Author |
Message |
kevin Mackin (mackin)
New member Username: mackin
Post Number: 1 Registered: 5-2008
| Posted on Friday, November 14, 2008 - 9:31 am: | |
With regards to the EMEA guidance notes this is a response I received recently from the USP in relation to using NIR for identification: You ask, I was wondering what the current thinking of the USP/FDA was with regards to the chemometrics that are used for treatment of the spectral data. There is no formal thinking alsong these specific lines. Anu chemometric tools used should be validated with respect to the algoritms employed, what is the impact of the result if errors occur in misllableng of spectra. for instance, or the wrong order of transforms are performed. So the user should be very well versed on what are proper working software functions versus a malfunctoning one. The control that the user should have over chemometric software should follow 21CFR Part 11 requirements. You should always have original raw untreated spectra in reterntion, there should be some audit capability of the software to monitor changes done to a final model / algorithmbydata / time stamp of change, the name of the userwho originally built a model versus the name of the user who changed the model input variables, authorized supervisorwho controls who has use of the model, who are model builders, who are model users, who is allowed to make changes etc... You ask, Are you aware of any USP/FDA guidelines with, respect to chemometrics? No, not to the extent that you have given for the EMEA guidelines. Generally most of the NIR community over here in the US beleive that these guideleines go to far in legislating what a user can do to obtain useful calibration models for the purposes described. Your statement, "that Principal Component Analysis should not be used for identification purposes and is more suitable for quantification. It also states that combined use of the Wavelength Correlation (WC) of the second derivative spectra and Maximum Wavelength Distance is most suitable for identification purposes," should not be included in a general guidance. It is up to the user to make these kinds of determinations. I would ignore statments like that if you can show a valid reason why PCA can in fact be used for identification purposes, or that wavwlwngth correlation can in fact be useful in second derivative. My opinion of this kind of legislation is that it is hogwash! You must procede with caution in creating NIR models for the purpose of identification of raw materials. Identification can be done in many modes. For instance to identify at a the level of quality of powder for size, you would assess the spectra for their sameness across wavelength space. If you want to assess the spectra for a property such as size or moisture. and not just fetermine its quality but comparison based on a numerical value such as particle size in micronsor moisture in % Water,then you must employ some kind of regression analysis. In both cases, you must have sufficient sample representation across the wavelength space, or principle component space in order to have a well trained model that will be capable of accepting normal material for manufacturing purposes,versus rejecting bad material not meeting the specification. Currenly, USP is drafting a general chapter on chemometrics, but does not provide guidance on how to employ the various algorithms, just introduces the use to each.. My best advice is utilize these packages no differently thn any other analytical tool, know their capabilites, know thier ability to be audited, how 21 CFR Part 11 compliant is the software etc.. know what are your objectives are, plan how to achieve them and be able to prove it. Finally, allways be able to demonstrate whay something fails based on the spectra, and if you cqan link that failure to process performance or to product performance. Every NIR failure (identification rejection) is an oportunity to understand your process! You really must be able to convince the manufacturing people to create the ability to continue to use the material out of the loop (non GMP manufactueing site), if you will to learn why the material will be acceptable to the process or the product. This is perhaps the biggest challenge we all have if we are going to ever get to process understanding. Good luck. Please feel free to contact me further if you any other questions. GER Gary E. Ritchie Scientific Fellow For PAT U.S. Pharmacopeia 12601 Twinbrook Parkway Rockville, MD 20852 Phone: 301-816-8353 email [email protected] |
iyas (iyas)
Intermediate Member Username: iyas
Post Number: 17 Registered: 7-2007
| Posted on Friday, November 14, 2008 - 1:39 am: | |
dear all ihave tried the the link http://www.emea.eu.int/pdfs/human/qwp/330901en.pdf but it was withdrawn, i think that the right link is http://www.emea.europa.eu/pdfs/human/qwp/330901en.pdf |
iyas (iyas)
Intermediate Member Username: iyas
Post Number: 16 Registered: 7-2007
| Posted on Friday, November 14, 2008 - 1:07 am: | |
if you please i have tried the http://www.icovps2008.org but it did not work i think the right website is http://www.icopvs2008.org/ kind regards |
Howard Mark (hlmark)
Senior Member Username: hlmark
Post Number: 203 Registered: 9-2001
| Posted on Thursday, November 13, 2008 - 10:38 am: | |
Gaith - an NIR-specific validation procedure was published a few years ago, as a two-part article: 1) J. Pharm. Biomed. Anal; 28; p.251-260 (2002) 1) J. Pharm. Biomed. Anal; 29; p.159-171 (2002) There is also a more recent general specification for validation of any chemometric-based analytical method, provided by ASTM, it is Practice E2617. I hope these are useful to you. \o/ /_\ |
Gaith Zubari (mgz)
New member Username: mgz
Post Number: 1 Registered: 11-2008
| Posted on Thursday, November 13, 2008 - 10:15 am: | |
how can I make validation to the NIR Methods? Is There are any Procedures for that? |
venkatarman (venkynir)
Senior Member Username: venkynir
Post Number: 52 Registered: 3-2004
| Posted on Tuesday, November 27, 2007 - 11:06 pm: | |
Dear Sharma ; NIR and chemometrics is one of the vital one for PAT in pharamacetical industries .Kindly go through web page http://www.icovps2008.org It may helpful to you. |
Sheelagh Halsey (shalsey)
New member Username: shalsey
Post Number: 3 Registered: 12-2006
| Posted on Tuesday, November 27, 2007 - 12:52 pm: | |
Typically 80-120% of the nominal assay value is used as the range for quantitative calibrations. |
Deepak Sharma (deepak)
New member Username: deepak
Post Number: 2 Registered: 11-2007
| Posted on Tuesday, November 27, 2007 - 12:28 pm: | |
Thanks to both of you for prompt response on my query. Pl. also suggest me on the range of samples required during a quantitation method development. Say,my specification limit for a product is NLT 99.0 %(99.0 to 100%), up to what range of samples are required to cover this spec .limit on NIR. Is there any fixed ratio in-between,or any kind of rule . How far I should go in the lower side of the specification limit to create a calibration model for quantitation method. What you recommend for this. (99.0 to 100% values are w.r.t Chromatographic Purity By Gas Chromatography.) Thanks |
Sheelagh Halsey (shalsey)
New member Username: shalsey
Post Number: 2 Registered: 12-2006
| Posted on Monday, November 26, 2007 - 12:09 pm: | |
There are also guidelines from the EMEA and PASG groups: http://www.emea.eu.int/pdfs/human/qwp/330901en.pdf http://www.pasg.org.uk/NIR/NIR_Guidelines_Oct_01.pdf Sheelagh |
Bruce H. Campbell (campclan)
Moderator Username: campclan
Post Number: 109 Registered: 4-2001
| Posted on Monday, November 26, 2007 - 11:06 am: | |
One of the first things to read is the part in the USP on validation. Bruce |
Deepak Sharma (deepak)
New member Username: deepak
Post Number: 1 Registered: 11-2007
| Posted on Monday, November 26, 2007 - 10:55 am: | |
Dear Sir I have recently joined your Forum. I am working in a pharmaceuitical organisation and using NIR for Identification and Quantification of Raw materials. Can you guide me on the method validation part of the instrument.What are nesessary parameters required to validate a new method. What are the requirements for regulatory point of view. Looking for a good response from you. regards |