5 replies to this topic

[poppysnoss]

Hello.

I have been tasked with dealing with measurement uncertainty. I hate it, hate it, hate it and just want to run in the opposite direction when I hear it mentioned.

I need to look at all our methods, so that we can work out UoM for our next audit. For quantitative tests, I am looking into RSD tables available from HPA. What on earth do I do with qualitative tests, e.g. detection of salmonella??? I have spent many hours lately trying to sort out when is needed but I think I'm going round in circles.

Also, do I need a separate procedure for measurement uncertainty, or will a short mention in the quality manual be suffice?

If anyone has even any small words of advice, I'd be eternally grateful!

[SaRaRa]

Well I am not sure if it can be much of a help but I 've made a small search. Maybe something of the following can prove to be useful.
Attention!!! Great linkstorm ahead!!!

Measurement uncertainty in microbiological cultivation methods

Internal quality control and uncertainty

Validating detection techniques

Considerations in the preparation of an experimental design for method verification

Risk characterisation

Uncertainty of measurement

Microbial food safety risk assessment

Validation in chemical measurement

The Salmonella Enteriditis risk assessment

The uncertainty of measurement

Food chemical risk analysis

Maybe more...

Good luck with reading!

Cheers!

[Charles.C]

Dear poppysnoss,

This can indeed be a highly circular subject.

The most frequent requirements IMEX for routine lab. micro. measurements are an ability to -

(1) Express the potential error in "counts", eg TVC plate counts
(2) Express the potential error in MPN statements
(3) Express the potential error in "not detected in Xg" type statements

There are a few standard statistical books aimed at micro.measurements if you hv tech.library access. There are also several standard refs which give nice worked examples if again you hv a library but limited use otherwise ??

Sarara's list above has some truly amazing stuff in it [ ] but (based only on the first one) tends to be a bit heavyweight (If the later ones are more directly useful, my humble apologies in advance, I definitely plan to work my way through them but slowly).

Anyway, in the context of 1 - 3 above, i suggest you hv a first look ar the review-type forum thread below which definitely covers 2-3 but quite can't remember if it explicitly looks at no.1 which is generally the simplest case of the 3 types. The books microorganisms in food nos 2,7 referred are probably the most frequently used starting texts due ease of understanding I think. A large portion of no2 is also on the IT somewhere, from memory, it is now superceded by no7(?) though a lot of it's basic content is still totally useful, just does not contain newer topics / applications.

http://www.ifsqn.com...?showtopic=4590

rgds / Charles.C

PS (= postscript ) - RSD = ????

HA = ???

added - I realised after writing the above that you were seeking more widely than just micro. stuff (hence Sarara's jumbo list of course). If you want to extend to things like weights and other QA variables, further encyclopaedias will be involved. Personally I never found that BRC auditors were competent (or interested) to go very deeply into such stuff. If you were being audited for ISO 17025 could be a different issue (no direct experience myself)

[poppysnoss]

Hello Sarara & Charles.

Many thanks for the links and advce. I'm sure I'll be able to find something in there to help. )

Charles - HPA is the Health Protection Agancy and they have RSD tables for UoM

http://www.hpa-stand...sop/pdf/rsd.xls

Thanks again, guys!

[SaRaRa]

You are welcome!

[Charles.C]

Dear poppysnoss,

Thks for the link and nice excel chart. I worked my way back to the qsop4 procedure which eventually revealed itself as a t-test, surprise surprise ! Their implicit suggestion to work with coefficients of variation is also valuable IMEX.

Nicely presented document except they don't warn you that typical obtained "errors" for plate counts may often be in the 100%+ level since no-one has the time, money or inclination to do a whole series of repeat measurements. Same comment regarding confidence limits for MPN, often enormous ranges, which makes their use for comparison to official standards in borderline cases of nMc situations practically unjustifiable (in fact at least one official method adds a factor to take account of such things).

Rgds / Charles.C