I was fired. Was I wrong?

[confusedchemist]

Walked in today and was walked out. Toxicology lab has been losing samples for a while now I suspect it was largely fiscal. Whatever.

What I want to pick your brains about is the reason they gave, was that I "failed" my 'training' in relation to processing properly LC/MS/MS chromatography data.

A month ago I was given a number of previously reported samples to analyze independently with Masshunter (Agilent LCMS software), to be later 'graded' against the reported values. I used the same methods they did and generated samples with values that were based on the calibration curve generated that day of the run. Basic stuff. However, I 'failed' to follow a particular practice they have been using, and I feel it is yielding in about 1 in 10 samples going out with false negative results, daily.

What I would like to do is for you to tell me, who is correct in the following scenario:

LCMS Curve is generated using increasing concentrations of analytes and their deuterated internal standards. Levels 1-9

The L1 is considered their Lower Limit of Detection
The L2 is considered the cut-off.
Any sample's value, with proper chromatography, above the cut-off is considered a positive result. Any below the cut-off is considered negative.

Basic stuff so far.

The calibration curve plots the values of analyte's response, in relation to the internal standard response. In theory the internal standard is added in equal concentration in all samples.

So, in essence, we are generating values based on the RATIO of the response to the internal standard's response, both measured by area-under-the-curve. Which, I believe, is the whole point of using internal standards in LCMS. The idea being that by using the ratio and not just the analyte's response alone, you can help weed out the effects of ion suppression. So that if something causes a lower response to the target analyte, it will likely also happen to the internal standard. So even if a sample randomly gets half the total response as it normally "should" due to ion suppression, it will still yield a correct concentration value due to the ratio being used to make the concentration value and not just the analytes response alone.

So far am I right? I think this is still page 1 of analytical chemistry textbooks, right?

Ok, so here is where they change it up. They claim that if the sample's response is LOWER than the L2's response, that even if the calculated concentration value is above the cut-off, it is to be called negative and not positive.

So for example:

Morphine Calibration Curve's L2 Analyte Response = 10,000 with a concentration of 100. (So cut-off is 100.)

Morphine of Test Sample's Response = 8,000 with a calculated concentration of 155.

I say the 155 is correct and the sample is positive because there was ion suppression. The use of an internal standard protected this sample from being called negative because its value was made from a ratio of both.

They say, the sample, despite being above the cut-off, is negative because the analyte's response was lower than the L2's.

I feel they are totally ignoring the existence of ion suppression in LCMS by this practice and are calling samples with properly calculated values, well above the established cut-off, negative. Thus resulting in them reporting to physicians a "false-negative" result.

This is a problem because we do urine toxicology and most of our samples are for drug compliance. As in, "is Jane Doe taking her meds or selling them?" I feel that because of their random "rule", they are reporting out that people who ARE taking their meds are not. And this will result in people getting fired from their doctors and from receiving their meds because of supposed 'non-compliance' when in reality it was they were just unlucky enough to have had their sample have ion suppression.

Who is right here? I got fired for calling those test values positive. What would you do? What is your call? Literally tens of thousands of people have been effected by this 'rule'. Is is scientifically sound?

Thanks guys!

[Hornet]

I'm really sorry you got fired, these aren't the kind of things that should get someone fired. I mean, all problems could be solved just by speaking to each other, once the laboratory has a common way to work everything will be fine, if no one talks than this kind of "problems" come out.

First, you should ALWAYS monitor your internal standard recovery by using a syringe/injection standard. You must monitor it because something can go wrong during preparation/purification/injection and you will need a internal standard recovery acceptance criteria.
There is no way that the analyte area in your sample will be the same of the one in the calibration run, it will be lower. By thinking this way your cut-off value is useless...you are basically quantifying with internal standard method but with external calibration limitations.


This is my criteria, someone could use a different one. In my opinion you didn't do anything wrong but i'm not your boss.

[James_Ball]

I don't feel your reasoning was wrong, a conscientious scientists will always look for the most accurate and defensible results possible. On the other hand, employers today, even in the laboratory space, are not always scientists. True scientists understand that most things are actually variables not fixed constants, but accountants and lawyers are very uncomfortable with limits that are not fixed and set in stone. A scientist understands that there is an inherent amount of inaccuracy incurred in the entire process that is additive with each additional step in that process. If the balance has a 2% possible variance and the volumetric flask has a 0.5% possible variance and the autosampler has a 1% possible variance and the mass spectrometer has a 5% possible variance, we know that as a total system we can not say our process can operate with an absolute possible variance of 1%, yet QC officials and marketers will insist we report to customers that our product will be X concentration +/-1%. If all the variances average out then we are within that reporting limit, if it happens that all the variances fall on the low side then we are far below what is expected.

Often the employer is not the scientist, and must use the information from differing sources to make decisions on performance. If it is determined that a response below 10,000 area counts is unreliable then no matter what the calculated concentration of the analyte is they will deem it unusable or not detected. Honestly though, there should also be a limit on the internal standard performance so that if it is also below a certain amount the whole sample result is deemed invalid and must be re-analyzed to make sure that the concentration value that is above the detection limit is valid or not. In the environmental side of things we base the detection limit and reporting limit on concentration, but also require upper and lower limits on deviation of the area response before the data is acceptable. We also include data flags such as J for Estimated concentration to tell the client that while an analyte was detected above the limit of detection concentration wise, there is some other factor that makes the accuracy of that result less than acceptable. As a scientist I agree with you that the data should be reported to the client, but with some type of qualifier to let them know the accuracy could be in doubt, but, management and marketing does not like giving such things to clients because it could give the appearance to the client, if that client is not scientifically knowledgeable about the analytical process, that the laboratory is not certain about its results which could instill doubts in the client as to the laboratories reliability. Clients who are scientifically minded welcome the added information about method performance and results that could possibly be skewed, but clients who are not scientifically minded are frightened by such things. It is a difficult balance between science and business and we as scientists often are faced with such problems as you experienced.

[MichaelVW]

Sorry about your luck.

I don't know anything about LC or toxicology.. but I'd be curious as to how they came up with their "cutoff." What do they do if the response is above the cutoff but the calculated result is less? I'm imagining running a low-level calibration verification standard at the same concentration as the cutoff... using the response criteria alone, it would fail about 50% of the time.

Depending on how the results are used... it seems weird that the decision point is at the bottom of the curve, anyhow.

[mckrause]

To echo a very important point that James made, in environmental GC-MS and LC-MS methods the area of the internal standard is used as a control on the validity of that specific injection. I hate defending USEPA environmental methods because in general they are garbage, but in this respect they nailed it.

Unfortunately, in LC-MS signal suppression is a real and constant issue. I agree; you use the ratio of the analyte to the internal standard to calculate the concentration. At the same time you need to ensure that the internal standard is within some finite tolerance (USEPA uses 50%-200% of the mean of the calibration standards for GC-MS). If the IS is out of tolerance the injection is not valid; if it is within toelrance you use the IS response to calculate the concentration. In my opinion your calculation was correct. Using 10,000 counts as a hard floor has multiple issues and completely ignores matrix caused signal suppression, which is a real and constant bug-a-boo in LC-MS.