Hypothesis: homomorphically encrypted site-level feature summaries will preserve external calibration of the IIM-CANCER score across decentralized myositis registries

Claim If centers share only homomorphically encrypted site-level feature summaries, the externally validated IIM-CANCER score will retain clinically useful calibration and discrimination across decentralized myositis registries without raw patient-level transfer.
Why this is plausible The key validation problem in cancer-associated myositis is not just model design; it is that high-quality cohorts are distributed across institutions with real privacy and governance constraints. Homomorphic encryption should let sites contribute sufficient statistics, calibration updates, and cross-site performance summaries while keeping raw data local.
Testable design
- Multi-center federation of adult IIM cohorts
- Compare raw pooled validation with encrypted sufficient-statistic validation
- Primary endpoints: AUROC, calibration slope, and Brier score
- Secondary endpoints: decision-curve net benefit, site participation rate, and refresh frequency
Falsification criteria The hypothesis fails if encryption materially degrades calibration or if operational overhead prevents practical multi-site validation.
References
- Cheon JH, Kim A, Kim M, Song Y. Homomorphic Encryption for Arithmetic of Approximate Numbers. Adv Cryptol (ASIACRYPT). 2018. DOI: 10.1007/978-3-319-70694-8_15
- Moons KGM, Altman DG, Reitsma JB, et al. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD). BMJ. 2015;350:g7594. DOI: 10.1136/bmj.g7594
- Vickers AJ, Elkin EB. Decision Curve Analysis: A Novel Method for Evaluating Prediction Models. Med Decis Making. 2006;26(6):565-574. DOI: 10.1177/0272989X06295361