Frequentist approach for estimation of false decision risks in conformity assessment based on measurement uncertainty of liquid chromatography analytical procedures

摘要

The measurement uncertainty (MU) related to analytical results can lead to false decisions in conformity assessment, such as accepting or rejecting incorrectly a medicine lot (consumer's and producer's risks, respectively). These risks can be global or specific. It is important to understand the different types of conformity decision risks, and the different approaches to estimate them to ensure the reliability of the analytical results. Thus, the aim of this work was to estimate the specific consumer's and producer's risks from the MU values of 64 liquid chromatography analytical procedures for antibiotic or antifungal assays, in order to evaluate their performances in conformity assessment. The specific risks of the analytical procedures were estimated by the frequentist approach following normal distribution using Microsoft Excel software, and in addition a spreadsheet was created to be available as supplementary material to estimate specific risks by this approach. Moreover, the global risks of the analytical procedures were estimated using Bayesian approach, assuming a uniform scenario of production process. And finally, the estimation of specific risks by Bayesian and frequentist approaches was compared. Only 39 % of the evaluated analytical procedures had MU within the recommended. When the result is close to the specification limit, the risk can be significant, in such cases, a strategy is to adopt guard bands to reduce or expand the specification limits, minimizing the risks. The spreadsheet created shows the risk of false decision for a MU value, considering results within and outside the specification limits, allowing to verify the risk according to the analytical result obtained. The global risks values were practically equal to the expanded uncertainty values, as there is no tendency of the production process between lots within or outside the specification, but once the analytical result is known, the frequentist approach provides a more reliable risk estimate (specific risk). The specific risks estimated by Bayesian and frequentist approaches were divergent by the influence of the production process information on the first approach, which may overestimate or underestimate the consumer's and producer's risks regarding the frequentist approach. Failures in medicine conformity assessment can cause much damage, therefore, preventive actions such as developing, evaluating and/or optimizing analytical procedures, are essential in order to guarantee measurement uncertainties below or equal to the target and adopt routine strategies to minimize the risk of false decisions in conformity assessment. (C) 2020 Elsevier B.V. All rights reserved.