Method Validation

[Pasha]

For performing method validations (HPLC), must you have minimum criteria set prior to the validation? If you have set minimum criteria in your study plan, and then it exceeds that minumum criteria, is it okay to then write you method with expanded acceptance criteria, or is this considered a failure of the validation and the method must be revised?

What is acceptable acceptance criteria for an analytical method - ICH guidleines indicate what portions of the validations must be performed but is there any kind of guidance that indicates what the acceptable criteria for the validation parameters would be (for example accuracy must be within 10% of target concentration etc.)?

[Mark Tracy]

Setting specs is a ticklish subject. You need to balance the external requirements (regulations, competitive pressures, yield, downstream processes, etc.) against the intrinsic capabilities of your assay method. Part of the balance is a statistical analysis of the method to see if it can deliver the external requirements. This is usually a committee effort, and the chemist is one member. Internal specs like system suitability can be handled within the lab, but external specs have to include more people.

Some specs are proxies for analytical quality. The manufacturing plant really does not care about the tailing factor of your peaks, but they do care about the accuracy and reliability of the assay results. Tailing affects peak resolution and the precision of the integration. Therefore you need an internal spec for tailing that guarantees that the assay is unaffected by tailing effects.

When you go into validation, the specs are preliminary based on R&D work. Final specs can only be set after you have enough data under production conditions, and validation is where you get that data.

I can't really offer rules of thumb because they are only a first approximation to the real specs.

[Pasha]

Okay, but what if you are only analyzing material that is purified (not in any biological matrix and has not required extraction ) - at what point do you say the method is not suitable? If you find in your validation that accuracy is at 80% is that acceptable for an analytical method? I thought for purified material the specs should be pretty tight.

[Mark Tracy]

It all depends on who wants to know. Most pharmaceutical formulators want assays accurate to 1 or 2 percent. You have to ask the customer (or the marketing department whose business it is to ask the customer).

Personally, I would not settle for ±20% precision or 80% accuracy for an assay of pure material, but that is a rhetorical case. I hope.

[tom jupille]

You have "the cart before the horse". Your analytical accuracy and precision need to be good enough to allow control of the process or product; in practice, that means significantly better than your control parameters.

If you are an organic chemical manufacturer selling an API specified to be >95% purity, you need precision on the order of ±1-2% (look at it this way, if you had ±10%, you would never know for sure if your product was in spec).

I'll repeat Mark's statement:

You have to ask the customer (or the marketing department whose business it is to ask the customer).