第二个问题，建议研发体系和商业化体系分开管理，因为二者要求有差别，如果不分开管理，研发阶段的很多活动会非常受限。研发体系的分界点可能每个公司每个产品不一样，监管部门更希望更早转移到商业化体系。一般最晚不晚于工艺验证前，如果可以在III期商业化规模成功放大后。

第1个问题有人回答了，我来回答第2个问题：

1.研发质量管理体系和商业化生产的质量管理体系需要分开。

2,如工艺开发阶段、临床一期、二期、三期、NDA）属于研发质量体系，而并不是商业化生产管理体系。

3,细分的话应当是临床生产与非临床生产。临床生产本质上是要按照GMP的原则进行。所谓的研发，更多的是指的早期研究和药学研究。

4.这个不能按照严格的阶段分：因为你即使进入了临床三期，同时可能也会进行药学研究。而应该按照场所或者情景分，比如说药学研究使用的是研发管理体系，而临床生产使用的是GMP体系。

关于第一个问题:  目视检查的限度可参考ISPE Cleaning Validation Lifecycle Applications, Methods, and Controls  6.3 Visual Inspection and Criteria

进行课件残留物限度研究，具体的标准和方法可参考以下内容：

6..1 Visible Residue Limit Studies
It is possible to conduct Visible Residue Limit (VRL) studies to determine the level of visible detection of residues
for many soilants and surfaces. A method to determine the detection level is by spiking decreasing amounts of each
residue onto testing coupons representative of equipment surfaces, allowing them to dry, and then having them
viewed by a group of observers. Multiple observers should view the residues under different light levels, from different
distances, and from different angles to mimic actual visual inspection conditions in order to provide a more rugged
visual limit.

"The lowest residue amount that is visible to all observers is the visual detection limit for that product." [3]

If the residues can be consistently observed at a known residue level, and this level is much lower than the cleaning
acceptance criteria, then VC provides a high degree of confidence that the equipment is sufficiently cleaned. Even
with a visual limit in place, VC is predominately not considered adequate by itself for establishing cleaning validation.
Direct surface sampling (e.g., swabs or TOC rinse measures) is also required [17]. However, VC could be used as a
criterion along with periodic sampling in a routine monitoring program after the cleaning validation is complete [79]. If
historical data shows that the visual limit is higher than what was obtained via coupon studies, launch an investigation
to verify. If the out of specification is correct, then raise the VRL.
For the VRL to have value, the data or results for the margin of safety for VI (distance between the safety limit and the
residue level represented by the VRL) must be sufficiently large to compensate for the variability between operators
performing the VI and also the inherent variability of the VI itself. Operators performing VRL determinations should
be qualified in the method. VI for VRL determinations should be performed by an operator and verified by a second
operator prior to the equipment being released for use. In addition, a periodic review of the controls is necessary after
the cleaning process has been qualified to ensure that performance has not been negatively impacted by increased
or new sources of variability and to confirm that a VRL is still a valid and justified approach.
If the VRL is at a level above the cleaning acceptance criteria, then VC has limited value in determining whether the
equipment is sufficiently cleaned  

我了解到的，目视检查没有验证的说法，但是需要对进行目视检查的人员进行资质确认（qualification），需要在规程中规定目视检查的方法。