bob jefferson

Hi Everyone

How is Limit of Detection calculated for microbiology methods?

Thanks

Bob

dave l

By putting low known levels of what ever you are trying to detect into a
sample and seeing if you can recover it.
Repeat this a number of times, if appropriate with various sample types, and
you'll have your limit.

Dave

bob jefferson

I think chemists use 3 standard deviations when looking at the lowest
level - Is this used at all in biology?

Bob

dave l

I don't, but then it depends on what you're detecting and for what purpose.

Dave

jorge1907

No guys - in initial consideration, it depends primarily on the method. Once
you've estab;lished what you could detect, you validate that this indeed holds
true.
Understand up fornt what your assumptions areand how thse limit your detection
potential- esp if this is a culture method. For example, if you incubate
aerobically, you miss the anaerobes. More importantly, one can only culture
5-10% of the bugs in the world so accept that.
There are more things in heaven and earth, Horatio,
Than are dreamt of in your philosophy

bob jefferson

I am working on working out a limit of detection for a cryptosporidium
method.

Using a flow cytometer I spiked raw and potable water with known numbers of
Oocysts.

I have results for potable water spiking with 3 Oocysts. 4 samples were
negative, 4 were detected and 3 were 2 Oocysts found and 2, 3 Oocysts found.

I am trying to work out limit of detection for both raw and potable water
but can only find a chemistry statistical package that does it based on
standard deviation. I dont think this is correct and cant find anyone to
explain how it is calculated.

In theory it is possible to detect one oocyst in a volume of water, another
theory is based on the average % recovery the more I think about it the more
tied up I become!

Thanks for your time on this.

Bob

n10

Hi Bob

Seems to me that your up against the uncertainty of measurent barrier in
your current quest

Perhapes this will help


http://www.campden.co.uk/whatsnew/news55.htm
UKAS approval for uncertainty measurement procedure

Many microbiology laboratories are under pressure to estimate the
uncertainty associated with their measurements, but until recently there has
been little guidance available on how this should be done. CCFRA has been
active in this area for some years, and recently published a review on the
subject (Uncertainty associated with microbiological measurement - Review
15). A general procedure for the estimation of uncertainty of measurement
for quantitative microbiological methods has been accepted by UKAS in a
recent audit, and we are able to offer consultancy to other laboratories to
help them respond to the demands of customers and accreditation bodies. If
you would like to discuss this, or if you need help in developing your own
procedures, please ...

Best N10

bob

Seems like your intuition is telling you that the answer is not simply
a formula. It requires context. I agree. If you state that a sample
has no cryptosporidium, that should be translatable to a p that
someone will die (get sick) from the error in that statement. That
requires understanding the ****ysis, not blindly using a formula.

A few "picky" comments below, and then a general comment at the end.

So each sample has an exact 3, rather than a statistical 3?

So you had 13 samples, containing 39 oocysts. You detected 16 of them,
or about 40%.


Now, from your understanding of the assay... Is that % dependent on
amount? Can you concentrate the sample? Any idea why % is low? Is it
possible that the oocysts stick to one or another material in the
assay, esp at low concentrations? Does your assay distinguish live vs
dead cells?


I would suggest that you discuss this with someone with this specific
concern. Your local water company might be a place to start. Perhaps
the CDC or EPA (if you are in the US). You want to become knowledgable
about what is done now. You must be able to state your results in a
way that can be compared with current procedures/standards. (That does
not preclude that you may be able to come up with a more stringent
condition than currently used, because your assay is better. But you
must start with the current procedures as a reference point.)


regards,

bob

bob jefferson

Strange you should mention that but thats another one of my problems,
Uncertainty of measurement!

The auditor has mentioned that it is generally accepted that UOM for
microbiology methods is based on reproducibility of data. Thats more than
was said for LOD!

I will read Review 15 with interest!

Thanks for letting me know about that.

Bob

dave l

Sound like UKAS to me
Presumably you must be enumerating cryptos rather than presence/absence.
If you were doing the latter then u of m would no be applicable.

Going back to your problem you've already shown that you can detect 3
cryptos x% of the time.
You can either leave it at that or state a low level (4 or 5 perhaps) at
which you detect cryptos as near to 100% as makes no odds. If it is UKAS
they may require ongoing ****ysis to show you are still able to detect
cryptos at the same level.

Dave