A Nine-Year Case Study: What Did ISR Tell Us?

Incurred Sample Reanalysis (ISR) was introduced over 15 years ago to the bioanalytical community and continues to be the industry standard to assess the reproducibility of bioanalytical data. Current industry guidance recommends testing ISR in scenarios including pivotal studies, first trial in subjects, and once per patient population. However, some of these scenarios (e.g., pivotal studies) can be ambiguous, leading to extensive analysis especially in long-running programs. This case study highlights ISR for the PK method for Drug A, a fully human IgG4-based monoclonal antibody tested for use in multiple tumor types, which was and continues to be performed in accordance with these recommendations, resulting in ISR drug concentration results for more than 2500 samples collected from participants in 16 studies (Phase I through Phase III) spanning approximately 9 years.
With this large dataset, we wanted to explore the potential sources of variability between the original and ISR results. Variables included patient population and disease type, analyst (both manual and automated), laboratory, and critical reagents. When evaluated across the 16 studies, by far the most significant contributor to variability was occasion-to-occasion analysis (>80%), with a minor contributor being change of a critical reagent. Other variables (indication, population, analyst, laboratory) did not significantly impact the variability observed in the ISR data.
Overall, the ISR passing rate for the program was excellent (~94%), with the original and ISR results demonstrating a strong linear correlation. This case study sheds light on and demonstrates two key points: (1) for the Drug A concentration method, there is little value in continuing to test ISR in additional cancer indications, and (2) questions the overall usefulness of ISR for its stated purpose as outlined in guidance documents. As determined by the data set, a majority of variability observed during ISR testing resulted from simply reanalyzing a sample in an assay that typically allows a 20% margin of error. While other analytical variables may change throughout the lifecycle of an assay, the impact is minimal, suggesting a well-controlled method will maintain a high-level of consistency over time.