Document evB626zL5n4BBj9o1mpyejjDE

CORPORATE ENVIRONMENTAL PROGRAMS GENERAL ELECTRIC COMPANY FAIRFIELD, CONNECTIOU 06431 October 11, 1984 TO: J.F. Brown J.C. Carnahan W.Y. Ligon Attached is a copy of the preliminary report from the dioxin and furan round robins conducted by EPA and NIEHS. I would appreciate your comments on same. SBH:cas Attachment cc: B.I. MacDonald S.B. Hamilton 783285 GENP 010474 / NON-OBJECTIVES OF THIS STUDY (Any attempt to treat this study as though* it had any of the following objectives, or to draw conclusions about these issuer will be discredited through publications.) 1 . To provide an opportunity to advertise commercially ones analytical shills. .2 To provide a training exercise. 3- To perform a quality check on contractors To provide an opportunity to discredit competitors. 5. To determine the lower limits of detection of different techniques for these analytes. 1 6 . To determine precision or replicability. 7. To determine whether it is physically possible to make a sample that can not be accurately analyzed for these analytes. REAL OBJECTIVES OF THIS STUDY 1 . To determine which of the various analytical approaches taken by different laboratories are able to lobtailn quantitatively accurate measurements of the levels 1 of tox1i'c dioxins and !1 furans in the ppt range, when the "right answer'* is not based on statistical evaluation of analytical data presumed to be reliable, but on amounts spiked. (Note: In order to accomplish this, the levels of analytes in the unspiked sample found by a given laboratory must be considered to be correct, for that laboratory'. ) | 2. To do so (1. above) using a natural and relevant matrix, 3* To determine, for at least some of th anaiytes, whether or |i not they are likely to be accurately determined in spite of the presence of selected interferences. Similarly] to attempt to detect which methodologies are more susceptable to interferences.- Since "real" samoles are usually so limited in size that II averaging replicate determinations is often impossible, and since the cost of these analyses is so great that replication is usually not done even when possible, to obtain at leas an impression of the likelyhood of obtaining a reasonable approximation of an accurate result without replication and de termalnati on of precisione 783286 G EN P 010475 iTfli <8P loSSt' 5* Most importantly, to identify aspects of this general analytical problem that need further examination, validation, method development, or whatever. To if possible d'esign specific experiments to recommend be supported/perfojrmed. This may even extend to considerations arising out of thej findings on the unspiked sample. j 6. Approximately nine sets of objectives that the participants hoped to accomplish. 1 i 783287 B N P 010475 ESTIMATED LIMITS OP QUANTIFICATION IN 5 g SAMPLE PRIOR TO RECEIVING THESE SAMPLES, 2378-TCDD 2378-TCDF I 2378- P C D D 23478-PCDF 123^78-HCDD I 23789-HCDF OCDD ..Columbia 5-6 5-6 5-6 5-6 5-6 5-6 5-6 Dow 4-12 4-12 4-12 4-12 20-28 20-28 20-28 Monsanto 1. 1 1 1 1 .* l.k 2 U.S.E.P.A. 1-50 1-50 20-120 20-120 40-140 40-140 110-400 N.Y.D.O.H. 5 5 10 10 18 18 25 Wrierht St. i*-5 k-5 8-10 8-10 8-10 8-10 12-15 LABORATORIES REPORTING AT LEAST SOME INSTRUMENT OUTAGE/DOWN TIME DURING THIS STUDYi Dow Univ. Nebraska N.Y.D.O.H. Columbia NFRL GENP 010477 Nl MC0O0 00 00 SOME NONUNIFORMITIES A. Methods of reporting results. 1. Stalling/Smith: Two numbers for sample #1. First value used by me in the comparison tables. 2. Rappe: Two complete data sets. Both treated separately. 3* Gross/Weerasinghei Concentrations for >C1^ calculated two different ways, giving very different results. One or the other way must be chosen. Harless: Values for >01^ reported as "Total PCDD',' etc. B. Starting level of experience/claims: Not all the labs have experience at determination of isomers other than 2,378- or higher chlorologues. Not all the labs use methods considered by themselves to be validated or suitable for all dioxins and furans. Not all labs consider their techniques to be isomer specific. Not all labs have extensive experience with furans. 783289 GENP 010478 SAMPLES REPORTED AS DAMAGED OR POTENTIALLY DAMAGED IN SHIPMENT For certain sample numbers a "spare" was aval lable (no problems at ray end.) For others, no spare was available (problem with ampule sealing, breakage, etc. at my end.) The lipid content corresponding to 5 g of adipose was known independently. The vo.lume content of each ampule was not knowd because CHCl^ evaporation was not identical for all. A radioactive marker at extremely low level was present in known!1 amouInt in each ampule. Thus in most cases it was possiblej to djeal with minor leaks or evaporation if it occurred. | i (1) Not reporting any loss or damage duriJg shipment: Columbia NFRL; N.Y.D.O.K.; E.P.A.s Wri^ght sjtate. (2 ) Reporting possible leak based on comparatiY|e volumes, later judged no significant leak had occurred (various criteria): Univ. Nebraska - #1 Monsanto - #1 (3 ) Ampule broken, leak certain, replacement able to be provided: Umea Univ, (#3 , # 8 ). [ W Leak, amount lost able to be determined by independent means at two sites, data corrected accordingly: Dow (#1 ) (5) Major leak, not replacable, n o t 'accurately measurable, data either not included or must be considered questionablei Dow (#3 ) Univ. Nebraska (#9) Monsanto (#5) , DATA ON SPECIFIC SAMPLES OR COMPONENTS NOT PROVIDED .E.P.A. data on penta- and hexachloro isomers not known to be isomer-specific, so reported as "total PCDDj*, "total PCDF' "total HCDD", "total HCDF" . Instrument breakdowns prevented N.Y.D.O.H. from completing measurements; HCDF and OCDD levels not reported. 783290 10d 7 n I % TIMS FACTORS IN THE STUDY A. Original deadline for receipt of data: 1 July, 198*+. Extended deadline (C.J.) i 8 July, 198*4-. Ultimate deadline after begging and pleading: Aug. 31 198** B. Dates data received by P.A. : Columbia NFRL Dow Monsanto E.P.A. Tunney's Pasture Univ. of Nebraska N.Y.D.O.H. Umea Univ. Wright State 16 June 1984 29 June 1984 2 July 1984 *5 July 1984 9 Aug 1984 20 Aug 1984 10 Sept 1984 17 Sept 1984 C. Time elapsed between shipping samples and receiving data: Columbia NFRL 48 days Dow l days Monsanto 64 days E.P.A. 67 days Tunney's Pasture Univ. of Nebraska 101 days 112 days N.Y.D.O.H. Ume Univ. 133 days 140 days Wright State D. Number of worker-days spent on workup/analysis as reported by participants! Columbia NFRL 5 workup, GC-MS not given. Dow Monsanto 26 total 6 total J I Others Not reported, Ume (given as 36-48 hr.per sample) 783291 010480 SOURCES OF STANDARDS 2,3*7.8-TCDD 2,3.7 8-TCDP 1,2,3.7.8-FCDD ' 2.3, ^.7.8-PCDF 1.2.3, ^.7.8-HCDD 1.2,3,7.8,9-HCDF OCDD Other CDF's P, Albro, N.I.E.H.S.,, Bell/Camahan, General Electric F. O'Keefe, N.Y.D.O.H. General Electric Cambridge Isotope Laboratories, Inc General Electric J. Ryan, Tunney*s Pasture General Electric ! All standards were donated for this study at no charge. All standards were received as solids with the weights recorded on the containers. The listed weights were taken to be correct whether they were or not. All disolution and dilution of standards were made using PG toluene, on the day they were used for spiking. All standard containers were stored, opened, Ind eventually disposed of, in a different building from where the adipose was processed Preparation and storage (very short-term) of spiking solutions was not done in the same laboratory where the adippse was spiked, All glassware was washed with Micro, DD water, PG acetone and PG methylene chloride, with a final toluene rinse just before use. Spiking solutions were prepared in volumetric flasks with glass stoppers, sealed with Tefloi^tape. All participating laboratories noticed that the HCDD level in the reference standard solution was about 1 OX higher than the other components. The same would apply to the spiked fat* t Standards were received along with evidence of identity and whatever information was available on purity. That none of the standards contained potentially troublesome levels of any of the other analytes was checked by GCfECD using a 25 meter column of SP 2100. (Since what was being checked was absence of any peak at specific elution positions, ability to resolve all possible isomers was not necessary.) None of the standards contained interfering (>1f*) leirels of other analytes, although some extraneous contaminants were sometime seen. 783292 oio.4 8 1 EXTRACTION OF THE ADIPOSE TISSUE The tissue was thawed and weighed. 'Most ofj the available tissue, a 600g portion, was combined with ^800g of anhydrous, Pesticide Grade sodium sulfate in a Sears food grinder purchased new for this project.(The grinder was washed with Micro, water, and methylene chloride before being used,)j Thej "dry" blend was reprocessed three times for uniformity, and had acquired a granular state at the end. | The blend was packed into two Pyrex columns, 20" by ^"diam., and elutriated first with a total of threJ liters of ethyl1 acetate containing 0.001^ BHT antioxidant. This v^as followed by 6 liters of methylene chloride. The extracts were combined and concentrated by rotary evaporation in vacuo at 37 The residue was taken up in chloroform and made to 900 ml. The weighjt of lipid from 600 g of human adipose tissue was 531*9 g* Thelre was no visible residue insoluble in the chloroform. Pilot studies indicated that this extraction procedure gives a yield ofj lipid not statis tically significantly different from that obtained by Soxlett- extraction for 16 hours. BLIND SPIKING PROCEDURE Each of ten flasks containing a Teflon stirring bar and glass stopper was weighed, loaded with 1/10 of jthe lipid extract, and weighed again. One at a time, each flask was spiked with 1 ml of a blind spiking solution provided j(also one -at-a-time by the person who made it (different from jthe person doing the spiking). Each flask was numbered to corresponk with the number on the spiking solution. Those samples numbered 3*5*7*9 and 10 also received 100 j&1 of a solution labeled general interference solution". Sample #9 additionally received 100 jdl of a solution labeled "Furan Interference Solution"? the I"interference" solutions also were of compositions unknown to the person doing the spiking. NOTE; There was no "#1" spiking solution, Nothing was added to flask # 1 , except the lipid. 783293 Ol f\A~ Each flask was again weighed, after which the glass stopper was sealed in place with Teflon tape. The flask contents were stirred magnetically until, based on previous experience, they should have been thoroughly mixed. They were then let stand with the stirrer off for 5 minutes. Aliquots (100 il) were sampled from the top, bottom, and middle of the flask and radio- assayed for Since the spiking solutions had all contained lit a -C-labeled hexachlorobiphenyl (2,2,,3i3 \ 6 ,6 '), uniformity of C concentration was taken to indicate uniformity of dispersal of the spiking solution throughout the flajsk. Analysis of variance indicated that none of the sampled showed a CV greater than 3*37% of the mean (predetermined criterion being 5$) E a c h flask was again weighed, and, again one at a time, aliquots were transferred on a weight basis to ampules. Each ampule received sample equivalent to 5 S original adipose tissue. That is, the 900 ml of lipid in c iloro-form represented 600 g of adipose. E a c h flask received l/lO of this, or 60 g eq, Aliquots removed for radioassay averaged 0 20 g of adipose equiv. per flask, leaving 59^8 g equiv, in X g tojtal weight. Then (5/598)X grams of solution was transferred! to each ampule. The ampules corresponding to a given sample number were labeled directly, taped, and loaded into a vacujom desiccator (a separate desiccator for each sample), 'Sach desiccator was connected through a water-backup trap and Drierite trap to its own aspirator. Chloroform was reduced under very slight vacuum (to avoid the possibility of bumping) somewhat, but it was found that this process would have taken intollerably long to go to completion and it was decided to leave most of the chloroform in the samples. The'samples were imbedded in Dry Ice (solid CO.) and the ampules sealed with a methane/ioxygiel|n torch. Since the samples had sat in the desiccators forj aboujt 2k> hours prior to sealing, and since care had been taken to I d e p1o s i t the solutions into the bottom of the ampules originally,:there was no detectable residue in the heated area that might h a v e ,c h a r r e d Sets of samples (#s 1 -1 0 ) were then packaged for shipment and delivered to Bob Harless of the U.S.E.P.A. J/ho handled the shipping. Each package also contained the reference standard solution and a descriptive cover sheet. 783294 g ENP 010481 SPECIFIC INTERFERENCES ! In addition to the analytes listed, samples 3 5 i7 9 *and 10 were also spiked with: (1) A mixture of polychlorinated diphenyl ethers, *t-8 chlorines, at 200 ppt total. i (2) Two different tetrachloro methoxy biphenyls, at 50 ppt each* (3) Aroclor 1016, at 10 ppb. This Aroclor chosen for its negligible furan contamination. W 1,2,3.^-TCDD at 50 ppt, (5) o,p- and p,p-DDE at 500 ppt each. In addition to the above, sample 9 was also spiked with: (1) 2 , 3 8-TCDF at 10.6 ppt. (2) 1,2,^,8,9-PCDF at 7-9 ppt. (35 1,2,3.^.6,9-HCDF at 2 6 . 8 ppt. I ji ?8329s PREPARATION OP SPIKING SOLUTIONS AND REFERENCE ,STANDARD The ampules containing dry dioxin and furan standards were opened one at a time for processing. The contents of each ampule was repeatedly leached into small portions of toluene, the "extracts" being transferred to a jvolumetric flask. Eventually each stock solution became $P ml in toluene, a'different concentration for .each standarjd in ^.ccordance with the recorded weight. In this way the stock solutions became of defined concentrations which subsequently were assumed to be correct, and it was not necessary to assume jthat no material remained unextracted from the original containers. Calculated aliquots of each stock solujtion jto give a final concentration of 1 .0 ug/ml in 5 nil final volume were combined in a 5 ml volumetric flask and mixed*, I1 thoroughly through ^0 inversions after making to volume with toluene. The pipettes . II** used for this were guaranteed * traceable to NBS, Other aliquots of the stock solutions were combined such that, after making to volume, 1 ml would contain 60 X the desired ppt concentration i* spiked fat extract (since 1 ml would be added to 60 g eq. adipose extract in each case). Before making to volume, each spiking solution also received 261000 14- I dpm of C-labeled hexachlorobiphenyl per ml. INTERFERENCE SOLUTIONS The "Furan Interference Solution" was made as above. The "General Interference Solution" was made on a dilrect weight basis for stock solutions, which were combined and diluted as above. These solutions did not contain ^ C , bui! were in toluene, were spiked into the lipid extracts at the same time as the spiking (analyte) solutions, and received identical mixing. * Note* Each participant received 1 ml of this single reference solution, which was only prepared once. 783296 G flSIP 010485 ADDITIONAL CONSIDERATIONS RELATED TO SPIKING E a c h spiking solution was carried to the lab where the spiking was done only after the previous spiking solution had been used, the flask sealed, and the residual spiking solution il removed from that lab and taken to storage Thus there was only one spiking solution in the "spiking lab' at any given time, and all previously spiked flasks of lipid were already sealed and stirring. This serai-paranoid procedure made it essentially impossible to spike the same sample twice or mislabel the flasks. No flask was opened for radioassay until the previously opened flask had passed the radioassay criterion !| and been resealed. Cross contamination after spiking should tI have been as close to impossible as we could manage One aspect of the blind spiking was tiiat, alt the end of the process, no one in the world except th!e perjson who made the spiking solutions (ine), and who was albne in the lab while making them, could know what or how much was spkked. This made subsequent security consistant with the former standards of the U.S.Army Biological Warfare Labs, whic i oug ht to be good 'enough for this study. LABORATORY CONTAMINATION Extraction of the adipose was done in a laboratory newly set up for sample workup, in a building nojfc previously used for making dioxin solutions or performing dioxin/furan analysis. However, it is not possible to be absolutely certain that contamination was completely absent from the total relevant environment. The procedures used should have ensured that if TI contamination did occur, it would appear as "background" in all samples. It may be worth noting that we were able to process liver samples in a similar manner in these laboratories 783297 P S P 0104!, right after finishing the human fat work. Blank liver samples did not show 2,3,7,8-TCDD at a 0.7 ppt detection limit relative to ^^C-TCDD internal standard. Unfortunately, those samples were not analyzed for other dioxins or furans. ?83298 CLEANUP: BASIC APPROACHES USED. A . Columbia National Fisheries Research Laboratory 1. Add internal standards (^C-TCDF & ^Cl-TCDF). 2. Sample in CH/CH2C12 through two K Silicate/silica gel columns, then through charcoal trap. Wash with two solvent mixtures 3 Back-elute with toluene. Concentrate# 55*^o vacuum. Through H^O^/silica gel in hexane 5-.6 Dry at room temp, with N 2 dissolve in 5 ul o-xylene, (Note: composit sample later treated by alumina* chrom. to remove chlorinated naphthalenes not removed by above.) .B. University of Nebraska 1. Evap. CHCl^ with hot water bath, weigh lipid. 2 Add internal standard (^C-TCDD). 3. Samples #l-#5i saponify in refluxing wet etlianolic KOH, 60 min. k. Dilute, extract with hexane k X 9 wash extract with water. .5. Samples #6-#10, reflux 2 hr. in hexane! 6 Partition against cone. H 2S0^. 7. Wash with water, dry under N 2 to 1 ml 8 . "Silica" column, eluted with 20# benzene. Exchange into hexane. 9. Alumina column. CCl^, 10?$ CH2C12 , 2 5# CH2C1 2' Last fraction saved, cone, under N2 into hexane. Monsanto 1. Add standards (13C-TCDD, 13C-PCDD, 13C-HCDD, 37nCil.-.TCDF, 37C1-PCDF, 37C1-HCDF), 2. Digest in cone. HC1, 1 hour. 3. Extract 2X hexane, dry with Na2S0^. Wash through H 2S0^ on silica with hexane w ash eluate with water, dry, cone. .5. Alumina column; and 5Q? CH2C12 . Sec. fraction to 1 ml. 6 Carbopack C/Celite, prewashed. Sequence of solvents, saving toluene fraction. Concentrate to 10 ul with dodecane keeper. 783299 GENP 010488 / Dow 1. Add standards (13C-TCDD, 13C-TCDF, '13c |h CDD, 1 3 C-0CDD.) 2. Partition hexane soln. against cone. H^SO^, 13-54- hr. 3 Organic phase through silica, NaOH on silica, silica, H 2S0^ " on silica, silicaj washed through with 5f benzene in hexane. Concentrate, dry, redissolve in hexane, 5 . AgMO^ on silica column, hexane eluant, then basic alumina. 80# CCl^, 75# CHgCl^. Evap. under N^, dissolve in CHCl^. 6 . Reverse Phase HPLC, fraction 6 to GC-MSj frJ 1-5 to: 7. Silica HPLC. E.P.A. ' 1. Add internal standards ( ^ C - T C D D , ^ C - Q C D D ) . 2. Evaporate chloroform with N 2 , 3 . Add ethanol: 4-5#KOH aq., 20:4-0. Stir at room temp, for 16 hours. . Partition into hexane using more ethanol. Combine two hexane washes of aq. phase. 5. Wash hexane extract with 2N K O H aq. 6 . Partition hexane phase against cone. H SO^ three times. Wash final hexane phase with water. Dry extract, concentrate to 1-2 ml. (D-K). ! 7. Neutral Alumina column. CCl^ (discard) and CH,C12 (sa v e ). Exchange solvent to hexane. 8. PX-2l/silica column. . Prewash, load, wash w i t`h hexane, CH2C1 2 , and benzene: C H 2C 1 2> 1 :1 . 9. Back elute with toluene. Concentrate just to dryness under N 2 at 60. 783300 GBNp 010489 F . Umea University (1) Add internal standards (^C-TCDD, ^C-TCDF, ^C-OCDD). (2) Dissolve in CH2Cl2/cyclohexane, pass through silica/ K silicate/Na2SOi4/silicate/silica/carbon. *(3) Wash with CH2Cl2/cyclohexane and with CH2C12/CH^0H/ benzene, 75* 20s 5 (*0 Back elute with toluene. (5) Concentrate at 35 (6) Pass through silicate/ H 2S0^ silica with hexane, onto acidic alumina. (?) Elute with hexane, 2$ CH2C12 in hexane, and ltl CH2C12/ cyclohexane. Last fraction saved for analysis. 783301 GENP 010490 0 ' 3. W.Y.D.O.H. (1) (2 ) Add internal standards (13C-TCDD, 37CI-TCDF, F c i - O C D D ) . Dilute with C^Clg partition against ^SO^, neutralize through N a 2C0^ and KOH/silica. Concentrate to 10 ml with boiling water bath. Exchange through cyclohexane into hexane. (3) Semi-automated cleanup. First onto acidic alixmina, wash with Jfo CH^Cl^; elute with 5^ CH^Cl^ orvto Charcoaj/Celi`te' (0 Wash with ICfc benzene in hexane: back elute with 5 xylene in hexane onto neutral alumina. I (5) he x a n e ( 6 ) Clean sample loop with benzenehexane, 1:1; flush rest of system with hexane before running next sample. H. H .P. B . c."o/*i * O) m Vrn nl JV ( '*C - T C O D ( Cl) P orA"i4"i o h rft^oca 4*c i \y lotilO ftjeof Uft,X U/a j Ii wt+lf \T0 x 0 M f W ft.4*tT * A r y 00 C l < r o M a . V e g r * 0>li e ,, F l a r i i / I J AO ft. + ! or |jo aC.4-*<. , 13. W *. %. *S0 , Z To C H t C l i f r o . t l/ O- I 783302 GENP 010 4 Q 7 INTERNAL STANDARDS USED (1) Columbia NFRL 50 ppt each, 13C-TCDF, 37C1-TCDF, 13C-0CDD. (2) Dow 3 ng 1`3-'C--TCDD, 200 pg 13C-TCDF, 1 ng 13C-HCDD, 5 ng 13C-0CDD. (3) Monsanto 1 Unspecified amounts, 13C-T0DD, 13C-PCDD, 13C-HCDD, 37C1-TCDF, 37C1-PCDF, 37C1-HCDP, 37C1-0CDD. (0 U.S.E.P.A.i 5 ng 13C-TCDD, 20 ng J1-3,'C-0CDD. (5) U. Nebraska1 90 pg ^3C-TGDD. (6 ) Ume Unlv. 1 17) C i 0.5 ng each, 13C-TGDD, 13C-TCDF, 13C-OCDD. ..cl,, ,5- i t O b , S 7d - T t l > F ; <SD N - Y . D . O . H : 3 7 7 f j ' 3C ' K D D y * 7Ci - T c . t > F / ) ,, ^ ' c i - d C P O , GLC COLUMNS USED ---------- (_!_)_CoLumbia-NERLi--30Mx0,..25mm,DB-^. (2) Dow1 30Mx0.25mm DB-5, and 60Mx0.25mm SE-5b (for HCDD) (^3)-- ;Monsantoi^-- ^ 6 0 M x ?-- mra_SP-=2-330__( C V ^ - C l ^ i IQMx ? mm DB-*> (OCDD) 783303 (0 U.S.E.P.A.i 60M SP-2330, 30M SP-2340, and 10M 0V-101 WCOT. (5) U, Nebraska1 30Mx0.32mm SP-23^0 and 60Mx0.25mm SP-2330. (6 ) Ume U n i v . SP-2330. 7) C. VHIA<) 01 ! IS ( x <> b B -5 3 0 H DP5 . (%y N . y . d .o .h : o H k o .z s * * s r z i ^ o 7ioio n IONIZATION TECHNIQUE AND MS RESOLUTION USED (1) Columbia NFRLi 55 eV E.I.j (Resoln. not given but quadrupole). (2) " Dow: E.I., resoln. not given. (3) Monsanto: (0 U .S .E .P .A .t E.I.i 1000. E .I .i 8000 (5) .U . Nebraska: 70 eV E.I. ; 7500. (6) Umea Univ.: E.I.? 5000/2000; and MNCI, 1000. (7) Ccjcndii L *. Air C X C l D H - C 0Cl j c o if i<* J0 0 0 0 6?) M .M D .o.H : Ef,r..* Pitjol , o4- q > , 783304 GENP 010493 KEY TO THE DATA TABLES 1= Columbia National Fisheries Research Laboratory 2= Dow Chemical 3= Monsanto 4 = U.S. Environmental Protection Agency 5= University of Nebraska 6= Health Protection Branch, Canada 7= N.Y. State Department of Health 8= Umea University, Sweden 9= Wright State University 783305 G KP 010494 BACKGROUND LEVELS REPORTED . 2.3,7,8-TCDD 2,3.7.8-TCDF 1.2,3.7.8-PCDD 2,3>4,7>B-PCDF 1 .2,3,4 ,7,878000 L,2,3,7,8.9-HCDF )CDD 1 17il8 5.10 17.24 21,26 0,9 <7 1790, 1900 * '2 1 2 ,1 6 2,2 17.23 20,26 4,5 <2 2800, 3700 3 . 14 <2 16 22 2 <3 2830 <fiPA 4 20 2a 29a 38a I4a <19a 2430 A/A* 5 35 . 30 <7 10/35b 3 <8 55 6 12 5 14 15 2 <1 1584 7 18 6 <7 <13 <1.6 -- ___ *First No, = level found 1 second number = corrected for sampl e leakage. Reported as Total TCDF, Total PCDD, Total PCDF, Total HCDD, Total HCDF respectively. *Level calculated two different ways gave two different results. 8 31.31 5.4 5.4o 35.24 5.4 5.18 6450, 7181 lotei All entries b e g i n n i n g with "<M represent _^no n eid etect ed ^ a t_th e-stated--detection--limits.-- 908L I 6V 0 W ^ a 0 fi OBSERVATIONS ON DATA FOR BACKGROUND SAfiiPLS #1 (1) The levels of OCDD were so out of range of the others that it would he meaningless to attempt to evaluate performance on the spiked samples (whose spiking level did not exceed 500 ppt) fr I this compound. However, the agreement among laboratories was very poor for this high level of OCDD in #1. SomeJapproaches obviously result in low recoveries of OCDD, while extremely high levels reported might either involve laboratory contamination or precipitation of this highly insoluble compound out of the standard solution during its storage in those labs. (2) Two labs reported levels of 2,37*8-TCDD in #1 much higher than levels seen in other labs. Only one of the two labs uses reflux saponification, but both use multiple exposure to strong base on columns. Could this be the cause, considering the high background of OCDD, HpCDD and what appears to be 1*.2, 3 167 8- iCDD? If this possibility can not be ruled out, even determination of 2,3.7i8-TCDD, which is at least 9Of* resistant to alkaline dechlorination under these conditions, may be compromised when there is a high background of higher chlorinated species in fat. (3) As will be seen subsequently, levels of several analytes, as reported by various labs, were higher in #1 tian in some of their other samples. Is this simply variability of recovery of other analytes relative to the internal standards, or can the other l| components, "interferences", etc. have a significant effect either on selective recoveries (when the non-CDD/CDF has a higher affinity for some adsorbant than some of the analytes but not others), or on mass spec, response (matrix effect)? Is t n s a problem? 783307 GENP 010496 LEVELS REPORTED FOR -2,3,7,8-TCDD, WITH THE LEVEL FOUND IN ample No. l) f i 12 3 4 5 2 0 oa 0 6* oa 3 35 39 46 44 45 4 5 6 6 15 30 * 5 5 11 ND 19 20 6 3 1 0 0 oa 7 oa 6 4 5 10 a 8 002 00 9 24 58 44 48 _* a 10 oa 0 0 oa oa SUBTRACTED A4, | 67 07 36 42 31 1 27 00 2 12 3 52 40 22 8 Spiked oa 0 45* 50 12 5 17 10 oa 0 55 oa 0 62 50 oa 0 * Sample leaked in shipment, data not reliably adjustable. Data as reported actually less than level reported for sample ff 1. All data rounded to integers. Corrected values below zero reported as zero. 80888Z O b s e r v a t i o n s o n r e p o r t e d 2 ,3 ,7 ,8-t c d d , s a m p l e s 2-10 (1) Labs reporting high levels for samples 5 and 7 , which contained tetrachloro methoxybiphenyls and 1 ,2 ,3 i*'"TCDD, did not report high values for sample 10 , which also contained these possible interference Moreover, these same labs reported high level's for sample **- and, in one case, 8, which did not contain these interferences. One lab had moderately low levels for all the samplesj, suggesting that their value for sample #1 was probably too high for some reason. Yet that lab had the lowest reported level for sample #1 of any of the labs, I It seems more lively that there was a problemj with the 0 internal standard in that lab. (2) The other lab showing at least several low values for 2,378TCDD did so only for the samples containing the interferences. This might reflect an effect of the interferences on recovery of this isomer, since this was also the only lab that did not use a ^ C - T C D D internal standard. 783309 GENP 01n<dOQ LEVELS REPORTED FOR 2,3,7,8-TCDF, WITH THE LEVEL FOUND IN Hi SUBTRACTED Sample No. 1 2 -3 i _5___ 6 . ...Z___ 2 6 5 5 1 0 oc 0 1 2 3 2 * oa 0 2 5 3 0 b 1 2 1 0 15 2 ^ 30 b 27* 5 19 20 I7a 35 30 26 22 6 0.5 0 0 2 0 0 0 7 12* 0 0 3 0 0 0 8 5000 02 0 9 12 10 n 20d a 19 15 10 b 002 oc 0 6 8 Spiked 8.9 15.12a 5.** 0 17.16 29.26 0,2 0.2 0.2 10.9 21.7 0 0 0 12.15 0.1 10.9 0 * Potential interference reported. Sample damaged in shipping. b Reported as Total TCDF. g Value measured was less than in Sample #1. d True level of "total TCDF" spiked, including 2,3^8, was 21.5 ppt. oiee8z 6617010 dM H9 COMMENTS ON VALUES REPORTED FOR 2,3,7.8-TCDF (1) One lab reporting apparent interference identified the problem as due to chlorinated naphthalenes, which could be removed by an additional cleanup step using alumina, (2) All of the values from one of the labs were almost exactly twice the right answers. This suggests either an error in calcu lation or a factor error in the assumed recovery. (3) Several labs had difficulty with sample #**-, which did not receive the interference mixtures. Other labs did not. This is difficult to account for. (*0 There was little background of 2,3>78-TCDF reported by most of the labs. In spite of that, false positives (amounts more than twice the reported background) were occasionally seen for unspiked samples, suggesting that "quantitative resolution", or ability to distinguish between twofold differences becomes very difficult in the 2 ppt range. 783311 G E N F 010500 LEVELS REPORTED FOR 1.2,3,7,8-PCDD, WITH LEVELS in SAMPLE jfl SUBTRACTED Sample No. 1 2 34 5 6 2 1* 0 2 2 0 0 3 31 21a 29 22 22/75b 14- 4 18 13 18 38 0l 5 12 1 oa 0 10/ 20b 2 6 31 20 21 19 07 7 13 15 14 3 0 1 8 1 1 4 0 0 14 9 0 0 0 0 __ a 0 10 8* 4 10 12 02 7 0 0 0 16 16 22 0 46 ro * 8 0,2 56.25 35.12 27,0 61,42 53.36 29,0 30,2 35,24 . Spiked 0 24.8 10.3 0 16,5 1 6 .5 0 0 10.3 -*-- Potential-- Interference'-Reported-.----------------------------------------- a Sample damaged In shipping* jl : Two different methods of calculating results gave two different answers. ZIZZ21 10 5 0 1 0 J M H O LEVELS REPORTED FOR 2.3,4,?, B-PCDF, WITH LEVELS IN SAMPLE //I SUBTRACTED Sample No. 2 3 k 5 6 7 8 9 10 1 8. 56 2k 5 k 32 ( l 3) 33 0 2 _3 00 37* 66 20 23 0 0* 02 30 27 01 18 33 02 k 5 66 3k .0 2 0 0 28 0. 5 0 0 5/15b 10 5/15b 20/55b 0 5/20b 6 0 2k 10 15 0 10 15 15 0 7 0 0 0 26 25 5^ 30 5^ 19 * Sample damaged in shipment a Possible interference reported. b Two methods of calculating results gave two different answers. c True level of "total PCDFM spiked, including 1.2.4,8,9. = 23*8 ppt. 8 Spiked 00 41,32 66,62 239 15.9 22,19 0,0 0 0 50.37 0,0 31.9 0 19.14 15.9 0,0 0 MHO tosio LEVELS OF 1,2,3,4,7,8-HCDD REPORTED, WITH SAMPLE SUBTRACTED Sample Mo. 1 2 3b :2 0,33 28 2b lb 3 0 0* 0 l 4 1.32 30 28 48 5 0.3 0 0* 0 6 1.3 0 0 0 7 66a 53 51 23 8 0,20 26 21 11 9 0000 10 0,36 31 30 2b 5 0 0 0 0 0 0 0 __ * 0 6 9 0 12 0 0 2b 17 0 lb * Sample damaged in shipping. g Interfrence reported. 7 8 Spiked 0 40.42 26.7 0 0,0 0 0 44,56 33.3 13 0 , 0 0 9 0,0 0 65 84,88 53.3 43 45. 33 26.7 19 0 , 0 0 6l 40,43 33-3 neesz 0S0l0dM 3D LEVELS REPORTED FOR 1,2,3,7,8,9-HCDP, WITH LEVELS IN SAMPLE ff\ SUBTRACTED yifS04 ** Sample No, 1 2 3 4 -5 6 ____ 2___ 2 000000 3 45 29a 25 58 0 12 4 000004 5 19 9 <8a 32 <10 0 6 000000 7 000000 8 ^9 49 39 46 50 25 22 0 0 19* a0 10 0 0 0 0 0 0 8 0 46,24 0,0 17.7 0,0 0,0 97.68 0,0 o,.o Spiked 0 33.2 0 11.1 0 0 53.' 0 0 * Potential interference reported. Sample damaged in shipment. , Actual level of "total HCDF" spiked, including 1 2,34,6.9. c 26,8 ** Not measured by this lab, Stee8L po s o io < iN a o COMMENTS ON LEVELS REPORTED FOR 1,2,3,7,8-PCDD (1) The contents of the "general interference solution", which did not include any other PCDD's, did cause positive interference in some of the labs. However, it appears to have caused a strong negative interference in some cases. The effect of variability in the apparent level in sample #1 made the two sets of data from one of the labs inconsistant with each other. Averaging the data would not have improved this problem. (2) "False Positives" occurred quite a few times, and even on occassion in samples that did not receive extraneous interferences. COMMENTS ON LEVELS REPORTED FOR 2,3, V,7,8-PCDF (1) Several labs experienced positive interference in the determin ation of this compound. However, there were four cases of "false negatives" reported also. There are only about 13 entries in the table (out of 36 possible) that are quite close to the "right answers" for the spiked samples. This was apparently one of the more challenging analytes. None of the participating labs lack at least one out-of-range answer, and most have several. COMMENTS ON LEVELS REPORTED FOR 1,2,3,^,7,8-HCDD (1) It must be kept in mind that the true level of this compound in the spiked samples is about 10X that indicated, because of the mislabeling of the standard. (2) This compound was apparently also destroyed by the reflux saponification procedure, and partially destroyed even by room temperature alkaline hydrolysis. Variability in the measured levels in sample #1 produced incompatible pairs of corrected results from one lab. (3) False positives occurred in only one labs false negatives were more common. (0 The "general interference" mixture apparently did not cause any problems with this analyte. 783316 GENP 010505 COMMENTS ON LEVELS OF 1,2,3,7,8,9-KCDF REPORTED (1) Recovery of this compound was low in two labs, suggesting the possibility of alkaline destruction (partial). (2) The contents of the "general interference" solution caused positive interference in two labs. The 1,2,3 i ^ 9 - H C D F also present in sample #9 only prevented correct measurement of 1,2,3*7.8,9 HCDF in one lab, and it was detected as an interference in that one also. (3) One lab had a calibration problem with this compound. (*0 In general, this analyte offered less problems than some, but its determination can experience positive interference from non-CDF*s with some of the techniques. 783317 GENP 010506 SOME GENERAL OBSERVATIONS (l) All of the labs except one had to decide how to quantitate the analytes for which they did not have an isotopic internal standard. Several different approaches were used. Comparing peak areas to those of external standards in separate injections ] could not compensate for partial and variable recoveries. Assuming that recoveries for a given sample would be the same for all analytes as was determined for 2,3f7i8-TCDD was generally not a f I I* valid assumption. When data for recoveries of several isotopic internal standards from the same sample were reported, it was obvious that different compounds showed significantly different recoveries, even in a given sample. Recoveries of C D F 's as a class usually differed, sometimes considerably, from recoveries of C D D 's as a class. It was possible to get very good numbers without having all 7 isotopic internal standards, but it was clear that better numbers resulted from the use of both CDD and CDF internal i standards than from only one class. Replicability' appeared to be somewhat better using El than using Cl, but data to evaluate this was minimal and this question should-be'tested in .some lab that has experience with and capability of doing both. (2) Mass spectral resolution* did not seem to matter much for 2,3i78-TCDD. What out-of-range high values were reported occurred in samples not given the interferences, and there is no obvious reason to assume they were due to insufficient MS resolution. This ti conclusion would not necessarilly apply to all of the analytes, however. The lab reporting use of El at the lowest reported MS resolution apoeared to have some non-CDF interference with the PCDF (3) The issue of "false positives" could not be addressed in this study for most of the analytes, because of the background in the human fat. This issue should be addressed in another study using a synthetic matrix that simulates human fat extract but lacks the background. ^ (*0 Some techniques are clearly incompatible jwith determination of both unsymmetrical TCDX's and higher chlorinated species. 'In particular, strong alkaline saponification and even multiple exposure to strongly basic adsorbents cause losses to varying extents when a human fat matrix is processed. | | |GE1NTP 010507 783318 (5) There was no indication in this study, looking at the limited data v/here replicates were reported, that the mean of two determinations was more likely to be close to the right answer than one or the other of the individual measurements. This may be misleading, if some of the participants reported individual results that were in fact the result of averaging replicate determinations. If that is not the case, then this study would suggest that increasing the cost of analysis by making multiple determinations on a single cleaned-up preparation is not likely to yield significantly improved numbers. (6) There were no sample/analyte combination's in this study for which no lab was able to generate a number very close to the right answer. JSample #3/PCDF came the closest to this of any (only one lab came very close to the right answer), and additional effort to improve reliability for this analyte might be desirable. (7) No lab in this study generated exclusively "right answers" throughout. Nothing worth doing can be done perfectly. (8) There was no one method conspicuously "better" than all other methods. This study would not support the conclusion that there is only one right way to approach this analytical problem. (8) Any future studies that take an approach similar to this one must find a better way to ship samples to avoid what appeared to be unpressurized storage compartment-related leakage problems. Also, "spares" must be available for all of the samples. In particular, the "background" sample should be replicated at least three times in every set (unless it is truely blank). None of these improvements were unrealized at the start of this experiment; they were merely physically impossible. (9) Although the "performance" on OCDD could not be evaluated in the same way as that for the other analytes, it is disturbing that levels reported for sample #1 were so very different between labs when the level was in the ppb range. Have we specialized in ppt level analysis to such an extent that higher levels can. no longer be handled? Perhaps this question also should be addressed in some future study. 783319 GENP 010508