certain inputs by a person.

In this process, generative AI uses machine learning and deep learning-based algorithms, generating content in various forms such as written materials, images, videos, music, and computer code.

Generative AI technology is also being utilized in tasks related to summarizing and composing clinical trial documents. Large Language Models (LLMs) based on natural language processing are trained on extensive clinical trial document data, understanding sentence meanings, extracting key information, or summarizing. This forms the basis for generating new documents like summaries of trial results or report writing.

While generative AI technology in the field of clinical trials can contribute to reducing the time and effort for researchers’ decision-making, concerns and risks exist.

In May 2023, the World Health Organization (WHO) expressed the view that while LLM in the medical field has various potential functionalities, stringent verification and supervision are necessary. This statement from WHO is noteworthy as it marks a transition from exploring the potential use of ChatGPT to the stage of applying and verifying it in clinical settings.

WHO’s call for strict caution in using LLM in the medical field stems from reasons such as potential biases and errors in training data, the possibility of generating and spreading false information, and the potential leakage of sensitive health-related information.

Issues of information protection and security also become factors causing hesitation in implementing generative AI. this is because clinical trial confidential information such as personal data and drug information may be included in the data used to build AI models.

However, now it is becoming increasingly important to recognize concerns about AI implementation and develop strategies such as improving data quality, reducing biases, and establishing relevant regulations and guidelines.

Considering the recent pace of advancement, it might be possible in a few years to input key conditions for clinical trials based on massive accumulated data and generate protocols (clinical trial plans). This could potentially lead to the generation of related documents like case records or consent forms based on these protocols.

Whether generative AI will lead to a singularity surpassing human intelligence is still unknown. However, it is clear that it will have a significant impact on the clinical trial industry in the future and its application areas will infinitely expand.


Your Partner for a Successful Journey

JNPMEDI has been putting effort on bringing disruption on healthcare industry via integrating digital technologies to clinical research & trial operations.