MR. TELLIARD: Our next speaker is Brian. MR. FOWLER: First of all, I would like. to thank Bill and the EPA for the conference, and I think it

Transcription

1 130 DETERMINATION OF 209 PCB CONGENERS AND CONGENER GROUPS BY HRGC/HRMS USING A SINGLE GC COLUMN: DETAILS OF EPA METHOD 1668A MR TELLIARD: Our next speaker is Brian Fowler Brian is with Axys, and as he has already indicated, he is going to be talking about Method 1668 and the 209 congeners and the ever-continuing story MR FOWLER: First of all, I would like to thank Bill and the EPA for the conference, and I think it is a privilege to be able to describe this method which I think is a real advance on congener-specific methods At Axys, we have been using various versions of 1668 for a number of years, and we are proud to have contributed to development of the method from the draft form through various forms to the Method 1668A The title of my talk might need a slight explanation Determination of 209 congeners and congener groups by HRGC, high-res GC, by HRMS, high-res mass spec, using a single GC column It is also a little misleading It says I am going to give details of Method 1668A Well, it is 130-odd pages of quite detailed text, and we don t have time to go through all the details So, what I will do is give an outline of the method and then discuss some of the more controversial parts of this method Now, 1668 has been developed initially as a method

2 131 for the determination of the toxic congeners Those congeners designated as toxic by WHO are indicated on this figure It includes the non-ortho substituted, the co-planar PCBs It also includes 12 mono-ortho-substituted PBCs that are biphenyls with a chlorine in one ortho position This is the list of 12 toxic congeners as listed by the or as designated by WHO in 1998 The original method was written around the older WHO list which also included two di-ortho-substituted congeners that currently aren t part of the controlled suite of congeners in 1668A I have mentioned that there has been evolution of this method, and I hope to explain how this has happened in this figure The draft method 1668 draft was published in 1995, modeled closely on the high-res dioxin method, 1613 It incorporated features from a number of labs, including ours, for the determination of co-planar PCBs, but it exploits the unique elution order that you see on the Octyl column The quantitation is using C 13 -labeled forms, analogs, of ten of the toxic congeners After this method had been in use for a while, it became apparent that the method had more potential than just for the determination of the toxics, and under contract, we did some method development We modified the GC conditions

3 132 and the high-res mass spec conditions to acquire all 209 So, that is from the mono through the deca biphenyls We also formulated standards in such a way that in five mixes, all congeners can be fully resolved That allows the retention time of each congener to be unambiguously determined Developed quantitation methods using isotope dilution against additional labeled standards that were added, and then tested the 209 congener method using aroclors At the Norfolk meeting two years ago, I presented the results of that, and the agreement was really quite impressive, considering that the reference data was generated by multi-column methods and fractionation methods So, with a single run, one could get quite good agreement With two runs that we had in this method, you got exceptional agreement between the published aroclor compositional data and this method Subsequently, this method is considered by New York/DC, as Larry just described, and features of this method were adopted for their program It became apparent that they needed only a single column, so they opted for the variant that used the Octyl column alone, and we have been using that version as one of the contract labs for the New York Harbor study

4 133 In the process of analyzing samples in this program, we accumulated experience of the performance of the method and were able to document how the method could be applied on a routine basis With discussions between the EPA and ourselves, features of this method were incorporated with additional standards and other calibration criteria that led to Method 1668A, the draft version The draft version was subject to a peer review A number of labs, I believe 11 labs and a number of agencies, reviewed the draft 1668A, and their comments were incorporated which has led to the current version of 1668A, the 1999 version So, that is the history of 1668A Just to reiterate about the draft method on which it is based, the draft is for the determination of 13 toxic congeners, tetra through hepta It is the high-res GC/high-res MS method, adopting many features of a dioxin protocol So, it is high sensitivity, high selectivity The high res is operated at 10,000 mass resolution, selected ion mode, electron impact That is the common mode that is used for dioxin analysis They have allowed many of the features of dioxin analysis to be adopted for PCBs with all advantages of the high precision that had been realized by that type of method

5 134 The draft method used two columns, an SPB-Octyl, 30 m, for 11 of the chlorobiphenyls plus the DB-1 for 2 of them The 10 labeled congeners were used as internal standards Now, moving on to the 1668A, this is GC high-res conditions expanded to acquire all 209 congeners The primary column is a 30 m SPB-Octyl, and that is used to report all 209 congeners A DB-1 is an option for confirmation where required Linearity and QC are based on 11 toxics So, although all 209 are used in some parts of the calibration, all the QC is just limited to the toxics This actually should say 12 toxics Sorry The 209 congener mix is used as a single point calibration to determine daily the response factor and retention time of all 209 in one cocktail Prior to extraction, samples are spiked with a labeled internal standard that includes 27 labeled congeners This extensive range of congeners includes components that are not part of the present New York method and is a major expansion of the draft method Here is probably a reason why the method is currently not in use, because we are still waiting for all 27 congeners to be available So, what I am describing is the method that will be run when all labeled forms are

6 135 available which I understand is very shortly The labeled standard includes label forms of all the toxics as well as the first and last eluters of each congener group After extraction, the lower extract is spiked with three more labels as cleanup standards Then, after cleanup, five additional labeled standards are added as instrument injection internal standards, also known as recovery standards Perhaps the controversial part of this method is the use of a single column and the use of the Octyl column, so I think it is worthwhile digressing for a minute on why that column is attractive and what operational consequences of it would be Now, since we know the retention time of all 209 congeners and the peak widths on various columns, you can simply model how efficient the columns would be for a congener analysis, and this indicates various separation criteria There is no column that separates all 209 congeners In all cases, there are co-elutions So, in this case, what we are saying is that peaks separated by 5 seconds, if they are considered resolved, an Octyl would give 164 resolved components A DB-5 would give 156 and so on Now, if you look across, on most columns, Octyl has a higher efficiency than other columns that are listed

7 136 here; however, not always At 8 seconds, DB-5 appears to have more resolution than an Octyl column, but I should also point out that that is a 60 m DB-5, and this is a 30 m Octyl So, instrument time is substantially longer here than here, and there is, since it is an expensive piece of equipment, it ties the instrument up unnecessarily using the longer column, and the advantages are relatively small for the increased run time As you can see, the number of resolved congeners falls as the separation criteria increase, but it is necessary to set some reasonable separation criteria to maintain a consistent report list So, let s just look at that issue for a moment Here we have a case where two components are separated by approximately 8 seconds This would be in the case for tetras Using typical peak widths, that would give about a 30 percent valley When those components are at equal abundance, they are obviously well separated When A is five times B, they are still practically separated So, we have adopted this as a practical working limit for resolution If I can compare that to the situation where they are only five seconds apart, there is about a 20 percent valley there When components are equally abundant, there is some separation, but it is not a practical separation,

8 137 because if the abundance of one component falls away, there is only this poor inflection which isn t a reliable integration point So, one would get very poor precision for the determination of that area with respect to that one Using this criterion, this doublet is actually treated as one peak, and I describe it as a peak domain This area is the sum of the two components, regardless of whether there is any separation by the integrator So, if you look at number of domains on an Octyl column to similar domains on a DB-5 and a DB-1, there is a similar situation Here we see that there are 159 practical domains on a 30 m Octyl and 160 domains on a DB-5 Slightly better, you would say, but remember, this is a 60 m column If we look at the actual number of resolved congeners, those individual compounds that have no coeluting isomers, Octyl is better than the other ones So, statistically so far, Octyl does look a good bet Now, the most critical part is how these columns handle the toxic congeners What I have listed out here is how many isomers co-elute with the toxics and how many higher homogs co-elute 1 + Cl means are there any congeners with one more chlorine co-eluting with a toxic, are there any congeners with two more chlorines co-eluting with toxics Immediately, this figure 2 looks like it is a

9 138 failure by Octyl This is the instance where the two hexa congeners, 156 and 157, co-elute They are both toxics, but, fortunately, those two congenersthose two isomershave the same toxic equivalent factors so that the method can quite legitimately handle these as one value, and the area of that can be used to derive a toxicity from the PCB extract The sum of these co-eluting higher homologues is 9 for Octyl, 13 for DB-5, and 12 for DB-1 So, once again, statistically, the Octyl column appears to have some advantages over the other columns, and for that reason, we have stayed with it Now, we have come up with a number of performance criteria This is one of the resolution examples that is shown in the method, and at the back, you probably can t see these numbers They are not particularly important What I wanted to point out is how a typical Octyl columnthis column had been in use for approximately two weeks when this chromatogram was acquired This pair has to be resolved with a valley of less than 40 percent For a new column, it is usually less than 5 percent There is another example published in the method where we demonstrate the required resolution for two octas, and the third really critical criteria is that two

10 139 congeners, 156 and 157 that I mentioned in the previous slide, do actually co-elute or elute within, I think it is, 3 seconds That is a characteristic of an Octyl column If they are resolved by the column, it is no longer a true Octyl column, and the elution order is not reliable So, there are two resolution requirements for the tetra and the octa and the co-elution requirement for 156 and 157 make a working performance spec that we have applied, and we have found it is practical for use for Octyl columns, and I think we have used 15 or 16 of them now, and all of them have met that spec When they show signs of degradation, there is clearly separation between 156 and 157, so it appears to be quite a useful indicator that the column is no longer serviceable So, that is the good part of Octyl, but I would admit that there are some problems with it It requires more maintenance than a DB-5 or a DB-1 or the typical more rugged columns that labs use However, it is not an unmanageable situation We find it is necessary to do a daily adjustment of head pressure to keep retention times within the right window and do front end trims to keep the resolution within spec One column will last for several hundred runs So, the cost of the column is not prohibitive, although they steadily wear out

11 140 The column also bleeds There is progressive loss of stationary phase So, during the service life, the head pressure may drop by more than a factor of two from its new value As the Octyl column is swept out, you start to see more of the column behind it which is more like the DB-1 which is another reason why the Octyl column degrades to something that has a typical DB-1 type performance The column is very oxygen sensitive, and it is important to exclude oxygen as much as possible Also, in order to do this analysis, it is necessary to take the column up to or above the maximum that is specified by the manufacturer which partly is responsible for the column s wearing out and giving a lot more bleed than other columns would So, it is a mixed blessing, but we find it a practical column to use There is another feature, while I have got this chromatogram on the screen here, that I should show These peaks are typical examples of resolved components There are minor peaks appearing at the bottom here, and these are fragment ions from higher boiling homologues I ll go backwards and forwards between these two slides for a moment When there isthe chlorine loss from a parent ion results in lower mass fragments So, for

12 141 instance, a hexachlorobiphenyl will give most of the signal for the hexachloro parent ion, but it will also generate fragment ions from loss of one and two chlorines Those two losses will generate some signal for the lower homologues, and it is important to monitor for the presence of a higher-boiling homologue at the retention time of a target congener When a higher homologue co-elutes with a lower one, it will cause some interference It can raise its detection limit, and it can be an interference Two important things to look for on the Octyl column are co-elution or a very close elution for 166 and 128 near PCB 126 and for intereferences to 169 by congeners 190 which isthis is a hexa 190 is a hepta, and 198 is an octa 198 can occur at substantially higher concentrations from 169, so it is important to monitor for this interference If this occurs, you have options You can run it on another column or actually clean the extract up to strip out the interferences 190 and 198 are di-ortho substituted and are readily separated from this one So, you could do it on a carbon column Our preference is just to run it through on another column, on a DB-1, when that is necessary To give you some feel for the extent of that interference, an m givesgives 4 to 6 percent of an M-Cl,

13 142 and it gives 1 to 15 percent of an M-2Cl in typical run conditions That is with the source set up to minimize fragmentation Mass resolution can t sort this interference problem for the M-Cl One would need 60,000 resolution to completely eliminate the M-Cl You would need 15,000 resolution to eliminate the M-2Cl That is doable, but we find that this interference is generally not limiting The main mechanism to avoid this problem is by selective stationary phases If the column can separate the target congeners well, then this interference is really not a problem And I think that we are all looking for the holy grail, a column that will separate all congeners without any of these co-elutions The method is calibrated, 1668A is calibrated this way First of all, we have the five native congener mixes That is five mixes, each of which is totally resolved in Octyl The mix is formulated so that it allows unique assignment of all retention times, and it is periodically used to confirm elution order The method is calibrated using the 12 toxic CBs and the C 13 -labeled level of chlorination, LOC, and labeled quantitation internal standards The linearity is done with a five or six-point ICAL using these, and the related QC which is the CALVER, IPR, and OPR also use these

14 143 This keeps the QC to a manageable level It would be quite onerous if we used all 209 congeners for ongoing QC The 209 congener mix is used as a resolution test to demonstrate that the column is working correctly, and, also, it is used every shift, every 12 hours, to determine retention times and response factor, and there is a continuous retention time and response factor update Since there are so many standards in the method, it is useful to plot them out to show how they are used The dark circles here are the injection internals, the recovery standards These are used as primary references for peak location These yellow triangles that don t show up too well are the cleanup standards They are designed to span a reasonable range of retention times The blue diamonds are the level of chlorination standards and the labeled forms of the toxics The intent of so many internal standards is to minimize the spread between the target compound and its reference, so at any point, there is only a 20 percent difference between the two So, the maximum relative retention time range will go from 08 to 12 using this scheme In general, the relative retention time range is very much smaller than that, so there is excellent

15 144 reliability of peak location, because there are so many standards to reference to Going back to calibration, the draft method had a fairly complicated scheme using different concentration ranges for tetras through hepta, and the total range was only a factor of 400, whereas the dynamic range or the concentration range of congeners in most samples is very much greater than that So, implicit in this limited range were many dilutions or reanalyses to stay within the calibrated range for the draft method That is because the draft method incorporated some very low abundance compounds such as 126, 77, and some very high abundance congeners, specifically, PCB 105, 118, and it would be regularly necessary to dilute things down by large factors Now, in 1668A, that has been addressed by increasing the dynamic range to really utilize the full working dynamic range of the mass spec The basic range is a range of 2000, but it is also specified to go to a lower point, CS2 which gives an extended range of a factor of 10,000 which should substantially reduce the need for dilution of samples to stay within the working linear range Considerations for reporting 209 congeners I ll go back to another chromatogram in just a moment It is important that the resolved individual congeners are selected to ensure that a consistent report list is

16 145 generated Because there are minor changes in chromatography can affect which compounds are reported, you don t want to be in a situation of only being able to report one compound when the column is fresh and, subsequently, have it coalesce with some other group So, the GC performance has been backed off so we actually are reporting 129 individual congeners off the single run Octyl column Unresolved pairs and groups of isomers are selected to allow for some column degradation The congeners are listed in the report by level of chlorination, usually just in order of IUPAC number They are listed from IUPAC 1 through 209, so we ll start at the monos and list out each resolved component And it is important to indicate how we deal with co-elutions I ll just go back to the example chromatogram again In this case, all these compounds are reported as individuals, but where there are co-elutions such as hereobviously, this peak domain is 28 and 20 At the listing for congener 20, there will be an indication that it co-elutes with 28 and a concentration of the domain entered On the line for 28, there will be no concentration entered and just the reference to its co-eluting partner Using that strategy, there is no duplication of congener list So, you can come up with a reliable homologue total

17 146 I ll go back to this one So, the concentration for unresolved isomer groups is reported only on the entry for the lowest congener number with no concentration given on lines of other multiple components The unique elution order of Octyl means that off this column, the co-elutions will be different than off other columns such as a DB-5, a DB-1 The strengths and weaknesses of the method This is a high sensitivity, high selectivity method, and it allows the toxic congeners and a wide range of other congeners to be accurately measured in complex samples with high confidence The method is calibrated using all 209 congeners which is an important point when looking at homologue totals There are no homologues that are targeted generally just within a retention time range Everything is targeted on a particular retention time, and every targeted component has to meet isotope ratio There are numerous internal standards built into the method and extensive built-in method performance to assure data users that the method is in control The similarity to 1613 method should allow this method to be readily implemented by other high-res mass spec users Weak points, the Octyl GC phase is less stable, and the column requires more attention than alternative

18 147 phases such as a DB-1 or a DB-5 On Octyl, there are occasionally co-eluting high homologues which can interfere and raise the detection limits for toxics, in which case, a second column may be required So, we have a working method, and it is, I think, useful at this point to point out the advantages of such an ambitious method, a full congener method The obvious comparison of a congener-specific method is to aroclor-based method Aroclors include only about 140 congeners, and many congener lists have been developed about those congeners present in aroclors Congeners absent from aroclors are indicators of other sources such as degradation of aroclors or specific sources We have found, by analysis of water, suspended sediments as part of the New York Harbor project that we routinely detect more than 140 congener domains, and those domains include many non-aroclor congeners The high sensitivity of the mass spec is a factor as well as the specificity of the Octyl column A review of the frequency of detected components, congeners in real samples, shows that it is difficult to find a congener that is never detected One interesting example is for congener 11, IUPAC 11, 3,3'-dichloro The 3 primes are not favored for substitutions, so you wouldn t expect it

19 148 to be found It is certainly in an aroclor It is noted as absent in aroclor, noted as less than 005 I suspect it is actually around this concentration, and this concentration is readily determined by 1668A In some instances, though, it rises from a trace amount of total PCB to almost 1 percent of the total PCB is this rare congener, and this anomalous abundance indicates some non-aroclor source of that congener That would be a kind of result that would not be generated by something based on components in aroclors So, I ll just reiterate some of the advantages of the full congener method, that you can generate a toxic equivalent from a PCB extract in a single run You can use alternate TEFs if people have specific interest Congener groups can be summed to give toxicity or some parameter from the congener totals For instance, for risk assessment purposes, the total of tetra through deca is an important concentration for PCBs You have reliable homologue totals, summing only those things that meet retention time, isotope ratio, and various other QC requirements with each congener specifically targeted For those who wish to link this data to historical aroclor data, you can use marker peaks to calculate aroclor

20 149 equivalents This method allows the determination of congeners not present in aroclors, such as dechlorinated and specific-process congeners And such a rich data set would be a great source of interest, I think, for people who do chemometric analysis To summarize, I think development of 1668A marks a significant advance in the PCB analytical methods and provides a sensitive, congener-specific method that includes all 209 congeners either as individual congeners or members of a group of resolved/unresolved isomers Currently, we are reporting 129 individuals for 159, I think, domains The WHO toxic congeners are determined at trace levels in the presence of the major congeners in an instrument run That is something I maybe didn t stress before is that these toxic components are minor, and they are analyzed in the presence of major components that may be present at several orders of magnitude higher, but they are all determined in one run The method has evolved to use the best technology currently available and is limited primarily by the congener separation and column performance The Octyl column is good, and we look forward to the day when an even better column is developed So, that concludes my presentation, and I would be

21 150 happy to answer questions Thank you QUESTION AND ANSWER SESSION SPEAKER: Hello I have two questions The first oneboth of these are in relation to the biosolids land application rule This method is specified in there The rule only requires the analysis of the coplanar PBCs which are the toxics, but it specifies the expanded Method 1668A, and it is unclear to me, does that

22 151 constitute a complication of the analysis? Does it make it more difficult to use 1668A, or does it make it easier, or would it affect the cost is what I am getting at MR FOWLER: No, it is not a complication at all The subset of congeners can be determined out of 1668A without any conflict SPEAKER: And the second question is really more to Bill In that same rule, the EPA definition of dioxins was expanded to include the coplanar PCBs, and that was, I guess, could be controversiali don t knowbut my question is that in the future, will analyses for dioxins also require both Method 1613 and 1668 in order to completely perform an analysis for dioxins, or is this a rule-specific definition MR TELLIARD: It is a rule right now, what they are looking at is they want to come up with the toxic equivalent loading, again, for land application So, they want to look at the standard dioxins which we have been doing and the, quote, toxic or coplanar PCBs from this method So, if you were going to run for 503, you don t need to run all of the congeners You could run just the toxics and the standard 1613 method So, you would run the predecessor method to this If you didn t for your purposes, for biosolidscan t say sludge anymoreyou

23 152 would just run 1613 and, basically, 1668 before it was revised where it just covered the toxics SPEAKER: Do you know ifcan these be combined into a single analytical run, or do they need to be like two separate analyses? MR TELLIARD: Yes, we are not going to make it easy SPEAKER: Bill, I am following up on that question there If you are interested in the TEQs for dioxins and the PCBs, you could very possibly have both run in a single analysis MR TELLIARD: Right SPEAKER: Since you don t need to do the totals on 1613 You still get a 2,3,7,8- and the coplanar PCB You could do this in a single run That is technically possible SPEAKER: I have one double-edged question and a second question The first one deals with your evaluation of the different columns I was wondering what the age of the columns were I would presume that they were all new columns that you were comparing to one another when it came to the relative resolutions, but it would be very interesting to know how well they do relative to one another when they are older columns as they become aged, since the Octyl column obviously ages much more rapidly

24 153 Secondly, on that same question is whether or not you used lab-generated samples or real-world field samples It is our experience that when you use real-world field samples, even though you go through quite a bit of cleanup, the columns age much more rapidly than if you use standards or aroclors or standards that you develop in the laboratory That is the first question The second one deals with the error involved in the TEQ congeners that you were discussing when you had the M-Cl and M-2Cl Have you calculated the error not in the concentration of the congeners that you are dealing with but the error in the TEQ that would be generated from those concentrations? Because that, ultimately, is what counts MR FOWLER: Okay As far as the age of columns that were employed in that comparison, these are all columns in good condition They had all been conditioned and, typically, in use for two weeks to four weeks in normal lab operations There is no doubt that an Octyl column ages more quickly than a DB-5, but that is something that we are able to handle at this point Getting on to your third point about the coelution of higher homologues with the toxics, if there is evidence of an interference, it should be eliminated by either injection on a separate column or a cleanup

25 154 MR TELLIARD: Anyone else? Thank you very much Oop, one more Hi, Harry MR MCCARTY: Brian, Harry McCarty You mentioned briefly the QC criteria looking like the dioxin methods I am assuming that there is an abundance ratio limit that was applied to the PCBs The two peaks have to agree within a certain amount Dale had called me some time ago with some questions about how we got to what we did in 1613 years ago What finally ended up in 1668? MR FOWLER: It is the same as for dioxin It is 15 percent MR MCCARTY: Was there some discussion of an allowance for a little outside of that based on what your daily standard was or MR FOWLER: If so, I didn t hear it We have been using 15 percent MR RUSHNECK: Yes, the identification criteria are the same in Method 1668A as they are in Method 1613 MR MCCARTY: It is the same thing we proposed in Finland, then Okay MR TELLIARD: Thank you MR FOWLER: Okay, thanks

26 155