Guidance for number of injections required.

[StephenO]

Is there any guidance on how many injections should be preformed on a sample when carrying out HPLC Analysis for the parameters below:

1. Assay
2. Dissolution
3. Uniformity of Dosage Units

Looking forward to your posts.

Regards
Stephen

[Terry]

I have a similar question on this topic, under what circumstance would double sample preparation be necessary?

[DR]

What we typically do -
1. Assay
2. Dissolution
3. Uniformity of Dosage Units

For routine analysis, single injection of single prep for all of the above after completion of initial blank(s), resolution solutions (if any) and 5 or 6 initial standards to establish system suitability. After each group of 5 (for some assays) to 12 (dissolution) samples, we inject at least one more standard, and sometimes a blank though I prefer using pairs of standards.

For validation (qualification of method, intermediate precision), plan on triplicate injections of triplicate preps for all samples. Standard routine is about the same as for routine runs.

Anyone else asking for duplicate injections, duplicate preps etc. is just begging lots of difficult questions about why this is needed and what should be done with results that differ.

[Consumer Products Guy]

We do duplicate injections, from one sample preparation. Our QA would have a fit if we suggested going to one injection. So our test procedures detail two injections, so we have to follow those, cGMP.

[StephenO]

Would two injections still be required for the Dissolution Sample and the Uniformity of Dosage Units Sample.


If so is this stated anywhere in the Pharmacopoeia's or by any regulatory body European or US.

Thanks

[HPLCaddict]

I'm not aware of any official guideline dealing with how many injections are required for certain tests.
By chance, I just talked to a colleague about this topic who attended an HPLC/GMP-Seminar. Practice in the different laboratories of the attendes was anything from single to triple injections. Statement of the lecturer regarding this: "Anything is legal, as long as it's in the SOP!".

My personal opinion: A MUST-HAVE is a decent SST matching the specific method at the beginning of each sequence. But even then single injections may fail, due to a bubble or whatever. For dissolution and content uniformity you've got several single values (6 resp. 10), so if there's a single outlier this does not immediately mean you're generating an OOS with all the hazzle involved. Furthermore, specifications usually are rather wide. If your dissolution specification is something like >85%, who really cares if you get a mean value of 95% or only 88% because of one single low value? In-spec is in-spec. If you're OOS because of one outlier, double injections won't help you anyway. Just omit the one injection that's an outlier? No way!
Therefore, concerning dissolution and CU, I vote for single injections.

Concerning assay determinations, it depends on the precision of the method vs. the specification limits. If you've got tight specifications like 98.0-102.0%, it might be wise to use multiple injections to narrow down the problem if you get outliers (system or sample related?).

[aceto_81]

We usually start with 5 injections of a reference, followed by samples, and always stop with a reference to check if the response hasn't changed over the time.
This has now been using for years, but no one knows where this comes from.

Now if we didn't prepare enough mobile phase, and our last reference isn't injected, we do need to restart everything according to our current practices.
But if we can clearly state that the mobile phase bottle was empty, do we really need to restart?
Or can we get away with an explanation?
Also do others always stop with a reference?

Thanks

Ace

[lmh]

aceto_81, I'm guessing you've answered your own question: if the idea of the final reference is to check that the response hasn't changed, then if you don't have that final run, you can't be sure the response hasn't changed. If the response might have changed, you don't know when, and so you have to reject all the samples and start again. The fact that you understand why the runs stopped doesn't help! If this happens often, you could of course put some reference runs at regular intervals amongst your samples, and then you only have to reject all data since the last good reference.

As regards replicate testing, I'd have thought the logical answer is: If you have validated your method and know the standard deviation of the measurements, then you can test the probability of a single measured value coming from an out-of-spec population. In practice you'd just narrow your acceptable range to account for the standard deviation (if the answer must be greater than 10, but the error is +/- 2, then it's quite likely that an answer of 11, "pass" actually came from a material that should have scored 9, so depending on what error-rate we are prepared to accept, we add some standard deviations to our minimum acceptable answer...). Of course you will get occasional out-of-spec answers because of outliers and random chance. If out-of-spec answers aren't very expensive, you can accept this. If they are catastrophically expensive, it makes sense to make replicate measurements (cheaper!) and look at the mean (which has a standard error, i.e. its standard deviation is root-n smaller than the standard deviation of individual measurements). If outliers are a problem, you could consider medians instead of means, but that's beyond my statistical knowledge. The main thing is to minimise the overall cost of achieving the necessary quality, balancing analytical costs against the costs caused by taking action on faulty analyses.

[aceto_81]

aceto_81, I'm guessing you've answered your own question. ...

Damn, I knew it...
But do other (pharma) companies always use a reference injection at the end of the sample list?
Or is this just as one states in the inhouse SOP's?


Ace

[StephenO]

We carrry out double injections of reference solution before and after (bracketing reference solutions) 6 injections of sample or at the end of the sample set if you havnt got 6 injections of samples solution.
The mean of the 4 bracketing reference solutions are then used in the calculation of the sample.

Regards

[DR]

We do duplicate injections, from one sample preparation. Our QA would have a fit if we suggested going to one injection. So our test procedures detail two injections, so we have to follow those, cGMP.

It's only CGMP if it says to do that in your own SOP. So, what does your QA suggest doing if you get two different results? How far apart do they have to be before the difference is considered to be "significant"? Is it only significant if one result passes and the other doesn't?...

I could go on pointing out how many more tiger traps are created than filled in by using duplicate injections. It is just not a good idea, in my opinion. It also sort of subverts the whole reason for method validation, instrument calibration checks and other items that are usually done so that you can have confidence in single injections.

[Consumer Products Guy]

We do duplicate injections, from one sample preparation. Our QA would have a fit if we suggested going to one injection. So our test procedures detail two injections, so we have to follow those, cGMP.

It's only CGMP if it says to do that in your own SOP.

We don't have SOPs that detail any specifics like number of injections. I myself think one injection is not sufficient. Our test methods state to average the results of the two injections, and that is the reported value. So that's what we do. If that's OOS, then an OOS investigation must be done.

[StephenO]

Our test methods state to average the results of the two injections, and that is the reported value. So that's what we do. If that's OOS, then an OOS investigation must be done.[/quote]


Do you calculate the Average response of the two injections our calculate each injection seperately?
Do you have a %CV criteria between the to injections responses or result, that has to be met?
What do you report if one of the injections is OOS and the Other is in specification?
If one injection gives OOS and the other is within specification and the average of the two is within the specification what is the procedure then?

[Consumer Products Guy]

I'll answer as best I can, in capiatls for clarity. I'm a separations scientist/industrial problem solver, and that includes test method development for topical OTC formualtions. QA and compliance are not my specialty field.

Do you calculate the Average response of the two injections or calculate each injection seperately?
THE CHOMATOGRAPHY SOFTWARE CALCULATES THE % ACTIVE FOR EACH INJECTION. WE AVERAGE THAT.

Do you have a %CV criteria between the to injections responses or result, that has to be met?
NO.

What do you report if one of the injections is OOS and the Other is in specification?
THE TEST METHODS DETAIL TO REPORT THE AVERAGE, OR TO AVERAGE THE PEAK AREAS ARE CALULATE MANUALLY. SO THE RESULT THAT IS REPORTED IS THAT AVERAGE.

If one injection gives OOS and the other is within specification and the average of the two is within the specification what is the procedure then?
THE TEST METHODS DETAIL TO REPORT THE AVERAGE, OR TO AVERAGE THE PEAK AREAS ARE CALULATE MANUALLY. SO THE RESULT THAT IS REPORTED IS THAT AVERAGE.

[StephenO]

Ok so how do you know which injection is representative of the sample, the OOS or the injection which is within the specification?

An instrument malfunction could of occured on either of the injections and you dont know which injection that is, how can you trust your result?

Regards