The Framework That Survives Bad Models: Human-AI Collaboration for Clinical Trials
Pub. online: 10 November 2026
Type: Case Study, Application, And/or Practice Article
Open Access
Open Access
1
Novartis Pharmaceuticals Corp., East Hanover, NJ.
2
Novartis Pharma AG, Basel, Switzerland.
Accepted
19 October 2025
19 October 2025
Published
10 November 2026
10 November 2026
Abstract
Artificial intelligence (AI) holds great promise for supporting clinical trials, from patient recruitment and endpoint assessment to treatment response prediction. However, deploying AI without safeguards poses significant risks, particularly when evaluating patient endpoints that directly impact trial conclusions. We compared two AI frameworks against human-only assessment for medical image-based disease evaluation, measuring cost, accuracy, robustness, and generalization ability. To stress-test these frameworks, we injected bad models—ranging from random guesses to naive predictions—to ensure that observed treatment effects remain valid even under severe model degradation. We evaluated the frameworks using two randomized controlled trials with endpoints derived from spinal X-ray images. Our findings indicate that using AI as a supporting reader (AI-SR) is the most suitable approach for clinical trials, as it meets all criteria across various model types, even with bad models. This method consistently provides reliable disease estimation, preserves clinical trial treatment effect estimates and conclusions, and retains these advantages when applied to different populations.
References
Baeten, D., Sieper, J., Braun, J., Baraliakos, X., Dougados, M., Emery, P., Deodhar, A., Porter, B., Martin, R., Andersson, M., Mpofu, S. and Richards, H. B. (2015). Secukinumab, an Interleukin-17A Inhibitor, in Ankylosing Spondylitis. New England Journal of Medicine 373(26) 2534–2548. https://doi.org/10.1056/NEJMoa1505066. https://www.nejm.org/doi/pdf/10.1056/NEJMoa1505066.
Deodhar, A., Blanco, R., Dokoupilová, E., Hall, S., Kameda, H., Kivitz, A. J., Poddubnyy, D., van de Sande, M., Wiksten, A. S., Porter, B. O. (2021). Improvement of signs and symptoms of nonradiographic axial spondyloarthritis in patients treated with secukinumab: primary results of a randomized, placebo-controlled phase III study. Arthritis & Rheumatology 73(1) 110–120.
Fournier, L., Costaridou, L., Bidaut, L., Michoux, N., Lecouvet, F. E., de Geus-Oei, L. -F., Boellaard, R., Oprea-Lager, D. E., Obuchowski, N. A., Caroli, A. (2021). Incorporating radiomics into clinical trials: expert consensus endorsed by the European Society of Radiology on considerations for data-driven compared to biologically driven quantitative biomarkers. European radiology 31 6001–6012.
Guo, D., Yang, D., Zhang, H., Song, J., Zhang, R., Xu, R., Zhu, Q., Ma, S., Wang, P., Bi, X. et al. (2025). DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. arXiv preprint arXiv:2501.12948.
Mo, Y., Chen, Y., Readie, A., Ligozio, G., Coroller, T. and Papież, B. W. (2023). Towards automatic scoring of spinal X-ray for ankylosing spondylitis. arXiv preprint arXiv:2308.05123.
Ronneberger, O., Fischer, P. and Brox, T. (2015). U-net: Convolutional networks for biomedical image segmentation. In Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 234–241. Springer.
Taylor, H. G., Wardle, T., Beswick, E. J. and Dawes, P. T. (1991). The relationship of clinical and laboratory measurements to radiological change in ankylosing spondylitis. British Journal of Rheumatology 30(5) 330–335. https://doi.org/10.1093/rheumatology/30.5.330. https://academic.oup.com/rheumatology/article-pdf/30/5/330/5041584/30-5-330.pdf.
Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M. -A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F. et al. (2023). Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971.
Van Der Heijde, D., Braun, J., Deodhar, A., Baraliakos, X., Landewé, R., Richards, H. B., Porter, B. and Readie, A. (2019). Modified stoke ankylosing spondylitis spinal score as an outcome measure to assess the impact of treatment on structural progression in ankylosing spondylitis. Rheumatology 58(3) 388–400.
Vera, P. and Makhlouf, M. (2023). BIOptimus: Pre-training an Optimal Biomedical Language Model with Curriculum Learning for Named Entity Recognition. arXiv:2308.08625. https://arxiv.org/abs/2308.08625.
