Issue With Validation of A Method

[mebecker1]

I have a method I need to validate. I am writing a method validation protocol and I really think there are issues with the method. The method is Determination of the Purity of an Alcohol using GC FID.

The purity of this alcohol SAMPLE material is approx 99.99% and is a raw material that is used in a manufacturing process. The method references USP 611, 621, and 467...467 does not apply because quantitation of the impurities is unnecessaryand I have red-lined that reference. The alcohols are Methanol and Isopropyl with a CoA indicating 99.99% purity.


The Methanol and Isopropyl alcohols are Category I, or ICH "assay." LOD and LOQ are not necessary for the validation. The selectivity is the issue.


1. All the injections are NEAT. This is not a mixture of alcohols.

2. The analytical chemist uses the sample as the system suitability (that is fine).

3. The analytical chemist uses the sample as the check standard. He uses this in case he decides to toss the last injection in the system suit.

3. The analytical chemist uses the sample as the sample.

4. Chemist was manually integrating the impurities until I suggested he lower the threshold from 30.00 to 0.05 and that worked beautfully., One problem solved.

The problem is he is using the sample for everything in the run based on the fact the purity is 99.99% He is not comparing it to anything. He has to compare it to something or FDA will reject the method and the validation.

The purity is determined by dividing the area of the peak for the alcohol of interest by the total area of the chromatogram. Using a Agilent JW DB 123-1346 which is 60 meters and really unnecessary for this analysis. But they want to stick with this column. That is fine, but this is a raw material being tested for use in a manufacturing process and must meet USP <1225>.

To validate the selectivity a standard has to be incorporated into this method. I just do not see that it is possible to ignore this.

The fact they use FID may lead to a problem with a hidden peak under that large peak the treat as the pure alcohol. I really think they need to run a standard. It will be difficult to procure a standard that is of higher purity than 99.99%. I do not think FDA will accept a validation that does not incorporate a standard for the selectivity and to show nothing is hidden under those alcohol peaks.

I understand this material is very pure but there has to be a way to incorporate a methanol and isopropyl standard

I think this method should be run on 2 columns and 2 systems : The Agilent 123-1334 which is 30 meters long and then on the 123-1346 which is 60 meters long. The column packings are identical only the lengths are different. And those alcohols come off at about 6 minutes. I think the 60 m column is overkill and each run takes approximately 25 minutes, but it is not prohibitive. I think that a standard has to be run to show the RT for the alcohol of interest and prove that there is nothing hidden under the peaks for the alcohols of interest, since the detector is FID. Have to say I do not think they will want to spend the 621.00 for the 123-1334. I know they want the methods validated as they stand, but do realize there have been mistakes. When methods are changed alarms begin to sound. I do not see how this can be accomplished without changing the method to include/incorporate a standard.

This has been an issue for a week now. They will have to prove to FDA that this material is pure and that the selectivity is accurate and correct by comparing it to something. I am not sure which way is the right way to go. I really think the method needs to be edited and amended. I do not think this will pass FDA criteria. I mean, this is a simple analysis. Area of alcohol/total area of chromatogram...but the validation has to prove selectivity and as it stands I do not believe that is the case.

Any ideas I would love to hear them.

[Peter Apps]

Focussing for the moment on the selectivity issues. There are two aspects; compounds that can be detected but that co-elute with the matrix peak, and compounds that cannot be detected by the FID.

There is no point at all in running the sample on two different lengths of column with the same stationary phase, since anything that co-elutes with the main peak on one column will co-elute on the other column as well. You need two columns with completely different phases.

A trickier problem is that the FID cannot detect water, which is a very likely impurity in an alcohol. You will need to do a separate determination of water content - Karl Fischer probably.

Peter

[Johnny Rod]

I'd say you need a standard otehrwise you've got area% (from your description) not an external standard assay. As Peter says you'd need to characterise this if it can't be bought with a suitable C of A stating its w/w content of analyte. Don't forget you need linearity and range if you're looking at ICH so you'll need to dilute the std a bit (say to 99% or whatever your spec limit is, or a little below your spec perhaps).

[mebecker1]

Karl Fischer is part of the analysis and it is done, But only for the IPA. It is not done for the MeOH. I mean there are a lot of problems here.

OOPS, My Bad...That is another problem KF is used for the IPA but not for the MeOH. I am glad you mentioned it because I have to.

We, meaning me and my team discussed this, and the up-shot was a different length would be okay. I was not convinced of that because I know of a similar assay that is being run with an OVV-101 and an OV-225. The peaks of interest come off 5 minutes apart. So, OKAY I will call Agilent and try and find a column with a different stationary phase that might work.

Okay TY. Will check for a different stationary phase and something faster. The analytical chemist hates waiting 25 minutes and I really think the analysis time is far too long. I have seen and integrated these chromatograms. I do think they will have to run a KF for all these alcohols. Other problem is EtOH is in the method but never analyzed for anything. EtOH is not used for anything.

I still think a NIST, Sigma Aldrich, or USP standard should be used. According to Category I, and ICH "assay" I do not have to quantitate the impurities. I do have to quantitate the main peak. But, if I do not have all the impurities I really cannot make that determination. I think that a standard has to be used to id that main peak.

I think I am trying to say, somehow I do not see this peak as being KNOWN unless a KNOWN is implemented. Something has to be used to actually ID the material. I have access to a CoA. I can get a CoA. It is my understanding the CoA states the alcohol is 99.99%.

Let me see what can do ...I MAY BE BACK...

[Peter Apps]

I would be very surprised if you can get a certified standard with a purity of higher than 99.9%. You then have the intractable problem that your sample has a higher purity than your standard, and you have to extrapolate off the top of the calibration line to get the mass fraction in your sample.

You could of course dilute the sample in something, and calibrate the diluted sample vs diluted standard.

What is the actual specification on the material - does it have to be 99.99% (in other words 99.98% is not good enough) ? or can it be (say) 99% ?. If the former then your only chance using GC is detection by isotope dilution mass spectrometry - no other technique that I can think of will give you the required accuracy.

Which of course begs the question - how is the 99.99% determination made in the first place ? If it is based on wet chemistry - oxidation with dichromate or something similar than you already have a primary method and you can safely forget about needing a standard.

Peter

Peter

[mebecker1]

I would be very surprised if you can get a certified standard with a purity of higher than 99.9%. You then have the intractable problem that your sample has a higher purity than your standard, and you have to extrapolate off the top of the calibration line to get the mass fraction in your sample.

You could of course dilute the sample in something, and calibrate the diluted sample vs diluted standard.

What is the actual specification on the material - does it have to be 99.99% (in other words 99.98% is not good enough) ? or can it be (say) 99% ?. If the former then your only chance using GC is detection by isotope dilution mass spectrometry - no other technique that I can think of will give you the required accuracy.

Which of course begs the question - how is the 99.99% determination made in the first place ? If it is based on wet chemistry - oxidation with dichromate or something similar than you already have a primary method and you can safely forget about needing a standard.

Peter

Peter

YES...That is part of the problem and that is why the analytical chemist does what he does. I am going to have to get the CoA. I am going to have to run these samples myself to make sure the validation works. But, I am positive the entire method will have to be red-lined and rewritten. THAT IS NOIT GOING TO MAKE PEOPLE HAPPY!!!! The only detector they have is FID.

The analytical chemist does not have the CoA. I have to go to the lab manager to get that. I see he has already red-lined the method and changed the RSD based on USP 621. This has to be done by USP and ICH so it is distillation or GC. and KF has to be done on ALL the alcohols. Method has to be consistent. This is a real mess. Not a difficult sample...But a real mess.. I just started this project 4 days ago and I cannot write the validation because the method sucks....

I will get the CoA today...these methods will be transferred to Indonesia...so they have to be right. FDA has already rejected the methods used in PR...and I am afraid this one just has to be rewritten and then validated...

[Peter Apps]

I disagree - this is a very difficult sample - demonstrating 99.99% purity on any organic substance is the kind of work that gets done in metrology labs, not in production QC. Your fisrt step must be to find out whether the 99.99% is an actual specification, or is it just a description (of questionable accuracy) of the material as it is ?

A further question - and answers to those in my previous post would be helpful - is it your job to make people happy

Peter

[krickos]

Hi

My first question is, is these alcohols used in final drug product or in process of manufacturing a drug substance? (quite different general requirements)

The references made in procedure in general seems non proper. As stated 467 do not apply, 611 is for liquids containing alcohol not "pure" alcohols, 621 is the only logical one (general chromatography text). These alcohols have their own monographs in USP intended for ingridients in final drug products not raw materials in drug substance prudution, however they do lack impurity testing. You may look at Ph Eur if they have anything if used in drug products.

As for using the sample for SSTs/standards? Why not use commercial availble 99,9% material and document/save/downlod CoAs? FDA approved of this in Q&As sessions when updated 467 chapter came into action.

A CoA indicating a purity of +99,9% do not necessary mean that specification is that thight. Normally monographs for normal QC testing do not get tighter than 99,0-101,0%. In process chemicals in drug product prodution tends to be lower.

And yes water content is a typical requirement (KF or TCD) for alcohols regardless if used in drug substance or drug product production.

[mebecker1]

i think i have figured out a way to approach this-take a look at what outline below and let me know what you think....

This is not a very difficult sample or a difficult analysis. This is the analysis of a raw material. 2 alcohols actually, IPA and MeOH. The analysis is: Purity of an Alcohol by GC,with FID. The purity of the analyte peak for each alcohol is determined by dividing the area for that peak by the total area in the chromatogram. It is not necessary to characterize any of the impurities.

I need to make sure there is nothing hidden under the analyte peak. So I am going to use anhydrous alcohols as the standards in the validation.

I have to write the validation protocol for the method and them I have to actually validate the method. This all has to be done very quickly and the department wants the methods validated as they are..but they are aware not all things are done correctly. I have the latitude to redline a method and correct it if I feel it is bad/wrong...There are many things wrong with the way this analysis is carried out. I do not want to destroy this analytical chemists career....But the fact that he uses manual integration and tweaks the integration to get 99.99% is just so wrong on so many levels...Not sure where to begin..

The methods developed reference USP 611, 627, 1225, and ICH R2, 467 has been red-lined and I see no connection since the only peak of interest is the peak for the alcohol. Quantitation of any impurities is irrelevant. That also leads to me using the SPECIAL CASE FOR VALIDATION AS OUTLINED IN 1225 WHICH REFERENCES % PURITY. In this SPECIAL CASE The result in % area calculation depends on adequate resolution and detection of all impurities in the sample.

I was able to show this chemist how to choose an appropriate threshold for automatic integration. Not sure why less than 1 had never occurred to him. It was the first thing that occurred to me. Also, explains why they use a 60 meter capillary column. Has a very large number of theoretical plates and yields enhance resolution and separation.

THIS SPECIAL CASE ALSO REFERENCES: Standards used in the validation must be ultra pure so there is no interference by impurities. I have to run the analysis on two columns with different stationary phases. I have those now. I contacted Agilent and will be using a 30 meter 0.32 id DB WAX column , 123-7033. The column they use is 123-1334 and is a JW DB 60 meter length 0.32 id. That should give me the 2 methodologies I need to validate.

There are 2 alcohol IPA and Methanol...I have now seen the CoAs and it was like pulling teeth to get them in my hands. I mean they are just kept under lock and key. These 2 alcohols are manufactured by Fisher there is not information regarding how the analysis for these chemicals is done.

I clearly see the analytical lab checks off 8 different attributes in the CoA that they must check off as being reported by Fisher. The purity of these alcohols is not 99.99%. The stated purities are as follows:

METHANOL: ASSAY: 99.8%, TEST VALUE: 99.9%, KF <0.10%
IPA: ASSAY: 99.5%, TEST VALUE: 99.9%, KF <0.20%

They only run a KF for the Methanol and they run an RI for the IPA. They do not care about the impurities and if they can get 99,99% from the GC they see their job as being done. They do not use the data from the KF for methanol when determining the purity of the alcohol peak. They only do it to validate the value in the CoA. I know GC/FID is not going to detect the water.

According to UP 1225 I have o use standards of very high purity when constructing the 2 methodologies so I have decided to go with anhydrous methanol and anhydrous 2-propanol with the same reported purity as the Fisher reagents the analytical lab is testing for use in their manufacturing process.

I have read a lot about standards and the requirement for characterization and I should be able to use these since "PURITY and DECLARED PURITY: a number indicating the content of the major constituent with respect to the respective testing method based on the water-free material"

I am also going to use the anhydrous alcohols as the SST for both methodologies.

I think this may work. Also, I have been told by the people who run this department that my validation protocol can include standards that are not used in the method and the method. That the method is still valid even if these standards are not used in the method. I have not actually found any reference regarding this directive n any of the USP or ICH guidelines. But, I have also been given the freedom to red-line anything in the method I feel is being done incorrectly.

So that s where things stand right now. I can get the anhydrous alcohols from Sigma Aldrich or a number of other places including NIST. I am pretty sure I will get a better CoA than Fisher provides the Analytical Lab. According to USP guidelines I do not have to characterize any of the standards they supply..

So I think in this way I can validate the method. I cannot be responsible for the fact that the analytical chemist tweaks the integration manually and actually gets a better value for the alcohol of interest than the CoA for that material indicates...I am not responsible for explaining that.

I think if I use anhydrous alcohols as my standards, check standards and SST that I do not have to worry about this assuming the separation and resolution for both columns is going to reveal anything else that might be in there.

[Peter Apps]

Have a closer look at the Sigma list of methanol grades - they do several with purities better than 99.9%. I am not sure that they will have a proper certificate of analysis for anything that is not specifically a reference material though.

Peter

[mebecker1]

Have a closer look at the Sigma list of methanol grades - they do several with purities better than 99.9%. I am not sure that they will have a proper certificate of analysis for anything that is not specifically a reference material though.

Peter

BTW-LOL @ The comment "Is it my job to make people happy"...Actually, that has been a big point of discussion in this group of validation consultants....So good point!!!!

THANK YOU FOR ALL THE HELP....

We are taking this approach: OUR JOB IS TO VALIDATE THE METHOD IN PLACE..And we are trying our damndest to not alter the methods in any way unless they are so bad we are forced to. If we are forced to do that then cannot validate a method that has a CR pending.

Okay I will take a better look at what Sigma offers. I have been looking for reference material. The only thing that bothers me about standards is a statement I read in USP that has to have been a serious mistake. I mean this statement was so bad I read it out loud to my entire group..the look on everyone's faces..was a Kodak moment. Something to the effect of USP standards do not have a designated purity and can be used without characterization all other standards from all other sources must be characterized and the CoA has to be verified.

I know for a fact that has to be wrong because I have worked with USP standards and have received clear documentation that states purity. The other 4 validation chemists said the exact same thing...Had to be a mistake in that printing of that paragraph...I am pretty sure NIST standards are acceptable as most branches of US government use them when presenting data in court. At least that has been my experience.

Now if I can get past the fact the I know that both 5890s have sticky injectors and that this new column is going to require some testing for implementation - I can order the Agilent Test Standard and validate the column is functioning. And I am going to install it on the 6890N not going to risk using this on the 5890s with the sticking injectors.

[mebecker1]

Have a closer look at the Sigma list of methanol grades - they do several with purities better than 99.9%. I am not sure that they will have a proper certificate of analysis for anything that is not specifically a reference material though.

Peter

Just a quick update: THANK YOU SOOOOOOOOOO Much. I found the perfect standard (LC/MS grade 99.9% purity Methanol and IPA). The Method is being red-lined for the EtOH, so no worries there. Looks like things will work out perfectly. Also, I now have another validation where I have to use USP 467 and I do have to quantify the impurities, organic VOCs. Just going to take same approach and I have a standar for this which easily exceeds the 99% purity stated in the CoA. So outside of all the new instrumentation issues things will work out.

This is the very first time I have ever actually asked anyone for help up here and I just wanted to express my appreciation for your time and expertise.

[krickos]

Somewhat beside the original topic but still intresting.

The only thing that bothers me about standards is a statement I read in USP that has to have been a serious mistake. I mean this statement was so bad I read it out loud to my entire group..the look on everyone's faces..was a Kodak moment. Something to the effect of USP standards do not have a designated purity and can be used without characterization all other standards from all other sources must be characterized and the CoA has to be verified.

I know for a fact that has to be wrong because I have worked with USP standards and have received clear documentation that states purity. The other 4 validation chemists said the exact same thing...Had to be a mistake in that printing of that paragraph...I am pretty sure NIST standards are acceptable as most branches of US government use them when presenting data in court. At least that has been my experience.

Well sometimes general texts may not keep up with specific chapters in pharmacopieas, might be one reason, simply overlooked, has happned before. Another may be that purity of reference for certain monographs is not that critical, if monograph refers to reference standard for IR identity and/or SSTs like resolution only, the purity (99,5 or 100,0%) is usually not critical compared to an assay. Still I agree, I would expect a designated purity on CoA.
Also, I agree that in my experiance FDA/EMEA accepts other recognized sources for reference standards unless generated in-house, especially if not availeble in pharmacopieas, NIST traceble standards for specific surface area (cement), latex particles (particle size distrubition) to mention some.