PDA TR60之6.2.4

6.2.4 Number of Lots for Stage 2 Process Performance Qualification (PPQ)
工艺验证第二阶段验（工艺性能确认）证所需批次数
The number of lots required for Stage 2 PPQ depends on the following:
工艺验证第二阶段所需批次数依赖于以下因素：
· Prior information about the process available from Stage 1 Process Design or quality history from similar processes. The more scientific evidence already available to establishes that the process is capable of consistently delivering quality product, the fewer the number of PPQ lots required.
来自于第一阶段（工艺设计）或相似工艺质量历史的以前的信息。在建立工艺（能持续产出优质产品）时，采用现存的科学证据越多， PPQ阶段所需的验证批次也就越少。
· Risk factors, including criticality of the product characteristics and extent of in-process quality control (e.g., PAT,100% inspection)
风险因素，包括产品特性的关键性和过程质量控制的程度（如： PAT， 100%全检）。
· Type of data: attributes (pass/fail) or variables (quantitative)
数据类型：属性型数据（合格/不合格）或者数值型数据（数量值）
· Statistical confidence desired
期望的统计学置信度
· Production rate (i.e., how often lots are produced). If only one commercial lot is produced per year, it will not be feasible to require a PPQ with a large number of lots.
生产率（也就是多长时间生产一批）。如果每年仅仅进行1批商业生产，这是不可行的，因为PPQ需要大量的批次。
Depending on the prior information and/or risk involved, it may not be necessary to determine the number of PPQ lots using statistical methods. The less information and confidence at the transition to Stage 2 (PPQ), the more advisable it is to use statistical methods to help determine the number of PPQ lots where feasible and meaningful. See the Appendix 8.1, Statistical Methods for Determining the Number of Lots, for statistical approaches to determine the number of lots. Regardless of the number selected and acceptance criteria used, the data collected during PPQ should be statistically analyzed to help understand process stability, capability, and within (intra-) and between (inter-) lot variation. Lots produced during Stage 1 under similar conditions as the PPQ lots may potentially be used to reduce the number of lots required at PPQ. This can be done using Bayesian statistical methods or by combining the Stage 1 data and Stage 2 PPQ results – if there are no significant differences in the data (50). The criteria for combining Stage 1 data and PPQ data should be specified before the PPQ lots are produced. These criteria would typically include such statistical comparisons as ANOVA (analysis of variance) to compare lot means, Levene/Brown-Forsythe or Bartlett’s test to compare the lot standard deviations, SPC charts, and equivalence tests to demonstrate that Stage 1 and PPQ data are similar (51).
依据所涉及的现有信息和/或风险，可能没有必要采用统计学方法来确定 PPQ 的批次数。在信息较少、置信度较低的情况下过渡到第 2 阶段时，采用统计方法来确定 PPQ 的批次数是明智且有意义的。见附录 8.1，采用统计方法确定验证的批次数。无论选择的数量和使用的验收标准是多少， PPQ 阶段内收集的数据均应进行统计分析以帮助理解工艺的稳定性、性能和批内和批间的变化。在第 1 阶段生产的批次（与 PPQ 阶段相似的条件下生产的）可能被用来减少 PPQ 阶段所需要的批次，这可以通过贝叶斯统计理论或合并第 1、 2 阶段的结果来（如果第 1、 2 阶段的数据没有显著差异的话）实现。合并第 1、 2 阶段数据的标准应在 PPQ 批次生产前规定。标准通常应包含一些统计比较（如：批次均值的方差分析、批次标准差的 Levene/Brown-Forsythe 或 Bartlett 检验、控制图和等效性实验）以证明第 1 阶段和 PPQ 阶段的数据是相似的。