INTERLABORATORY METHOD VALIDATION STUDY FOR DIOXIN
AN INTERIM REPORT
by Robert G. Heath
Human Effects Monitoring Branch
OPPf OTS, EPA
January 5, 1979
-------
Interlaboratory Method Validation Study for Dioxin
A. Introduction and Scope
B. Study Design
C. Results: Tables and Graphs
Standards
. Beef Fat
Human Milk
Types and Frequencies of Errors
D. Statistical Analysis of Lab C Beef Fat Reports.
The regression equation and confidence limits.
Confidence limits for a predicted report value
for a given spiking level.
Spiking and extraction precision vs. quantitation
(GC-MS). precision
Detection limit characteristics
E. Estimation of a "true" TCDD level/ with statistical
confidence limits, from spiking study results (Lab C
Beef Fat example)
F. Discussion
-------
Acknowledgements page to be prepared.
ii
-------
A. Introduction and Scope
i
The Interlaboratory Method Validation Study for Dioxin
was undertaken to measure the accuracy and precision with
which 2,3,7,8-tetrachlorodiobenzo-p-dioxin (TCDD), when
added to beef fat and human milk at low parts-per-trill ion
concentrations, can be extracted and quantified by methods
of gas chromatography-mass spectrometry (GC-MS). Method
validation also included quantitation of equivalent amounts
of TCDD standards. In particular, the study was undertaken
to develop regression statistics for converting reported
TCDD concentrations to "best estimates" of actual (but
unknown) concentrations and for expressing the reliability
of such estimates in terms of statistical confidence limits.
The study was also intended to determine the lowest concen-
tration of TCDD that was identified with practicable consis-
tency and the frequency of "false positive" and "false
negative" reports.
All samples were prepared and extracted at the EPA Pesticide
Monitoring Laboratory, Bay St. Louis, Mississippi. Analyti-
cal laboratories participating in all or part of the GC-MS
quantitation were those of Dow Chemical 'Company, Harvard
\
University, University of Nebraska, Wright State University
and the EPA Health Effects Research Laboratory (HERL),
-I-
-------
Research Triangle Park, North Carolina. Analytical labora-
tories are identified only as laboratory A,B,C,D, or E
throughout the report; alphabetical order is independent
of the above laboratory order. The number of samples, by
type, quantified by each laboratory is shown in Table A-l.
-2-
-------
TABLE A-l.
Number of Samples, by Type, Quantified by Participants
Sample Description
Acid/base cleanup Neutral Extraction Lab
Laboratory
A
B
C
D°
E
Standard
26 .
25
26 + 3
0
1
Beef fat
26
26
26
6
0
Human milk
26
26
0
0
11
Beef fat
0
0
16
16
0
Totals
78
77
71
22
12
Type totals
81
84
63
32
260
-3-
-------
B. Study Design
Beef fat and human milk samples were "spiked" with Cl
TCDD at levels ranging from 0 to 81 ppt. Standards were
prepared so as to contain equivalent amounts of the chemical,
Samples were prepared from one of two pools of rendered
beef fat (pools F and G) and one of two pools of human milk
(pools M and N). Fat was from cattle without potential
exposure to dioxin; the milk had been collected in regions
where use of pesticides potentially contaminated with TCDD
was incidental. Pools were constructed using equal amounts
of fat or milk from each animal or donor, using a separate
set of animals or donors for construction of each pool.
Eleven samples each were prepared from, fat pool F and milk
pool Mthe major pools. The samples from each pool were
spiked individually at 0, 0.5, 1, 4, 9, 16, 25, 36, 49, 64
or 81 ppt. The samples were then extracted by procedures
developed/refined at PML, and the extract was divided
into the required numbers of equal aliquots for shipment
to the analytical laboratories. Spiking levels, excepting
0.5 ppt, were systematically incremented as the squares of
the digits 0 through 9 to provide close spacing at low
levels and a moderate, systematic increase in spacing with
increasing levels.
-4-
-------
To test the precision of the extraction methodology, five
samples each were prepared from the minor pools (fat pool G
and milk pool N). These samples were spiked individually at
0, 9, 25, 49, and 81 ppt, extracted by the same procedures
used for samples from pools F and- M, and divided into
the required number of equal aliquots for shipment to the
analytical laboratories,, Thus, fat pools F and G and milk
pools K and N provide replicate samples at the above levels
of spiking for testing extraction precision.
To test the precision of GC-MS quantitation for comparison
with that of extraction, laboratories were provided two
aliquots of the G- and N-pool extracts, along with two
aliquots from each of the matching extracts from pools F
and N, so as to obtain duplicate analyses of the same
extract. Labs also received four standards at each
equivalent of 0, 9, 25, 49, and 81 ppt, as well as single
standards at 0.5, 1, 4, 16, 36, and 64 ppt. Standards are
denoted as S.
All samplesfat, milk and standardswere required to be
prepared and shipped in random order, and laboratories were
to analyze the samples in the order in which they were
received. Samples were identified only by shipment number,
so that laboratories knew neither the type of sample
nor the TCDD level at the time of analysis.
-5-
-------
A variation in the above procedure was developed for a
set of beef fat samples analyzed at Lab D. The fat
samples in that set were spiked at £ML but were extracted
at Lab D using a neutral extraction procedure rather than
the acid/base procedure utilized throughout the study.
Accuracy (the degree of constant tendency to either under-
report or over-report the true level) and precision (varia-
tion among repeated measurements of the same extract) have
been measured by methods of regression analysis; comparisons
of extraction vs quantitation precision are by analysis of
variance based on those spiking levels for which there were
duplicate analyses of replicate extractions. The analytical
schedule is presented in Table B-l.
-6-
-------
Table B-l.
Design Diagram for Phase II Dioxin Study
TCED
Level Beef Fat
(ppt) Measurements
0
0
1/2
1
4
9
9
16
25
25
36
49
49
64
81
81
Pool
Code
F
G
F
F
F
F
G
F
F
G
F
F
G
F
F
G
Lab A
2
2
1
1
1
2
2
1
2
2
1
2
2
1
2
2
Lab B etc.
2
2
1
1
1
2
2
1
2
2
1
2
2
1
2
2
Human Milk
Measurements
Pool
Code
M
N
M
M
M
M
N
M
M
N
M
M
N
M
M
N
Lab A
2
2
1
1
1
2
2
1
2
2
1
2
2
1
2
2
Lab B etc.
2
2
1
1
1
2
2
1
2
2
1
2
2
1
2
2
Standard
Measurements
Pool
Code
S
S
S
S
S
S
S
S
S
S
S
Lab A
4
1
1
1
4
1
4
1
4
1
4
Lab B etc.
4
1
1
1
4
1
4
1
4
1
4
-7-
-------
C. General Results
Analytical results for the quantitation of standards are
presented in Tables C-l through C-3. Figures C-l through
C-3 (each figure follows its respective table) show the
plotted results and the least squares regression lines and
equations for reported values on spiked values. The theore-
tical line y=x,. for perfect extraction and quantitation is
also shown for comparison.
»
Equivalent results for beef fat samples are presented in
Tables and Figures C-4 through C-9, and those for human milk
appear in Tables and Figures C-10 through C-12.
An explanation of the types of reporting errors and an
enumeration of those errors are presented in Tables C-13
through C-16. In this report, the reporting of a positive
value in an unspiked sample is identifed as a "False Positive1
/
(FP), and a positive report given when the detection limit
exceeds the level of spiking is identified as a "false
positive" (fp). A "false not detected" (fnd) is defined as
a report of "nd" when, in fact, the level of detection is
less "than the level of spiking.
As might have been expected, the highest frequency of errors
occurred at spiking levels below 9 ppt.
-8-
-------
Table C-l.
Dioxin Phase II: Interlaboratory Quantitation Study
Type of Sample: Standard (5g equivalent)
Preparation Lab: PML
Quantitation Lab: Lab A
Sample ID
Study
S-0
^
S~°2
s-o3
s-o4
S-0. 5
S-l
S-4
S-9,
S-92
S~93
S-94
S-16
S-25,
S-252
S-253
S-254
S-36
S-49-L
S-492
S-493
S-494
S-64
S-81-L
S-8
S-813
S-81,,
PML
ST-0
ST-0
ST-0
ST-0
ST-.5
ST-1
ST-4
ST-9
ST-9
ST-9
ST-9
ST-16
ST-25
ST-25
ST-25
ST-25
ST-36
ST-49
ST-49
ST-49
ST-49
ST-64
ST-81
ST-81
ST-81
ST-81
Ship-
ment
2
40
45
49
- 15
19
10
21
28
46
47
22
16
42
48
52
5
13
32
44
50
17
12
29
43
51
Recov.
37
ClJ/(%)
88
101
83
94
95 -
141
77
68
93
113
69
86
93
90
64
131
81
67
109
98
98
107
46
98
102
106
TCDD Levels (ppt) Detection
Added
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
Reported
320 322 Avg. 320
nd
\
nd
2
nd
1
nd
2
7
4
7
10
6
nd
19
6
18
16
23
26
18 . '
68
nd
44
40
44
56
Limit
322
0.3
6
2
2
1
2
2
2
2
7
3
2
0.7
3
2
2
4
2
2
1
4
0.5
2
2
2
3
-9-
-------
Figure C-l
Standards
Lab A .(322 m/e)
120
110
100
90
BI
c.
& 6r
Ig
p.
CJ
1-
c 50
E-:
D
J^
40
30
20
10
0
*
'
/
r "*
BEE IOUO
w
4*
> -**
_^^
-^~
*^ * ,
- '
.
.
^
^
<^
.:,;
. I
.
"" " ''"V"
J
s
\s
^^
1
.
. - '
*
r
' jf
y
^
f
-^~ ' .A
r
,
.
^
jf
*
yj
. *"
y- '
i
f
^^^
_fl
4 *S
.H/ jf
t\* J
y
.,^
^^^
_^^
1
f
- '
jf
I S
f
y
! ^
y
'
»
v
'
JP
/"
i ' '
.
*
-ll ^-
JT f^ -. , , r
rfi- ^ »
» *jr^ I
>^
I
f
--
10 20 30 40 50 60 70 80
x = TCDB added (ppt) -10- '
-------
Table C-2.
Dioxin Phase II: Interlaboratory Quant i tat ion Study
Type of-Sample: Standard (5g equivalent)
Preparation Lab: PML
Quantitation Lab: Lab B
Sample ID
Study
S"°l
s-o2
s-o3
s-o4
S-0.5
S-l
S-4
S-9,
S-92
S-93
S~94
S-16
S-^
S-252
S-253
S-254
S-36
S-49J
S-492
S-493
S-494
S-64
S-8^
S-8
S-813
S-81,
PML
ST-0 '
ST-0
ST-0
ST-0
ST-.5
ST-1
ST-4
ST-9
ST-9
ST-9
ST-9
ST-16
ST-25
ST-25
ST-25
ST-25
ST-36
ST-49
ST-49
ST-49
ST-49
ST-64
ST-81
ST-81
ST-81
ST-81
Recov. TCDD Levels (ppt)
Ship-
ment
2
40
45
49
15
19
10 .
.21
28
46
47
22
16
42
48
52
5
13
32
44
50
17
22
29
43
51
,7 Reported
Cl (%) Added
0
0
0
0
0.5
1
4
9
9
9
9
16.
25
25
25
25
36
49
49
49
49
64
81
81
81
81
320
nd
nd
3
nd
nd
nd
-
4
9
4
10
10
16
17
7
29
6
13
52
40
8
28
59
21
77
83
322
nd
nd
nd
2
nd
nd
-
5
6
6
7
10
22
13
19
21
12
.22
.29
18
25
33.
50
23
47
43
Avg.
nd
nd
-
-
nd
nd
-
4.5
7.5
5
8.5
10
19
15
13
25
9
17.5
40.5
29
16.5
30.5
54.5
22
62
63
Detection
Limit
320
4
2
2
3
3
4
-
3
4
2
3
2
4
2
1
3
4
5
5
1
2
3
5
1
6
3
322
3
3
2
1
4
5
-
2
1
2
1
1
4
1
2
1
5
3
1
1
1
2
2
1
2
2
-11-
-------
Figure C-2a
Standards
Lab B (322 m/e)
120
1 T /S
110
100
- on
80
-^^
.
C-.
fc in
^ 70
e
o>
u
o
u
fiO
o ./
*^
A
CJ
H
50
40
n A
30
20
10
<
0
; :
«
/
y**?
T^
L>
BEE loiio-
f
.-.y. ...
f
* -4-"^
^^
.0 2
.
^
x^
_r^*
,f
V
1"^
^x**^
0 3
.
X
V
»
*
~^~
0 ^
TCDD
,
*^
r
jf
.
w
_^^^
--^^^ »
*0 5
added (pp
vyr
'
; - -
t
-
*^
f
-1
J
.^
_^^
^^"^
'
0 6
t)
«. W
y V"
^V X^
xv~ ^ .
s
^ . .
_J
0
t "^ _d
_r^^'^
^"^
0 7
x
x
x
v y '
.
f
7 _*-^pl^
^^^
^^^
___^
0 8
-12
.
^
^
*
s
1
S
s
f
s
-
^ *
^g^
^^^
\
1
t
1
"
- - -
"
.
0
!-
-------
Figure C-2b
Standards
Lab B (320 m/e)
120f
110!
100
90
?r
ftC
W
5C
4C
^ f\
3C
?n
£\J
1 /^
10
.
r\
\)
*
$
*£
r ^ ^
BEE 10X10
_
. .
.
-
*"
^
LO :
.
.
,
§... y
, ,
_s
jf ' '
s
^
to^*
Q
V
JO 3
. . .
- . '
jf
j
j
/"
^^
^~ .
^
A
0 4
TCDD
.
O
J
.f
Jr
J
-*
V
^^
^
^f
s^
rt
V
A
0 5'
added (pi
^
4
J
^
'
^
.r
,S^
^^
3 6C
»t)
r
-CiT >
^
^
^
jS
^^
^*
^^
0,
'
) 7
|
s
4
f
J
f
.?
,
_± *
- jt
\j* J^
^^
^T
'
'
_ .
0 8
^
'. -
i
fr
*
f
^ ^ *
- .
A
V
jfV _/^
'"V jf^
J^
i -X^
X^ , ,
-------
Table C-3.
Dioxin Phase II: Interlaboratory Quantitation Study
Type of Saraple: Standard (5g equivalent)
Preparation Lab; PML
Quantitation Lab: Lab C
Sample 3D
Study
s-o1
S-°2
s-o3
^
S-0.5
S-l
S-4
S-9,
S-92
S-93
S~94
S-16
S-25l
S-252
S-253
S-254
S-36
S^j^
S-492
S-493
S-494
S-64
S-81-L
S-8
S-813
S-81A
EML
ST-0
ST-0
ST-0
ST-0
ST-.5
ST-1
ST-4
ST-9
ST-9
ST-9
ST-9
ST-16
ST-25
ST-25
ST-25
ST-25
ST-36
ST-49
ST-49
ST-49
ST-49
ST-64
ST-81
ST-81
ST-81
ST-81
Recov. TCDD Levels (ppt)
Ship- -_
ment Cl
2
48
57
62
15
19
10
21
56
60
64 .
22
16
55
63
69
5
13
51
52
70
17
12
49
59
67
(%) added
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
Reported
320 322 Avg.
nd
nd
nd
nd
nd
nd
nd
nd
11
9
6
16
21
17
25
24
34
41
51
47
47
65
76
77
80
80
Detection
Limit
320 322
10
4
4
4
6
12
6
10
5
4
4
10
4
4
4
4
16
12
3
9
5
6
6
3
3
5
-14-
-------
120
110
100
90.
80
70
^ 60
50
40
30
20
10
Figure C-3 - Standards . j
Lab C (322 ra/e) ^/jg/t^U-'
i .
^W-V'i^v^M^M^H'HB-^M^VB
.
f
* X
x^
^
H^Bl^l*k*««B^^^^^^««MW
4
* X
X" X
j^ X
X
:
II ! . "I
"--' .
.
X
" X
X
1^ -^
X* S j
, x"
^VWM^B^MIHII-IVH^^M
!
. . .
*
A
X" X^
x -x1^
^ x^
x x^
1
9
^
«'
* s
' ' - /
t .x ^
x^
1 1 -!! l.^^^-^_
^^^^^M^^-MMAf^^M^MI^^H
1
:
A
f.* y
^» .X
w
rf' ^ ^y
~r*s s /^
V r* S&
S
.
.^ _r
-^
"^/ X
X
x X"
X
Xt
X^ '
_-
^^^^^^^^ II 1 1
^*v-4*vp^«^A4M*-*nw^M n»
.
\
' 1
,
X
>
1 rf* X^
^/^
_ j T^^
rff jjT
jr ^r o
. ..___X-...
+ * /-
X
X
r -
;
I
1
I
. _ .
-'
.
. _ ,. .
i-
- -
.
- - -
BE joiic 10 20 30 40 50 60 70 80
TCDD added (ppt) -15- . /
-------
Table C-4.
Dioxin Phase II: Interlaboratory Quantitation Study
lype of Sample: Beef Fat (5g sample)
Extraction Lab: JML; Method: Acid/base
Quantitation Lab: Lab A
Sample ID
Study
F-°l
F-02
GH^
G-O,
F-0. 5
F-l
F-4
F-9,
F-92
G-91
G-92
F-16
F-25,
F-252
G-251
G-252
F-36
F-49-L
F-492
G-491
G-492
F-64
F-8^
F-812
G-811
G-812
PML
FE
FE
GB
GB
FG
FC
FI
FK
FK
GC
GC
FL
FD
FD
GE
GE
FA
FH
FH
GA
GA
FB
FJ
FJ
GD
GD
Ship-
ment
11
37
24
39
1
7
6
25
41
20
30
26
3
34
18
38
8
27
33
14
35
9
23
31
4
36
Recov.
ICDD Levels (ppt) Detection
37 Reported Limit
Cl {%)
71
72
73
53
69
72
70
64
75
93
68
72
71
77
73
68
80
73
99
87
68
101
53
89
69
71
Added
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
320 322 Avg. 320
103
nd
nd
2
12
3
28
4
9
2
nd
8
no
22
5
8
34
18
7
6
32.
86
25
22
110
40
322
7
5
4
2
0.4
2
5
2
1
1
1
4
1
6
5
1
3
2
2
0.8
5
3
2
4
1
2
-16-
-------
C-A
Beef Fat Samples
Lab A .(322 m/e)
120
110
100
__j
« 90
80
^7Z{-
~X- 1-
o
0)
4J
o 60
cu
g
50
AO
20
10
II
7(1
80
-------
Table C-5.
Dioxin Phase II: Interlaboratory Quantitation Study
Type of Sample: Beef Fat (5g sample)
Extraction Lab; PML; Method: Acid/base
Quantitation Lab; Lab B (T\ st '
Sample 3D Recov.
Ship- 3,
Study PML -ment Cl (%)
F-0. FE 11
F-02 FE 37
GrO, GB 24
G-02 GB 39
F-0.5 FG 1
F-l FC 7
F-4 FI 6
FK 25
F-92 FK 41
G-^ GC 20
G-92 GC 30
F-16 FL 26
FD 3
F-252 FD 34
G-251 GE 18
G-252 GE 38
F-36 FA 8
F-491 FH 27
F-492 FH 33
G-49, GA 14
G-492 GA 35
F-64 FB 9
F-8^ FJ 23
F-812 FJ 31
G-81, GD 4
G-812 GD 36
TCDD Levels (ppt)
Reported
Added 320 322 Avg.
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
4
13
nd
9
7
nd
7
7
6
8
7
13
12
12
18
12
27
21
35
38
15
54
42
27
60
39
7
6
10
5
12
nd
20
13
13
21
11
18
10
17
27
.8
29
22
26
43
28
25
34
32
64
32
5.5
9.5
-
7
9.5
nd
13.5
10
9.5
14.5
9
15.5
11
14.5
22.5
10
28
21.5
30.5
40.5
21.5
39.5
38
29.5
62
35.5
Detection
Limit
320 322
4 -
2 .
4
2
4
4
5
5
2
7
2
7
4
2
6
3
5
3
3
7
2
8
5
2
9
2
2
3
4
2
3
4
5
3
2
5
2
.2
.2
3
6
2
7
2
2
8
2
2
1
3
5
2
-18-
-------
Figure C-5a
Beef Far Samples
Lab B (322 m/e)
120
*
iin
J.J.U
inn
> 90
Q f\
O* '
^*S
4-J
^i_ / ^^
»^^ -
«**
QJ
fin
ou
VH
O
r *j
50
/ rt
40
an
JU
20
10
rk
i . f
.
.
6*^
^j "
. -
-
X" "
*
- ,
.
X"
J-**p^^ - -
-
.
y*
^
.
W
. -X"
*"
^^^
*?"
r X^
">
.^^^
-
*
^
r
f
*
*
-
_»* "".-rf
-------
Figure C-5b
120
110
100
90
Beef Fat Samples
Lab B (320 m/e)
80
^», TI ^ i
CL
O.
70
^
o
Q)
I
Q
Q
60
50
40
30
20
10
lil. .^f^..
I -
BEX 10H
oLO
20
30
40
50
60
70
80
-------
Table C-6.
Dioxin Phase II: Interlaboratory Quantitation Study
Type of Sample: Beef Fat (5g sample)
Extraction Lab: EML,° Method: Acid/base
Quantitation Lab: Lab C kl/L
Sample ID Recov. TCDD Levels (ppt) Detection
Ship- Reported Limit
Study PML ment Gl (%) Added 320 322 Avg. 320 322
F^
F-O,
G-0,
J.
G-02
F-O. 5
F-l
F-4
F-9,
F-92
G-9,
G-92
F-16
F-25,
F-252
G-251
G-252
.F-36
F-491
F-492
G-49:
G-492
F-64
F-81-L
F-812
G-811
G-812
FE
FE
GB
GB
FG
FC
FI
FK
FK
GC
GC
FL
FD
FD
GE
GE
FA
FH
FH
GA
GA
FB
FJ
FJ
GD
GD
11
54
24
65
1
7
6
25
68
20 .
50
26
3
47
18
71
8
27
58
14
66
9
23
61
4
53
0
0
0
0
0.5
1
4
9
9
9
9
. 16
25
25
25
25
36
49
49
49
49
6*4
81
81
81
81
nd
nd
nd
nd
.nd
nd
nd
9
11
nd
12
19
(63)*
24
26
27
31
50
48
39
48
58
76
74
76
77
8
5
6
6
14
10
18
4
5
14
5
10
10
2
12
4
6
6
4
10
6
4
6
4
14
3
*Aberrant value both included and e "'xSed in calculations.
-21-
-------
o
a.
o)
M
O
HQi
IOC?
90-
HO-j
70 -
50 ~
40-
30-
9n
zu
10 -
o"-
1 x:
Figure C-
.
- .
... . ~
£
^
7*-*
X^ X*
V^
_r _i^
-V^ "^
) *^
V
1
6a
uf-p.
. : .
.-.__
__
1 ' :
ei _X^
j£
Ss
j?
r
jf f
^
W '
' ! i . -4
'
'
i i
'
-
. ; i .
:
e
&
*-**
9/f
j^n
jr»
/f
tf ' i .
r^ .
h)
Lab C
.......
. ' i I
t
i
.-
; : -
- . 1 1 .
! -
J J
jr
/jf
f~jr '
-^ '-9
P^ '
50 U
(322 m/e)
r ..- 1 7^
[n =. I/;
= -
f * '
**
J x
X1 jX"
' X
j^
^yc
^ ^^
u >
TCDD adde
t,L
|J-»r
/
i*
f
f jf
* _r
-T* J^"
^ X"
M X"
jX"
/=-
o 6
d (ppt)
v'
jA H*
__: ^
y£ .,<
JX
-3^«
f
-
o n
.
-
f
s
? ^ y
,x 'V-r
"
-------
Figure C-6b
v&t- >x-> got,
Lab C (322 m/e)
(n - 18)
OU01 33E
12 Or
Lift
lOSJ
90
i
80-
0.
C-
^w
o
2 701
w
o ~
CL
0)
K
Q I
O
5 60-
m
cn_
401
.
"^ni
Of>_
i ft-
/
0-
; :
***VM^^^^^V^^*V*M^H«M
,
- - »
....... I
; . .
- . .
jf
.f
s \
JT J
f
*
^
f
i
' . : . i
i-= ' ' '
.;.,..
' '-
r-. . .
HiV^4^MI^^Mfl^H««^^>Vl^^B^
. . (
.
: . .
..-.-.
. . -
. .
jf
- jT
jT
S
.f a
S P /
^ **-H
'
r-y
te 2
i ' . i j '. '
. !-: r
vv#^hrt«l^^H^lM«i^VMi^HABA
' 1
| _ .-
_ _._
.
: i . - i
...
. i
i . . . i 1
1
-
' -^
jf
j^ ^
jS S
''-f* ' _f
Jr\ +
s */
s ' _^»
- jf
s ....
6 5
~ '-'-
'". . i . . -
..'.
: ;
; __.
1 ....
' 1 . 1 . 1 ; ;
. : 1 . ' '
'
' .
-^
JT
jjr f
J^j* '
j^ f
jr v»M^H**a^M^^Hi^^M^MP-
*
*
J
f
f
f _4
-e. ,/L
jT
*-S
*r
o
.
TCDD added (ppt)
-23-
-------
Table C-7.
Dicocin Phase II: Interlaboratory Quant itat ion Study
Type of Sample: Beef Fat (2.5g sample)
Extraction Lab: Lab D; Method: Neutral extraction
Quantitation Lab: Lab C
Sample ID Recov. TCDD Levels (ppt) Detection
Ship- ~7 Reported Limit
Study FML ment Cl (%) Mded 320 322 Avg. 320 322
0 nd 6
0 nd 6 '
0.5 nd 6
1 nd 6
4 17 6
9 nd 8
9 10 5
16 12 6
25 24 5
25 25 10
36 31 6
49 45 9
49 70 .5
64 52 5
81 76. . 8
S-7 35 81 70 . 3
-24-
F'-°l
G'-O,
F'-0.5
F'-l
F'-4
G--9,
F'-16
F'-251
G'-251
F'-36
G'-491
F'-64
F'"811
S-12
S-l
S-4
S-9
S-2
S-10
S-13
S-3
S-6
S-ll
S-16
S-15
S-8
S-14
S-5
41
28
31
37
29
38
42
30
34
40
45
44
36
43
33
-------
120>
110
100
90
80
c.
c-
o-
0)
4J
J-l
.o
C.
0)-
i
70
60
50
40
30
20
10
Figure C-
. .
i -
-
r* ' '
-^JZ.
7
i
x-
**T7/.._
S *
. .^i -
Mr 1o»cLU *
^XjX
^
r V
/
^^^^M^^V^^HMMPM^h^HM
Be<
Lai
Neutral
. t
' -
. .
. . . , .
. :
fir
(S
.^
^ w
^ .^^^__^_
0 30 7»(
»f Fat Sai
D C (322 i
Extractic
t .
,
.
i .
ss
S.S
1 .1.1 !
cples
n/e)
an, 2.5 gn
. . i -
. i * i
-
: . - 1 .
A
*
. ^ ^r
f 'j^
* . ^>
:f
:
1 »H^ -
i P
^ ^,
^_
*
^
' ' '* _r
r 1 r^
^S^
jr
^^M«V^^H^^>IM«^h^lK^^^^^
50 bU 7
TCDD added (ppt)
_.. |
i
~\
1
I
' .
.
,
f
f
' f
\ '
' * J
^
» s
/TV i* i _r
r « r^1^
SvT' ^ _. S^
a** r ><
-<_ « * ' _X^
O - J^
iV * - jS\
* _T '
f 4"*. V^ i M.
f- X- i
I
. .
. ___ _
J 80
-25- /
-------
Table C-8.
Dioxin Phase II: Interlaboratory Quantitation Study
Type of Sample: Beef Fat
Extraction Lab: Lab D; Method: Neutral extraction
Quantitation Lab: Lab D .1 /^/^)
Unfi/*^^
Sample ID
Study PML
F'-OI
G'-OI
F'-0.5
F'-l
F'-4
F'^
G--9,
F'-16 .
F'-251
G'-251
F'-36
F'-^
G'-49l
F'-64
F'-Sl,
1
S-12
S-l
S-4
S-9
S-2
S-10
S-13
S-3
S-6
S-ll
S-16
S-15
S-8
S-14
S-5
S-7
Recov.
Ship- -_
ment Cl (%)
41
28 53
31-1
31-2
37-1
37-2
29 57
38
42-1 81
42-2
30
34 38
40
45
44
36
43
33
35
TCDD Levels (ppt)
Reported
Added 320 322 Avg.
0
0
0.5
1
4
9
9
9
16
25
25
36
49
49
64
81
81
nd
nd
nd
2.5
nd
0.9
42
19
nd
11
25
29
nd
32
62
49
50
89
nd
23
nd
2.3
nd
45
17
nd
8.5
19
26
25
41
73
49
54
155
86
nd
(23)
2.4
nd
44
18
11
22
28
(25) .
37
68
49
52
(155)
88
Detection
. Limit
320 322
23
35
16
0.8
13
0.4
16
15
16
9.2
12
62
8.5
22
25
34
-
24
7.7
6.9
9.2
0.5
2.2
15
15
46
15
12
10
25
33
10
8.5
70
15
-26-
-------
120
110
Figure C-8
Beef Fat SaniDies
Lab D (Avg. of*320, 322 m/e)
Neutral Extraction
100
90
80
1-
70
60
50
40
30
20
10
0
BEE JOHO
10
20
30
TCDD addei
ed (ppt)
60
70
80
-27- . /:
-------
Table C-9.
Dioxin Phase II: Inter laboratory Quant i tat ion Study
Type of Sample: Beef Fat
Extraction Lab: IML; Method: Acid/base
Quantitation Lab: Lab D
Sample ID
Study PML
F-On
1
F 0
2
G-0,
1
G-0,
2
F-t).5
F-l
F-4
F-9,
1
^92
G-B.
1
G-9,
2
F-16
F-25.
1
F-25,
2
G-25
1
G-25,
2
F-36
F-49,
1
F-49,
2
G-49-j^
G49
2
F-64
F-81n
1
F-812
G-81,
1
G-81,
FE
FE
GB
GB
FG
PC
FI
FK
FK
GC
GC
FL
FD
FD
GE
GE
FA
FH
FH
GA
GA
FB
FJ
FJ
GD
GD
Recov.
Ship- 3?
ment Cl (%)
SO-5
SO-4
SO-1
83
SO-3
SO-6
SO-2
TCDD
Added
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
Levels .(ppt) Detection
Reported Limit
320 322 Avg. 320 322
10
(nd)
13
(8.3)
29
35
32
(34)
71
(83)
114
(145)
107
(100)
nd
(nd)
nd
nd
nd
nd
nd,nd
nd,36
16
(14)
44
(37)
.
68
(73)
143
(154)
nd 3 66
(11) (15)
6
6
14
10
nd 4 6.1 (3.1)
8.3 5.0 (7.0)
24 10 4.4
9.-B (8)
37 3.0 5.0
(8.0) (7.0)
12
4
6
74 10 10
(14) (13)
6
127 (17)
10
(17)
-28-
-------
' Figure C-9
120
Beef Fat Samples
Lab D
Acid/Base Cleanup (2.5
iio:
_
*
.100
90
8CF
7C
~jr.\
50
40
30
20
^\~~
H
10 £
f V /
O 322
Avg. j^w/
Regression based on averages
BEE
TP"
20
30
40 50 60
TCDD added (ppt)
70
80
-29-
-------
Table C-10.
Dioxin Phase II: Interlaboratory Quantitation Study
Type of Sample: Human Milk (10 g sample)
Extraction Lab: EML; Method: Acid/base
Quantitation Lab: Lab A / 1 c
Sample ID
Study
M-0,
JL
M-O,
&
N-0,
^
N~°2
M-0. 5
»
M-l
M-4
M-91
H-92
N-9!
^t
N-92
M-16
M-25T
M-252
N-251
N-252
M-36
M-49
M-49
N-49
N-49
M-64
M-81
M-81
N-81
N-81
FML
AB
AB
BB
BB
AD
AE
AF
AC
AC '
EC
BC
AG
AH
AH
BD
BD
AI
AA
AA
BA
BA
AJ
AK
AK
BE
BE
Ship-
ment
71
77
58
60
54
59
63
53
55
61
64
56
70
76
66
74
69
72
78
57
62
68
67
75
73
79
Recov.
37
cr'(%)
103
92
80
103
99
91
92
93
89
69
107
79
99
90
81
104
96
110
119
100
98
85
89
104
136
121
TCDD
Added
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
Levels (ppt)
Beported
320 322 Avg.
1
4
nd
2
. nd
nd
1
3
5
5
"4"
5
15
11
15
7
28
29
29'
36
39
36
31
21
114
24
Detection
Limit
320 322
1
2
2
2
1
1
1
1
1
4
........ -j.
1
2
2
1
1
1
1
3
1
' 1
2
1
2
2
1
-30-
-------
Figure C-10
120
110
100
Hunan Milk Samples
Lab A (322 m/e)
90
80
B.
C.
o 70
o
e.
0)
I
g
60
50
40
30
20
10
-9£^._
-H
._ 5-
BCC 10110-
TCDD added (ppt)
-31-. /
-------
Table C-ll.
Dioxin Phase II: Interlaboratory Quantitation Study
Type of Sample: Human Milk (10 g sample)
Extraction Lab: PML; Method: Acid/base
Quantitation Lab: Lab B C\ s
Sample ID
Study
M-01
M-°2
N-OI
N~°2
M-»0.5
M-l
M-4
M-9,
M-92
N-9,
N-92
M-16
M=25X
M-252
N-25,
N-252
M-36
M-49
M-49
N-49
N-49 .
M-64
M-81
M-81
N-81
N-81
PML
AB
AB
BB
BB
AD
AE
AF
AC
AC
BC
BC
AG
AH
AH
BD
BD
AI
AA
AA
BA
BA
AJ
AK
AK
BE
BE
Ship-
ment
71
77
58
60
54
59
63
53 '
55
61
64
56 .
70
76
66
74
69
72
78
57
62
68
67
75
73
79
Recov. TCDD Levels (ppt)
Detection
-- Reported
ClJ/(%) Added -
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
320
-
-
2
2
-
-
nd
3
3
2
7
7
24
29
-
23
45
49
17
-
-
-
-
-
-
322
10
8
17
15
3
7
4
7
7
10
9
9
26
35
34
32
38
69
53
21
10
82
no
97
no
96
Avg.
(10)
(8)
9.5
8.5
(3)
(7)
(4)
.5
5
6
(9)
8
16.5
29.5
31.5 '
(32)
30.5
57
51
19
(10)
(82)
(110)
(97)
(110)
(96)
320
-
-
2
2
-
-
2
2
2
2
2
1
4
1
-
1
3
4
1
-
-
-
-
-
-
Limit
322
2
2
3
3
1
3
1
2
2
2
2
2
2
2
2
2
2
2
2
2
1
3
3
2
2
2
-32-
-------
Figure C-lla
Human Milk Samples
Lab B (322 m/3)
1
:
»
-
. '
. : : .
: "
.
__: 1
,
&-,
6
O ' 7*-Q
_ x )c<
s\- / f-f
y - i
_/; ''£j
-J .
f
BEE 10HD-
' :
! - . . -
,
OX
X x
X ^
/ s
7 :
-O
'
.0 2
.
. -
-^v -j
V* f
, x
_r ^
X y
V' x^\
>^
0 3
. . . -
- -
. . . '
. .
X
jr
jr "^ ' '
s
.
, . [
0 4
.
. -
.r
X
/
J O
Jr O
Jr
Jf
.f
J>-
s-^
0 5(
TCDD adc
: . .
....
:
' " ' X^
jr
'_r
, jT
J'
^
J
jf ...
' /"
'
^
f
3 ^
led (?Pt)
1
.
f^»
f** **. s
*^w s
x
_r
.4 X i
- - -^- x -
y:
jf
X
x
x
X
(**
j/-
* .
/" : -
lO 7
x
in. _r
X^
**\ * x^
X
JC X^
J'
* A
if
'
X
X
X
r
' -
.
0 8
' +1
jf
s
O/
/
/ : . .
s .
f
.X
f
f
f
/
-
- -
'
- -
,
.
-
,
0
-33- /.
120
110
100
90
80
iJ
c.
e-
o '
o.
60
50
40
30
20
-------
]
120-
110
100
90
J
£ 80
a.
^s
c
a;
JJ
u
I 70
0)
w
o
a
£
P 60
50
AO
30
20
10
j
(
0
Figure C-]
- -
i
ji
*
T jf
X"
" _A
O
.
-
A '
*J X
+ "^f ' .
X jr
r -X
S
f
.»
o
Hu
La
ji
y
_
*
^^
- ^x^
^ JL
man Milk
b B (320
j
X
^Sr
x^
^x
^
*f
^^
,f
^
Samples
m/e) ^
. .
. ,-. . .
X
V'
f
X
f
' .X^
X"
^X-
_y
X
I/
X
X
x1
^
J^
.r
x^
X
X
jX^
^x
' -
.X
r
^
i
1 i
I
'
t
;
i
1 ""
BEZ 1011010 20 30 N 40 50 60 70 80 ,:
TCDD added (ppt) -:>«- /
-------
Table C-12.
Dioxin Phase II: Interlaboratory Quantitation Study
Type of Sample:
Extraction Lab:
Quantitation Lab:
Human Milk + 1 standard
FML; Method: Acid/base
Lab E
Sample ID
Study ML
Recov.
Ship- ,7
ment Cr'(%)
TCDD Levels (ppt)
Reported
Added 320 322 Avg.
Detection
Limit Avg.
320 322
M-OI
M-02
M-4
-
M-l
M-4
M-9X
M-92
M-16
M-25
S-l
AB
AB
A2
AD
AI
AE
AF
AC
AC
AG
AH
STD-1
HMT-3
-10
-6
-4
-9
-11
-2
-1
-7
-5
-8
-12
100+
66
95
-64
100
100
77
100+
100+
100+
100+
100
0
0
0.2
0.5
0.8
1
4
9
.9
16
25
1
1.5
0.6
9
0
0.9
1.4
6
14
5.5 .
29
34
1.9
0.3
0.1
1
0.2
0.3
0.4
0.6
3
1
2.1
2
0.2
-35-
-------
OJ
a.
c.
a
. yy
jf .
&
.2 .
J
/
f
f
/
V f-
.X
x
i
Lab
j
/
_^_ _
f
»
*
*
Human M
E (Avg.
<
/
f
f
t
*-'
*
.
V aaioziaLO 20 30 AC
«
ilk Sampl(
of 320, 3:
=s
12 m/e) >
,
x
l X
,
- X
x^
x^
. / .
X
,/
f
jf
f
' '
#
f
'
X
/
f
f
/
X
r
f .
X
y
tff
rrfi
- r>» y/
"v /
'' X
* f
jf
.f
f
'
./
'X
*
' -
50 bU /
r
_i£_
9
EsJ
y^
|
'
.,
,f
d: .f
/
f '
f
f !
^ ;
*
X"
f
t
- ,
t
1
) :
__^__^_
"
. .
. ... _
..
U BO -
-1C. /
-------
TABLE C-13.
Types of "Valid" and "Erroneous" Values
Reported for Spiked Sarnples
Laboratory
Report
Spiking Category
Sample
Itot Spiked
Sample Spiked
S < DL I/
S >.DL 2/
not detected (nd)
positive value
valid valid false nd (fnd)
False Positive (FP) false positive (fp) valid
V Spiking level less than the detection limit.
2/ Spiking level greater than or equal to the detection limit.
-37-
-------
TABLE C-14.
Incidence of Reporting Errors' for Standards
Number of Measurements (n) and Errors, by Types
Sample Sample Spiked
Lab m/e
A 322
B 320
B 322
C 322
D
Avg:
E 320, 322
Totals:
Not Spiked
(n) FP
(4) Tl\
(4) f If
(4) WJ/
(4) 0
(0)
(0)
(16) 3
Spike <
(n)
(3)
(2)
(2)
(3)
(0)
(1)
(11)
fp
0
0
0
0
-
0
0
9 ppt
fnd.
0
0
0
0
0
0
Spike >
(n)
(19)
(19)
(19)
(19)
(0)
(0)
(76)
fP
0
0
0
0
-
-
0
9 ppt
fnd
&
0
0
0
-
-
2
I/ See Table C-13 for error definitions
-38-
-------
TABLE C-15.
Incidence of Reporting Errors for Beef Fat Samples'
Number of Measurements (n) and Errors, by Type
Sample
Sample Spiked
Not Spiked
Lab m/e
A 322
B 320
322
C 3222/
C (NE)-' 322
D (NE) 320
D (NE) 322
D 320
D 322
E
Totals:
Acid/base
NE
(n)
(4)
(4)
(4)
(4)
(2)
(2)
(2)
(2)
(2)
(0)
(20)
(6)
FP
2
3
4
0
0
1
1
1
0
10
1
Spike < 9
(n)
(3)
(3)
(3)
(3)
(3)
(8)
(7)
(2)
(4)
(0)
(18)
(18)
fP
2
2
2
0
1
2
3
1
0
7
6
PPt
fnd
0
0
0
0
0
0
0
0
0
0
0
Spike > 9
(n)
(19)
(19)
(19)
(19)
(11)
(14)
(13)
(10)
(8)
(0)
-
(94)
(38)
fP
0
0
0
0
0
1
1
2
0
-
2
2
ppt
fnd
&
0
0
^
<&
0
0
0
0
-
1
1
I/ See Table C-13. for error definitions
2/ NE denotes neutral extraction; otherwise/ acid/base
cleanup utilized
-39-
-------
TABLE C-16.
Incidence of Reporting Errors' for Human Milk Samples/
Number of Measurements (n) and Errors, by Types
Sample Sample Spiked
Not Spiked
Lab m/e
A 322
B 320
322
C
D
Avg:
E 320, 322
Totals:
(n)
(4)
(2)
(4)
(0)
(0)
(2)
(12)
FP
3
2 "
4
-
-
2
11
Spike <
(n)
(3)
(1)
(3)
(0)
(0)
(5)
(12)
fp
0
0
2
-
-
0
2
9 ppt
fnd
1
1
-
1
3
Spike > 9 ppt
(n)
(19)
(11)
(19)
(0)
(0)
(4)
. (53)
fP
0.
0
0
-
0
0
fnd
Ax
if
0
-
0
0
I/ See Table C-13 for error definitions
2/ All extractions utilized acid/base cleanup
-40-
-------
D. Statistical Analysis of Laboratory C Beef Pat and
Standard Reports
For practicality, a detailed statistical analysis of analyti-
cal results is presented only for Laboratory C in order to
determine the optimum known accuracy and precision that can
currently be achieved in quantifying low ppt levels of TCDD
in samples of the types analysed. (Complete statistical
analyses of the results of other laboratories can be conducted
if determined advisable.) Laboratory C quantified only stan-
dards and beef fat samples; therefore, the exact reliability
of the analytical method for human milk is currently
speculative.
Two types of upper and lower 95% confidence limits have
been calculated for the regressions of reported values (Y)
on spiking levels (X), as shown for standards . iriTigure D-l
and for beef fat in Figures D-2 and D-3. First are the 95%
confidence limits for the line itself, as are graphed by the
pairs of lines closest to the regression line in the above
Figures. These limits are interpreted as follows: The true
regression line (as would be determined if the experiment
were repeated a countless number of times under the same
conditions) lies within the confidence limits unless the
test results are sufficiently unusual to be among those
expected to occur less than 5% of the time.
-41-
-------
The second set of confidence limits, depicted by the pair of
lines furthest from the regression line, predict the 95%
confidence limits for the result of a single analysis at a
particular spiking level. Interpretation is as follows:
The result (reported value) of a single analysis of a
standard or beef fat sample spiked at a given level can be
predicted to fall between the 95% confidence lines unless
the analytical result (which includes extraction as well as
GC-MS guantitation) is one sufficiently unusual to be expected
to occur approximately 5% of the time.
In calculating the regression lines and confidence limits,
values of "nd" have been excluded. For Lab C all spiking
levels below 9 ppt were reported as nd, and therefore, the
lower limit of guantitation in this study must be considered
to fall somewhere between 5 and 9 ppt. Lab C gave no
erroneous reports (i.e., no reports classified as FP, fp or
fnd) for standards or for 5g beef fat samples when extraction
utilized acid/base cleanup.
The calculated regression line for standards lies very close
to the theoretical line, the slope of 0.983 being essentially
egual to the theoretical slope of 1 and any point on the line
being from 1 to 2 ppt less than the spiking level (Figure
D-l). The 95% confidence limits for a predicted result of a
single analysis fall only 6 to 7 ppt above and below the re-
gression line. Thus, accuracy can be expressed as a negative
-42-
-------
bias averaging about 2 ppt over the range of levels tested,
and precision in terms of the confidence limits for the line
and for predicted results of individual standards. There
was no apparent tendency for increased variability among
reported values at higher (or lower) spiking levels, i.e.,
variance about regression was apparently independent of the
spiking level (See Table C-3 to compare values).
The results of the beef fat analyses were slightly more
variable than those for standards, as might be expected. In
particular, one value was an apparent outlier (reported value
63 ppt; spiking level 25 ppt) and has been both excluded
(Figure D-2) and included (Figure D-3) in calculations. The
rationale for excluding the value is based on a discussion
with the principal investigator at Lab C; he was reasonably
certain that on-the-spot calculation of separate measurements
of the same GC-MS run would have revealed a discrepancy and
the sample would have been rerun. Thus, exclusion of the
value assumes a laboratory procedural modification to
eliminate the possibility of reoccurance. The reported
value of the sample's duplicate was 24 ppt. (A second
samplethat spiked at 64 pptwas originally reported as 32
ppt. Recalculation without knowledge of the spiking level
revealed an arithmetic error, resulting in a revised value
of 58 ppt, which has been used in calculations.)
-43-
-------
Figure D-l
Standards
Lab C (322 m/e)
a.
D,
Theoretical line, y»X
Regression line:
y « .98x - 1.30
95% conf. limits for regression line
95% conf. limits for individual analyses
-------
1
110
Figure D-
f
1
1
100 i
90
80
70
60
50
4C
30
20
10
1 *
y
-^~
s ^
/*- S ^
s / if
y jr ^ j
^ ' f *
r ^ S /
^ / X
-" / S
2
Beef I
'at Sample
s, Lab C
rheoretical line, Y = X -"""
Regression line, Y = 0.89X 4
95% confidence limits:
regression line """"""""N^
individual samples ^
-
y ,
S
s ^
S jSp /
r fjf /
*F f
y /
1
s
s"
^
f s
-3*- 7%?
s _/m./
/ ^ V
f S
"" y
s
S
\
\
\
\
\
\
s
^
ff
s _{_fs
Jr S S
# Ss
ss " /
f jr
V
/^
r
(n = 17):
""""^t, -'
^^*s», _
"*^V
2.7
\
\T
_\T
\
\
--^
V
\
\
\
\
V
\
\
\
\
\
L ' H
\ \
\ \
\ \
\ \
\ "
\ '
\
\
-Y-
\
\
V
\
\
\
V
\
\
\
5g, acid;
/base, 322
^.
>S 1
"V i
Tt '
n i
\
\
V
\
\ \
\
\
\
\
\
>t
\
V
\
\
\
\
\
1
i'
\ . ! .
\.
-\
\
\
Ik .
V *
\
r\ ^
S jm]Jf
imjr *t7s
^^*a * y s A
r i t ^ Y" si
I y s Syt
1 ' \if 'S '
Y m\ S^S
A T\''S S
S \ LfSs s
y' ytivv' s
y f\s\s *
s "sXjf
r/jf X\ ' ih ' '
s'jrs
*r^*S
f J*
^
J^
r
,
S*l*
'Xf
:
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
, \
\ ^
\ (.
\ ^
\ s r
\ " *
\ / y
'\f S
/ 5f y *s
\? s
.f u^T* J?
r r^ '^
J^ .T ^
s S
*y^ jr~~~
r s
1 m/e
-
;
r
:
»'
St
jT
-^r ^
1 St S
s~
/. m
Sf*
^ ' ' f
j S ft
\** r r
* * ' r
~/^- s s* -
s / -A S /
f ^ ^f^ -^ ^
-? s \^ y
* s S
f. * .^'~s'
* / s f
/f f. f~
_r s ^
S s y
s
s ' ' jf
s
*~
;
i
1
j
1
: ;
T": "
-
-
*"
- -
-
"\
BEE 10110 10
20
30 40 50
TCDP added (net)
60
70
80 -45-
-------
/.-Eat Camples, .Lab C (n=18) : 5e;, acid/base,.322 m/e ,
110
100
T.
o
o
Theoretical line: Y = X
Regression line: Y «* 0.84X +7.1
95% confidence limits:
regression line
individual samples
40 50 60
TCTtn added--(Dot)-
30
20
10
-------
Figure E-2
Estimated True TCDD Level in Beef Fat: Lab C
(From regression analysis of 18 spiked samples)
OtIOT 33E
_
_:
. .. J
-
-
EstiH
- ^
X
- 95% c
4-vmfi
-
-
1
1
i
i
i /
I - /
' /
/
f
s
/'
.
X"
/
S ' '
s
_r
S
/ ~ '
, -
//
//
//
/s
//
//
//
; 1 . .
- - 1 '
ated true
K
» (y - 7.
onfidence
level in
u, .. .
' - - *, '. >
-A
' ' f
* '
A
J>
S
S " -
/"
&
#
y/
//_
/s
. . .
levels
D/0.84 *
limits f(
5g sanpl<
»
X
iff
f
.
J ;
1
//
J/
/&
//
// '
s/
S/ - '
/y
//
f .
- - t i .
i
s
/
s
/
j
^**^
-
"TT
Q ^"-~-
^
X
\
V
'
X
; r^
\i /
f
^ ; /
^^\f
f Lf
/ /
r /
* /
i tfm.m
s
s
7
\
^
\
' \
. : V
V
X y
V
' ' //
/f
/f
//
J/
l&
' //
J&
f
>
jf
jf
s
/f
#
rff
SS
s/
//
\
V
V
V
V
X.
^>.
^^.
^^.
S
^
s
i
^
^^v
X
'f ^~
y X
/ \
/ ^
/ \
jf
s
A
- i .
.
.
//
#
//
//
//
//
\ I #
\ //
^ y/
\ //
\\ JT
\ & '
\t *
yr
'- J?
jf
jf
/S
/
w
//
Jf
(f
#
7
JT
S
Jf
s
,
/r
/a^
if"
&
,/
jfr /
y ^
.
s
y
f
y
s
". : I
1
! ' i
i i
'
i -- j
110
100
90
4J
o.
**s
V*
Q
Q
e)
M
I
ft)
jJ
o
4J
(0
u
80
70
60
50
40 n:
30
20
10
10-
y = Reported TCDD level (ppt)
-80-
--9D
-------
Figure E-l
Estimated True TCDD Level in Beef Fat: Lab C
(From regression analysis of 17 'spiked samples)
1104
i
-
_
.
90
^^. *
&,
a 80 ~
.,-(
§ 3
70 ~
.u
"T3
itffe.
^Hkh
*j
05 -
II
* 50 -
An _
""
30 I
rt y\
20 ~
Estimated true
X=(Y-.2.7
95% confidence
true level i
>
-1
: '
.
level:
limits fo
n 5g sanp
^"
j
j/ ~ ~f -
s
/
s
X J*'
- - 1 ./ J-
--.--. \, f
" ' i jr
\ f
' ' f
; / \ f j
/ \~ f ^
S 1 AT
' \f '
i f
f
' -j'
y y^
f
s
S, s \
f J ' /
S . J s
jf :
jff ; S
jr ! x
1 jr ! /
MVH^ ^^'^VW^WAV^M
i ^
f
^H
'
r
le ^V
^
/
^
t
V
\ .. /
~ -T
71
X"
f
f >
/ f
/ s
y-
JT
^v y^
X ^
\
y< \ ^
/ \ff
\ ' f
\ J^
/^^"^^Mv* f j*
s )" '
/ , ,
/ /
/ f
f
S
/ ' S
s
7* >
yL. .-£
\r It
yv /
r^_^h
f /;
f /
jr /
r /
.
/
/
S
f ' *
f /
f /
s
/
s
f . '
f
^ B-*^B~«^^-^^B
.
t
1
^^W^^^^^^IB^^^^H^KV*
W^^V^W^^O^»^BP^M«l«^HMh
/
f
S
**
S t
f /
/L 1
- , , X\
f
.
-
*
_r
J
f
f
S
'
1
1
1 1
«^^^pv^p^^^v«^^«^-^^^i*^
1 . ... : 1 !.- \ i j
V = Reported TCDD level (ppt) ~49"1 >
-------
calculations (Figure E-l) (see part D). When that value is
included, the confidence limits are about four times as wide
(Figure E-2). For example, the estimated actual value for a
reported value of 40 ppt is 42 ppt with 95% confidence
limits of 35 to 49 ppt when the aberrant 63 ppt has been
excluded from regression calculations. When included, the
estimated true level for the same reported value (40 ppt) is
39 ppt with confidence limits of 12 to 66 ppt. Again, the
essentiality of developing procedures to detect and correct
aberrant results at the onset is emphasized.
The confidence limits presented in the above graph are for a
single extraction and GC-MS quantitation of a sample. Confi-
dence limits can be narrowed if 2 or more independent extrac-
tions and quantitations of the cause sample are performed and
the reported values averaged. This approach may be of value
in applied TCDD residue evaluations.
-48-
-------
E. Estimation of Actual TCDD Levels, with Confidence
0
Limits, from Spiking Results: Lab C
.Using regression statistics developed from spiking study
data, a "best estimate" of the "true" level of TCDD in an
unspiked sample can be derived from the reported level, as
well as statistical confidence limits for that estimate.
The. procedure is that of estimating the value of the
independent variable (X) for a measurement of the dependent
variable (Y), in this case the reported TCDD level in a
sample. (The basic approach is used, for example, in
estimating LD50's in dose-response studies.) Confidence
limits for such estimates tend to be broad relative to these
for the regression line per se.
Figures E-l and E-2 present estimated "true" values of TGDD
/\
(X) for reported values (Y) ranging from approximately 9 to
80 ppt. In both figures the reported values (Y) now appear
on the horizontal axis and the estimated "true" values (X)
on the vertical axis. The slope of the'new line is the
reciprocal of the slope of the regression of Y on X. The
new regression equations appear on the graphs.
The 95% confidence limits for estimates 'of actual values
range from 6 to 7 ppt above and below the predicted value
when the aberrant value of 63 ppt is eliminated from the
-47-
-------
F. Discussion
This study has demonstrated that for standards and spiked
samples of beef fat, the extraction and guantitation methodo-
logy exist to quantify TCDD at levels as low as 9 ppt with
practicable accuracy and precision. This conclusion assumes
that extraction methods ar-e exactly those used at PML and
quantitation utilizes procedures and instrumentation indenti-
cal or equivalent to that of Lab C. Otherwise, practicable
precision has not been fully demonstrated.
The reliability of the above methodology for quantifying TCDD
in human milk is yet to be determined. However, based on
the fact that milk results are essentially as precise as
those for beef fat among laboratories that performed both
sets of analyses, a quantitation problem is not anticipated.
None-the-less, the procedure needs to be verified with
further .testing.
Lab C has indicated that their instrumentation may be
capable of quantifying TCDD levels below 9 ppt in samples of
the type used in the study. Reports of False Positives
(positive TCDD values reported for unspiked samples) by
laboratories other than Lab C may present a basic problem
when attempting to quantify in the range of 0 to possibly 8
ppt. Additional analyses of spiked samples are necessary to
determine if a quantification level below 9 ppt can be
achieved.
-51-
-------
Reprinted from Analytical Chemistry, 1981,53,1902.
Copyright © 1981 by the American Chemical Society and reprinted by permission of the copyright owner.
Method Validation for the Determination of
Tetrachlorodlbenzodioxin at the Low Parts-per-Trillion Level
M. L. Gross,* Tung Sun, P. A. Lyon, S. F. Wojlnski, and D. R. Hllker
Department of Chemistry, University of NebraskaLincoln, Lincoln, Nebraska 68588
A. E. Dupuy, Jr.
USEPA, OPTS/OTS, EED, Field Studies Branch, Toxicant Analysis Center, Bay St. Louis, Mississippi 39529
R. G. Heath*
USEPA, OPP/OTS, HED, Health Effects Branch, 401 M St., S.W., Washington, D.C. 20460
This statistically designed study Is directed at determining the
precision and accuracy obtainable In the quantltatlon of
2,3,7,8-tetrachtorodibenzodioxin (TCDD) In standard solutions
and In fortified beef adipose tissue. The TCDD was extracted
after digestion of the adipose tissue In ethanollc potassium
hydroxide, and the resulting solution was cleaned up using a
concentrated sulfurlc acid wash and short column liquid
chromatography. The analysis was conducted with packed
column gas chromatography Interfaced to a high-resolution
mass spectrometer operating at a mass resolution of
10000-12000 (10% peak width definition). Quantltatlon was
achieved by employing an Internal standard method which
Involved 2,3,7,8-TCDD labeled with "Cl. The results were
submitted to comprehensive statistical analysis In order to
determine "best estimates" of concentrations In actual sam-
ples and to express the reliability of such estimates In terms
of statistical confidence limits.
Attention has been focused recently on the need for eval-
uation of data taken in environmental analysis 0). One aspect
of evaluation is a quality assurance program which should
include proficiency testing. This is a report of a blind study
of the proficiency achievable for the analysis of 2,3,7,8-
tetrachlorodibenzo-p-dioxin. The study also included com-
parison of replicate results with those obtained in other lab-
oratories. Those results are available in another report (2).
This study was undertaken to measure the accuracy and
precision with which 2,3,7,8-tetrachlorodibenzodioxin (TCDD),
when added to beef fat at low parts-per-triUion concentrations,
could be extracted and quantified using packed column gas
chromatography-high-resolution mass spectrometry (GC-
HRMS). Method validation also included quantitation of
equivalent amounts of TCDD in standard solutions.
The extraction and analysis employed in this study stem
from the pioneering work of Baughman and Meselson (3) and
later O'Keefe, Meselson, and Baughman (4) who employed
high-resolution dual ion monitoring with direct probe intro-
duction of the sample. The disadvantage of this procedure
has been overcome by scientists at Dow Chemical by devel-
oping packed column GC for sample introduction to the mass
spectrometer (5). Although this approach is specific for the
general class of TCDDs, it does not permit separation of all
TCDD isomers. However, some can be distinguished (6); for
example, 2,3,7,8-TCDD can be resolved from 1,3,6,8-TCDD
using the gas chromatography employed in this study.
To meet the needs of environmental monitoring at the low
parts-per-trillion range, particularly for samples relating to
certain types of combustion (7-11), isomer specific methods
have been developed. One approach is to use capillary column
GC coupled with either low (8,12,13) or high-resolution (14)
mass spectrometry. A single column can be used to separate
2,3,7,8-TCDD from all the other 22 isomers (13). A second
approach involves a combination of high-pressure liquid
chromatography and packed column GC/low-resolution MS
-------
ANALYTICAL CHEMISTRY, VOL. 53, NO. 12, OCTOBER 1981 1903
(6). This latter method is totally isomer specific.
Nevertheless, the methods evaluated here have advantages.
The extraction and cleanup are relatively simple, and the
GC/MS procedure is efficient Therefore, the analysis is rapid
and suitable for carefully controlled monitoring studies and
for sample screening.
This extensive blind study is the first to be conducted for
parts-per-trillion levels which has made use of rigorous sta-
tistical design and analysis of the data. Recently, analytical
data for various chlorinated dioxins have been submitted to
a similar statistical treatment (75). However, those results
were not obtained hi a blind study.
The specific goals of the investigation are to develop re-
gression statistics for converting reported TCDD concentra-
tions to "best estimates" of actual (but unknown) concen-
trations and to express the reliability of such estimates in
terms of statistical confidence limits. In addition, we intend
to determine the lowest concentration of TCDD detectable
with practicable consistency and the frequency of "false
positive" and "false negative" reports.
EXPERIMENTAL SECTION
Sample Extraction and Cleanup. The adipose tissue samples
were extracted by methods described previously (14). Briefly,
beef adipose was rendered to obtain a representative sample that
was free of connective and other tissue. A 10-g sample was spiked
directly with 20 ng of 37C1-TCDD and a known amount of
nonenriched or "native" TCDD. After addition of 15 mL of
distilled water, 20 mL of ethyl alcohol, and 40 mL of 45% po-
tassium hydroxide, the mixture was refluxed with stirring for 2.5
h. After being cooled, the resulting homogenous solution was
extracted 4 times with 25 mL of hexane each, and the hexane
extracts were combined.
The combined hexane extracts were washed with base solution
followed by four 50-mL portions of concentrated sulfuric acid.
The hexane layer was then washed with distilled water, neu-
tralized, and dried by passing it through a glass column packed
with anhydrous sodium carbonate. The hexane was then con-
centrated in preparation for liquid chromatography.
The concentrate was transferred to a small alumina column
(4.5 cm X 0.5 cm) prepared hi a disposable Pasteur pipet. The
column was eluted with 6 mL of carbon tetrachloride (discarded)
followed by 4 mL of methylene chloride which was collected in
a distillation receiver. The solvent was evaporated through a
micro-Snyder column, and two separate 2-mL portions of hexane
were added and also evaporated. The residue was taken up in
3 mL of hexane and rechromatographed on a second column as
just described. After the methylene chloride was evaporated, 2
mL of benzene was added, the solution concentrated to 0.100 mL,
transferred quantitatively to Chromaflex tube, and carefully
concentrated to 0.060 mL using a stream of dry nitrogen. This
extract was split into two equal portions and each was sealed in
3 mm i.d. X 7 cm glass tubing and stored in a freezer until the
GC/MS analysis. Samples were shipped in dry ice cooled con-
tainers.
GC/MS Analysis. The sample extracts were analyzed at the
University of Nebraska in two batches (see below) using a Pye
Unicam Series 104 gas chromatograph which was coupled directly
to a Kratos MS-50 high-resolution mass spectrometer. Data
acquistion was accomplished by use of a 40K word Nicolet Model
1180 computer which was interfaced to the peak switching cir-
cuitry of the mass spectrometer.
An aliquot representing 20-30% of the sample was injected
(injection at 250 °C) on a 183 cm x 0.64 cm o.d. GC column
containing 3% OV-3 on Supelcoport (from Supelco, Inc., Belle-
fonte, PA). A helium flow rate of about 20 cm3/m was used at
a column temperature of 250 °C for 1 min and then the tem-
perature increased linearly to 300 °C at a rate of 10 deg/min. The
column was maintained at 300 °C for 7 m. The eluent from the
column (principally solvent) was vented to the atmosphere for
the first 1 min and then the entire gas flow was admitted directly
to the mass spectrometer source. The interface consisted of 75
cm X 0.03 cm i.d. glass-lined stainless steel tubing connected to
a 7.5 cm X 0.15 mm i.d. glass restrictor and an electrically non-
321.8936
327.8848
Figure 1. Actual signal output (lower trace) and smoothed output (51
point sliding window) for high-resolution dual ion monitoring of an
adipose extract fortified with TCDD at 16 pptr. The signal on the left
Is for m/z 321.8936 (larger signal) and on the right Is for mlz
327.8848, the internal standard. The reported concentration is 12 pptr.
conducting glass coil. The transfer line and the glass restrictor
and coil were held at 250 and 220 °C, respectively. The retention
time, determined by using standard solutions of 2,3,7,8-TCDD,
was 4.4 and 5.3 min for the first and second batches, respectively,
of sample analyses. The 2,3,7,8-TCDD and any coeluting isomers
were quantitated by duel ion monitoring using real-time peak
matching at a mass resolution of 10000-12000 (10% valley). One
channel was centered at m/z 327.8848 ("Cl-TCDD, the internal
standard) and the other at m/z 321.8936 (the most abundant
molecular ion of TCDD having natural isotopic abundances). The
complete peak profiles were acquired by scanning at a frequency
of 2 Hz over a masss range of 300 ppm (0.097 atomic mass unit).
The output for about 120 sweeps was accumulated hi the memory
of the computer, submitted to a smoothing routine, and then
plotted on an X-Y recorder (see Figure 1 for a typical output).
The signal averaging was started at the first appearance of internal
standard signal observed on the peak matching oscilloscope (11).
Data Handling. The concentrations of TCDD were deter-
mined from a calibration plot using the ratio of the maximum
peak heights at m/z 327.8848 and 321.8936. The calibration data
were acquired by using solutions of the native TCDD and ^Cl-
TCDD which were analyzed at a rate of about one every two to
three unkowns. The calibration was linear over a range of 5-150
pg of native TCDD. If no signal was observed at m/z 321.8936,
the detection limit was calculated to be 2.5 times the noise am-
plitude (a 2.5:1 signal-to-noise criterion).
The percent recovery was calculated knowing the size aliquot
removed from the original extact and measuring the absolute
intensity for the isotopically labeled internal standard (m/z
327.8848). The response of the GC/MS to the internal standard
was determined by injecting known amounts of the standard.
Further confidence for the assignment of the peak profile as
TCDD can be obtained by injecting a second aliquot and mon-
itoring m/z 320 and m/z 322. This was not done for these samples.
The internal standard method (using m/z 322 and m/z 328 peak
intensities) is preferred for quantitation.
The results were transmitted without knowledge of the code
to the office of the coordinator of the Dioxin Monitoring Program,
United States Environmental Protection Agency, Washington,
DC. The data were then decoded and forwarded to one of the
authors (R.G.H) for statistical analysis using standard methods
(16).
RESULTS AND DISCUSSION
Study Design. Beef fat samples were "spiked" with native
2,3,7,8-TCDD at levels ranging from 0 to 81 pptr (parts per
trillion). Standards were prepared so as to contain equivalent
amounts of the chemical. Samples were developed from one
-------
1904 ANALYTICAL CHEMISTRY, VOL. 53, NO. 12, OCTOBER 1981
Table I. Sample Design for Validation Study
beef fat standards
TCDD
level0
0
0
'/2
1
4
9
9
16
25
25
36
49
49
64
81
81
Parts per trillion.
pool
code
F
G
F
F
F
F
G
F
F
G
F
F
G
F
F
G
no. of
extracts
2
2
1
1
1
2
2
1
2
2
1
2
2
1
2
2
pool
code
S
S
S
S
S
S
S
S
S
S
no. of
solns
1
1
1
4
1
4
1
4
1
4
of two pools of rendered beef fat (pools F and G) taken from
cattle without potential exposure to TCDD. Pools were
constructed by using equal amounts of fat from each animal.
Tissue from a separate set of animals was used for construction
of each pool (see Table I).
Eleven samples were prepared from fat pool F, the major
pool. The samples were spiked individually with native
2,3,7,8-TCDD and the internal standard. They were then
extracted as described in the Experimental Section. Spiking
levels, excepting 0.5 pptr, were systematically incremented
as the squares of the digits 0 through 9 to provide close spacing
at low levels and a moderate, systematic increase in spacing
with increasing levels.
To test the precision of the extraction methodology, we
prepared five samples from fat pool G, the minor pool (see
Table I). These samples were extracted by the same proce-
dures used for samples from Pool F. Thus, fat pools F and
G provided replicate samples at the above levels of spiking
for testing extraction precision.
To test the precision of GC-HRMS quantitation for com-
parison with that of extraction, the analysis laboratory was
provided two aliquots of the G-pool extracts, along with two
aliquots from the matching extracts from pool F, so as to
obtain duplicate analysis of the same extract. The laboratory
also received four replicates of some of the standard solutions
(see Table I).
All beef fat samples and standards were prepared or ex-
tracted and shipped hi random order. In all, 52 samples were
involved. They were transmitted to the University of Ne-
braska in four batches each containing about equal numbers
of extracts. The first 27 samples were analyzed along with
suitable standards and other samples over a 5-day period. The
remaining 25 samples were analyzed over a 6-day period 4x/2
months later. The samples were selected for analysis in
random order. Samples were identified only by shipment
number, so that neither the type of sample nor the TCDD level
was known at the time of quantitation.
Accuracy and precision were measured by methods of re-
gression analysis (16). The comparisons of extraction vs.
quantitation precision have been made by analysis of variance
based on those spiking levels for which there were duplicate
analyses of replicate extractions.
Three possible types of analytical reporting errors are
recognized in this study. The reporting of a positive value
in an unspiked sample is identified as a "false positive" (FP),
and a positive report given when the detection limit exceeds
the level of spiking is identified as a "false positive" (fp). A
9O
THEORETICAL LJN£,Y
REGRESSION LINE;
Y = .98X - I.3O
95% CONF. LIMITS
FOR REGRESSION LINEx
95% CONF. LIMITS
FOR INOMDUAL ANALYSESN
STANDARD SOLUTIONS
0 IO 2O 3O 4O 50 60 70 80
TCDO ADDED (ppt)
Figure 2. Reported concentration vs. concentration of TCDO actually
added to standard solutions.
9O
SO
70
'a
- 60
3O
20
IO
O
THEORETICAL UNE,Y=X-
REGRESSION LINE, Y=O.89X
95 % CONFIDENCE LIMITS:
REGRESSION LIMITS LINES
INDIVIDUAL SAMPLES,,
BEEF FAT-5g.
02030405060 70 80
TCDD ADDED (ppt)
Figure 3. Reported concentrations vs. concentrations of TCDD actual/
added to beef adipose. The aberrant value (see text) has been in-
cluded.
"false not detected" (fhd) is defined as a report of "nd" when,
in fact, the level of detection is less than the level of spiking.
Statistical Analysis. Analytical results for the quanti-
tation of standards and beef fat samples and the percent
recoveries are presented in Tables n and in, respectively. The
plotted results for standard solutions, the least-squares re-
gression line, and equation for reported values are given in
Figure 2. Figure 3 is a similar presentation for beef fat data
but excludes an aberrant report of 63 pptr for a sample spiked
at 25 pptr (see below). In each of the above figures, the
theoretical line, y = x, representing perfect extraction and
quantitation, is included for comparative purposes.
Two types of upper and lower 95% confidence limits or
"bounds" have been calculated for the least-squares regressions
-------
ANALYTICAL CHEMISTRY, VOL. 53, NO. 12, OCTOBER 1981 1905
Table II. Results for Analysis of Standard Solutions
(5 g Equivalent) for TCDD
analysis results
study
code"
S-0,
S-02
S-03
S-04
S-0.5
S-l
S-4
S-9,
S-9,
S-9,
S-94
S-l 6
S-25,
S-252
S-253
S-254
S-36
S-49,
S-492
S-493
S-494
S-64
S-81,
S-812
S-813
S-814
TCDD
added b
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
TCDD
reported6
nd
nd
nd
nd
nd
nd
nd
nd
11
9
6
16
21
17
25
24
34
41
51
47
47
65
76
77
80
80
detection
limit6
10
4
4
4
6
12
6
10
5
4
4
10
4
4
4
4
16
12
3
9
5
6
6
3
3
5
%
recovery c
93
85
85
85
100
94
104
76
80
95
100
94
99
85
95
75
110
130
90
90
80
90
104
100
75
70
0 Code refers to sample numbers used for the entire
study at TAG and at UN-L. 6 Parts per trillion. c Aver-
age recovery 92 ± 13% (standard deviation). The report-
ed concentrations are corrected for recovery in the
internal standard calculations method.
of reported values (y) on spiking levels (x), as shown for
standard solutions (Figure 2) and for beef fat (Figure 3). First
are the 95% confidence bounds for the regression line itself,
as are graphed by the pair of lines closest to the regression
line in each of the figures. These bounds are interpreted as
follows: the true regression line (as would be determined if
the study were repeated a countless number of times under
the same conditions) lies within the confidence limits unless
the analytical results are sufficiently unusual to be among
those expected to.occur less than 5% of the time.
The second set of confidence bounds, depicted by the outer
pair of lines, constitute the 95% confidence limits for the result
of a single analysis at a particular spiking level. The inter-
pretation is as follows: the result (reported value) of a single
analysis of a standard or beef fat sample spiked at a given level
can be predicted to fall between the 95% confidence bounds
unless the analytical result (which includes extraction as well
as GC-HRMS quantitation) is among those sufficiently unu-
sual to be expected less than 5% of the time.
In calculation of the regression lines and confidence limits,
values of "nd" have been excluded. All spiking levels below
9 pptr were reported as "nd". Therefore, the lower limit of
quantitation in this study must be considered to fall some-
where between 5 and 9 ppt.
The analysis laboratory gave no erroneous reports (i.e., no
reports classified as FP, fp, or fnd) for standard solutions or
for 5-g beef fat samples extracted utilizing acid/base cleanup.
The calculated regression line for standards solutions lies
very close to the theoretical line, the slope of 0.981 being
essentially equal to the theoretical slope of unity and any point
on the line being from 1 to 2 pptr less than the spiking level.
The 95% confidence limits for a predicted result of a single
analysis fall only 7 pptr above arid below the regression line.
Thus, accuracy can be estimated as a negative bias averaging
Table HI. Results for Analysis of Extracts of
5 g of Beef Fat for TCDD
analysis results
study
code"
F-0,
F-02
G-0,
G-02
F-0.5
F-l
F-4
F-9,
F-92
G-9,
G-92
F-l 6
F-25,
F-25,
G-25,
G-25,
F-36
F-4 9,
F-492
G-49,
G-492
F-64
F-81,
F-812
G-81,
G-812
TCDD
added6
0
0
0
0
0.5
1
4
9
9
9
9
16
25
25
25
25
36
49
49
49
49
64
81
81
81
81
TCDD
reported6
nd
nd
nd
nd
nd
nd
nd
9
11
nd
12
19
(63)d
24
26
27
31
50
48
39
48
58
76
74
76
77
detection
limit6
8
5
6
6
14
10
18
4
5
14
5
10
10
2
12
4
6
6
4
10
6
4
6
4
14
3
%
recovery c
67
90
75
70
93
79
92
66
75
84
85
63
67
100
75
75
68
72
85
120
75
79
61
85
57
75
a Code refers to sample numbers used for entire study.
6 Parts per trillion. c Average recovery 78 ± 14% (stand-
ard deviation). The reported concentrations are cor-
rected for recovery in the internal standard calculations
method. d Aberrant value (see text).
about 2 pptr over the range of levels tested, and precision can
be expressed in terms of the confidence limits for the line and
for predicted results of individual standards. There was no
apparent tendency for increased variability among reported
values at higher (or lower) spiking levels, i.e., variance about
regression was apparently independent of the spiking level
(sees Table II and III).
The rationale for excluding the aberrant value of 63 pptr
for beef fat is as follows. The workers in the MS lab were
suspicious of the first analysis and reanalyzed this sample
obtaining a value of 24 pptr (true value, 25 pptr). However,
calculations of TCDD concentrations were not made on-the-
spot; therefore, the discrepancy was not realized until a later
time when a third analysis was not possible. The 24 pptr result
was not reported because the signal-to-noise ratio was worse
than the first determination. We are reasonably certain that
if calculations were made immediately, the discrepancy would
have been realized and a third analysis conducted.
The calculated regression line for beef fat, based on 17
samples (aberrant value excluded), lies close to the theoretical
line y = x over the range of spiking levels tested (see Figure
3). Its slope of 0.89 (95% confidence limits 0.84-0.95) differs
slightly from the theoretical slope of 1.0, the calculated line
intersecting the theoretical line at approximately 25 pptr.
Thus, bias, although small, is a function of the spiking level,
ranging from approximately +2 pptr for samples spiked at
9 pptr to -6 pptr for those spiked at 81 pptr. Precision was
comparable to that for standards: the mean squares for de-
viation from regression are 5.51 pptr2 for standards and 5.15
pptr2 for fat (5 df each).
Analysis of variance of the differences between reported
and spiked TCDD levels for replicate fat samples and du-
plicate aliquots of the same extract shows that variation be-
tween replicates is not significantly greater than that for
-------
1906 ANALYTICAL CHEMISTRY, VOL. 53, NO. 12, OCTOBER 1981
duplicate aliquots (i.e., within replicates). The mean square
for replicate samples was 10.68 pptr2 and that for duplicate
aliquots was 7.92 pptr2 CF4>6df = 1.35; n.s.). This result dem-
onstrates the reliability of the extraction methodology at low
parts-per-trillion levels.
Estimation of True TCDD Levels in Beef Fat. By use
of regression statistics developed from spiking study data, a
"best estimate" of the true level of TCDD in an unspiked beef
fat sample can be derived from the reported level, as well as
statistical confidence limits for that estimate. The method
is that of inverse prediction for estimating the value of the
independent variable (the true TCDD level) for a given
measurement of the dependent variable (the reported level).
The estimated true values of TCDD (x) for the reported
values (y) ranging from approximately 9 to 80 pptr has been
obtained by replotting the reported values (y) on the hori-
zontal axis and the estimated true values (x) on the vertical
axis. The slope of the new line is the reciprocal of the slope
of the regression ofy on x. Two new regression equations (eq
1 and 2) are obtained for the data excluding and including
the aberrant value, respectively.
x = (y - 2.7)/0.89
x = (y - 7.D/0.84
(1)
(2)
The 95% confidence limits for estimates of true values range
from 6 to 7 pptr above and below the predicted value when
the aberrant value of 63 pptr is eliminated from the calcu-
lations. When that value is included, the confidence limits
are about 4 times as wide. For example, the estimated actual
value for a reported value of 40 pptr is 42 pptr with 95%
confidence limits of 35-49 pptr when the aberrant 63 pptr has
been excluded from regression calculations. When included,
the estimated true level for the same reported value (40 pptr)
is 39 pptr with confidence limits of 12-66 pptr. It is essential
that the procedure to detect and correct aberrant results be
implemented at the time of analysis.
The confidence limits presented are for a single extraction
and GC/MS quantitation of a sample. Confidence limits can
be narrowed if two or more independent extractions and
quantitations of the same sample are performed and the re-
proted values averaged. This approach may be of value in
TCDD residue evaluations.
In conclusion, this study is a demonstration that for
standards and spiked samples of beef fat, the extraction and
quantitation methodology exist to quantify TCDD at levels
as low as 9 pptr in 5-g samples with practicable accuracy and
precision. This conclusion applies if extraction methods are
exactly those used at TAG and for quantitation procedures
and instrumentation equivalent to that of the University of
Nebraska.
Improvements in the detection limit can be expected by
making use of more thermally stable GC columns to reduce
chemical noise and by turning to capillary column GC/MS
to increase the instantaneous concentration of the TCDD in
the ion source.
ACKNOWLEDGMENT
We thank Mary Frankenberry for help with the statistical
analysis.
LITERATURE CITED
(1) ACS Committee on Environmental Improvement and Subcommittee on
Environmental Analytical Chemistry Anal. Chem. 1980, 52, 2242.
(2) Heath, R. G. "Intel-laboratory Method Validation Study for Dloxin, An
Interim Report", United States EPA, Human Effects Monitoring Branch,
Office of Toxic Substances, Washington, DC, 1979.
(3) Baughman, R. W.; Meselson, M. S. EHP Environ. Health Persepct.
1973, 5, 27.
(4) O'Keefe, P. W.; Meselson, M. S.; Baughman, R. W. J. Assoc. Off.
Anal. Chem. 1978, 6), 621.
(S) Shadoff, L a.; Hummel, R. A. Bkmed. Mass Spectrom. 1978, 5, 1.
(6) Nestrick, T. J.; Lamparski, L. L.; Stehl, R H. Anal. Chem. 1979, 51,
2273.
(7) Olie, K.; Vermeulen, P. L.; Hutzinger, O. Chemosphere 1977, 6. 455.
(8) Buser, H. R.; Bosshardt, H. P.; Rappe, C. Chemosphere 1978, 7, 165.
(9) Eiceman, G. A.; Clement, R. E.; Karasek, F. W. Anal. Chem. 1979,
5t, 2343.
(10) Bumb, R. R.; Crummett, W. B.; Cutle, S. S.; Gledhill, J. R.; Hummel, R.
H.; Kagel, R. O.; Lamparski, L. L.; Luoma, E. V.; Miller, D. L; Nestrick,
T. J.; Shadoff, L. A.; Stehl, R. H.; Woods, J. S. Science 1980, 210,
385.
(11) Kimbte, B. J.; Gross, M. L. Science 1980, 207, 59.
(12) Buser, H. R. Anal. Chem. 1977, 49, 918.
(13) Buser, H. R.; Rappe, C. Anal. Chem. 1980, 52, 2257.
(14) Harless, R. L.; Oswald, E. 0.; Wilkinson, M. K.; Dupuv, A. E., Jr.;
McDaniel, D. 0.; Tai, Han Anal. Chem. 1980, 52, 1239.
(15) Langhorst, M. L.; Shadoff, L. A. Anal. Cham. 1980, 52, 2037.
(16) Snedecor, G.; Cochran, W. "Statistical Methods", 6th Ed., Iowa State
University Press: Ames, IA, 1967.
RECEIVED for review February 19, 1981. Accepted July 22,
1981. This project has been funded at least in part with
Federal Funds from the U.S. E.P.A. under Contract No. 68-
01-4305. The content of this publication does not necessarily
reflect the views of policies of the U.S. E.P.A., nor does
mention of trade names, commerical products, or organizations
imply endorsements by the U.S. Government.
------- |