Confirming human antibody responses to a therapeutic monoclonal antibody using a statistical approach

Confirmatory assays for immunogenicity testing typically involve testing a sample in the presence or absence of excess drug. A decrease in assay signal in the presence of drug is taken to indicate the presence of anti-drug antibodies (ADAb) and the sample is confirmed positive. While there is widespread acceptance of the principle, there are currently no published guidelines for determining how much the signal should be reduced for a sample to be confirmed positive. In this report we address this issue using a novel approach employing a Student's t-test. The basic premise is to assess if an observed decrease in signal in the presence of drug is greater than what might be expected to occur as a result of the normal variation in the system. A key component of the method involves being able to capture and measure all of the normal variation. This requires a modification of commonly employed methods of sample preparation. We validated the method and tested samples from a clinical study. In addition, we reanalyzed the data to see what would have been the outcome had we used two other common approaches for confirmatory assays, one based on a minimum percent decrease in signal to confirm positivity (arbitrarily set at 25%), and one requiring a minimum drop in signal, set by a low quality control (QC) sample. The t-test approach proved superior over a wide range of assay signals and appeared to result in fewer false negatives.