As regulatory agencies increasingly focus on the processes and workflows behind the Trial Master File (TMF), not just on what it contains, the pressure has increased on sponsors to ensure theirs is always in a good state.
Whereas once organizations might have waited until an inspection to ensure all documentation was uploaded and met quality control (QC) requirements, they must now navigate more guidelines about how they store documents and data.i Inspectors want to know not just what is in the TMF, but how it got there, how it is being managed and who has oversight.
With the TMF integral to demonstrating good clinical practice (GCP) in a clinical trial, it is critical that companies can show they are uploading documents to the TMF in a timely fashion and are keeping track of what needs to be in there at each stage.ii
Managing the workload
There are several reasons organizations have been more reactive in the past, including resource shortages, document owners’ inexperience with the TMF and the labor-intensive work required to maintain its integrity.
While still in the early stages of development, artificial intelligence is already helping to navigate this workload and reduce the demand on resources, as well as the risk of inspection findings.
[Author’s note: The examples in this blog draw on more mature AI tools such as machine learning and natural language processing (NLP), not the newer large language model (LLM) approach to AI.]
Our internally generated data shows that between 9 and 12 percent of quality issues are due to documents being filed incorrectly. AI can help classify documents in the TMF, alleviating pressure on those unfamiliar with the filing structure, and reducing the likelihood of misfiles.
Another exciting area for streamlining processes is in risk-based QC, where AI can be directed to do additional checks on documents that are easier to classify, reducing the requirement for resource-heavy manual QC. Even where some manual QC may still be needed, AI can help those tasked with checking the documents by offering suggestions.
Providing better oversight
Small to mid-sized companies don’t have the human resources to manage all their TMF documents, and this has historically forced them to focus on catching the biggest problems.
This has inevitably meant some documents do not have appropriate oversight and document owners unfamiliar with the TMF may misfile them, which could lead to an inspection finding.
AI again offers solutions, providing document owners with classification suggestions as they upload, reducing the need for manual QC down the track.
Automated QC for non-key documents, such as checking for missing pages or other metadata, will also give them oversight where there would otherwise be none.
This enables an improvement in overall TMF health without a corresponding increase in spend on staffing, which most small to mid-size organizations can ill afford.
Better use of expertise
AI also has the capacity to empower TMF personnel in their careers. By automating manually intensive rote work and providing pre-trained suggestions (e.g., for document classification), AI tools can help free busy study teams from time-consuming queries and remediation effort to focus on improving TMF quality, completeness, and timeliness.
For example, TMF teams at sponsor companies could focus more on TMFs that are outsourced, since responsibility for the TMF remains with the sponsor even if outsourced to a CROi. Another key area deserving more expert attention is site engagement, for example to provide additional training and guidance to critical site staff on uploading TMF documentation.
Early warning signals
All TMF teams should be doing periodic QC or completeness quality checking, not least because it is a priority for the regulators. In its guidance on good clinical practice for clinical trials, for example, the UK’s Medicines and Healthcare products Regulatory Agency (MHRA) says: “The complete TMF is the basis for inspection and all the documents in it must be made available to the inspectors.”iii
But with documents coming from so many different sources, sometimes related data or documents that should be added to the TMF are overlooked. One example is Form FDA 1572, which includes the list of investigators. This form can look fine when it is added to the TMF but the information about the investigators involved may not have been captured.
It is a laborious process to go through each of these documents manually as part of periodic QC to ensure details such as the investigator’s CV, processional license, financial disclosure form and training documentation have been included. This is particularly problematic with big studies or companies with lots of studies, where typically completeness checks are only carried out on each site once a year.
Enabling AI to review the documents instead could provide real-time early warning of missing data, ensuring issues are picked up much sooner in the file’s development, and considerably reducing manual resourcing involved in completeness checks.
While it has some way to go, AI would eventually be able to conduct sophisticated QC checks and provide the TMF team with a to-do list to complete the file, giving companies far greater oversight.
Way of the future
The early signs of what AI can do to improve an organization’s handling of the TMF are promising. With further advances in algorithms and, crucially, the right expertise directing the building of AI models, it is clear there is great potential to reduce the number of mundane tasks assigned to TMF teams. This should enable them to focus on critical areas while helping to ensure that the Trial Master File is in the best possible health at all times.
About the author: Aaron Grant is VP of Solutions Consulting at Cencora PharmaLex (formerly Phlexglobal), where he is focused on helping clients to solve challenges through a mix of people, process and technology.
[i] Guideline on the content, management and archiving of the clinical trial master file (paper and/or electronic), EMA, 2018. https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-content-management-archiving-clinical-trial-master-file-paper/electronic_en.pdf
[ii] GCP Inspections: Expectations and the dos and don’ts for hosting, MHRA Inspectorate, March 2020. https://mhrainspectorate.blog.gov.uk/2020/03/10/gcp-inspections-expectations-and-the-dos-and-donts-for-hosting/
[iii] Guidance: Good clinical practice for clinical trials, MHRA, updated April 2023. https://www.gov.uk/guidance/good-clinical-practice-for-clinical-trials
This blog is intended to communicate PharmaLex's capabilities which are backed by the author’s expertise. However, PharmaLex GmbH and its parent, Cencora, Inc., strongly encourage readers to review the references provided with this blog and all available information related to the topics mentioned herein and to rely on their own experience and expertise in making decisions related thereto as the blog may contain certain marketing statements and does not constitute legal advice.
