Establishing LLOQ and Reporting Range

Discussions about IC and related topics

22 posts Page 1 of 2
Hi Experts! I'm looking for some knowledge on determining limits of quantitation.

Background: I've very recently taken over running a Metrohm 940 Professional IC Vario that is set up for simultaneous detection of cations and anions. The person who was running the instrument knew less about it than I do (which is saying a lot), so there are no established protocols or procedures for anything regarding it. My training consisted of "put the vial in the rack and press the play button".

Purpose: The lab is looking for the concentrations of standard inorganic cations and anions in samples we are sent from the field (we do a water extraction on the samples). The testing we do is not governed by any regulating body, although we are currently trying for ISO 17025. The method that has been cited on past reports is ASTM D4327, which is similar to EPA 9056A (part of SW-846).

Problem: In my opinion, in order to have any integrity as a lab and any confidence in the reports we send, we need to establish some kind of QA/QC for what we're doing. The first thing I'm attempting to tackle is defining a reportable range, and in order to do that I need to determine an LLOQ. However, given that I have no defined or regulated method, I'm struggling with how to do this. I also don't have a good grasp of statistics, despite my best efforts, so I get lost just trying to Google things. For example, I don't want to report a value of 0.005 ppm because I don't trust it, but I need to be able to provide a reason as to why I don't trust it.

Questions:
1. To determine the LLOQ, can I just pick a really low concentration, make a standard and run it several times, and then calculate something like %RSD and say, if %RSD is less than X we're good to go?

2. I've read in several places that the LLOQ can be calculated as 10 time the std dev of the blank. But I don't understand how that's possible for IC, because the blank run is a graph, not a single data point. How do you calculate std dev of multiple graphs?

3. Each anion/cation of interest will have its own LLOQ, correct? And it is standard practice to set this value as the lowest point of a calibration curve?

4. What other questions should I be asking?
Dear Kristin

MagIC Net has an automatically calculated result for the limit of detection ( RS.'Analysis name'.'Component name'.LDTCCONF95). Search for "Limit of detection" in help. Here you find all the calculations applied when using a linear calibration.

As an alternative you may apply the 'Signal to Noise' ration. For the you take the peak height of the lowest calibration level. For the noise you may take the Noise value measured for a defined time in the chromatogram (TP.'Command number'.NOISE Measured noise) or you may get the noise value by checking the height of the noise near to the peak of intrest (zoom to the baseline and estimate the noise level).

Both methods will give you the LOD of your working range (calibration range) and not an absolute LOD of the Instrument.
Dr. Markus Laeubli
Manager Marketing Support IC
(retired)
Metrohm AG
9101 Herisau
Switzerland
Markus,

Since I have very little understanding of the MagIC Net software (and I often find the manual confusing), let me make sure I'm understanding this correctly.

Under Method>Evaluation>Results>Statistics, I can define the LDTCCONF95 result for all components. Then when I run a sample using this method, the limit of detection will be calculated based off the calibration curve used for that sample?

Once I have this result, I can use it as the lowest point for a new calibration curve? Will calculating a new calibration curve then change the LOD?
Dear Kirstin

No Problem.

You need to run a multi-point calibration with (I would recommend) 5 levels.

The result of LDTCCONF95 in the last calibration runs or following sample runs give the LOD based on this calibration curve.

Only if you need to go below of your actual lowest calibration level, you may use the result for an approximation for a lower lowest calibration.

I would recommend to apply a calibration range where your expected lowest concentration corresponds to the lowest calibration level.
Dr. Markus Laeubli
Manager Marketing Support IC
(retired)
Metrohm AG
9101 Herisau
Switzerland
Thanks Markus! This is helpful.

If my LOD comes out as a lower concentration than my lowest calibration standard, are results between these two values reportable? Or is it standard practice to only report concentrations that are in between the lowest and highest calibration standards?
For example: say the calculated LOD is 0.01ppm, and my lowest calibration standard is 0.05ppm. If when running a sample the instrument gives me a value of 0.03ppm, can I report this number, or do I just say the result is <0.05ppm?
Dear Kristin

Typically we say, that only results within a calibration range should be reported.
Therefore the correct answer to your question is "you should not report 0.03 ppm".

Anyway, I usually checked my calibration curve visually. I then check the offset as well as the position of the calibration points. if the offset is a quite bit smaller than the expected signal for the LOD peak AND the calibration points all are close to the curve (here the RSD from the calibration information should be below approx <1.5 to 2%), I would report such a result.
But this requires some experience.
Dr. Markus Laeubli
Manager Marketing Support IC
(retired)
Metrohm AG
9101 Herisau
Switzerland
Kristin.Striber wrote:
Thanks Markus! This is helpful.

If my LOD comes out as a lower concentration than my lowest calibration standard, are results between these two values reportable? Or is it standard practice to only report concentrations that are in between the lowest and highest calibration standards?
For example: say the calculated LOD is 0.01ppm, and my lowest calibration standard is 0.05ppm. If when running a sample the instrument gives me a value of 0.03ppm, can I report this number, or do I just say the result is <0.05ppm?


TNI Guidance on Instrument Calibration
GUID-3-110-Rev0
December 5, 2018

Data reported below the lowest calibration standard concentration must be qualified.

https://www.nelac-institute.org/docs/gu ... 666708.pdf
anionman - thank you! there was some interesting and helpful information in that document.

Are most calibration curves for anion chromatography (looking at the 7 standard inorganic anions) best described by a linear or quadratic fit? Or does it depend on the concentrations?

Since I am not following an established or regulated method, I'm thinking to accept 6-point calibration curves that have an R2 value of 0.995 or better with a %RSD of 5.0% or less. Does this seem reasonable? Are there other statistical values I should be looking at?

Are there any methods that suggest averaging multiple calibrations to obtain a final curve?
Anion calibrations are linear. It could get non linear if you get column overloading in high concentrations. You don't want that. The method I follow( 300.0 )requires that you only use a linear calibration.
The initial demonstration of linearity must use sufficient standards to insure that the resulting curve is linear.

https://www.epa.gov/sites/default/files ... 1_1993.pdf

Notice that the only requirement for the calibration is that each standard be within 10% of target value. The method only requires 3 calibration points and that works because it is linear.
anionman wrote:
Anion calibrations are linear. It could get non linear if you get column overloading in high concentrations.


Dear anionman

Anion calibrations show inherently a certain nonlinearity. When you apply a three-point calibration you definitely will not realize that. Especially as the correlation coefficient of the special type of non-linearity shows a value close to 1.

If the calibration range is within one order of magnitude, a linear calbration is acceptable. For ranges larger than one order of magnitude, we recommend quadratic calibration as this gets a much better fit. An alternative is the high-low calibration. Here the we work with 2 or more calibration curves, where the highest point of the low-calibration ist the lowest point of the high calibration curve.
Dr. Markus Laeubli
Manager Marketing Support IC
(retired)
Metrohm AG
9101 Herisau
Switzerland
Kristin.Striber wrote:
Are most calibration curves for anion chromatography (looking at the 7 standard inorganic anions) best described by a linear or quadratic fit? Or does it depend on the concentrations?

Since I am not following an established or regulated method, I'm thinking to accept 6-point calibration curves that have an R2 value of 0.995 or better with a %RSD of 5.0% or less. Does this seem reasonable? Are there other statistical values I should be looking at?

Are there any methods that suggest averaging multiple calibrations to obtain a final curve?


Dear Kristin
Just follow the protocoll of your institution.
Quite often a linear calibration is required. Then you just have to follow this.
You typically will reach R2 >= 0.995.

This comes from a special effect of the R2 calculation. As an example: If you take a small sector of a cycle, place a couple of points on this curve and run them trhough R2 calculation. The result will be R2 = 1.00000. I do not believe that you accept a cyrcle segment to be linear.
The non-linearity of an anion calibration looks quite similar to a cyrcle segnment. Therfore, R2 will always be quite close to 1.

When you run 6 calibration levels you will definitly get acepptable results.
Dr. Markus Laeubli
Manager Marketing Support IC
(retired)
Metrohm AG
9101 Herisau
Switzerland
The difficulty is that my institution does not have any kind of established protocol. The only resource I was provided with was the ASTM D4327 standard. I'm trying to write SOPs for the method, calibration, and a QA/QC procedure, and my desperation is what prompted my posting on this forum.

I did read some articles that discussed how R2 values may not be the best way to judge a calibration curve, which is why I also thought to include parameters for %RSD.

My current thought for a calibration curve is six points that span the range of 10ppm to 0.05ppm, although I may leave out the highest concentration. From what a gather from your posts, perhaps I should use a quadratic instead?
Markus Laeubli, Metrohm wrote:
Dear anionman

Anion calibrations show inherently a certain nonlinearity. When you apply a three-point calibration you definitely will not realize that. Especially as the correlation coefficient of the special type of non-linearity shows a value close to 1.

If the calibration range is within one order of magnitude, a linear calbration is acceptable. For ranges larger than one order of magnitude, we recommend quadratic calibration as this gets a much better fit. An alternative is the high-low calibration. Here the we work with 2 or more calibration curves, where the highest point of the low-calibration ist the lowest point of the high calibration curve.

The quality control section of U.S. EPA Method 300.0
(Section 9.0) requires a demonstration of linearity, MDLs,and acceptable instrument performance by the analysis of a QCS prior to performing analyses using the method.
The method linearity using the Dionex IonPac AS18
column was determined over a seven-point calibration
range. MDLs for each of the anions in U.S. EPA Method
300.0 Part A were determined by performing seven
replicate injections of deionized water, fortified at a
concentration of three to five times the estimated instrument detection limits. Table 2 shows the standards used to calculate the MDLs and concentrations of the QCS. Table 3 shows the linear concentration ranges investigated...

Table 3. Linearity, MDLs, and retention time and peak area precisions obtained using the Dionex IonPac AS18 column.

Analyte Range mg/L linearity
Fluoride 0.1–100 0.9991
Chloride 0.2–200 0.9999
Nitrite-N 0.1–100 0.9992
Bromide 0.1–100 0.9999
Nitrate-N 0.1–100 0.9999
Phosphate-P 0.1–100 0.9999
Sulfate 0.2–200 0.9998

https://www.thermoscientific.com/conten ... 372-EN.pdf

I have noticed that fluoride and nitrite can get non linear, but others stay linear to the levels I need to see.
anionman wrote:
Table 3. Linearity, MDLs, and retention time and peak area precisions obtained using the Dionex IonPac AS18 column.

Analyte Range mg/L linearity
Fluoride 0.1–100 0.9991
Chloride 0.2–200 0.9999
Nitrite-N 0.1–100 0.9992
Bromide 0.1–100 0.9999
Nitrate-N 0.1–100 0.9999
Phosphate-P 0.1–100 0.9999
Sulfate 0.2–200 0.9998


Dear anionman
this ia an old story and an old discussion.
But the linearity/non linearity does not depend of the columntype nore the manufacturer.
Linear calibration gives very good R2 values (see my earlier post to Kristin). But if you run multipoint calibrations in the above levels. Run them through linear and quadratic fit, you will find a better fit of the curve (%RSD) and therefore better results with the quadratic fit. But if you need to follow a certain standard, linear fit might be required.
Dr. Markus Laeubli
Manager Marketing Support IC
(retired)
Metrohm AG
9101 Herisau
Switzerland
The method 300.0 requires proof of linearity so to follow it the calibration must be linear. The method also specifies reporting in the linear range so if we showed it wasn't linear we would be in trouble. I've read that non linearity is more evident with carbonate eluents and that's one reason we don't use it. I have always used KOH. The OP didn't mention what eluant she uses.

TNI 2016 required at least a 5 point calibration. Method 300.0 also lists MDLs. In mg/L:

Fluoride 0.01
Chloride 0.02
Nitrite-N 0.004
Bromide 0.01
Nitrate-N 0.002
o-Phosphate-P 0.003
Sulfate 0.02

I can see lower than 0.02 for Cl and sulfate though. It depends on the suppressor. A new suppressor has a noisy baseline, but it smooths out a lot as in breaks in over several weeks.
22 posts Page 1 of 2

Who is online

In total there is 1 user online :: 0 registered, 0 hidden and 1 guest (based on users active over the past 5 minutes)
Most users ever online was 1117 on Mon Jan 31, 2022 2:50 pm

Users browsing this forum: No registered users and 1 guest

Latest Blog Posts from Separation Science

Separation Science offers free learning from the experts covering methods, applications, webinars, eSeminars, videos, tutorials for users of liquid chromatography, gas chromatography, mass spectrometry, sample preparation and related analytical techniques.

Subscribe to our eNewsletter with daily, weekly or monthly updates: Food & Beverage, Environmental, (Bio)Pharmaceutical, Bioclinical, Liquid Chromatography, Gas Chromatography and Mass Spectrometry.

Liquid Chromatography

Gas Chromatography

Mass Spectrometry