Use of Innovative Methods in Long-Term Follow-up Observational Studies: An AI-Enabled Targeted Literature Review
Author(s)
Joydeep Sarkar, PhD1, Ian Bonzani, BSc, PhD2.
1IQVIA, Singapore, Singapore, 2IQVIA, London, United Kingdom.
1IQVIA, Singapore, Singapore, 2IQVIA, London, United Kingdom.
OBJECTIVES: Long-term follow-up (LTFU) studies following clinical trials are becoming increasingly important, especially with new treatment modalities like cell and gene therapy (CAGT). Given the long duration, breadth of data collection, these studies have unique challenges - collecting clinical & safety data across multiple care settings, repeated collection of patient-reported outcomes (PRO), continued patient engagement and difficulty with site visits. Advances in digital technology and patient engagement tools coupled with changes in patient data access and regulatory landscape has enabled new approaches. This study was a targeted literature review (LR) of the evidence behind the evolving use of new approaches in LTFU studies.
METHODS: The LR was limited to last 5 years to account for new entrance of solutions. The abstracts extracted from PubMed were refined using AI to answer different topics. Prompts for the AI-models were validated to reach >95% accuracy over a random sample (25 studies) before use.
RESULTS: PubMed search identified 2286 studies. Average length of follow-up was 5 years while less than 25% had a 10-year follow-up or more. Only 5% of the studies were related to CAGT. Filtering for studies using decentralized approaches yielded no results. Similarly, patient mediated EMR data access also yielded no results. 194 studies used existing registry data. 118 studies collected repeated PROs - quality of life or indication specific symptom severity scales. Use of digital tools (ePRO) vs paper to administer PROs could not be determined based on abstract or free-full text when available. Use of financial compensation to support recruitment or retention in LTFU studies yielded no results. However, use of financial compensation has been reported in other clinical studies in the same period (717 studies).
CONCLUSIONS: The review highlights the growing adoption and potential of new study approaches, but published data on effectiveness of such methods in LTFU studies is limited.
METHODS: The LR was limited to last 5 years to account for new entrance of solutions. The abstracts extracted from PubMed were refined using AI to answer different topics. Prompts for the AI-models were validated to reach >95% accuracy over a random sample (25 studies) before use.
RESULTS: PubMed search identified 2286 studies. Average length of follow-up was 5 years while less than 25% had a 10-year follow-up or more. Only 5% of the studies were related to CAGT. Filtering for studies using decentralized approaches yielded no results. Similarly, patient mediated EMR data access also yielded no results. 194 studies used existing registry data. 118 studies collected repeated PROs - quality of life or indication specific symptom severity scales. Use of digital tools (ePRO) vs paper to administer PROs could not be determined based on abstract or free-full text when available. Use of financial compensation to support recruitment or retention in LTFU studies yielded no results. However, use of financial compensation has been reported in other clinical studies in the same period (717 studies).
CONCLUSIONS: The review highlights the growing adoption and potential of new study approaches, but published data on effectiveness of such methods in LTFU studies is limited.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
SA100
Topic
Patient-Centered Research, Real World Data & Information Systems, Study Approaches
Topic Subcategory
Literature Review & Synthesis, Registries
Disease
Genetic, Regenerative & Curative Therapies, Rare & Orphan Diseases