LOQ>Sample>LOD help

[tayzyboy]

Hi all, we have a method with LOD at 0.1nM and LLOQ at 2nM .
We have validated the method fully. However when analysing "real" samples we found that quite a lot of our concentrations are in this 0.1-2nM range.

We tried re-evaluating the LLOQ at 0.5nM but our accuracy is roughly 50% out. (though precision is good).

It is no-longer an option to re-develop the method. (real samples are all gone).

Is there any way we can report something other than <LOQ for this data??

For example something like "1.2nM +/- 50%" . as this is obviously at a higher concentration than "0.2nM +/- 50%".

We are not trying to cover up the fact that our method is not really suitable for the samples, but it would be nice if we could give a better indication of concentration than simply "<LOQ"

Anyone have any ideas or past experience of this problem?

Thanks

James

[DR]

I wrote conditional formats in Excel to handle these. In one assay (one of Rod's), we had to report X if it was >0.5µg/mL. LOD was ~0.2µg/mL. For values in between the LOQ & LOD, the format would report just that ("between 0.2µg/mL & 0.5µg/mL"). If the result was <LOD, it would report "<0.2µg/mL".

[<0.2]"<0.2µg/mL X";[<0.5]"between 0.2µg/mL & 0.5µg/mL X";0.00"µg/mL X"

[ravenwork]

It is legitimate to report detections for these samples. The detections are real since they occurred above your validated LOD. Stating that there was a detection without assigning some number to it is somewhat absurd. Your method gives a number, but that number must be flagged as being below your LOQ. If you clearly identify the numbers as being below the LOQ, then all is well.

In environmental testing, we do have to be careful with this. Many clients would much rather have us just report <LOQ, rather than designate a highly inaccurate result, even though that result is a real detection. They have their own regulations and clients to deal with so we accomodate them, even though our retained, raw, data clearly shows the detections. Other clients aren't as squeamish about the truth and want the flagged numbers.

Evan L. Cooper

[tom jupille]

it would be nice if we could give a better indication of concentration than simply "<LOQ"

The catch is that you can't. By definition, LLOQ is the level below which you cannot reliably measure concentration. As Evan suggested, how you report is ultimately driven by client requirements/preferences, but be careful, because if you give a number, no matter how hedged with ranges and disclaimers, someone somewhere will jump on the opportunity to misinterpret it.

[Bruce Hamilton]

Hi all, we have a method with LOD at 0.1nM and LLOQ at 2nM .
...
We tried re-evaluating the LLOQ at 0.5nM but our accuracy is roughly 50% out. (though precision is good).

Is there any way we can report something other than <LOQ for this data??

It's quite common to report results as " between LOD and LLOQ ", however my question would be - Why is the difference between the two so large?. I would look at the chromatograms and try to determine why your LLOQ has to be so much higher than typical for LOD/LLOQ ratios.

You need to be confident that chromatography is consistent, and you understand why the LLOQ has been set so high.

If your are truly desperate, and are prepared to sign your name to numbers that may be little better than those in a telephone directory,
you could report numerical values with weasel-worded qualifications, and " for information only ". I would still be loathe to issue any numbers for results below 3x the LOD, unless I understood why the LLOQ was so high.

In my world, "For Infrormation Only" is really the last desperate option when a research client wants to ascertain if any trends exist, and will not make any quality decision using the data, or report it unqualified as valid data from the method. You have to understand your client's imperatives, it's about risk management and trust.

Some quality critters will probably go into cardiac arrest, but some around here have accepted that FIO data can be useful. YMMV.

Bruce Hamilton

[tayzyboy]

Thanks,

Lots of things to think about there. - Especially the "people will jump on a number" comment . We've had that in the past, it can cause problems.

I'll carefully consider all the advice.

Cheers everyone

James

[gbalock]

You need to be confident that chromatography is consistent, and you understand why the LLOQ has been set so high.
This is what Bruce said and this leads me to the question, What are your criteria for determining LOD & LLOQ? The ICH guidelines in the Pharma industry say that a peak is detectable when it is three times the noise and LLOQ is when the peak is ten times the noise. If your LOD is truly three tiems the noise at 0.1nM, then your LLOQ should be at three times the LOD or 0.3nM. Also, recovery at very low levels can be plus or minus 50%--this isn't the assay. So, if your precision is good at low levels (RSD <15%) report the numbers, they're real

[barry]

You need to be confident that chromatography is consistent, and you understand why the LLOQ has been set so high.
This is what Bruce said and this leads me to the question, What are your criteria for determining LOD & LLOQ? The ICH guidelines in the Pharma industry say that a peak is detectable when it is three times the noise and LLOQ is when the peak is ten times the noise. If your LOD is truly three tiems the noise at 0.1nM, then your LLOQ should be at three times the LOD or 0.3nM. Also, recovery at very low levels can be plus or minus 50%--this isn't the assay. So, if your precision is good at low levels (RSD <15%) report the numbers, they're real

I was agree with you, i alway set LLOQ as 10 three times the noise (if the precision is good enough).