最近看到一个工艺验证方案中混合均匀性的评价还是取7个点，计算RSD<5.0%。从实际经历说说这个评价标准的不合理性。仅个人观点，也许有很多不对之处。

    2014年做的一个10mg的胶囊（内容总重约120mg）。大概历程如下：工艺验证第一批，混合均匀性评价：取7个点，计算RSD<5.0%，合格。然后充填胶囊，胶囊的含量均匀性也完全合格（内控95-105%，货架标准90-110%）。车间做完后所有都看似很成功，但仔细看数据就不对了，混合含量均匀性7个数值中有两个差异超过10%，按照RSD<5.0%仍然合格，按照的我的理解均匀性应该与最终制剂相关，既然有超过10%的差异，成品也有超过10%的可能，QBT思想成品检验合格，但不代表所有的样品都合格。

    经过各种努力说服QA、车间修改验证方案3点内容：（1）取样量要准确，取样量不能过多，取样要反映成品的均匀性，太多有混合作用，有掩盖成品不匀的风险；（2）增加充填过程中含量均匀性的检测，大概每隔30分钟取样检测；（3）混合均匀7个点，RSD<5.0%，没有依据反驳暂时先按这个标准，如果充填过程中含量均匀性不合格再重新调整。按照这个思路再做一批充填过程中含量均匀性就不合格了。重新调整工艺后把RSD调整为3.0%。重新工艺验证，基本都能满足要求。这才放心。

  查了一下相关资料。重点看一下ISPE对混合均匀性理解。可能完全按指导原则并不能做出好药，关键在于思考和理解，不过多沉溺于现有的文件。如有偏见，请指出，共同学习、进步。

15. FDA recently announced the withdrawal of its draft guidance for industry on Powder Blends and Finished Dosage Units — Stratified In-Process Dosage Unit Sampling and Assessment. What were the Agency’s major concerns with this guidance?
FDA’s major concern was that Sections V and VII of the withdrawn draft guidance no longer represented the Agency’s current thinking, as explained below.
Section V (Exhibit/Validation Batch Powder Mix Homogeneity) recommended that at least three replicate samples be taken from at least ten locations in the powder blender, but that only one of the three replicates be evaluated to assess powder blend uniformity. The Agency currently recommends that all replicate samples taken from various locations in the blender be evaluated to perform a statistically valid analysis. This analysis can demonstrate that variability attributable to sample location is not significant and that the powder blend is homogenous. Statistical tools are available to ascertain both the number of replicates and the number of sampling locations across the blender that should be analyzed to conduct a valid analysis.
Section VII (Routine Manufacturing Batch Testing Methods) acceptance criteria designated to the Standard Criteria Method and the Marginal Criteria Method were based upon the limits published in the United States Pharmacopeia (USP) General Chapter <905> Uniformity of Dosage Units. However, the procedures and acceptance criteria in USP <905> are not a statistical sampling plan and so the results of the procedures should not be extrapolated to larger populations. Therefore, because the procedure and acceptance criteria prescribed in section VII provided only limited statistical assurance that batches of drug products met appropriate specifications and statistical quality control criteria, FDA no longer supports their use for batch release. Currently, there are several standard statistical practices (see references) that, if used correctly, can help to ensure compliance with the current good manufacturing practice (CGMP) regulations, including 21 CFR 211.110 Sampling and testing of in-process materials and drug products, 21 CFR 211.160 General Requirements [Subpart I, Laboratory Controls],and 21 CFR 211.165 Testing and release [of the finished drug product] for distribution.1