Issue link: https://www.e-digitaleditions.com/i/1302985
Pharmaceutical Technology Regulatory Sourcebook October 2020 13 Table VI. Suitable statistical tools for traditional approach process performance qualification (PPQ) data analysis. Statistical tool Intra batch Batch-to-batch Considerations Acceptance criteria Considerations Acceptance criteria Descriptive statistic (e.g., mean, median, standard deviation, coefficient of variation (CV), range, etc.) Basis for evaluation that must be integrated with other tools Results within specification. Generally, CV% between samples is used to evaluate data Basis for evaluation that must be integrated with other tools Results within specification. Generally, CV% between batches is used to evaluate data Histograms Possibility to evaluate distribution shape, process centering, and process variation (suggested link with capability evaluation). Consider if enough data are available for intra batch evaluation Output should be normal distributed and centered around the mean of the specification Possibility to evaluate distribution shape, process centering, and process variation (suggested link with capability evaluation). To have sufficient data suggested to group all samples result from all batches Output should be normal distributed and centered around the mean of the specification Process capability (Cpk) Possibility to evaluate ability of a process to provide stable outcome within established limits. Appropriate population and normal distribution must be taken into account. Consider if enough data are available for intra batch evaluation. Depending on criticality of process different limits can be set, however, if Cpk ≥ 1.33, the process is generally considered centered and under control Possibility to evaluate ability of a process to provide stable outcome within established limits. Appropriate population (at least 30 data points) and normal distribution must be taken into account. To have sufficient data suggested to group all samples result from all batches Depending on criticality of process different limits can be set, however, if Cpk ≥ 1.33, the process is generally considered centered and under control Control charts, trends and shifts Generally, it is not applicable in PPQ (not enough batches) Analysis of variance (ANOVA) (nested anova) Generally, it is not applicable intra batch Possibility to compare means and variance, between batches Depending on the selected confidence level Table VII. Statistical tools for continued process verification. Statistical tool Considerations Acceptance criteria Descriptive statistic (mean, median, standard deviation, coefficient of variation, range, etc.) Basis for evaluation that must be integrated with other tools Results within specification Histograms Possibility to evaluate distribution shape, process centering and process variation Generally, it is not applicable if not linked to capability evaluation Trends and shifts Respectively at least 7 consecutive data points rising or falling and at least 7 consecutive data points below or above the mean (7) Trends and shifts should be avoided; however, they can be considered not critical if within warning limits (see control chart) Process capability (Cpk and Ppk) Possibility to evaluate ability of a process to provide stable outcome within established limits. Appropriate population into account. Depending on criticality of process different limits can be set, however, if Cpk ≥ 1.33, the process is generally considered centered and under control Control charts Possibility to evaluate trends, shifts and outcome of the process in comparison with specification limits; additional "warning limits" (8) can be applied to control charts (i.e. control limits) to increase the process control in order to better characterize process general behavior and to identify any out of control point. All data must be within warning limits and warning limits must be within specification limits. PCA (principal component analysis) Outliers, trend and shift detection Inside Hotelling's T2 limit; inside DModX limit Multivariate control charts Possibility to evaluate process trends and shifts in comparison with specification limits constructed by means of all critical process parameter (CPP) contribution. Possibility of early fault detection and correction. Critical quality attribute can be predicted on the bases of all CPP. All data must be within warning limits and warning limits must be within specification limits.