ADVANCING CLINICAL DEVELOPMENT AND EVIDENCE GENERATION THROUGH CAUSAL AI: A TARGETED REVIEW OF APPLICATIONS AND IMPLICATIONS FOR REGULATORY-GRADE EVIDENCE
Author(s)
Jackie Vanderpuye-Orgle, MSc, PhD1, Komaleshwari Rani, MSc2, Ashley E. Tate, PhD3, Rafael Coderch Lanau, Sr., MSc4;
1PAREXEL, Durham, NC, USA, 2Parexel International, London, United Kingdom, 3Parexel International, Amsterdam, Netherlands, 4Parexel International, Austin, TX, USA
1PAREXEL, Durham, NC, USA, 2Parexel International, London, United Kingdom, 3Parexel International, Amsterdam, Netherlands, 4Parexel International, Austin, TX, USA
OBJECTIVES: This review aimed to (1) map the landscape of causal AI use in clinical development and RWE; (2) evaluate methodological transparency and quality; (3) identify cases demonstrating meaningful impact on evidence generation; and (4) highlight opportunities and challenges for broader regulatory adoption.
METHODS: A targeted literature search was conducted across peer reviewed journals, conference proceedings, and pre-print repositories (2015-2025). Pharmaceutical research studies were included if they applied causal methods, i.e. propensity score-based approaches, doubly-robust machine learning, Bayesian causal modeling, or causal discovery. Data extraction focused on study purpose, analytic methods, data sources, validation strategies, and relevance to decision-making. Findings were synthesized narratively.
RESULTS: A total of 36 studies met inclusion[KR1] criteria. Applications clustered into four areas: (1) clinical trial optimization, including subgroup identification and target trial emulation; (2) external control arm development using RWE; (3) observational treatment effect estimation using advanced causal estimators; and (4) causal discovery and predictive algorithms. While methodological sophistication has increased, reporting quality and documentation of assumptions varied substantially. A limited number of studies aligned methods with regulatory guidance, though several demonstrated improved bias reduction, efficiency gains, or enhanced model inputs for HTA.
CONCLUSIONS: Causal AI is rapidly expanding and shows promise for strengthening both clinical development and RWE. However, broader adoption will require stronger methodological standards, improved transparency, and clearer regulatory guidance. As tools mature, causal AI may become central to generating regulatory grade evidence.
METHODS: A targeted literature search was conducted across peer reviewed journals, conference proceedings, and pre-print repositories (2015-2025). Pharmaceutical research studies were included if they applied causal methods, i.e. propensity score-based approaches, doubly-robust machine learning, Bayesian causal modeling, or causal discovery. Data extraction focused on study purpose, analytic methods, data sources, validation strategies, and relevance to decision-making. Findings were synthesized narratively.
RESULTS: A total of 36 studies met inclusion[KR1] criteria. Applications clustered into four areas: (1) clinical trial optimization, including subgroup identification and target trial emulation; (2) external control arm development using RWE; (3) observational treatment effect estimation using advanced causal estimators; and (4) causal discovery and predictive algorithms. While methodological sophistication has increased, reporting quality and documentation of assumptions varied substantially. A limited number of studies aligned methods with regulatory guidance, though several demonstrated improved bias reduction, efficiency gains, or enhanced model inputs for HTA.
CONCLUSIONS: Causal AI is rapidly expanding and shows promise for strengthening both clinical development and RWE. However, broader adoption will require stronger methodological standards, improved transparency, and clearer regulatory guidance. As tools mature, causal AI may become central to generating regulatory grade evidence.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR85
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics, Confounding, Selection Bias Correction, Causal Inference
Disease
No Additional Disease & Conditions/Specialized Treatment Areas