PII-053 - A NOVEL MODELING APPROACH FOR PREDICTING THE TIME COURSE OF ANTI-DRUG ANTIBODIES OF BIOLOGICS: APPLICATION TO AXATILIMAB AS AN EXAMPLE

R. Wada¹, L. Cheng², Y. Yang², X. Liu², H. Li¹, X. Chen², J. Sheng³; ¹QuanTx Consulting, Mountain View, CA, USA, ²Incyte Corporation, Wilmington, DE, USA, ³Incyte Research Institute, Wilmington, DE, USA.

Background: One challenge with biologics is the development of antidrug antibodies (ADAs), which potentially impacts pharmacokinetics, pharmacodynamics, safety, and efficacy. Several approaches have been utilized to model the time-course of ADA titer, including the zero-inflated Poisson (ZIP) model (Bonate et al. J Pharmacokinet Pharmacodyn. 2009). The objectives of our analysis were to use a modified ZIP model to categorize and characterize the time course of ADA titer following exposure to biologics and to identify potential predictors of onset and extent.

Methods: This analysis used axatilimab monotherapy data from 311 patients in 3 clinical studies (NCT03238027, NCT03604692, NCT04710576). A total of 82 patients with treatment-emergent ADA titer values ≥2 and 608 titer values were included. ADA incidence was categorized as persistent (all positive ADA tests after the first positive test), transient (positive ADA test followed by only negative ADA tests), or recurrent (positive ADA test followed by negative ADA test followed by positive ADA test). The modified ZIP model includes 2 parts (Figure): the first part is a logistic model accounting for the probability of a negative titer; the second is a Poisson model for the probability of observing a given titer number as log2(titer value). Various predictors were evaluated for ADA onset and extent.

Results: ADA incidence with axatilimab monotherapy was categorized as persistent (n=55; 67%), transient (n=22; 27%), or recurrent (n=5; 6%). The modified ZIP model characterized the time course of ADA titer across individual patients. The estimated time to first positive titer was 66 (10-90th percentile, 43–166) days. The estimated initial titer was 2.2, with a doubling time of 26 (range, 9.2–60) days. The estimated maximum titer was 19 (range, 2.3–1410). The only significant covariate on the titer extent was the area under the concentration curve in the previous 4 weeks. The slope of the relationship was negative, indicating lower titer levels associated with increasing exposure.

Conclusion: A novel ADA model, including a saturable effect of time on the probability of positive or negative ADA test results, was developed. The utility of the model was demonstrated using axatilimab immunogenicity data and may facilitate ADA analysis in future biologics development.