5 or 6 injections for system suitability

[Alfred88]

Dear all:
The discussion regarding this topic is very beneficial!

We normally run 6 injections of standard for SS before running any other sample. Our SOP states that we should check the first six injections if they meet the requirements; however, in practice, usually we set up the instrument to run (in sequences) overnight, so we are willing to take the risk that the first 6 injections will not pass the RSD<3.0%. In fact, I remember we had to throw away (at least once) a sequence run because it did not meet this requirement.

Back to the original question, 5 is not the same as 6, so 'ym3142' may need to justify the deviation, or else redo the validation!

Alfred.

[Peter Apps]

Hi jdlh

If the system fails system suitability you need to look at why - even if the answer that you come up with is that the poor rsd was just in the upper tail of the normal distribution of rsds. Re-running sets of QC samples until you get a set that passes is the same as monkeys with typewriters hammering out the complete works of Shakespear if you give them long enough - random processes will make a rubbish system produce a set of good results every now and then.

Peter

[jzt]

At the minimum, this is a deviation from the protocol (you need to explain to QA why you didn’t follow the protocol).

To save yourself from having to repeat the validation, you need to have strong justifications. In our company, if your work is to support early phase projects (before end of phase 2), and the RSD for the 5 injections was very tight, you may be able to say your validation data are still valid.

In our drug substance assay methods, the RSD (n=5) for both peak area and retention time should be no more than 1.0%. One can not support a specification acceptance criterion of “98.0% to 102.0%â€

[Bruce Hamilton]

[quote="jzt"]At the minimum, this is a deviation from the protocol (you need to explain to QA why you didn’t follow the protocol).

In our drug substance assay methods, the RSD (n=5) for both peak area and retention time should be no more than 1.0%. One can not support a specification acceptance criterion of “98.0% to 102.0%â€

[Bruce Hamilton]

I'm not sure about the 2% RSD versus 98-102% comment though. From my vague recollection, there are plenty of USP monographs that specify API assay limits 98-102% with sample replicate SD of 2%.
The 98-102% are the limits according to that method, and should accomodate any expected analytical variations. It seems OK to me, but I could well be wrong, though.

Apologies for following up on my own post. I should point out that their are good economic incentives for manufacturers to have validated methods with improved precision compared to pharmacopoeia.

Many APIs are expensive, and if a generic manufacturer uses an analytical method that has 1% instead of 2% RSD, then they can formulate their product with 99% API, rather than 100%.

The incentive used to be much greater when some limits were 95 - 105%, a generic manufacturer with validated methods and stability data could formulate to 97%.

I'm not suggesting that the poster above works for such a company, but I have.

Bruce Hamilton

[sridharkam]

Dear Ym,
5 or 6 is not the issue. But the issue is that you have deviated from the protocol

[DR]

double post

[DR]

Apologies for following up on my own post. I should point out that their are good economic incentives for manufacturers to have validated methods with improved precision compared to pharmacopoeia.

Many APIs are expensive, and if a generic manufacturer uses an analytical method that has 1% instead of 2% RSD, then they can formulate their product with 99% API, rather than 100%.

The incentive used to be much greater when some limits were 95 - 105%, a generic manufacturer with validated methods and stability data could formulate to 97%.

I'm not suggesting that the poster above works for such a company, but I have.

Bruce Hamilton

Given the six-sigma climate of late, I doubt that such practices are actually tolerated. If a series of lots were consistently on the low end of assay, quality issues would certainly surface on stability (unless a company has the luxury of either a very short shelf life or a rock stable compound). Personally, I hate seeing a 97% release assay because I know that it will be that much closer after 3months at H4 for a less than perfectly stable product... If any of you find yourselves working for a company that is pound-foolish enough to target low release assays, I hope your résumés are up to date.

[Peter Apps]

Hi DR

A few years ago I was in a lab that did QC analyses on a liquid that had to have a certain component at a certain concentration (no names no pack drill !). The target was 20 % (v/v) with a range of 18 - 22. Being simple analysts we could never understnad why they were consistently below the target at about 18.5 %, and we even considered offering to trouble-shoot their mixing process for them. It took months before it dawned on us that it made sense for them to dilute the active as much as they could while still leaving a little margin for error !.

Peter

[DR]

I pity the company that gets a FD483 for doing that. They'll have to recall existing lots, file process ammendments, make changes, revalidate etc. etc. etc. if the FDA decides that they're doing that intentionally and gets pissy about it ("misbranding" is in the preamble to the FD&C act that paved the way for the creation of the FDA).

[jzt]

Bruce wrote:
[I'm not sure about the 2% RSD versus 98-102% comment though. From my vague recollection, there are plenty of USP monographs that specify API assay limits 98-102% with sample replicate SD of 2%.
The 98-102% are the limits according to that method, and should accomodate any expected analytical variations. It seems OK to me, but I could well be wrong, though.

Many USP monographs use RSD of 2% doesn't mean that is the appropriate criterion. here's an example:

For the 5 injections of standard solution (instrument precision, not even considering weighing or solution preparation errors), the following values were obtained for individual injections: 97.5, 99.3, 99.7, 100.5, and 103%. The average is 100.0%, with RSD of 2.0%. This run will still pass system suitability. However, this means even if your sample has 100% potency, there is a good chance that the analysis result may be 97.5% or 103%, which will be out of specification of 98%-102%.

Today's HPLC instruments have no problem meeting the requirement of RSD < 1.0% anyway. why not give yourself more assurance and really use the system suitability to catch potential issues. In fact, we start to get worried if we consistently observe RSD > 0.5% during method development stage.

by the way, I have not had the opportunity to work for a generic drug company. on average only 10% of my work is under GMP.

[syx]

I have a substance that less stable in mobile phase due to oxidation and hydrolysis, the concentration of the API will be decreased significantly in 2 hours. The chromatographic method is long enough, more than 20 minutes. Could we use 3 point system suitability test for this case?

[ym3142]

syx,
the suggestion from me is to find a solution system which would not cause the hydrolysis and oxidation, such as change the pH , add anti-oxidant, keep in low temp and dark; use fresh samples. (may be you have to find the life time , and prepare the standard from time to time like every hour.

[ym3142]

JZT,
I agree with you. You did a good job to dig up what is under RSD 2%. I also agree in most situation our methods can reach RSD 1%. Unfortunately propably, it is not worth the time and trouble to pursue. One of the principle in method development is not OVERDEVELOPED.
In the true life, there are many elements which can cause RSD >1% though the method itself can give RSD <1% in 90% of the cases. Those elements includes: column is not stablized enough; column is not as good as when it is new; column is slightly contaminated; temperature fluctuation; reason-unkonwn base line drifting; electrical spike; LC vial issues, injector issues, air bubble isuues; detector issues and so on;
As you show, two out of 5 number are beyond +/_ 2%. but the validated method will be valid because we specify the recovery is within +/_2% in accuracy test calculated based on the average of the standard.
In The worst case, each tablet potency is "97.5, 99.3, 99.7, 100.5, and 103%. I would consider this is high quality drug product since you do take 500mg if you take five tablets(LC 100mg/tab). I will believe the therapeutical efficacy is as good as this drug intends to knowing the situation could be so great difference when people take the medicine.
"