Hypothesis: Among SLE patients treated with anifrolumab, active lupus nephritis and low absolute lymphocyte count will explain more between-patient heterogeneity in herpes zoster risk than prednisone dose alone, indicating that immune-state context is a stronger effect modifier than steroid exposure by itself.

Why this matters: Current bedside reasoning often over-focuses on glucocorticoid dose. But nephritis-level immune dysregulation, cumulative co-immunosuppression, and lymphocyte depletion may better capture the host context in which VZV reactivation emerges after interferon-receptor blockade.

Testable prediction: In a pooled observational dataset or post-marketing pharmacovigilance-linked registry, a model including nephritis status, absolute lymphocyte count, vaccine status, and concurrent MMF/CYC will show materially better calibration/discrimination for herpes zoster than a simpler model centered on prednisone dose and age. Pre-specify improvement by delta-AUC, calibration slope, and decision-curve net benefit.

Suggested design: Multicenter registry validation with internal-external cross-validation across sites. Secondary analysis: interaction term for nephritis x lymphopenia.

Falsifier: If steroid dose alone performs as well as or better than the richer immune-context model, the hypothesis fails.

References:

1. Morand EF et al. N Engl J Med. 2020;382:211-221. DOI:10.1056/NEJMoa1912196
2. Restrepo-Escobar M et al. J Clin Rheumatol. 2012;18:225. DOI:10.1097/RHU.0b013e31825a255e
3. Chen HH et al. Lupus. 2013;22:238-244. DOI:10.1177/0961203312470186
4. Steyerberg EW. Clinical Prediction Models. 2nd ed. Springer; 2019. DOI:10.1007/978-3-030-16399-0