Which ICP test ? ATI or Triton

[TastyScrants]

In a study we (@taricha, @Rick Mathew ) are conducting of ICP vendors, ATI flunked the RODI test in which I spiked an RODI sample with a little Instant Ocean. ATI detected nothing while ICP-Analysis.com detected the elements correctly.

As for the aquarium water test, both ATI and Triton do OK with the elements that are present at more than 1 ppm. The trace element (<1 ppm) analysis from both vendors seems very imprecise, +/- 100% or worse for some elements and results that can range from detected to non-detected.

Take the ICP analysis for trace elements as a rough estimate. @Randy Holmes-Farley has recently commented about this in regards to managing reef aquarium water. I forget which post it was. Maybe he won’t mind repeating himself.

I’m currently trying to find out if any of the manufacturers laboratories have any recognised accreditation. During your study did you find any information related to any accreditation of their processes or their initial method validation?

 

[Dan_P]

I’m currently trying to find out if any of the manufacturers laboratories have any recognised accreditation. During your study did you find any information related to any accreditation of their processes or their initial method validation?

Did not even try. Previous published attempt by Ross to garner information from Triton showed just how secretive they are about their ICP method.

None of the vendors supply accuracy or precision information for the analysis they provide.

 

[TastyScrants]

Did not even try. Previous published attempt by Ross to garner information from Triton showed just how secretive they are about their ICP method.

None of the vendors supply accuracy or precision information for the analysis they provide.

That’s a shame.

I did find some LOD and LOQ data on the ATI website (here). They had correlation data too but no information on what other method theirs was measured against so it’s worthless information. It’s probably just the data supplied by the manufacturer for whichever instrument they’re using. I could only assume that the correlation data is to a reference method considering the high order of ICP in general, but who knows!

For all we know the data could be being fudged by their algorithms to sell their respective dosing materials (I’m not picking on Triton here I’m talking about all of them).

How do we even know the instrument is regularly cleaned? When I used to perform ICP-MS in a clinical lab the flow path had to be regularly cleaned with nitric acid.

I’m getting Theranos vibes.

 

[periocillin]

triton

 

[Dan_P]

That’s a shame.

I did find some LOD and LOQ data on the ATI website (here). They had correlation data too but no information on what other method theirs was measured against so it’s worthless information. It’s probably just the data supplied by the manufacturer for whichever instrument they’re using. I could only assume that the correlation data is to a reference method considering the high order of ICP in general, but who knows!

For all we know the data could be being fudged by their algorithms to sell their respective dosing materials (I’m not picking on Triton here I’m talking about all of them).

How do we even know the instrument is regularly cleaned? When I used to perform ICP-MS in a clinical lab the flow path had to be regularly cleaned with nitric acid.

I’m getting Theranos vibes.

LOL!

I toyed with the idea of a random number generator as a business model, but our ICP vendor study data doesn’t support it I am now entertaining another idea

For around $50 (US), we should not expect expect the accuracy and precision of a Hanna Checker in ICP OES data.

If this notion reflects reality, then we must adjust our view about how we use the data. This data might be useless as a reference point, such as confirming Hanna Checker results. For elements with high concentration, >1 ppm, the data might be useful in judging whether the elements in aquarium water are in the approximate abundance that we expect. We would look for elements that are way out of line. For trace elements, < 1 ppm, we might do the same thing with the caveat that a second test is necessary to confirm the need for action.

We are still collecting data and will try to develop some ideas about how we might use noisy data.

 

[Christoph]

Hi everyone,

I am the scientist behind oceamo, and also the person who develops methods and runs the spectrometers (among other instruments). If you are having questions regarding the "everyday practice" in our lab i am happy to answer them.

With ICP it is very difficult to give "global" LLOD/LLOQ data, since those values always depend on sample composition. For example higher phosphorus content will influence our ability to measure iodine, since both emission lines are very close together at 178 nm. One has to be careful since some published LLOD/LLOQ data is very unrealistic and far beyond the capabilities of ICP-OES. For most important trace elements we are in the 0,2-2 ppb area regarding LLOD. Just let me know if you are interested in data of specific elements.

Of course we validate our methods with control samples and keep a close eye on linearity and stable machine operation (often also using internal standardization). Also we do measure most trace elements at several wavelengths (if the element has several sensitive emission lines), so we can rule out interference by other elements, which is critical in ICP-OES.

Machines are serviced regularly and key components (like the plasma torch) are replaced regularly, since measuring seawater is quite "stressful" for the equipment.

Of course the autosampler/sample uptake system is rinsed between samples to avoid carryover from sample to sample.

Control samples and ring trials showed that we produce high quality data, and are not a "random number generator"

Sure coming close to the LLOD RSDs are higher, which is the case for every analytical procedure. That is the reason why it does not make sense to report values with several decimals after the comma.

Best regards,
Christoph

 

[Randy Holmes-Farley]

Yes, thanks. You repeated exactly what I read in one of your recent posts. So how the heck do you decide on how much to dose when the ocean varies so much and the ICP results can be all over the map for some elements?

Well, perhaps it's as simple as dose some and see if it helps.

Even if the ocean were a fixed value, we regularly see great reef tanks with undetectable levels of one or more trace elements and/or elevations in others.

I just don't think exact values are needed, but might be a guide to what is known to be acceptable (ignoring issues relating to how the chemical form that is unknown to us impacts bioavailability).

 

[Randy Holmes-Farley]

Hi everyone,

I am the scientist behind oceamo, and also the person who develops methods and runs the spectrometers (among other instruments). If you are having questions regarding the "everyday practice" in our lab i am happy to answer them.

With ICP it is very difficult to give "global" LLOD/LLOQ data, since those values always depend on sample composition. For example higher phosphorus content will influence our ability to measure iodine, since both emission lines are very close together at 178 nm. One has to be careful since some published LLOD/LLOQ data is very unrealistic and far beyond the capabilities of ICP-OES. For most important trace elements we are in the 0,2-2 ppb area regarding LLOD. Just let me know if you are interested in data of specific elements.

Of course we validate our methods with control samples and keep a close eye on linearity and stable machine operation (often also using internal standardization). Also we do measure most trace elements at several wavelengths (if the element has several sensitive emission lines), so we can rule out interference by other elements, which is critical in ICP-OES.

Machines are serviced regularly and key components (like the plasma torch) are replaced regularly, since measuring seawater is quite "stressful" for the equipment.

Of course the autosampler/sample uptake system is rinsed between samples to avoid carryover from sample to sample.

Control samples and ring trials showed that we produce high quality data, and are not a "random number generator"

Sure coming close to the LLOD RSDs are higher, which is the case for every analytical procedure. That is the reason why it does not make sense to report values with several decimals after the comma.

Best regards,
Christoph

Thanks for chiming in Christoph.

 

[Rick Mathew]

Hi everyone,

I am the scientist behind oceamo, and also the person who develops methods and runs the spectrometers (among other instruments). If you are having questions regarding the "everyday practice" in our lab i am happy to answer them.

With ICP it is very difficult to give "global" LLOD/LLOQ data, since those values always depend on sample composition. For example higher phosphorus content will influence our ability to measure iodine, since both emission lines are very close together at 178 nm. One has to be careful since some published LLOD/LLOQ data is very unrealistic and far beyond the capabilities of ICP-OES. For most important trace elements we are in the 0,2-2 ppb area regarding LLOD. Just let me know if you are interested in data of specific elements.

Of course we validate our methods with control samples and keep a close eye on linearity and stable machine operation (often also using internal standardization). Also we do measure most trace elements at several wavelengths (if the element has several sensitive emission lines), so we can rule out interference by other elements, which is critical in ICP-OES.

Machines are serviced regularly and key components (like the plasma torch) are replaced regularly, since measuring seawater is quite "stressful" for the equipment.

Of course the autosampler/sample uptake system is rinsed between samples to avoid carryover from sample to sample.

Control samples and ring trials showed that we produce high quality data, and are not a "random number generator"

Sure coming close to the LLOD RSDs are higher, which is the case for every analytical procedure. That is the reason why it does not make sense to report values with several decimals after the comma.

Best regards,
Christoph

Christoph thanks for jumping into the conversation...Great to have someone who actually runs the test protocols available to answer questions...So here goes my first one....

When you say control samples and "ring" samples are you referring to known standards and intra or inter lab testing of know samples...where the same sample is sent to multiple labs and or a reference lab? When you mention that it produces high quality data how would you quantify "high quality"

Again thanks for jumping in I think you will have a lot to contribute to the conversation.

Rick

 

[Christoph]

Hi Rick,

"ring samples" are sent to multiple labs that do not know the composition. The samples are measured and reported back. Afterwards you get notified, how close you were to the correct values. Unfortunately those "ring samples" are not available for seawater (but for freshwater), but the results still give you a very good indication of how well you (thats the most important part) and your machines (which is second in importance) are performing.

We also do checks with seawater control samples that are produced in-house from certified standard materials that is purchased from vendors.

"High quality" is difficult do define, you are totally right What i meant is that we are producing data that, to the best of my knowledge, is close enough to the "real" value to be significant and useable for the reef aquarist. For example if we report elevated copper, we are very certain copper is actually elevated, and the signal is not an interference from other elements emissions.

Best regards,
Christoph

 

[Rick Mathew]

Hi Rick,

"ring samples" are sent to multiple labs that do not know the composition. The samples are measured and reported back. Afterwards you get notified, how close you were to the correct values. Unfortunately those "ring samples" are not available for seawater (but for freshwater), but the results still give you a very good indication of how well you (thats the most important part) and your machines (which is second in importance) are performing.

We also do checks with seawater control samples that are produced in-house from certified standard materials that is purchased from vendors.

"High quality" is difficult do define, you are totally right What i meant is that we are producing data that, to the best of my knowledge, is close enough to the "real" value to be significant and useable for the reef aquarist. For example if we report elevated copper, we are very certain copper is actually elevated, and the signal is not an interference from other elements emissions.

Best regards,
Christoph

Christoph: Thanks for the very quick reply...This is very helpful information in understanding your philosophy....Very well explained!

My next question is related more to the precision ( Repeatability and Reproducibility)

If you measure the same sample multiple times over a period of time on what factors (LoD, Interference...) are the variability dependent upon?... How do you quantify that variability ...Like ± % or ± some value?

 

[Dan_P]

Hi everyone,

I am the scientist behind oceamo, and also the person who develops methods and runs the spectrometers (among other instruments). If you are having questions regarding the "everyday practice" in our lab i am happy to answer them.

With ICP it is very difficult to give "global" LLOD/LLOQ data, since those values always depend on sample composition. For example higher phosphorus content will influence our ability to measure iodine, since both emission lines are very close together at 178 nm. One has to be careful since some published LLOD/LLOQ data is very unrealistic and far beyond the capabilities of ICP-OES. For most important trace elements we are in the 0,2-2 ppb area regarding LLOD. Just let me know if you are interested in data of specific elements.

Of course we validate our methods with control samples and keep a close eye on linearity and stable machine operation (often also using internal standardization). Also we do measure most trace elements at several wavelengths (if the element has several sensitive emission lines), so we can rule out interference by other elements, which is critical in ICP-OES.

Machines are serviced regularly and key components (like the plasma torch) are replaced regularly, since measuring seawater is quite "stressful" for the equipment.

Of course the autosampler/sample uptake system is rinsed between samples to avoid carryover from sample to sample.

Control samples and ring trials showed that we produce high quality data, and are not a "random number generator"

Sure coming close to the LLOD RSDs are higher, which is the case for every analytical procedure. That is the reason why it does not make sense to report values with several decimals after the comma.

Best regards,
Christoph

Thanks Christoph for joining the conversation. Dan

 

[Dan_P]

Hi everyone,

I am the scientist behind oceamo, and also the person who develops methods and runs the spectrometers (among other instruments). If you are having questions regarding the "everyday practice" in our lab i am happy to answer them.

With ICP it is very difficult to give "global" LLOD/LLOQ data, since those values always depend on sample composition. For example higher phosphorus content will influence our ability to measure iodine, since both emission lines are very close together at 178 nm. One has to be careful since some published LLOD/LLOQ data is very unrealistic and far beyond the capabilities of ICP-OES. For most important trace elements we are in the 0,2-2 ppb area regarding LLOD. Just let me know if you are interested in data of specific elements.

Of course we validate our methods with control samples and keep a close eye on linearity and stable machine operation (often also using internal standardization). Also we do measure most trace elements at several wavelengths (if the element has several sensitive emission lines), so we can rule out interference by other elements, which is critical in ICP-OES.

Machines are serviced regularly and key components (like the plasma torch) are replaced regularly, since measuring seawater is quite "stressful" for the equipment.

Of course the autosampler/sample uptake system is rinsed between samples to avoid carryover from sample to sample.

Control samples and ring trials showed that we produce high quality data, and are not a "random number generator"

Sure coming close to the LLOD RSDs are higher, which is the case for every analytical procedure. That is the reason why it does not make sense to report values with several decimals after the comma.

Best regards,
Christoph

How large is the salinity effect? Would the ICP results of a sample of 35 ppt aquarium water be different from the same water diluted to 20 ppt with RO water and, of course, the concentrations adjusted for the dilution?

 

[Christoph]

Sure! Im happy to provide some insight. Unfortunately im quite shport onm time, so follow up answers might take some time.

To Ricks question:

Every sample is routinely measured in 3 replicates. From those replicates you get a %RSD (procentual relative standard deviation), which is of course higher when you are measuring low concentrations, because noise already plays a role, while at high signal intensities the noise does not contribute to a significant extent.

This %RSD furthermore reflects the variability youre having in your sample introduction system (peristaltic pump flow, nebulization,...)

This value gives a good indication about the error margin your are facing that is coming from the machine and the physical limits. It does not reflect the total possible error, which also includes the used standards, purity of reagents, volumetric uncertainties, uncertainties of balances, etc... for this reason it is very difficult to state a valid ± %. I can understand it would be nice to have those errors (or a 95% confidence interval) shown in the report, but i have not yet found a reliable and practical way to do so.

Best regards, Christoph

 

[Christoph]

How large is the salinity effect? Would the ICP results of a sample of 35 ppt aquarium water be different from the same water diluted to 20 ppt with RO water and, of course, the concentrations adjusted for the dilution?

If we would receive a 20 ppt seawater sample i would notify the customer that lower accuracy is to be expected. It would not be "unuseable", but we expect our samples to be in the 30-37 ppt range. This is why customers should always contact me in case of "special" samples like brackish water. Other sequences/methods might be more suitable to analyze those samples.

 

[Rick Mathew]

Sure! Im happy to provide some insight. Unfortunately im quite shport onm time, so follow up answers might take some time.

To Ricks question:

Every sample is routinely measured in 3 replicates. From those replicates you get a %RSD (procentual relative standard deviation), which is of course higher when you are measuring low concentrations, because noise already plays a role, while at high signal intensities the noise does not contribute to a significant extent.

This %RSD furthermore reflects the variability youre having in your sample introduction system (peristaltic pump flow, nebulization,...)

This value gives a good indication about the error margin your are facing that is coming from the machine and the physical limits. It does not reflect the total possible error, which also includes the used standards, purity of reagents, volumetric uncertainties, uncertainties of balances, etc... for this reason it is very difficult to state a valid ± %. I can understand it would be nice to have those errors (or a 95% confidence interval) shown in the report, but i have not yet found a reliable and practical way to do so.

Best regards, Christoph

Wow...Awesome, honest and helpful answer!...Thanks...Will most likely have addition questions as I ponder some more...Once again thanks for joining the discussion!

Rick

 

[Dan_P]

If we would receive a 20 ppt seawater sample i would notify the customer that lower accuracy is to be expected. It would not be "unuseable", but we expect our samples to be in the 30-37 ppt range. This is why customers should always contact me in case of "special" samples like brackish water. Other sequences/methods might be more suitable to analyze those samples.

This is interesting. Thank you!

 

[ThePhoReefer]

I used to use triton until usps lost my test so I lost my money… and unable to get new one. Triton is a risk cus if usps lost the Mail basically say good bye to your test, they won’t give u a new one if it’s lost. I’m disappointed in customer service overall. Switched to ati, surprised they include rodi test too. Also heard they have great customer service.

 

[taricha]

hey @Christoph
Could you share a raw data shot of an ICP-OES spectrum from a (typical) saltwater sample? Just raw Counts Per Second vs wavelength would be very interesting to me, @Rick Mathew , @Dan_P - even if you didn't tell us what element values were later derived from it.

(It's probably too information-dense for a picture, a spreadsheet or something maybe?)

We know there's considerable statistical heavy lifting that goes on between the raw data and the generated element values in order to tame it into something meaningful, but seeing the raw would be enlightening.

 

[Christoph]

Hello Taricha,

im not aware of a possibility to get a "full spectrum" dataset from our processing software. We just measure intensity at dedicated emission wavelenghts.

Also bear in mind that spectral data (if you get hold of it) changes significantly with measurement parameters (axial/radial view, nebulizer flow, RF power). For example atomic lines would be more pronounced with low RF and high nebulizer flow, while the opposite is true for ionic lines (especially the UV ones). So i doubt the data would be enlightening.

Best regards,
Christoph