Purpose: Pre-existing antibodies pose unique challenges to standard immunogenicity assays which rely heavily on statistical calculations from individuals assumed negative to the therapeutic in question. Although an increased number of articles have been published on this topic, it is important to understand each therapeutics’ characteristics to a create scientifically justifiable approach. Two case studies are presented using different strategies to overcome the challenges in designing an appropriate assay with meaningful cut point factors.
Methods: Two case studies utilized a homogeneous bridging electrochemiluminescence (ECL) immunogenicity assay with biotin and Sulfo-TAG labeled biotherapeutics. Both assays were optimized to ensure response seen in human samples were specific to the therapeutic, not a function of assay reagents. Pre-existing antibodies were suspected for both compounds. Once methods were optimized, individuals were screened for antibody prevalence. In case 1, prevalence was 99 %, RLU signal ranged from 793 to 55743 with a median signal of 16493. In case 2, prevalence was 70 %, RLU, signal ranged from 102 to 4612 with a median signal of 250. Due to high prevalence for both programs, the traditional screen assay was excluded, and a titer comparison of pre- and post-dose was selected for first tier. Remaining questions for each program included: best approach for negative control pool (NC), strategy for plate specific titration cut point (TCP) factor, criteria for potentially positive treatment-induced titer, threshold, and confirmation cut point (CCP) design.
Results: With 99% prevalence in case 1, a pseudo-negative control from an alternative species was selected. Surrogate matrix (NC) and human serum diluted to a “negative level” were compared and found statistically similar using an F-Test. The TCP determination also focused on a “negative signal level”. Fifty individuals were serially diluted, and a predicted regression curve was calculated for each to determine the plateau background. Mean negative signal was 93.5 RLU, and the upper limit for 95% confidence interval was 96.1. This factor minus the mean NC resulted in a floating TCP of 14.6. To characterize treatment-induced titer, a twostep dilutional increase, and a statistically significant titer increase were assessed. End point titer (EPT) levels for 50 individuals ranged from 1 to 2855, with a median of 611 (Pre-MRD). With the possibility for large sample titers, serial two step or three step dilution evaluation was not practical. Alternatively, a statistically significant increase was calculated based on titer variability for individuals. Two EPT for each of 50 individuals were used to calculate 50 ratios (assumed pre-dose and post-dose). The upper 90th confidence interval was used giving a factor of 1.19 which is applied to each subject’s pre-dose titer. CCP was calculated by diluting 20 individuals to their “negative level” and assaying with and without therapeutic. Inhibition was evaluated statistically using a 1% false positive rate which resulted in CCP = 21.3%.
Case 2 with 70% prevalence allowed screening numerous individuals to create an NC pool with signal comparable to assay buffer (AB). A floating TCP factor was evaluated based on variability of both NC and AB. During development, twelve assays with 6 replicates each NC and AB, showed similar signal with mean NC = 111 RLU, %CV of 4.3, and mean AB = 106 with %CV 7.5. F-Test confirmed they were statistically similar along with a correlation factor 0.90. Since an NC pool would be difficult to recreate, AB variability was selected for TCP factor, and a floating TCP based on 0.1% Parametric Cut Point was calculated to be 22 in validation. To characterize treatment-induced titer, both the two step dilutional increase, and statistically significant titer increase were assessed. End point titer levels of 20 individuals during development ranged from 1.1 to 97.3 with a median titer of 20.2 (Pre-MRD). A statistically significant ratio was determined as in case study 1 from 20 individuals for a factor of 1.75. Based on the low titers observed during development, a multiplicative ratio of 1.75 was deemed too sensitive, and the greater than two step serial dilution increase was selected for potential positive. CCP followed the same logic in diluting 20 individuals to a negative level, and then assaying with and without therapeutic to evaluate the % inhibition statistically with a 1 % false positive rate.
Conclusion: Two Immunogenicity assays were successfully validated with appropriate cut point determinations using different approaches based on their unique situations.
Ana Lara Rojo– Richmond, Virginia