The Toxpod

When is a Drug Not a Drug?

October 02, 2019 Season 2 Episode 9
The Toxpod
When is a Drug Not a Drug?
Chapters
00:03:22
Immunoassay
00:05:13
Chromatography
00:08:29
UV-VIS Spectroscopy
00:11:06
Mass Spectrometry
00:23:38
Is the threshold model the best approach?
00:31:54
Signal to noise
The Toxpod
When is a Drug Not a Drug?
Oct 02, 2019 Season 2 Episode 9
Tim Scott & Peter Stockham
We talk about drug identification and the various guidelines that help us do it
Show Notes Transcript Chapter Markers

In this episode, we dive into the techniques, practices, and guidelines used in forensic toxicology for drug identification. Do we have the balance right?

Some of the available identification guidelines:
World Anti-Doping Agency (WADA)
European Workplace Drug Testing Society (EWDTS)
Organisation of Scientific Area Committees (OSAC)
Gesellschaft für Toxikologische und Forensische Chemie (GTFCh)

That article Pete mentioned on amphetamine isomers:
Brown, D. H., et al. (2016). "Beta-methylphenylethylamines: common fragmentation pathways with amphetamines in electrospray ionization collision-induced dissociation." Drug Test Anal 8(3-4): 344-350.

Contact us at thetoxpod@sa.gov.au


The Toxpod is a production of Forensic Science SA and the South Australian Attorney General's Department. The opinions expressed by the hosts are their own and do not necessarily reflect the views of their employer.

Tim:
0:01
Hello and welcome to The Toxpod. I'm Tim Scott.
Peter:
0:14
I'm Peter Stockham.
Tim:
0:15
And today we're going to do an episode on drug identification.
Peter:
0:19
Mass spec guidelines, chromatographic guidelines and that sort of thing.
Tim:
0:22
Yeah, there's a whole range of different tests that we use in forensic toxicology to try and screen for drugs, confirm drugs, and some of them are better than others at confirming drugs. What makes a confirmation? How much information do you need in order to say that a drug is definitely there? It really depends what your purpose is in identifying and reporting those drugs. If the purpose of the test is just to give some sort of presumptive information to a crime scene investigator or something like that, it's probably fine to just be mostly certain that the drug is there. You don't have to be 100% certain, but if you're reporting a drug where the presence of the drug is going to be a crime in itself, well then you better be sure that that is the drug that you're detecting.
Peter:
1:09
Yeah.
:
1:09
As an analytical toxicologist, you kind of have an idea in your head of what constitutes a good confirmation, you sort of think, well I'll know it if I see it, but when you actually come to write it down on paper, what exactly do you need in order to say it's identified? It's actually much harder than you think.
Peter:
1:25
Yeah. And you have to do that between laboratories and get it standardized. It's even harder.
:
1:28
Yeah. Different labs are using different instruments.
Peter:
1:31
So different organizations involved with analytical chemistry will have their own guidelines. So we have ones from WADA, we have ones from forensic laboratories around...
:
1:37
So that's, WADA is the world anti-doping agency.
Peter:
1:41
Yeah. And we've got one from the GTFCh, um, ones from the EWDTS, which is...
:
1:48
Uh, European workplace drug testing society.
Peter:
1:50
Something like that. And we've got our own for, for Australia, our forensic laboratories have got guidelines that we follow and even individual standards for specific, um, different work types, like urine testing for example, will have their own guidelines for identification.
Tim:
2:06
And the Americans are coming out with some more guidelines. So there was a talk at TIAFT about the...
Tim:
2:11
Yeah Marilyn Huestis gave a talk about the OSAC guidelines they've currently got out for consultation. But I think in any standard they still say you have to be an experienced mass spectroscopist before you can, uh, you can't just use the guidelines alone. You've got to use a bit of experience and knowledge of your field.
Tim:
2:26
Yeah it's like any kind of rules, isn't it? It's really, the people who write these kinds of guidelines really want everyone to comply with not just the letter of the law, but the spirit of the guidelines as well. But you can't put that into the letter obviously.
Peter:
2:38
No. So you might see that a drug passes the guidelines, but just because it passes the guidelines, it doesn't mean it's definitively that particular drug. It may be something else.
Tim:
2:47
Yeah. Or, on the other hand, sometimes you might be 99% certain that a drug is actually there, but it just doesn't pass the guidelines for some technical reason. But you're almost certain that it's it.
Peter:
3:00
Yeah there's the old, um, if you've got a full scan spectra a lot of guidelines say if there's a contaminant of a certain height of the base peak, like five or 10%, depending on the guideline, then you have to explain that peak. But sometimes you can't explain that peak. So does that mean that that drug is not there? Of course it doesn't. It means it is there, but you have to qualify it somehow.
:
3:18
Yeah. That's where the judgment comes in. So just going back a step then, there are some techniques which are not very good for confirmation of drugs at all, like immunoassay for example. So that's an excellent screening technique, but it's not going to tell you which drug is there. So some guidelines will count an immunoassay test towards part of a confirmation. You know, it's not a confirmation in itself, but if you add that on to some other things, maybe some chromatographic tests and mass spec tests and so on.
Peter:
3:46
So that includes immunoassay, maybe even outside the laboratory as well. So oral fluid, onsite testing.
:
3:52
Yeah, but...
Peter:
3:53
But in the lab is the main one we're going to talk about here, I guess.
:
3:55
Yeah, but I'm not, even in the lab immunoassay, I'm not very keen on that forming part of the confirmation because there's just such a high chance of false positive results. So from a compound that's got nothing to do with your analyte of interest, but then there's also the fact that it can pick up other compounds in that class of drugs. I don't think it really should be part of a confirmation.
Peter:
4:16
But you do use it to direct your analyses though. That's typically what you do with immunoassays is you analyze it by immunoassay, then follow on with the correct analytical technique after that. So it already is sort of qualifying as an identification, isn't it?
:
4:30
Well not really. I don't think so. It's not forming part of the final identification, which I think, even though it directs your analysis that, that is, should be the purpose of it. Just to direct your analysis, not to form any final part of the confirmation.
Peter:
4:43
Yeah. Okay.
Tim:
4:44
That's my opinion. But there's, some of the guidelines disagree with that.
Peter:
4:47
No, I agree with that cause it's redundant because it, of course it was positive because you did the subsequent test. So it shouldn't count.
Tim:
4:52
Yeah, only the ones that are positive are going through for the subsequent test anyway.
Peter:
4:54
Yeah. So in some fields they use colorimetric tests or even infrared spectra to...
Tim:
5:00
Yeah, FTIR.
Peter:
5:02
Yeah FTIR, but that's in a different field altogether, that's more like in illicit drugs where they seize compounds. These colorimetric tested initially direct their analyses. But um, I guess it's similar to what we do with ELISA, or immunoassays.
Tim:
5:14
And then you have chromatographic tests, which are obviously better. So you're separating out the compounds now, but still, even despite reproducible retention times, even despite possibly running on two different columns, whether it's GC or LC, most toxicologists and most guidelines wouldn't regard that as a confirmation in itself. Just a retention time or two retention times.
Peter:
5:36
So alcohol is one of the few drugs I think where it generally is accepted not to worry about mass spectrometry.
Tim:
5:43
Yeah, I think that's true, but I..
Peter:
5:45
It's always done on two columns.
Tim:
5:47
Yeah. But why can't you confirm another drug on just two, by two retention times? Why is alcohol excluded from some of these guidelines in needing a mass spec identification? To me it's just a purely practical thing.
Peter:
6:01
Yeah, it's the way that laboratories are set up and some labs might not even have a mass spec, so maybe it's more structural rather than scientifically based.
Tim:
6:08
Yeah, I think it definitely is and, but you could make a scientific argument of, well, ethanol is a very small molecule. The chance of something else coeluting with it is quite small because if you test a lot of these other small molecules, alcohols, ketones, things like that, you can probably exclude all the things that are gonna coelute with it on both columns.
Peter:
6:28
Yeah, it has to coelute on both columns. So that's the big thing. That's what discriminates them.
Tim:
6:32
Yeah. But still, I think you could, you could almost make that argument for any drug.
Peter:
6:36
So you could just do it for, without mass spec. Just have GC NPD with dual columns.
Tim:
6:41
Yeah. It's really just a practical consideration that ethanol is excluded because why couldn't acetone be excluded? Some labs measure acetone.
Peter:
6:48
Yes.
Tim:
6:48
Well you could make exactly the same argument for acetone, but it doesn't get excluded from any guidelines.
Peter:
6:53
We're not trying to undermine the whole basics of drugs and driving, alcohol and driving practices. I think it's been pretty well established that the practice of using gas chromatography, two columns for alcohol is pretty well accepted.
Tim:
7:06
Yeah, I just like poking holes in things.
Peter:
7:09
Yeah. Stop that.
Tim:
7:09
Have you noticed that?
Peter:
7:10
I have. You're wrecking things.
Tim:
7:12
Well, the thing about ethanol though as well though, that's, kind of supports the way that it's done is that mass spec doesn't add a lot of information because it's such a small molecule, you don't actually get many masses that come from it in your mass spec and they're not really that...
Peter:
7:27
Characteristic?
Tim:
7:28
Characteristic of ethanol. They could come from other things as well. So does mass spec really add anything? It's, I guess it's just the nature of small molecules that makes them hard to meet the identification criteria that we've got here. In some ways. But then in other ways it's easier because there's probably less things that are gonna interfere with them.
Peter:
7:45
What if you used high resolution GC mass spec?
Tim:
7:49
For ethanol?
Peter:
7:51
Yeah.
Tim:
7:51
Wow.
Peter:
7:52
It wouldn't make any difference, would it?
Tim:
7:53
No, I don't think so.
Peter:
7:53
Because they're the same, same molecular formula? All of these small molecules.
Tim:
7:57
That would be a huge waste of an accurate mass spec. So then if chromatography is not enough on its own, you need some kind of detector on the end, right? And then normally we're talking about mass spec detectors, but you can also have a UV vis detector.
Peter:
8:14
Well you could have a TLC plate at the end of your column and then put that onto a UV detector.
Tim:
8:19
What, would you have like a conveyor belt of TLC plates moving past the end as the columns dripping out?
Peter:
8:23
You could. I'm not saying it's a good idea.
Tim:
8:27
I've yet to see that. So Mass Spec is really the gold standard in identification, but let's just talk a little bit about UV vis identification. So rather than based on mass, it's based on chromophores in the molecules.
Peter:
8:42
I think it's accepted that there are flaws in using UV as identification because you can have many different, compounds will have the same chromophores or even metabolites will have the same chromophores. So it's quite difficult to separate them.
Tim:
8:53
Yeah, it can be actually very useful, that fact because sometimes you can identify metabolites in an assay just because they have the same UV spectrum as the parent. Like you might not know that they're there, but you find a compound that's got a different retention time to the drug, but it's got exactly the same UV spectrum.
Peter:
9:08
That might be handy once, but then they're just a nuisance aren't they?
Tim:
9:10
Yeah. Most of the time they're a nuisance. But they're, if you're doing that particular kind of thing, they can be handy or they can be handy in the sense of if you identify a metabolite that may lead you on to screen for the parent, maybe the parent you don't see in your standard assay. So that can be useful. But most of the time, as you say they are a bit of a nuisance and that for that reason and for other reasons as well, UV vis isn't the greatest in terms of identification.
Peter:
9:35
It's still acceptable in a lot of standards though isn't it.
Tim:
9:38
Well, I think it's gotta be because a lot of labs use UV vis spectroscopy.
Peter:
9:42
Including us.
Tim:
9:44
Yeah. Yeah. So I think just because a new technique comes out, which is better, it doesn't mean that you can, you know, necessarily dispense with all the techniques that have gone before. They're still useful. But in order to get a good confirmation from UV vis you need really good chromatography, so you need good retention time matching and you really need to be aware of potential interferences.
Peter:
10:05
Yeah, So possible things that need to be assessed during your validation phase, I guess.
Tim:
10:09
Yeah. Other drugs of the same class especially, which might be very similar in structure and have a similar retention time. So with UV detection, when you get down to quite a low concentration, it starts to look a little bit fuzzy, you know? Yeah. But you don't do ratios like you might do in a mass Spec, you know, MRM or SIM or something like that. You're not doing ratios of different wavelengths generally you're just looking at the overall profile of the wavelength. Is it visually similar? But when does it become too dissimilar? It's very difficult to tell that.
Peter:
10:43
Yeah you could base it on percentage match if your library matching program does that. But where do you decide that cut off? It's difficult.
Tim:
10:50
Yeah. And you can do derivative spectra as well. Sometimes that's a little bit of a help. I've found that's mostly a help in ruling things out rather than ruling things in, because sometimes it's just too noisy anyway.
Peter:
11:03
Yeah the inflections at a different spot, so it shows that it's not the same.
Tim:
11:07
So let's talk about mass spec then.
Peter:
11:08
Yeah, let's get onto the good stuff. So most standards start out with the requirement for a number of diagnostic ions or characteristic ions. And so sometimes we say, uh, some, some even say it shouldn't be a common ion like a 58 or a 91, which is quite common for some molecules, but sometimes that's all you've got.
Tim:
11:25
I know it's a bit rough to say you can't use those because I think of something like amphetamine, it's such a small molecule to start with and its fragments really are pretty generic.
Peter:
11:35
And they're the same as methamphet too.
Tim:
11:38
Yeah. So when you're validating a method, you do a lot of selectivity experiments. You should run some blank matrices of whatever the sample type is that you're analyzing to make sure there's no endogenous compounds that are going to give you a false positive. But you should also run a bunch of other drugs, common drugs that you might see to make sure none of them are going to interfere.
Peter:
11:58
Uh, there was, probably about five years ago, I heard a talk at a conference where we were, they were talking about isomers of amphetamine and methamphetamine and on LCMS it was very difficult to tell them apart. had the same fragments, same retention time. And I'm pretty sure they'd pass most people's, uh, guidelines for MS ID criteria.
Tim:
12:19
Yeah, that's interesting, isn't it?
Peter:
12:20
Exactly the same fragmentation. So this was M2PPA, I think, methyl-2-phenylpropylamine. And for amphetamine, its isomer is betamethyl phenethylamine. So it's just got the uh, phenyl group in a different spot.
Tim:
12:35
Yeah. But unless you know about these things, you don't know to guard against them. Right? Cause you can't test for every single compound in selectivity experiments.
Peter:
12:45
No, it's just important to be aware of them though, because these compounds, are in sports supplements that apparently increase your metabolism and they're good for people who are working out, so...
Tim:
12:52
Yeah, and especially in the case of these low molecular weight amine type compounds, there's just so many of them around that have the same, almost exactly the same structure. So they're going to elute very closely. The Mass Spec might be either very similar or even identical. But that's just an example obviously, any, this could happen for any drug. The higher in mass that you go and the more sort of complicated structures you get, maybe the less likely it is that this kind of thing is going to happen.
Peter:
13:21
Yeah, well it's just the higher the molecular weight, the less likelihood you're going to have an isomer, I guess. For example, there's, when LCMS was first coming out probably 10 years ago or so, there was the example of tramadol and ODM venlafaxine have got the same formula and very similar fragment ions and it was just, you have to be very aware of these two compounds and make sure they separate on your, cause they're very common drugs.
Tim:
13:44
Yeah, there's a, there's quite a lot of these kinds of pairs aren't there. You've gotta be aware of them. They might not elute at the same time on your method. It depends completely on your chromatographic method, whether or not they elute the same.
Peter:
13:56
So how do you put that in a standard? Make sure all isomers do not coelute.
Tim:
14:01
Well that's, that's the really tough thing about when it comes down to putting these things in writing in standards. Cause I don't think you can have a standard that completely guards against possible coelution because you didn't, might not be aware of that other compound existing. So in standard selectivity experiments, you, people will run quite a few drugs. Could we do more? Probably yes we could. But do we need to do more than what we're currently doing? I don't really think so.
Peter:
14:27
It depends on the lab I guess. But we try hard don't we, everyone?
Tim:
14:30
I think if you, it's not the number of compounds that you're running in selectivity experiments that's the issue. It's whether you're getting the right ones, you know you've got to run the right compounds to see if they're coeluting, the ones that have the same mass or the ones that are going to have a similar retention time.
Peter:
14:46
Or are used in a relevant context, in the same sort of context.
Tim:
14:49
Yeah, the ones that you might find in your samples.
Peter:
14:52
So of course there's different sorts of instrumentation, different sorts of acquisition. So you can have full scan spectra, you can have selected ion monitoring, so we're only just looking at a couple of ions. Then you move on to MSMS. Then you include things like high resolution mass spec and it can get quite complex.
Tim:
15:08
So guidelines will often break up the mass spec identification principles into low res and high res Mass Spec because for low resolution, so we're talking about nominal mass resolution here.
Peter:
15:21
So your typical GC MSD or...
Tim:
15:24
Yeah, obviously the bigger the window of your mass there, the less specific it is. When you go to high res, you've got a very narrow window. You're talking about milli Daltons.
Peter:
15:33
So the resolution's the important thing isn't it? So on a single quadripole even though it looks like a stick, a little line on a page, the real true mass is a very wide peak that goes across an entire mass unit.
Tim:
15:46
Yeah. We should probably think about our mass spec like that. Even though it doesn't show it like that on the page, we really should think about it like that. There is some uncertainty associated with that apex of that mass peak. Just the same as there is for a chromatographic peak.
Peter:
15:59
Yeah. So they call that centroided data. So yeah, you always have to think about that. And so for a quadrupole, if you have a look at the raw data, it's quite scary.
Tim:
16:10
Just how wide that window is.
Peter:
16:11
That's right. And if you have two drugs coeluting then you see how uh, yeah, how uncertain you could be about identification based on mass. But you know, it's only nominal mass, so that's fine.
Tim:
16:19
Yeah. And we're only talking about parent mass there, the M+H. Then you obviously go to fragments because no ones, you should never confirm based on just the parent mass alone and retention time.
Peter:
16:30
Yeah so they all say you need at least a few diagnostic ions, right?
Tim:
16:34
Yeah. So then, then you're gonna fragment that parent mass and then once you get those extra few ions, the chances of something else having the same fragment ions are much, much less.
Peter:
16:44
Yes. Unless it's amphetamine and the other drug I talked about earlier.
Tim:
16:47
Yeah, so there's a couple that do fragment in exactly the same way.
Peter:
16:50
That's why you need good chromatography.
Tim:
16:52
Yeah. Well that raises another point actually about retention time matching because if you get an overloaded peak, if you're saturating both the column and the detector, it's going to be a really broad peak and it's going to have a flat top for a long time. And so retention time matching for that is not really appropriate. What can you do to try and match that retention time? I mean you can just inject less or dilute it or whatever. You could also try re extracting that, if you've, if you're using a full scan mass spec for example, you could try re extracting that peak with the, an isotopologue.
Peter:
17:26
Ah, okay.
Tim:
17:27
So that you'll get, you won't be saturating the detector so much and you might get a bit of a sharper peak. It's still going to be really broad because the column has done that, you can't help the saturation on the column, but you might get a better peak apex. Sometimes better than re-injecting it or reanalyzing it.
Peter:
17:43
For diagnostic ions, you can either have a full scan spectrum or you can have an instrument which in, which just monitors a set number of ions, so selected ion monitoring, often called. In a lot of guidelines they discriminate against full scan spectra I think. So if you've got a full scan spectrum, you might have 10 ions and you compare that to a library with the spectrum there's got, that's got 10 ions. But then the, a lot of guidelines say if you've got a, as part of your mass spectrum, if there's a peak that's not part of the library spectrum or your authentic spectrum, then you have to account for that. But if you've got, if you're using SIM, you don't even see that extra ion. So you get much easier to... It's unfair, isn't it?
Tim:
18:22
Yeah. It's like you take your full scan mass spectra and if you just put both your hands over the ends of it and cover up all the other things that you don't want to see, you've basically got a SIM or an MRM spectra there.
Peter:
18:33
Exactly.
Tim:
18:34
Yeah. So I think that's a bit unfair, but that's in a lot of standards, including when the one that we adhere to.
Tim:
18:39
Yeah, but it isn't, I think the principle of it though, I agree it's a bit unfair, but the principle of it is that once you know some information, you can't then ignore it. Like if you see that extra ion, then you just can't ignore the fact that it's there. If you don't see it, fine, you haven't seen it, but once you've seen it, you know because that extra ion that could actually lead to it being another compound, right? Like that could mean that it's another compound.
Peter:
19:04
It could be.
Tim:
19:04
Yeah, and so sometimes it'll be different guidelines will be, you know that that interfering ion has to be less than 10% or less than 50% or whatever, but especially if you've got full scan, accurate mass spectra and you've got 10 ions there, they're all accurate mass and you've got one little other ion that's interfering, which you can't account for, that, that's where it's, with the guidelines, as we were saying before, it's very hard to put these things into writing because you want to make sure that people are doing it properly. But there are always going to be the odd case where you just think, oh, I know this one's confirmed. It's just not meeting the guidelines for some strange reason.
Peter:
19:40
It's impossible for it not to be confirmed, but...
Tim:
19:42
You could run that again on a, you could run exactly the same sample then on a triple quadrupole for example, and confirm it.
Peter:
19:49
Yep. Or just switch your instrument to SIM.
Tim:
19:51
Yeah, I think the more you see the better though really it's, on the whole, I think doing, even though you don't want to do full scan spectra in every instance, right, it's gotta be fit for purpose and sometimes it's just not practical to do that. But on the whole, I do think full scan spectra is better and it gives you, I think over time if you're monitoring full scan spectra, it really gives you a good picture of what's happening for drugs in the, um, well in the mass spectrometer, I guess we're talking about in this case.
Peter:
20:20
But some guidelines do say that if you have a, if you have a certain number of ions, then you don't have to worry about measuring ion ratios. So if you're doing SIM or MRMs, then you have to have ion ratios to confirm. But if you have diagnostic ions, enough of them, then you don't necessarily have to get the ion ratios right. And I think that's a practical aspect because trying to get ion ratios off a full scan mass spectrum, it's just a nuisance, it would really hold up analytical work.
Tim:
20:44
But what we really want is for the software that controls various instruments to be able to do exactly what we want it to do in a very quick way in order to show us that it's meeting these kinds of identification guidelines.
Peter:
20:55
Yeah, so the software should have a drop down menu to say which MS Guideline are you adhering to?
Tim:
21:00
That would be great.
Peter:
21:02
Yeah, it would be.
Tim:
21:02
But then do we trust the software? Well at, this brings up another thing, Pete with library matches. So...
Peter:
21:08
Ah, that's what I was going to talk about.
Tim:
21:08
In any kind of detection, whether we're talking about IR, whether we're talking about UV, MS, there's, they've always got libraries that you can match to and there's various algorithms that you can use, forward and reverse matching and things like that.
Peter:
21:23
So when is a drug not a drug? When it's 65%, 75% match?
Tim:
21:27
Yeah. I Dunno. Maybe this is just me being cautious. I've always been a little bit skeptical of library matches because unless it's like a, a fantastic match and then it'll probably give you like a 99% match or something like that. But if it's anything that's kind of on that border where it's getting down to 70, 60, it's really touch and go as to whether it's it or not I think.
Peter:
21:46
And that's why in a lot of guidelines it says the laboratory has to verify their own score to determine whether a drug is confirmed or not.
Tim:
21:54
Yeah, well its, cause it's not like you can set a number in a guideline because all the different instruments and software have different...
Peter:
21:58
But then the laboratory has exactly the same trouble. So how do they verify it?
Tim:
22:02
Yeah, I don't know. It's extremely difficult I think to verify that. But the thing about using those kinds of scores that you get from databases is that it does make things a lot quicker. If you have to rely on an analyst's judgement every single time for every single confirmation that really slows down your process.
Peter:
22:19
There is a risk of contextual bias coming in there.
Tim:
22:23
Yeah. Well this is where if you're using MRMs for example, on a triple quadrupole instrument, you can set up your reports very easily to just have, you know, pass, fail and it will highlight which ones pass and which ones don't and so on. There's very little thought involved at all into whether something is confirmed.
Peter:
22:39
Yeah. As long as you trust the guidelines as being the ideal confirmation.
Tim:
22:44
You mean in terms of the ratios, the windows on the ratios and things like that.
Peter:
22:48
Yeah because they do have different ratios depending on which guidelines you look at.
Tim:
22:51
Yeah, I think they're all sort of trying to get at the same thing. Like I think no matter which guidelines you read, like we've mentioned these half a dozen different organizations and there's lots more guidelines out there as well than the ones we've mentioned. I think they're all trying to get at the same thing. You can see what they're trying to do, but obviously it's written by different people in different contexts. And so the wording's a little different. Some of the parameters are a little different.
Peter:
23:12
And sometimes the wording is not that precise and can be misinterpreted or interpreted in two different ways.
Tim:
23:18
Yeah, well actually that's what I like about the, the process they're going through with the OSAC guidelines, they're sort of putting out a draft document for comment from basically anyone it seems. Hopefully they will get a lot of feedback from people in a lot of different contexts to say, Hey, the way you've worded this here means that this is a loophole that you know, could be exploited or something like that. So with these guidelines though, Pete, I think pretty much all the guidelines that I've read, the identification guidelines, seem to work on a kind of threshold approach where you've got to get up to a certain threshold. Maybe they use like a point system or maybe they don't, but it's just you've got to have this and this and this. Once you get above that certain threshold, then it's confirmed. It's a positive and basically anything else you get after that is just gravy. It doesn't really matter.
Peter:
24:03
Super confirmed.
Tim:
24:03
Yeah. It's just all regarded as confirmed. You could think of this as equivalent to like cutoff levels in a urine standard or something like that where once you get up to that positive level, then the urine is positive for that drug and it doesn't really matter how positive it is, it's just positive or negative, but I wonder if it would be possible to have a different kind of approach to identification, which is based more on probability where, because we know like as we've been saying before, a low resolution MRM confirmation of amphetamine is not really the same as a high resolution full scan MS of Diazepam, which has 10 ions there. We know that they're not the same.
Peter:
24:44
Yeah.
Tim:
24:44
Right? That there's more chance of that amphetamine being something else than the diazepam being something else, but that's not reflected in any kinds of guidelines because they both, once you get over that threshold, they both meet the criteria. I wonder if it'd be possible to have, assign some kind of probability to the assessment of that.
Peter:
25:03
That's the general trend of forensic science, isn't it? To go towards Bayesian type probability. I have no idea how to apply it in toxicology.
Tim:
25:11
Yeah. I don't...
Peter:
25:12
I've been lucky enough not to think about that too much.
Tim:
25:14
Yeah. I don't really have the statistical background to be able to sort of take this idea anywhere, but if you think about a urine cut off level, again, for example, so I'm talking about quantitative cutoff here. If you report a negative result, right, that's just below the cut off. We know because there's a measurement uncertainty associated with that result, we know that there is a chance, a very small chance, there's a chance that, that the true result in that sample is actually above that cutoff.
Peter:
25:40
Yeah.
Tim:
25:40
So surely the same is true with confirmations. If you confirm that amphetamine based on a couple of transitions in a low resolution mass spectrometer, we know that the chance of it being something else is very, very small, but there is still a chance. Right? Can you assess that probability that it is something else and then compare that to other techniques and other compounds and so on.
Peter:
26:03
So in that sort of calculation of probability you need things like a weighting factor, like um how good your chromatography is, so how narrow your peaks are for example.
Tim:
26:11
I guess so.
Peter:
26:12
And number of plates in your column.
Tim:
26:13
Yeah. You probably need to take into account like what common interfering ions might be present. This is where, you'd probably need a huge amount of data to be able to do this kind of statistical analysis, but just on the principle of it, aren't we being a little bit misleading in suggesting that something is definitely confirmed when we know that there is some uncertainty about that. Just like we know with our quantitative results, we know that there's a possibility that we haven't got the true result there. Don't we also know that for confirmations, that there's a possibility that it could be something else, even though it's tiny?
Peter:
26:44
Well we had enough trouble trying to work out how to do measurement uncertainty of drugs. It took us long enough to get that under control. But this, maybe this is the next thing we have to get under control.
Tim:
26:52
Because when you're adhering to these guidelines that we've been talking about, once you get above that threshold, basically all identifications are equal, but we know actually some identifications are more equal than others.
Peter:
27:03
Yeah. And if, say if you have, even if you have for example, Diazepam, and you've got some MRMs for Diazepam, your ratios are slightly out, but they're still within the guidelines. Maybe there's a weighting associated with that.
Tim:
27:15
Yeah, see I guess this is where it could be useful is in saving work, saving kind of rework and things where you're really trying to get to that threshold and maybe you're just below that threshold and if you do some more work, obviously you could get it up to that threshold, but is it really necessary to do that work? Or what about, here's another example, where you've got a high concentration of a drug and so you're monitoring various transitions, but one of the transitions saturates. Your highest transition, so this might be your quantitation ion for example.
Peter:
27:47
So how many transitions you got?
Tim:
27:49
Let's say you've got three.
Peter:
27:50
Okay. Yep.
Tim:
27:51
One of them saturates on the detector and so it throws out the ratios.
Peter:
27:55
Yep.
Tim:
27:56
But you know why it's failing. It's because one of them has saturated. You know, that's it. Do you really need to then go and re-inject at a lower volume or you know, do, you can, there's various ways you can get around it to then try and get it above the threshold. But do you really need to?
Peter:
28:11
You could still do that now, but just use different ions. So if instead of three ions you monitored five ions, you knew one of them was out, then you can just do the calculation on three other ions.
Tim:
28:20
Well, interesting you say that. No, I don't think you can according to some guidelines, I think they say you can't exclude uh an ion, although I don't know, in that particular case where you know the reason why it's failing. I guess someone could argue though, do you really know that that's the reason? Like you're assuming that's the reason because you're using your judgement, but do you know for sure. It gets very complicated.
Peter:
28:40
You're talking about the probability say, um, if a drug has more ions, for example, so methamphets only got three or four fragment ions, whereas diazepam...
Tim:
28:50
And not very characteristic.
Peter:
28:51
Yeah, yeah, not very, diazepam's got what'd you say 10 ions?
Tim:
28:55
Well, something like that.
Peter:
28:56
But what if there was an isomer of diazepam? Another example might be protriptyline and nortriptyline. I think they're isomers and there spectra have exactly the same ions. There might be a few differences, but pretty, very, very, very similar. So even though there's, you'd think there'd be a much lower probability of that identification being wrong, it's not necessarily the case because you can have isomers that have the same fragments or even some metabolites will have the same fragments.
Tim:
29:21
Yeah. So then doesn't that all factor into the probability equation then of, what's the likelihood of there being, cause that's what it's all about. What's the likelihood of there being another compound which has the same retention time, which has virtually the same mass spectra, but maybe for some reason the ion ratio is a little bit out or something. What's the odds of that happening? Obviously it's the odds of that happening. A better for a very simple molecule like amphetamine, which doesn't have very characteristic ions.
Peter:
29:49
Yeah and it'd be different depending on your chromatography quality as well. So all these things could weigh into it perhaps, so higher the molecular weight, the more weighting you get.
Tim:
29:59
Yeah. And then see here's the thing, would it be specific to every individual method in every lab? Like you know, you'd have different methods would have different things coeluting and...
Peter:
30:10
Well we have that for measurement, uncertainty for quantitative analysis.
Tim:
30:12
Yeah I suppose so, every lab seems to find a way to do it.
Peter:
30:15
Let's just make life harder for everyone out there.
Tim:
30:17
Well it could make life easier though. That's...
Peter:
30:19
Do you think?
Tim:
30:19
That's what I think, if you, if our clients would be content, I mean that's what it really comes down to. What, what are the court's willing to accept? But courts are often willing to accept probabilistic evidence. I mean that's most forensic evidence is like that.
Peter:
30:33
Yeah, they're used to it aren't they?
Tim:
30:33
Toxicology, we tend to think of it as more black and white, but we know that it's not always so black and white. So is there a better approach to convey that to the courts? I don't know if there is or not. I don't know if it's possible to do this, but one of you out there, if you have a background in statistics, might like to take this idea further.
Peter:
30:52
I guess in a way we do have a probabilistic approach to it because in some guidelines, high resolution mass spectrometry accounts for a higher points weighting. So that's, isn't that sort of taking points into consideration there or probability?
Tim:
31:07
Yes, I guess it is in a way, but it's still the threshold model where once you get above a, the threshold is set quite low. Well, I don't want to say it's low, it's high in terms of our confidence in it, but it's low in terms of what we can possibly do on our instruments and so, and but it's that set that low on purpose because not all labs are going to be able to achieve that and that's fine. That's the way it is.
Peter:
31:29
And it's probably not necessary to go to that high, I mean...
Tim:
31:31
No it's not necessary, but if you can go further and you do go further, shouldn't you be able to reflect that in your results somehow that you have a higher degree of certainty in that result.
Peter:
31:43
So in a points system, like maybe the OSAC guidelines, you might have four points for methamphetamine, but for Diazepam you might have 25, just because there's so many ions.
Tim:
31:54
One other thing that's often talked about in identification guidelines is about signal to noise. So a chromatographic peak might have to have a signal to noise of three to one. That's pretty common. That's probably the most common.
Peter:
32:07
So what's, what's noise then?
Tim:
32:09
Yeah, exactly.
Peter:
32:10
Is that chemical noise or is it electronic noise?
Tim:
32:12
Yeah. In a high resolution mass Spec, you get very little noise.
Peter:
32:16
Yeah. You can have very clean chromatogram. But what if there's a compound right next to it, which is chromatographically separate that is half the height of the target on that you're looking at.
Tim:
32:26
Yeah.
Peter:
32:27
That's not noise. You know that that's, that's a compound don't you?
Tim:
32:30
Yeah. But when, when do you know that, I guess? Because if you imagine you've got a chromatogram and it's, you've got that sort of noisy baseline and then you've got your analyte peak coming up in the middle. That noise that we're talking about that your, your peak has to be three to one of that noise. Is that talking about the maximum noise or the average noise? Is it just the noise in that window, retention time window that you're looking at? Or is it across the whole chromatogram?
Peter:
32:56
Is it calculated, um, across a certain number of points? There's the RMS noise.
Tim:
33:02
Yeah. Often these things aren't defined. And as you say, if you've got a, if you've got a particularly high bit of noise that's coming out right next to your peak, maybe it is another compound. But how do you know until it gets to a certain threshold? It's really hard to know if that's actually a compound unless you know what the compound is, I suppose.
Peter:
33:17
So we're trying to be, trying to make it less subjective, but really I think it still is quite, isn't it?
Tim:
33:23
Well, yeah, I think that is a perfect example of where it's really hard to put these things in writing in guidelines.
Peter:
33:29
Yeah.
Tim:
33:29
You'd have to write an essay. People probably have written essays on that, but you've gotta make guidelines simple so that people can follow them. If they're too long, this is like uh standard operating procedures in your laboratories. If they're too long, people just won't read them or they'll get lost. They'll never be able to find what they're looking for. You've got to make them simple enough to read and follow, but complex enough to take into account all of these kind of things that we're talking about.
Peter:
33:53
Yeah, it's pretty hard.
Tim:
33:53
And actually one thing that should be noted, even though the OSAC guidelines at the moment are out for comment as we mentioned, but no matter what guidelines your lab is following, I think this should always be an open conversation to be able to provide feedback to whatever the organization is, whether that's GTFCh or TOXSAG in Australia, whatever the organization is, you should always feel free to provide feedback, I think. If you're seeing something in the guidelines that's actually not lining up so well in practice that, that kind of feedback is really valuable.
Peter:
34:21
And it might not be, it might just be uh written a bit unclearly, it might not necessarily be what they mean. So it's always good to clarify it.
Tim:
34:30
Yeah, so identifications pretty complex, but in general we do a really good job in forensic tox. We don't want to give the idea Pete that we're not doing a good job right, as a field?
Peter:
34:39
No we're very responsible I think you'd say.
Tim:
34:41
Yeah, I, I think if anything, I guess my point in a lot of this is that maybe we're too cautious about things. Maybe we're too conservative in saying whether something's there or not. Maybe we could be a little more...
Peter:
34:52
Bringing in contextual bias into it as well. Should we either, ever be conservative? I mean, if, if it's a drug facilitated assault case and a drug is there, are we more likely to confirm it then we would be for a driver's case where we don't want to wrongly convict someone.
Tim:
35:08
Yeah, that's a, that's a huge issue, isn't it? That whole thing of contextual bias. So if you have any comments about any of this stuff we've talked about today or anything else to do with identification?
Peter:
35:18
Or complaints?
Tim:
35:18
Yeah, sure. Complaints. We'll take complaints. Send them into the toxpod@sa.gov.au.
Peter:
35:24
Thanks for listening to us.
Tim:
35:26
See you next time.
Immunoassay
Chromatography
UV-VIS Spectroscopy
Mass Spectrometry
Is the threshold model the best approach?
Signal to noise
×

Listen to this podcast on