Claim

A federated validation pipeline based on homomorphically encrypted pooled sufficient statistics can preserve clinically useful external validation of pegloticase safety models—especially those using G6PD status, pre-infusion urate, and reaction history—without sharing raw patient-level infusion records across sites.

Rationale

Pegloticase cohorts are fragmented across infusion centers, and safety events are uncommon enough that single-center models are unstable. Cross-site pooling is scientifically necessary but operationally difficult because infusion logs, laboratory trajectories, and genomic screening data are privacy-sensitive. For many validation tasks, what is needed is not raw data but secure aggregation of counts, moments, contingency tables, and calibration components.

Testable design

• Participating gout or rheumatology centers compute local sufficient statistics for predefined validation targets
• Encrypt and aggregate those statistics under an FHE workflow
• Reconstruct pooled AUROC, calibration slope/intercept, Brier score, and decision-curve summaries centrally
• Compare encrypted-validation outputs against a conventional pooled analysis in a consented benchmark subset

Falsification criterion

If encrypted pooled validation materially degrades calibration assessment or ranking performance beyond a prespecified tolerance margin versus conventional pooled analysis, the hypothesis is false.

Why it matters

If confirmed, this would offer a practical DeSci path for validating rare-event infusion-safety models across decentralized sites while minimizing data-transfer risk and governance friction.

References

1. Sundy JS, Baraf HSB, Yood RA, et al. JAMA. 2011;306(7):711-720. DOI: 10.1001/jama.2011.1169
2. FitzGerald JD, Dalbeth N, Mikuls T, et al. Arthritis Care Res (Hoboken). 2020;72(6):744-760. DOI: 10.1002/acr.24180
3. Dowlin N, Gilad-Bachrach R, Laine K, et al. CryptoNets: Applying neural networks to encrypted data with high throughput and accuracy. ICML. 2016. PMLR 48:201-210.
4. Blatt M, Gusev A, Polyakov Y, Goldwasser S. Secure large-scale genome-wide association studies using homomorphic encryption. Proc Natl Acad Sci U S A. 2020;117(21):11608-11613. DOI: 10.1073/pnas.1918257117