Aiming to enhance regulatory submissions with GenAI | Takeda Stories
Aiming to enhance regulatory submissions with GenAI
When it comes to regulatory submissions for medicines, quality matters. A recent survey proves this point. In 2022, the American Medical Writers Association conducted a global survey of 32 regulatory reviewers. This survey revealed that unclear, inaccurate and overly lengthy documents can significantly hinder a reviewer’s ability to assess an applicant’s submission, which may delay the approval process and may create a negative bias toward future submissions from the applicant1. Improved regulatory timelines could help bring important therapies to patients more quickly.
So how can pharma companies such as Takeda improve the quality of our submissions? The answer may lie in generative artificial intelligence (GenAI).
“We have been developing a GenAI tool in-house,” says our head of Medical Writing, Dylan Harris. “Takeda has a history of innovation lasting more than 240 years, and we feel that developing a tailored solution meets our needs while offering the flexibility to scale and expand seamlessly.”
Dylan Harris, head of Medical Writing at Takeda
Dylan leads a team of 100-plus full-time medical writers and contractors in the U.S., China and Japan, who are divided into four groups aligned with our therapy area units – Oncology, Neuroscience and Gastroenterology and Inflammation – as well as our Marketed Products division. His team is testing the prototype to speed up the compilation of a regulatory dossier by automating the analysis of clinical study data, and to make this information clearer, more concise and more consistent.
“Speed is obviously a key benefit,” Dylan adds. “We estimate that the analysis and tabulation process which took our medical writers an average of 20 weeks to complete manually can likely be cut in half.”
The tool may also give us a higher level of analysis. “It appears that the prototype can tell us what’s significant and what’s not,” Dylan explains. “It can potentially provide briefer and clearer information and eliminate certain types of errors, which could directly address the regulators’ concerns and may reduce the likelihood of requests for further information.”
He explains that we have deliberately started testing the prototype with section 11 of the U.S. Code of Federal Regulations clinical study report template, which covers the efficacy analysis of the Food & Drug Administration requirements. “It’s the most challenging section,” he says. “If we get it right here, we believe it will work in other sections.”
Our medical writers are already seeing benefits from testing the prototype. One, Zack Wildemore, reports: “It summarized 70 pages of data in two sentences almost immediately – and the accuracy was remarkable.” Another, Tanya Bollenbach, says: “The prototype is capable of producing factual text in a writing style that is easy to read.”
“The process of writing and reviewing documents for health authorities is time-consuming and prone to human fickleness and error. GenAI, with its capabilities to understand, generate and refine text, presents an opportunity to revolutionize this process.”
Gregory Cuppan, McCullan-Cuppan consultancy
Gregory Cuppan, medical writing expert
Gregory is quick to caution, however, that the technology is still in its infancy, and that important issues still need consideration – how to deal with unstructured source data and the handling of sensitive patient data and proprietary information, for example, as well as accountability and transparency.
Yael Gozin, head of Nonclinical Writing and Systems
“Access control and ensuring data privacy are paramount,” he writes. “It will be crucial to establish clear guidelines on the extent of GenAI involvement and to ensure that human oversight remains integral to the process.”
Human oversight is central to the evolution of the Takeda tool. Yael Gozin, head of Nonclinical Writing and Systems, says: “Our goal is not to replace human expertise but to enhance it by integrating it with advanced machine capabilities, delivering superior outcomes.”
John Chan, head, ShinrAI Center for Artificial Intelligence and Machine Learning at Takeda
John Chan, head of our ShinrAI Center for Artificial Intelligence and Machine Learning [ShinrAI is the Japanese word for trust, reliance and confidence], adds: “Our writers will still be writing, editing, reviewing and verifying the data analyses, but the overall writing and revision time will be much shorter.”
If trained well, GenAI has the potential to make submissions faster, more reliable and more consistent. And testing, which is due to complete by the end of February 2025, has so far shown that we are on the right track.
*McCulley-Cuppan consultancy receives compensation from Takeda as a vendor
