Singleton, duplicate, or triplicate?

[Consumer Products Guy]

Poll: How many replicates (sample preparations) do you typically do for pharmaceutical products, like to test a batch? And how many injections?

[juddc]

1 replicate preparation for every 5 or fewer samples; replicate injections for samples are treated similarly (post system suitability with an N=6).

If I have one sample, I prepare it twice and inject one replicate twice. If I have 10, I prep 2 replicates and inject 2 randomly selected samples twice.

All of this applies to samples that are familiar being assayed via routine methods. For other situations, things might be different.

[DR]

Dissolution samples 1 inj. per time point per dissolution vessel. n= either 6 or 12 vessels depending on the protocol purpose.

Blend uniformity samples n=6 or n=10 samples, 1 injection of ea.

Content Uniformity n=10, 1 inj. per.

Composite assay, related substance assay n=1 taken from a pool of numerous comingled dosage units.

The only time we routinely do multiple injections of multiple preps of a given sample is when we're validating a method or verifying performance of an instrument.

[AdrianF]

I have raised a similar topic recently
http://www.sepsci.com/chromforum/viewto ... highlight=

If duplicate assays have show adequate reproducibility it is reasonable to move to single testing.

[jzt]

duplicate preparation of one sample may provide safeguard against sample preparation/dilution errors, or to demonstrate no significant error. however, I don't see the point of making multiple injections of the same preparation. system suitability already showed that there's no significant difference in repeated injections of the same solution (usually with the standard), why do you still need to demonstrate this with the sample preparation? if you only do this to a selected few (even claiming to be random), it is hard to justify why you treat some samples differently than the others.

[JA]

jzt,
If duplicate sample preparations provide a means to check against the occurance of sample prep or dilution errors, wouldn't duplicate injections of each help to check against random injection bias?

For a standard with N=6, the RSD where five results are tightly packed and one is a bit "skewiff" could still pass specification. I've seen instances where one of the six replicates was omitted due to being an outlier.

Is the point of it the trade off between a few more OOS investigations versus a degree of time saving for each run? Perhaps our repeatability criteria need to be tightened up?