Aniek Sies  Principal Scientist at Johnson & Johnson

Aniek Sies is a Principal Scientist in Manufacturing Science and Technology (MSAT) in Johnson & Johnson, working on the application of statistical methods to manufacturing risk assessment and regulatory decision-making. She holds a PhD with a focus on statistical methodology for estimating optimal treatment regimes.

A statistical risk-based framework for sunsetting one-sided release tests: from limited data to regulatory approval

Manufacturing control strategy changes such as removing routine release tests (“sunsetting”) require a defensible quantitative assessment of residual quality risk. While this may appear straightforward, it is statistically challenging due to the irreversible nature of the decision and the frequent reliance on limited historical batch data that are often heavily left censored.

I present a pragmatic statistical framework designed to quantify the probability of missing future out-of-specification batches. It integrates well-established statistical tools, including process capability analysis with confidence bounds to account for finite sample uncertainty, predictive probability statements, and Bayesian posterior predictive modelling.

The framework is illustrated using an established drug product for which sunsetting of release testing was approved by both FDA and EMA. The discussion subsequently focuses on early lifecycle scenarios with limited and highly censored data. Throughout, the emphasis is on statistical judgment: making assumptions explicit, acknowledging identifiability limits, and supporting transparent and defensible decision-making when data are far from ideal.

Swetlana Berger Associate Director Biostatistics of UCB

Applying DoE to Bioassays: Faster Optimization, Relevant Performance, Better Decisions

Bioassays are a key analytical technology from basic research to drug discovery in the biopharmaceutical industry. Because these assays are complex and often involve multiple handling steps, optimization should be guided by the intended use (fit-for-purpose assay) and focused on the performance characteristics that matter most for that use—for example precision, accuracy, sensitivity, dynamic range, selectivity, robustness, and throughput.

For process optimization, Design of Experiments (DoE) has long been established as a more powerful strategy than a One-Factor-at-a-Time approach. DoE enables systematic exploration of factors and their interactions, helping teams converge efficiently on settings that improve the targeted performance characteristics within the expected operating window. Despite these advantages, DoE is still not used as broadly as it could be, often due to perceived costs, effort, and complexity.

We will share how implementing DoE as a regular technique for bioassay optimization can be accelerated by (1) aligning experiments to the assay’s intended use and critical performance characteristics, (2) identifying and enabling a key user group, and (3) providing JMP-specific training using the JMP DoE platform.

Michele Firmo  Data Scientist at F Hoffman-La Roche Ltd

Statistically Derived MAD for TOST: A Combined Accuracy and Precision Approach for PSD Co-validation Studies

Within the Synthetic Molecules (SM) drug development process, robust methods to evaluate Particle Size Distribution (PSD) for drug substances are essential. When manufacturing occurs across multiple sites, these analytical methods must be transferred effectively.

The Two One-Sided t-Test (TOST) is the standard statistical framework for assessing site-to-site comparability. However, its application requires a Maximum Allowable Distance (MAD) to be predefined and justified. While available guidance documents provide clear limits for repeatability, specific thresholds for site-to-site comparison are lacking. Consequently, a MAD is typically established based on expert judgment, which can be difficult to defend during regulatory review.

We propose a novel methodology for deriving statistically justifiable MAD values by leveraging the Combined Accuracy and Precision (CAP) framework from ICH Q2(R2). The CAP approach evaluates the joint contributions of accuracy and precision to the total variability of a quantity of interest, defined here as the 90% confidence interval of the difference between site means. Our approach reverses the CAP logic: starting from the regulatory bound on method repeatability, we partition the accuracy and precision contributions to total variability. By subtracting the reference site’s method precision, this process allows derives an acceptance limit for the site-to-site difference of means. This value is subsequently utilized as the MAD for the TOST, ensuring the MAD is directly linked to the established regulatory acceptance criteria for repeatability.

We demonstrate an application of this technique using synthetic datasets to illustrate its performance and limitations across various scenarios. Finally, we provide a comparative analysis of this approach against other methods commonly employed by analytical scientists to assess site-to-site comparability.