Tacrolimus Pharmacogenomics

Testable

hypothesis: if multiple autoimmune centers contribute only encrypted sufficient statistics for tacrolimus troughs, genotype counts, dose, eGFR, and inhibitor exposure, then a homomorphic-encryption pipeline can externally validate CYP3A5-guided dosing models with calibration slope and discrimination close to plaintext pooling.

Primary endpoint: absolute difference in calibration slope between encrypted and plaintext meta-analysis below a prespecified noninferiority margin.

Secondary endpoints: AUC difference, privacy leakage audit, runtime overhead, and heterogeneity tolerance across sites. This is clinically relevant because tacrolimus has high interpatient variability and precision-medicine validation is most useful when it can be done without exposing raw genotype or lab data. References: Froelicher D et al., Nat Commun 2021, DOI 10.1038/s41467-021-25972-y; Liu Z et al., Ann Intern Med 2015, DOI 10.7326/M14-1030; Niioka T et al., Clin Exp Nephrol 2015, DOI 10.3109/00498254.2015.1045571.

Testable

hypothesis: in lupus nephritis and related autoimmune nephritis cohorts, a prespecified model that combines CYP3A5 phenotype, starting dose normalized to weight, early tacrolimus trough trajectory, and CYP3A inhibitor/inducer exposure will identify patients who fail to reach target exposure by day 7 better than weight-based dosing alone.

Primary endpoint: failure to reach at least 80% of the prespecified target trough by day 7 or protocol-defined need for major dose escalation.

Secondary endpoints: calibration slope, AUC, Brier score, and decision-curve net benefit. Mechanistic basis: CYP3A5 expressers generally need higher dose intensity to achieve similar troughs, while renal/hepatic context and interacting drugs modify the real-world expression of that genotype effect. References: Birdwell KA et al., Clin Pharmacol Ther 2015, DOI 10.1002/cpt.113; Niioka T et al., Clin Exp Nephrol 2015, DOI 10.3109/00498254.2015.1045571; Wei L et al. 2025, DOI 10.12182/20250560509.