Inhalation

INH1024

Issue link: https://www.e-digitaleditions.com/i/1527415

Contents of this Issue

Navigation

Page 26 of 38

Inhalation OctOber 2024 25 variability. Appropriate product and process develop- ment serve that goal. Once a batch is manufactured, however, the closeness of its actual properties to the goal must be demon- strated—usually through sampling and testing of finished product. For example, delivered dose from the sampled units needs to be measured, then the data assessed and evaluated. ere are several ways to do this. For dose uniformity, the FDA guidance for nasal products [5] recommends use of a "counting test" with the following steps: • Collect doses from 10 units at the beginning of a batch and 10 units from the end. • Count the number of determinations within and outside of the 80-120% label claim interval and 75-125% label claim interval. • Based on the counted instances, either pass the batch, go to second-tier testing or, ultimately, reject the batch per the instructions in the guidance. e more recent FDA MDI/DPI guidance [2] includes a type of Parametric Tolerance Interval Test (PTIT) for controlling dose uniformity, which directs the sponsor to: "Establish test parameters (e.g., sample size, tolerance interval factor (k factor)) and accep- tance criteria that will ensure, to a confidence level of 95 percent, that at least 90 percent of the units in a batch (i.e., the coverage) will meet the established upper and lower limits (i.e., 80-120 percent of TDD [target deliv- ered dose]) [2]. A PTIT is, statistically speaking, a more sound approach to evaluating and demonstrating batch quality, compared to a "counting" test. Among PTIT benefits is its sharper operating characteris- tic (OC) curve, which allows a clearer separation of higher-quality from lower-quality batches. Another advantage of a PTIT is its scalability, so that the sponsor could apply the same quality standard on a larger sample-size test (e.g., for validation purposes, requiring a higher level of confidence) as a smaller sample-size test (e.g., used for batch release, which requires a lower level of confidence for routine test- ing). Statistically, such adjustment is straightforward for a parametric test [6] but not for a counting test. However, it is also critical to recognize that this statis- tical approach is no different than statistical methods and statistical thinking put forth as a limiting qual- ity sampling plan from ISO standards. ese con- nections are important because a procedure might seem new or different, but if it is analogous to current standards, that should be acknowledged to ensure the use of such a procedure is in alignment with current standards. For example, a limiting quality sampling plan, as recommended and implemented in ISO standards, is not typically recommended for routine sampling and release testing. • Align quality standards and statistical methods for assessing adherence to those standards to maximize patient benefit with respect to safety and efficacy as well as maximizing availability of good product. • Result in, potentially, faster regulatory approv- als—because clearer processes and sampling plans should make it easier to demonstrate compliance with regulatory standards, translating to faster reg- ulatory approvals. • Enable a clearer understanding of how changes in a manufacturing or analytical process directly relate to critical quality attributes, allowing the sponsor company to better evaluate risk of any deviation and—importantly for preventing drug shortages—quickly and comprehensively respond to changes in the supply chain. During product and process development, the sponsor should also study how the most criti- cal design process parameters relate to each other as well as how they relate to the output (demon- stration test results). is would streamline inves- tigations of any out-of-specification result and identification of root causes. An example e sponsor's goal is to manufacture every batch as close as possible to "target" (e.g., with mean deliv- ered dose at 100% of the label claim) and with low 1 Descriptive statistic for a tested sample of product on a single occasion was being incorrectly used to draw inference to untested product (i.e., the population) 2 USP has explicitly stated descriptive statistics should only be used to make decisions about samples tested 3 Lack of statistical clarity from USP 4 Failure to provide clarity has resulted in misunderstanding and misuse of a USP standard 5 USP formalized a product quality standard in the form of a descriptive statistical metric with acceptance limits that a product must pass "whenever tested" 6 Industry and the FDA seem to have misinterpreted the meaning and intent of USP standards as batch release criteria 7 No explicit connection to product population characteristics necessary to meet stated requirements, i.e., no clear articulation of characteristics of distribution that would ensure passing whenever tested 8 The FDA's actions have inadvertently fostered misunderstanding and misuse of the standard by approving and allowing use of this descriptive statistic as an inferential release test Table 1 Historical and cultural influences that have led to the misuse of descriptive statistic as an inferential release test

Articles in this issue

view archives of Inhalation - INH1024