Reply to “Understanding the Assumptions on Anifrolumab Dose Regimen Selection for a Phase 3 Trial in Lupus Nephritis”

We thank Dr Krishna for their considered review of our paper describing anifrolumab dose regimen selection for the phase 3 lupus nephritis (LN) IRIS trial. ¹

We agree that matching systemic exposure in LN to nonrenal systemic lupus erythematosus (SLE) may not extrapolate to equivalent efficacy. However, dose selection was based on both exposure and modeling the intensified regimen's impact on clinical response (figure 6). ¹ We identified a regimen within the explored limits (constrained by available clinical safety evidence) that best addressed underexposure due to proteinuria‐driven clearance in LN. ²

Using urine protein‐creatinine ratio (UPCR) allowed our model to remain relatively simple, yet adequately evaluate both efficacy‐related response and disease impact on pharmacokinetics. In our Discussion, we acknowledged that UPCR cannot distinguish active inflammation from renal damage. ¹ However, for modeling, continuous UPCR retains more information about the system than a dichotomized endpoint based on a fixed threshold, such as complete renal response. We also acknowledged extrapolation beyond the observed dosing regimen as a limitation, though our model‐predicted UPCR levels are within the observed range from TULIP‐LN. This analysis aimed to support anifrolumab phase 3 dose selection, rather than optimize the therapeutic index. No exposure–safety relationship was identified in the large pooled SLE dataset and doses of 1000 mg were well tolerated in a phase 2 trial; ³ thus, we believe it sufficient to account for safety by constraining regimen boundaries based on available data.

TULIP‐LN discontinuation was influenced by the study protocol (e.g., glucocorticoid tapering requirements) in addition to treatment effects. As TULIP‐LN early stopping criteria were not included in the IRIS study design, we curated the dataset by treating such events as right‐censored, rather than dropout events. This is detailed in the Supplementary Methods. ¹

The presence of multiple local optima is a constant consideration in parameter estimation of nonlinear mixed‐effects models. While likelihood profiling could provide additional information about the local identifiability and uncertainty of the estimate, it would not establish that the estimate corresponds to a unique global optimum. ⁴ Here, the same optimum was found across a range of initial parameter values. Confidence in the final model was based on successful convergence, model diagnostics, the plausibility of the parameter estimate, and bootstrap‐derived confidence intervals.

Our modeling supports the selection of 6 months of intensified anifrolumab dosing for the phase 3 IRIS trial, during which the efficacy and tolerability of this treatment regimen will be assessed in patients with LN.

FUNDING

This work was funded by AstraZeneca.

CONFLICT OF INTEREST

JA, WT, and CL are employees of and own stock in AstraZeneca. JA, YLC, RT, and CL hold intellectual property/patents in AstraZeneca. YLC and RT are former employees of AstraZeneca. YLC owns stock in AstraZeneca. LB is an employee of AstraZeneca. DJ is a consultant for Amgen, AstraZeneca, Aurinia, BMS, CSL Vifor, GSK, Novartis, Otsuka, Roche, and Takeda; has received Grant/Research Support for CSL Vifor and GSK; declared receipt of speaker's fees from Amgen, CSL Vifor, GSK, and Otsuka; holds advisory board membership for Alentis, GSK, Hansa, and Novartis; and owns stock in Aurinia.