Reporting Impurities

[Chris]

Does anyone have guidelines on how to report impurities for stability studies.

We set up a stability method by HPLC for one of our clients with 12 specified impurities with each being at approximately 1% of the active concentration. We changed from a direct area normalisation calculation to a %w/w calculation using relative response factors as we demonstrated that the area normalisation technique was not accurate.

What are the advantages and disadvantages of reporting by %w/w or by % of taotal purity, and can both of these methods be classed as acceptable for stability purposes?

[maris]

1)are all 12 specified impurities also the degradation products?

2) It's not a good practice to change the method or the calculation for more accurate during the stability study - you loss the ability to follow after the trends - in cGMP you have to use validated stability indicating method - only in this case you can be confident with the results.

3)The reporting of impurities should be done according to corresponding ICH/FDA guidances, use as references : ICHQ 3A(R) and ICH Q3B(R), for stability see ICH Q1A(R2).

[Chris]

The impurities are a combination of degradation products and process related impurities.

We changed the method and performed the validation prior to commencing the validation study.

I had a look at some of the ICH documents and it does mention reporting impurities as % of the active substance, however does this mean as % of the nominal active content or the actual measured active content?

[Chris]

When reporting impurities as % of the active is this better as % of the nominal or as % of the actual content?

[GoldLeader]

All-
I'm also looking for some advice from the veteran pharma chemists out there. We are having the same discussion as Chris right now about reporting imps as % theoretical or actual. For example, my active label claim is 100 mg/mL. If my assay is 98.0%, then do I report my impurities based on 100 mg/mL or 99 mg/mL? I.E., I have an impurity at 1 mg/mL. Do I report this as 1/100*100 = 1.00%; or 1/98*100 = 1.02%? The impact may seem negligible unless my spec is NMT 1.00% (the first result passes, the second fails). Looking forward to some opinions.
-Gold Leader

[ravi]

My opinion is tha you should do it at % of Claim (nominal) than actual as you could be biasing your results. I've some labs calculate all their peaks based upon their main standard peak (or reference deg/imp peak, if available). You need to determine if your extinction co-efficients are comparable first (and apply the appropriate RRFs).

This is the best way to determine if your degradate is not further degrading (or that the growth is realistic).

[BenNC]

I spent several years at a CRO and got the chance to see a lot of different methods from a lot of different sponsors. In that time, all of the impurity methods that I saw used the nominal label claim of the sample rather than the measured.

Of course there is always the err to the side of caution argument for impurities (when in doubt overestimate)

[DR]

I spent several years at a CRO and got the chance to see a lot of different methods from a lot of different sponsors. In that time, all of the impurity methods that I saw used the nominal label claim of the sample rather than the measured.

Of course there is always the err to the side of caution argument for impurities (when in doubt overestimate)

In my experience, the only time I've deviated from this is if we keep a development batch on stability and the initial assay values were pretty far off the target, or if we're doing some sort of excipient study and there's a fair amount of variation from sample to sample because they were all hand weighed. Then it makes some sense to relate RS peaks to the actual active concentration.

[ravi]

....In my experience, the only time I've deviated from this is if we keep a development batch on stability and the initial assay values were pretty far off the target, or if we're doing some sort of excipient study and there's a fair amount of variation from sample to sample because they were all hand weighed. Then it makes some sense to relate RS peaks to the actual active concentration.

Would you then compare it versus initial value or the actual value in the sample? Do you think that one would get a more accurate value with this type of normalisation?

[BenNC]

Another consideration might be the actual accuracy of you impurities assay. If you only require 90-110% or 95-105% recovery on your validation, then you are not going to be able to add any real accuracy by normalizing to the measured assay.

Along that same reasoning, if you see 1-2% variability in the recovery of your potency sample, then you may introduce more error into your impurity.

Keep it simple.

[Chris]

Thanks for your help.

We have recently performed some parallel testing with another lab on the same method and we have observed a difference in the total impurity values, they typically observe around 18% (by area normalisation) compared to our 20% (again by area normalisation).

There is a group of 7 impurities that do not show excellent resolution and we report by dropping perpendiculars to the baseline whilst the other lab perform valley to valley. The other lab developed the method as they manufacture the product but have not validated it for impurities, we will be validating the method for impurities over the next couple of weeks so this should highlight the correct integration technique.

However as a rule of thumb where baseline separation is not achieved should the valley to valley or dropping perpendiculars to the baseline technique be used?

I have always considered that it is far safer to report the worse case scenario as by performing valley to valley on several consequtive peaks you are excluding peak areas that are generated by the impurity components.

Thanks
Chris

[GoldLeader]

Chris-
I agree with you that proper integration technique would be dropping verticals instead of v-v. We consider it this way: Any point in our sample chromatogram with a response greater than that of the corresponding blank injection must be from the response of some related compound and not from the sample matrix. By drawing v-v baseline under a group of seven peaks you are artificially creating a "baseline hump" that would be absent in the blank. Therefore, this baseline hump (regardless of which specific peaks comprise it) must be accounted for in the total impurity report. Hope this is clear without pictures...

-Gold Leader

[Chris]

Thanks that is the same justification that I have given especially as the specificity analysis did not generate a disturbance under the seven peaks in question.