1
1
U.
S.
ENVIRONMENTAL
PROTECTION
AGENCY
2
3
4
26TH
ANNUAL
CONFERENCE
ON
ANALYSIS
OF
POLLUTANTS
IN
THE
5
ENVIRONMENT
6
7
C
O
N
F
E
R
E
N
C
E
8
9
10
May
14,
2003
11
12
13
CHICAGO,
ILLINOIS
14
15
16
17
18
19
20
21
22
23
24
25
2
1
U.
S.
ENVIRONMENTAL
PROTECTION
AGENCY
2
PUBLIC
MEETING
3
CHICAGO
HOLIDAY
INN
MART
PLAZA
4
May
1,
2003
5
MR.
TELLIARD:
This
6
public
meeting
is
being
sponsored
by
7
the
Office
of
Water,
Office
of
Science
8
and
Technology.
The
notice
on
the
9
door...
I
don't
know
if
most
of
you
saw
10
that...
said
please
check
all
sharp
11
objects
and
firearms.
I
hope
you
did
12
that.

13
I
am
going
to
spend
about
ten
14
minutes
kind
of
giving
an
overview
of
15
what
we
did,
why
we
did
it,
and
where
16
we
are
at,
and
then
we
will
start
a
17
list
of
speakers
who
want
to
come
up.

18
Most
of
these,
from
what
I
can
tell,

19
are
basically
to
congratulate
the
20
Agency
on
their
tremendous
effort
at
21
getting
this
out
and
how
pleased
they
22
all
are
with
it.
There
are
a
few
23
dissenters
whose
names
I
won't
24
mention.

25
304(
h),
under
the
Clean
Water
3
1
Act,
304(
h)
says
that
the
Agency
will
2
promulgate,
pontificate,
and
implement
3
methods
for
the
measurement
of
4
compliance,
and
in
so
doing,
we
have,

5
under
this,
specified
methods
as
6
published
in
the
Code
of
Federal
7
Regulations
under
Part
136.
8
It
is
important
to
understand
9
the
things
we
are
talking
about
today
10
apply
only
to
Part
136
or
the
11
wastewater
methods.
They
do
not
apply
12
to
Safe
Drinking
Water
methods
as
13
published
under
141.
So,
we
are
only
14
talking
about
wastewater
methods
in
15
Part
136
today.
So,
the
comments
need
16
to
kind
of
be
directed
that
way.

17
In
June,
1999,
the
Agency
18
published
Method
1631B,
the
analysis
19
of
mercury
at
trace
level,
and
a
20
number
of
regulated
industries
felt
21
that
the
method
was
too
tight,
and
22
they
didn't
like
a
number
of
things
in
23
it,
they
wanted
some
things
added,
and
24
after
very
cordial
negotiations,
we
25
came
down
and
resolved
most
of
the
4
1
methods
issues,
and
one
of
the
issues
2
that
came
out
of
the
settlement
3
agreement
was
an
agreement
to
review
4
the
quantitation
limits
and
detection
5
limits
as
specified
in
Part
136.

6
We
had
actually
started
working
7
on
this
a
little
earlier,
but
some
8
people
felt
it
was
moving
at
the
speed
9
of
molasses
and
wanted
to
speed
it
up.

10
Now,
you
all
know
detection
11
limits.
We
have
some
50
different
12
ways
of
expressing
detection
and
13
quantitation.
There
is
the
MDL,
there
14
is
the
IDE,
the
IQE,
the
CAQ,
the
15
PQL,
the
political
quantitation
level,

16
all
of
which
have
been
out
there
17
floating
around
by
various
agencies
18
and
organizations.
Well,
finally,
we
19
are
going
to
try
to
bring
this
to
20
conclusion
by
moving
ahead
and
21
resolving
some
of
the
applications
as
22
it
fits
into
the
overall
program.

23
The
settlement
agreement
says
24
that
EPA
must
assess
existing
Agency
25
and
alternative
procedures
for
5
1
determining
detection
and
quantitation
2
limits
under
the
Clean
Water
Act.
At
3
a
minimum,
we
had
to
evaluate
the
4
method
detection
limit
procedure
as
in
5
Part
136,
Appendix
B,
and
the
minimum
6
level
for
quantification
included
in
7
Method
1631.
EPA
also
must
submit
its
8
assessment
for
formal
peer
review
9
which
we
have
done,
and
we
had
it
10
peer
reviewed
by
two
chemists
and
two
11
statisticians.
This
peer
review
was
12
completed
in
August,
2002.

13
We
have
had
three
Federal
14
Register
notices
regarding
this
15
particular
document,
the
one
notifying
16
the
world
of
this
meeting
today
as
17
well
as
the
publication
on
March
12th
18
of
the
document
describing
the
EPA's
19
assessment,
and
also
the
notice
20
concerning
changes
to
40
CFR
Part
136,

21
as
reflected
in
the
document.

22
General
terms,
method
detection
23
limit
basically
represents
the
lowest
24
concentration
that
allows
for
a
25
difference
between
the
sample
that
6
1
contains
a
substance
and
one
that
does
2
not.
That
is
basically
the
one
I
got
3
out
of
the
8th
grade
science
book.

4
Quantification
limits
are
basically
5
used,
in
our
case,
in
the
ML,

6
represents
the
lowest
concentration
7
that
can
be
measured
with
some
degree
8
of
confidence
and
can
actually
tell
9
how
much
is
there,
and,
of
course,
the
10
big
glitch
here
is
the
question
of
11
degree
of
confidence.

12
When
we
looked
at
the
various
13
approaches,
there
are
50
different
14
terms
that
we
find
floating
around
for
15
describing
detection
and
quantitation,

16
and
there
is
a
general
lack
of
17
consensus
among
various
Government
18
organizations,
FDA,
NASA,
us,
the
Army
19
Corps
of
Engineers,
all
of
which
have
20
their
own
little
ways.
Of
course,

21
then,
the
consensus
standard
22
organizations,
ASTM,
AOAC,
Standard
23
Methods
all
have
their
approaches
24
also.

25
So,
establishing
detection
and
7
1
quantitation
limits
is
not
what
you
2
would
exactly
call
an
exact
science,

3
and
this
has
continued
in
the
latest
4
articles
in
a
number
of
journals
which
5
continue
to
debate
on
large
is
small
6
and
how
small
is
large,
which
was
the
7
latest
article
I
read.

8
Most
of
the
concepts
are
broken
9
into
two
groups,
one,
basically,
that
10
the
error
is
effectively
constant
at
11
the
low
concentration
range,
and
12
limits
are
based
on
errors
observed
in
13
replicate
measurements
made
at
a
14
single
low
concentration,
and
that
is
15
basically
what
Curry
has
done,
ACS,

16
ISO,
IUPAC,
and
EPA.
The
other
17
approach
is
basically
that
the
error
18
varies
as
a
function
of
concentration,

19
and
limits
are
based
on
the
error
20
observed
at
replicate
measurements
21
gathered
across
the
range
of
detection
22
and
quantitation,
and
examples
like
23
this
are
USAFAMA
and
ASTM
in
their
24
procedures.

25
Now,
one
of
the
reasons
we
8
1
would
like
to
come
to
some
sort
of
2
consensus...
and
the
consensus
does
not
3
mean
that
there
is
a
single
answer.

4
It
means
that
we
agree
that
things
5
within
a
certain
range
are
acceptable
6
to
the
Agency
simply
so
that
when
we
7
get
down
to
the
question
of
comparing
8
data,
there
is
some
reference
between
9
how
we
got
A
and
B
and
C.
There
is
10
an
effort
and
continues
to
be
an
11
effort
so
that
as
we
spend
$
1
trillion
12
for
this
set
of
data
that
some
other
13
agency
or
some
other
organization
can
14
use
that
data.
So,
therefore,
we
15
would
like
to
have
data
of
known
16
quality.

17
Historical
use
of
the
MDL
and
18
ML,
it
is
basically
used
to
verify
19
laboratory
performance
and
this
takes
20
many
forms.
Some
are
contract
21
requirements,
some
are
State
QA
22
requirements,
and,
basically,
the
23
laboratories
are
required
to
24
demonstrate
their
ability
to
measure
25
at
the
detection
limit
and
also
to
9
1
measure
at
the
level
of
quantitation.

2
Many
laboratories
use
the
MDL
3
procedures
as
a
means
of
demonstrating
4
and
advertising
their
ability.
We
can
5
measure
0.02
fentagrams.
Who
cares?

6
But
that
is
part
of
the
thing.

7
Demonstrate
the
acceptability
of
8
laboratories
as
modification,
and
EPA
9
uses
the
MDLs
published
in
approved
10
test
methods
as
a
standard,
but
it
11
does
allow
flexibility
across
the
area
12
if
people
want
to
use
another
13
approach.

14
Now,
on
some
of
these,
for
15
example,
in
our
contracts,
we
require
16
that
the
contract
laboratory
17
demonstrate
that
they
can
reach
the
18
MDL,
and,
actually,
in
our
contracts,

19
we
require
that
they
do
it
once,
and
20
then,
from
then
on,
they
have
to,

21
quote,
calibrate
and
demonstrate
at
22
the
ML
but
not
at
the
MDL.

23
Although
it
is
not
our
intent,

24
when
we
started
this
thing,
to
use
the
25
MDL
or
the
ML
as
a
regulatory
limit,
10
1
we
have,
in
fact,
published
a
number
2
of
different
articles,
statements,

3
guidance,
that
said
we
don't
think
4
that
the
MDL
should
be
a
regulatory
5
number.
That
does
not
mean
it
has
6
not
shown
up
in
permits
or
in
rules
7
across
the
country,
but
that
was
our
8
recommendation.

9
And
the
other
concern
is
what
10
do
you
report,
reporting
all
those
11
values
found
above
the
ML
or
above
the
12
MDL,
reporting
only
those
values
found
13
above
the
MDL
but
flagging
those
that
14
were
less
than
the
ML
but
above
the
15
MDL,
reporting
all
the
values
found
16
but
flagging
those
that
were
found
17
below
MDL,
and
all
of
the
above.
So,

18
you
have
various
levels
of
reporting
19
and
various
formats
for
reporting.

20
In
the
assessment,
we
had
a
21
study
plan
which
basically
reflected
22
the
agreement
that
was
signed.
This
23
draft
was
put
together
by
December,

24
2001
and
subject
to
peer
review
and
25
finalized
in
April.
The
information
11
1
used
in
the
assessment
was
to
look
at
2
basically
three
different
literature
3
researches
that
we
did,
one
in
'
97,

4
one
in
'
98,
and
one
in
2001.

5
The
goal
was
to
identify
the
6
concepts,
procedures,
and
issues
that
7
we
were
going
to
consider
in
the
8
assessment,
and
more
than
100
9
documents
describing
detection
and
10
quantitation
concepts
and
issues
were
11
identified.

12
The
actual
data
sets
that
were
13
used
in
the
study
were,
one,
the
EPA
14
detection
and
quantitation
limit
study
15
which
was
affectionate
referred
to
as
16
the
6000
data
set,
and,
basically,

17
that
was
16
different
con...
10
EPA
18
method
at
16
different
concentration
19
run
with
7
replicates.

20
The
second
study
was
the
GC/
MS
21
threshold
study
which,
again,
looked
22
at
the
performance
of
ability
to
23
measure
and
measure
the
effect
of
24
having
a
background
correction
on
or
25
off
on
the
GC/
MS
and
its
impact
on
12
1
the
data.

2
The
American
Automobile
3
Manufacturers
study
looking
at
4
estimating
the
alternate
minimum
level,

5
and
that
involved
nine
laboratories
6
using
Method
2000.7,
245.2.

7
The
data
base
developed
for
8
Method
1631
and
1638
interlab
studies,

9
and
that
was
used
to,
again,
calculate
10
and
compare
the
various
procedures.

11
Concepts
and
procedures
12
considered.
EPA
considered
detection
13
and
quantitation
terms,
concerns,

14
concepts,
or
procedures
that
were
15
advanced
or
published
in
the
16
literature
and
by
various
EPA
offices,

17
other
government
agencies,
non­

18
government
agencies,
and
others.
Many
19
of
the
concepts
considered
were
not
20
fully
evaluated,
because
they
were
not
21
fully
supported
by
corresponding
22
detection
and
a
definition
of
23
detection
procedure;
did
not
reflect
24
the
entire
measurement
process,
that
25
is
to
say,
extraction,
concentration,
13
1
and
the
instrumentation;
were
not
2
uniquely
designed
for
a
single
3
program;
were
no
longer
advanced
by
4
the
organization
or
group,
for
5
example,
the
alternate
minimum
level
6
had
fallen
by
the
wayside,
so
it
was
7
not
considered.

8
EPA
evaluated
four
sets
of
9
concepts
that
were
widely
referenced
10
and
generally
reflected
the
diversity
11
of
the
concepts
advanced
to
date,

12
Method
Detection
Limit,
the
MDL;
the
13
ML;
the
Interlaboratory
Detection
14
Estimate,
the
IDE;
the
Interlaboratory
15
Quantitation
Estimate,
the
IQE
by
16
ASTM;
the
Limit
of
Detection
and
Limit
17
of
Quantitation
by
ACS;
the
Critical
18
Value
and
Minimum
Detectable
Value
19
which
is
the
one
put
forward
by
IUPAC
20
and
ISO.
So,
these
were
the
21
procedures
and
concepts
that
were
22
evaluated
in
the
study.

23
In
addition,
EPA
considered
six
24
specific
issues
that
were
raised
by
25
the
litigants
in
the
settlement
14
1
agreement,
and
these
were
the
criteria
2
for
the
selection
of
appropriate
use
3
of
statistical
models,
the
methodology
4
for
parametric
estimates;
statistical
5
tolerance
and
prediction;
and
the
6
criteria
for
design
of
detection
and
7
quantitation
studies;
including
the
8
selection
of
the
concentration
level;

9
interlaboratory
variability;
and
10
incorporation
of
elements
of
11
probability
design.
These
were
added
12
as,
we
felt,
reflected
in
the
13
settlement
agreement.

14
Additional
issues
that
were
15
also
considered
as
a
result
of
the
16
peer
review
and
other
comments
were
17
the
matrix
effects,
minimumization
of
18
false
negatives
and
false
positives,

19
cost
and
ease
of
implementation,
how
20
well
detection
and
quantitation
limits
21
published
in
the
method
reflect
22
individual
laboratory
capability.
So,

23
these
were
added
to
the
previous
list.

24
EPA
developed
six
criteria
25
based
on
the
considerations
of
the
15
1
issues.
One,
the
criteria
should
be
2
scientifically
valid,
can
be
tested,
3
has
been
peer
reviewed,
has
a
known
4
error
rate,
is
supported
by
a
well
5
defined
procedure,
and
has
gained
6
widespread
acceptance.
Two,
should
7
address
demonstrated
expectations
in
8
the
laboratory
in
method
performance,

9
including
routine
variability.
Should
10
be
supported
by
a
practical
and
11
affordable
procedure
that
a
single
lab
12
can
use
to
evaluate
method
13
performance.
Detection
concepts
should
14
provide
a
99
percent
confidence
that
15
the
substance
will
be
detected.

16
Quantification
concepts
should
identify
17
the
concentration
that
gives
a
18
recognizable
signal
and
is
consistent
19
with
the
capabilities
of
the
method
20
when
performed
by
an
experienced
staff
21
in
a
well­
operated
laboratory.

22
TRACK
02:
...
come
up
to
this
23
podium
and
to
give
comment
on
the
24
proposed
regulatory
changes,
and
this
25
is
an
opportunity
for
us
to
take
16
1
comment.
We
are
not
going
to
debate,

2
argue,
fight,
or
carry
on
over
the
3
comments.
These
are
comments
that
4
people
feel
represent
their
5
organization
or
their
personal
6
opinions,
and
they
will
be
taken
that
7
way.

8
So,
with
that...
the
other
thing
9
is
I
should
point
out
for
the
10
speakers,
please
speak
into
the
11
microphone.
We
are
recording
this,

12
and
after
it
is
done
being
recorded,

13
we
will
have
a
stenographer
type
it
14
up,
and
we
will
include
this
in
the
15
court
record.
So,
your
comments
are
16
formal
comments,
and
they
will
be
17
going
into
the
record.
So,
please
18
speak
clearly
and
into
the
microphone.

19
SPEAKER:
And
our
first
20
speaker
this
morning
is
Richard
21
Burrows,
Severn
Trent
Laboratories.

22
MR.
BURROWS:
I
would
23
like
to
thank
Bill
for
the
opportunity
24
he
has
given
us
to
speak
here
on
a
25
subject
that
is
of
profound
interest
17
1
to
the
environmental
lab
testing
2
community.
I
am
actually
speaking
on
3
behalf
of
the
American
Council
of
4
Independent
Labs.

5
This
presentation,
we
are
going
6
to
cover
a
number
of
issues.
I
am
7
going
to
start
off
with
a
discussion
8
of
whether
we
think
the
MDL
procedure
9
works
or
not.
You
can
probably
guess
10
that
we
think
it
does
have
a
few
11
failings,
so
we
are
going
to
describe
12
what
those
are,
talk
about
why
the
MDL
13
procedure
sometimes
fails,
what
14
alternatives
there
are,
and
how
all
of
15
this
eventually
relates
to
the
ML
or
16
the
quantitation
limit.

17
I
am
going
to
focus
on
18
demonstration
of
the
detection
limit
19
for
routine
analysis
at
an
20
environmental
testing
lab.
That
is
21
probably
99
percent
of
the
MDL
studies
22
that
are
actually
done.
However,
the
23
procedures
that
we
are
describing
24
could
also
be
used
for
secondary
25
detection
limit
for
a
new
method
at
18
1
Part
136
if
they
were
given
a
bit
of
2
a
multi­
lab
flavor,
and
we
will
3
continue
to
develop
our
ideas
for
how
4
that
could
be
done.

5
The
MDL
is
really
a
critical
6
concept
to
environmental
testing
labs,

7
because
it
is
used
not
just
for
tests
8
performed
under
Part
136
but
for
all
9
environmental
analysis
programs,

10
drinking
water,
wastewater,
solid
11
waste.
Even
if
the
particular
agency
12
doesn't
agree
with
the
MDL,
we
still
13
have
to
do
the
MDL,
because
it
is
14
typically
mandated
at
the
State
level.

15
If
the
MDL
revisions
are
16
promulgated,
we
expect
the
revised
17
regulations
will
become
an
immediate
18
requirement
for
most
of
our
work.

19
When
we
take
a
look
at
what
20
really
controls
what
your
detection
21
limit
is,
it
seemed
to
us
that
there
22
were
two
fundamentally
different
types
23
of
tests.
One
type
of
test,
it
is
24
really
the
variation
of
the
blank
that
25
controls
the
detection
limit,
and
that
19
1
would
be
for
a
test
such
as
2
inductively
coupled
plasma
optical
3
emissions
spectroscopy
where
you
always
4
get
a
result.
It
may
be
a
negative
5
result,
but
you
always
get
a
numerical
6
result
in
any
test
for
a
blank
or
for
7
a
sample.

8
On
the
other
hand,
there
are
9
some
tests
where,
most
of
the
time,

10
the
result
that
you
get
for
the
blank
11
is
simply
ND.
These
are
tests
where
12
the
results
are
filtered
through
a
13
threshold
or
through
requirements
to
14
detect
a
group
of
peaks
in
the
same
15
retention
time
window,
qualifier
ions
16
may
be
within
the
same...
within
a
17
given
ion
ratio,
and
in
this
case,
it
18
is
really
not
so
much
the
blank
but
19
more
the
ability
to
qualitatively
20
identify
an
analyte
when
it
is
present
21
that
controls
the
detection
limit.

22
It
is
a
policy
decision
that
23
has
been
made
that
the
MDL
is
going
24
to
be
an
estimate
of
the
measured
25
concentration
at
which
there
is
99
20
1
percent
confidence
that
a
given
2
analyte
is
present
in
a
given
sample
3
matrix.
We
think
that
maybe
it
would
4
be
a
better
policy
decision
to
define
5
the
MDL
as
a
concentration
where
there
6
is
actually
99
percent
confidence
of
7
actually
detecting
a
given
analyte,

8
and
I
will
come
back
to
that.

9
For
all
this
last
20
years
or
10
so
that
we
have
been
doing
MDLs,
we
11
have
been
operating
with
this
theory
12
or
hypothesis
that
this
level
at
which
13
you
do
have
confidence
distinguishing
14
the
measured
result
from
the
blank
is
15
determined
by
multiplying
the
standard
16
deviation
of
seven
or
more
spiked
17
replicate
analyses
by
the
Student
t
18
value.
That
is
the
theory,
and
the
19
next
normal
step
for
a
theory
is
to
20
test
it.

21
Now,
if
you
take
a
look
at
the
22
EPA's
technical
support
document
which
23
is
a
valuable
read,
I
would
say,
for
24
anybody
who
hasn't
seen
it
and
is
25
available
in
the
EPA's
docket,
they
21
1
identified
161
relevant
publications,
2
but
as
far
as
we
could
tell,
none
of
3
these
included
a
really
convincing
4
test
or
demonstration
that
the
MDL
5
procedure
really
worked,
in
other
6
words,
that
it
did
identify
the
7
concentration
where,
for
subsequent
8
routine
analysis,
you
could
distinguish
9
a
measured
result
from
a
blank.
So,
we
10
thought
we
would
try
and
test
it.

11
Bill
mentioned
the
Episode
6000
12
data
set
that
EPA
had
generated.

13
Here,
for
a
number
of
different
14
analytical
techniques,
there
were
seven
15
or
eight
replicates
performed
at
a
16
whole
number
of
different
17
concentrations,
very
useful
data
set.

18
I
will
pick
one
example
out
of
here,

19
Method
200.8
which
is
ICP/
MS
analysis
20
for
chromium,
and
these
are
the
21
results
that
were
obtained
for
this
22
data
set:

23
The
EPA
calculates
their
MDL
as
24
0.07
g/
l,
and
from
that,
calculated
25
an
ML
of
0.2.
If
we
take
a
look
at
22
1
this
data,
we
can
see
that
the
2
recovery
at
low
values
and
including
3
at
the
MDL
is
extremely
high
and
that
4
the
mean
value
that
is
obtained
in
5
this
first
group
of
spiking
6
concentrations
is
all
about
the
same.
7
In
other
words,
what
is
going
on
here
8
is
that
there
is
some
kind
of
bias
9
that
is
causing
an
excessive
positive
10
result
relative
to
the
spiking
level.

11
It
could
be
contamination
in
12
the
blank.
It
could
be
some
kind
of
13
calibration
effect.
We
don't
really
14
know
what
it
is,
but
what
this
bias
15
results
in
is
the
fact
that,
really,

16
if
you
have
a
true
concentration
at
17
0.07
g/
l,
you
don't
have
any
18
confidence
at
all
of
distinguishing
19
this
result
from
a
blank,
and
looking
20
at
this
data,
we
think
the
MDL
really,

21
given
this
data
set,
ought
to
be
about
22
1
or
15
times
higher
than
the
MDL
23
that
is
calculated
using
the
EPA's
24
procedure.

25
Why
is
that
happening?
Well,
23
1
the
MDL
makes
some
assumptions.
One
2
assumption
that
it
makes
is
that
when
3
you
take
a
look
at
the
distribution
of
4
results
from
a
group
of
MDL
5
replicates,
the
distribution
of
the
6
blank
is
going
to
be
the
same
and
7
that
the
distribution
of
the
blank
is
8
going
to
be
centered
around
zero,
and
9
from
that,
you
calculate
the
MDL
here
10
with
a
99
confidence
level
that
you
11
are
going
to
be
able
to
distinguish
a
12
result
from
the
blank.

13
So,
what
happens
if
the
blanks
14
actually
have
positive
bias?
What
15
happens
is
that
the
MDL
stays
exactly
16
where
it
is,
because
what
you
are
17
calculating
it
from,
the
distribution
18
of
the
MDL
replicates,
stays
the
same.

19
MDL
stays
just
where
it
is.
But
now,

20
if
you
look
at
the
distribution
of
the
21
blanks,
it
may
actually
encompass
or
22
even
be
higher
than
the
MDL
value.

23
So,
as
a
result,
now,
if
we
24
look
at
our
blank
results,
this
shaded
25
area
where
we
get
results
above
the
24
1
MDL,
these
are
all
our
false
2
positives.
MDL
really
ought
to
be
3
there.

4
So,
if
we
now
think
about
well,

5
how
could
we
incorporate
this
blank
6
bias
into
the
MDL
equation
such
that
7
we
get
a
more
sensible
result
for
the
8
MDL,
one
way
of
doing
it
is
to
9
include
the
mean
of
the
blanks
in
the
10
calculation,
and
here
we
have
the
mean
11
of
the
blanks
plus
the
standard
12
deviation
times
Student
t.
So,
I
13
think
that
gives
us
a
bit
of
an
14
improvement.

15
The
other
variable
in
that
16
equation
is
the
standard
deviation.

17
So,
another
critical
point
is,
is
the
18
estimate
of
the
true
population
19
standard
deviation
good?
And
there
is
20
some
reason
to
believe
that
it
might
21
be.

22
For
several
years
now,
the
USGS
23
has
been
touting
what
they
call
the
24
long­
term
MDL.
EPA
does
mention
this
25
in
their
technical
support
document,

25
1
and
they
say
that
they
will
go
on
to
2
discuss
the
issues
raised
by
the
long­
3
term
MDL,
but
it
is
not
really
very
4
clearly
discussed
in
that
technical
5
support
document,
unfortunately.
So,

6
the
USGS
long­
term
MDL
involves
a
7
large
number
of
replicates,
an
8
extended
period
of
time,
multiple
9
instruments,
operators,
calibrations,

10
and
sample
preparation
events.

11
We
could
imagine
that
the
12
standard
deviation
generated
from
that
13
would
be
higher
than
a
very
short­
term
14
set
of
measurements.
The
ASTM
IQE
15
makes
some
similar
specifications
that
16
it
is
critical
that
ordinary
within­

17
lab
variation
must
be
allowed
to
18
affect
the
measurement
process,
as
19
happens
in
routine
measurement,
and
20
that,
ideally,
each
measurement
used
21
in
the
calculation
of
the
detection
22
limit
would
involve
a
different
23
analyst,
a
different
day,
random
24
orders,
maximum
number
of
qualified
25
analysts
possible,
maximum
number
of
26
1
qualified
measurement
systems.

2
This
is
in
contrast
to
the
MDL
3
where
the
assumption
is
made,
perhaps
4
incorrectly
sometimes,
that
the
idea
5
is
that
all
the
replicates
should
be
6
done
in
a
short
period
of
time,
and
7
this
first
statement
here
is
from
the
8
technical
support
document.
It
says
9
when
the
MDL
procedure
is
followed
as
10
intended,
in
other
words,
the
MDL
is
11
determined
by
an
experienced
analyst
12
on
each
device
or
instrument,
so,

13
individual
device.
And
the
second
14
statement
comes
from
the
actual
15
Federal
Register
notice.
It
tells
you
16
to
prepare
a
sufficient
quantity
of
17
matrix
to
provide
samples
for
a
18
minimum
of
seven
analyses.
That
19
certainly
implies
that
you
are
going
20
to
do
them
all
at
the
same
time.

21
So,
the
assumption
is
made
that
22
the
Office
of
Water
procedure
23
specifies
that
the
replicates
should
24
be
done
on
one
day
by
one
analyst,

25
one
prep
event,
one
calibration.

27
1
Well,
does
that
matter?
We
2
thought
we
would
try
and
test
that.
3
So,
we
gathered
a
group
of
data.
We
4
have
an
MDL
for
this
set
of
data
5
which
is
a
typical
40
CFR
MDL,
single
6
instrument/
single
analyst
MDL.
We
7
compared
this
to
the
standard
8
deviation
that
we
calculated
from
200
9
routine
method
blanks,
and
we
also
10
compared
it
to
19
method
blanks
that
11
were
performed
in
a
single
batch
on
a
12
single
day.

13
One
example
I
pulled
out
of
14
this
data
set,
the
lead,
the
MDL
was
15
1.63
g/
l.
The
long­
term
blank
16
standard
deviation
was
2.3,
and
the
17
short­
term
blank
standard
deviation
was
18
0.4.
In
other
words,
the
long­
term
19
standard
deviation
was
5.5
times
20
bigger
than
the
short­
term
standard
21
deviation.
When
we
use
our
equation
22
using
the
mean
plus
Student's
t
times
23
the
standard
deviation
of
long­
term
24
blanks,
we
get
a
result
of
5.5.
So,

25
we
get
a
result
here
that
is
quite
28
1
different
from
the
MDL
that
was
2
calculated
by
the
short­
term
procedure.

3
Well,
let's
graphically
take
a
4
look
at
what
is
going
on
here.
Here,

5
we
have
got
our
MDL
replicates
again,
6
relatively
small
standard
deviation.

7
We
calculate
the
MDL
from
these.
It
8
is
down
here,
and
when
we
actually
9
look
at
our
distribution
of
routine
10
blanks,
the
distribution
looks
like
11
this.
It
is
a
much
wider
12
distribution.
The
MDL,
we
think,

13
ought
to
be
out
here,
and
now,
this
14
shaded
area
here
represents
the
false
15
positives.

16
What
does
it
mean
in
the
real
17
world?
Well,
we
take
a
look
at
of
18
those
270
blank
results
we
had
for
19
lead.
10.7
percent
of
them
were
20
greater
than
the
MDL.
If
we
used
our
21
equation
of
Student's
t
times
the
22
standard
deviation
long­
term
blank
plus
23
the
mean
of
the
blanks,
2.6
were
24
greater
than
the
MDL,
and
we
also
25
calculated
using
three
times
the
29
1
standard
deviation
which
has
been
2
touted
in
some
publications,
and
now
3
we
had
0.4
percent
of
the
results
were
4
greater
than
the
MDL.

5
Expanding
that
to
a
bigger
data
6
set,
we
took
six
labs'
worth
of
data
7
for
22
elements
and
did
the
same
8
thing,
and
overall,
we
had
14.2
9
percent
of
the
blank
results
were
10
greater
than
the
MDL.
Of
course,
this
11
means
that
you
are
going
to
get...
so,

12
these
are
false
positives,
and
this
is
13
going
to
be
happening
in
the
samples
14
as
well
as
in
the
blanks.
So,
14.2
15
percent
sounds
pretty
bad,
and
16
remember
that
this
14
percent
is
for
17
each
individual
element.
So,
that
is
18
an
average
of
three
false
positives
19
per
analysis.

20
When
we
used
our
equation
of
21
the
long­
term
standard
deviation
of
22
the
blanks
plus
the
mean,
we
had
1.8
23
percent
of
the
results
were
greater
24
than
the...
TRACK
02
ENDS.

25
TRACK
03...
than
the
short­
term
30
1
standard
deviations.

2
So,
again,
to
look
at
this
3
graphically,
we
have
got
our
MDL
4
replicates
here,
relatively
small
5
standard
deviation.
Our
short­
term
6
blanks
also
have
small
standard
7
deviation.
The
MDL,
using
the
Office
8
of
Water
procedures
calculated
out
9
here,
and
here
is
our
distribution
of
10
the
long­
term
blanks
and
where
we
11
think
the
MDL
should
be.

12
I
want
to
define
a
couple
of
13
terms
here.
The
LC
and
LD
are
terms
14
that
were
originally
defined
by
Curry
15
back
many
years
ago
before
I
even
16
started
to
think
about
detection
17
limits
or
even
environmental
chemistry,

18
but
they
are
internationally
accepted
19
terms
now
for
the
critical
level
and
20
the
detection
limit
or
internationally
21
accepted
maybe
excepting
the
United
22
States.

23
The
critical
level
is
set
at
24
the
standard
deviation
of
blank
times
25
a
constant
related
to
the
confidence
31
1
level
desired,
and
the
LC
is
exactly
2
what
the
current
MDL
procedure
3
targets.
LD
described
the
level
at
4
which
you
have
got
detection
5
capability,
and
LD
would
be
the
level
6
at
which
you
have
got
95
or
99
7
percent
confidence
of
actually
8
detecting
a
true
value
that
was
in
a
9
sample.

10
We
can
derive
both
LC
and
LD
11
from
these
long­
term
blank
12
measurements,
making
the
assumption
of
13
constant
variance
across
this
range.

14
LD
is
simply
two
times
higher
than
LC.

15
Let's
take
a
look
at
what
this
16
looks
like
in
a
distribution.
If
we
17
have
a
true
concentration
at
LD,
then
18
our
distribution
is
comfortably
19
different
from
the
distribution
of
the
20
blanks.
In
other
words,
if
we
have
a
21
true
concentration
at
LD,
then
our
22
result
is
something
that
we
are
going
23
to
have
confidence
that
we
can
24
distinguish
from
a
blank,
so
we
can
25
say
it
is
a
true
result.

32
1
If
we
look
at
the
distribution
2
of
a
true
value
at
LC,
then
here,
if
3
the
value
is...
50
percent
of
the
time,

4
you
are
going
to
get
a
result
that
is
5
greater
than
our
distribution
of
the
6
blanks.
So,
at
LC,
only
50
percent
7
of
the
time
are
we
going
to
be
able
8
to
say
that
we
have
a
true
detection.

9
We
would
recommend
using
LD
as
10
the
detection
limit
or
the
MDL
rather
11
than
LC
but
reporting
data
down
as
low
12
as
LC.
This
means
now
that
the
13
detection
limit
would
become
what
most
14
people
actually
think
of
the
detection
15
limit
as,
the
lowest
limit
of
reliable
16
detection.
It
is
quite
difficult
to
17
explain
to
people
that
your
method
is
18
not
necessarily
able
to
detect
19
something
at
the
method
detection
20
limit.
That
just
gets
people
very
21
confused.

22
EPA
covers
this
concept
a
23
little
bit
in
the
technical
support
24
document.
The
one
objection
that
is
25
raised
is
that
if
LD
is
used
as
a
33
1
censoring
limit,
then
50
percent
of
2
the
results
of
the
true
value
of
LD
3
would
be
censored,
and
that
is
why
it
4
is
important
that
while
we
would
5
report
LD
as
our
detection
limit,
we
6
would
continue
to
report
results
down
7
as
low
as
LC
so
that
we
will
get
the
8
best
of
both
worlds,
essentially.
We
9
wouldn't
be
censoring
results
that
we
10
know
are
really
true.
Yet,
we
would
11
be
informing
the
data
user
what
the
12
true
detection
capability
of
our
13
method
was
so
that
a
correct
decision
14
could
be
made
about
whether
this
15
method
was
suitable
for
the
data
use
16
or
not.

17
Unfortunately,
things
are
never
18
quite
that
simple,
and
as
I
mentioned
19
early
on,
some
methods,
there
is
a
20
lack
of
instrument
response
at
21
the...
in
the
blanks
as
a
result
of
22
threshold
or
method
required
analyte
23
identification
criteria.
Therefore,

24
for
these
methods,
there
is
no
way
you
25
can
use
the
standard
deviation
of
the
34
1
blanks
to
evaluate
the
MDL.

2
Looking
at
this
graphically
3
again
here,
distribution
of
the
blanks
4
is
everything
is
as
zero.
Everything
5
is
non­
detect
for
the
blanks.
But
if
6
we
have
our
set
of
spikes
at
LD,
the
7
detection
limit,
then
we
would
have
a
8
distribution
as
shown
here.
And
we
9
would
calculate
an
MDL
from
this
10
distribution
of
blanks
at
a
level
that
11
there
may
indeed
be
very
little
12
confidence
that
you
can
detect
13
anything.

14
We
went
back
to
EPA's
Episode
15
6000
data
set
and
had
a
look
at
16
Method
524.2,
a
GC/
MS
method,
and
we
17
see
here
that
72
out
of
the
81
18
analytes
have
7
replicates
that
are
19
all
non­
detect
at
a
concentration
at
20
or
above
the
MDL,
and
some
of
the
21
analytes
have
7
replicates
that
are
22
all
non­
detect
at
more
than
3.18
times
23
the
MDL,
in
other
words,
all
non­

24
detects
at
a
level
which
would
25
normally
be
described
as
the
35
1
quantitation
limit.
So,
obviously,

2
that
is
not
making
a
lot
of
sense
3
here.

4
I
think
that,
for
these
tests
5
where
there
is
filtering
the
MDL
as
it
6
is
currently
determined
is
a
pretty
7
meaningless
number.
It
is
not
the
8
level
at
which
you
have
got
confidence
9
in
distinguishing
something
from
a
10
blank.
It
is
not
the
level
at
which
11
you
have
got
confidence
that
you
can
12
actually
detect
something.
It
really
13
doesn't
appear
to
be
a
number
that
14
really
means
anything.

15
So,
how
do
we
discover
what
the
16
detection
limit
is
for
these
types
of
17
tests
that
have
got
thresholds
and
18
filtering?
We
believe
a
statistical
19
approach
is
going
to
be
extremely
20
difficult
or
is
not
going
to
work
well
21
because
of
the
very
varied
22
requirements
for
analyte
detection.

23
First
of
all,
you
have
got
to
24
detect
a
chromatographic
peak.
Then,

25
you
have
got
to
consider
you
might
36
1
need
to
have
co­
elution
of
qualifier
2
ions.
You
might
need
to
meet
certain
3
ion
ratios.
You
might
have
to
do
4
pattern
recognition.
A
human
might
5
have
to
do
pattern
recognition
of
6
maybe
an
aroclor
pattern
or
a
mass
7
spectrum,
and
to
do
a
statistical
8
analysis
of
this
becomes
almost
9
impossible.

10
So,
we
considered
an
empirical
11
approach,
and
this
type
of
empirical
12
approach
is
included
in
the
EPA's
13
proposed
MDL
procedure
as
a
technique
14
for
making
an
initial
estimate
of
the
15
MDL.
We
think
that
the
initial
16
estimate
does
a
far
better
job
than
17
the
statistical
calculation
that
would
18
follow
it.

19
So,
essentially,
you
simply
20
spike
at
increasingly
lower
21
concentrations
until
a
detection
22
limit...
until
the
detection
fails,

23
until
you
fail
to
detect
the
compound.

24
Very
important
here
that
routine
25
conditions
for
detection
are
used.

37
1
You
shouldn't
just
being
trying
to
2
pull
something
out
of
the
noise
that
3
you
wouldn't
try
to
pull
out
of
the
4
noise
in
a
regular
sample.

5
Once
you
had
determined
the
6
lowest
limit
that
you
could
truly
7
detect
something,
you
would
repeat
8
this
spike
of
this
concentration
in
9
several
batches
to
verify
and
then
10
continue
those
verifications
on
a
11
quarterly
basis.

12
Again,
we
don't
think
that
the
13
data
should
be
censored
at
this
lowest
14
limit
of
detection
that
we
have
15
determined.
Otherwise,
we
would
throw
16
away
at
least
50
percent,
even
17
assuming
100
percent
recovery,
or
18
maybe
even
more
than
50
percent
if
we
19
don't
have
100
percent
recovery
of
our
20
data
if
we
censored
at
LD.

21
For
these
tests,
again,
because
22
we
don't
have
any
standard
deviation
23
of
the
blank,
LC
is
a
difficult
number
24
to
really
properly
determine
or
even
25
define,
but
we
think
right
now
a
38
1
reasonable
lowest
limit
to
report
2
consistent
with
the
current
procedures
3
would
be
to
calculate
LC
the
way
that
4
the
current
MDL
is
calculated,
and
we
5
are
open
for
other
suggestions
for
the
6
lowest
limit
to
report
the
censoring
7
limit.

8
So,
again,
now
to
go
on
and
9
discuss
how
this
all
relates
to
the
10
minimum
level
or
the
quantitation
11
limit,
EPA
provides
three
varied
12
definitions
of
the
minimum
level.
One
13
is
the
lowest
level
at
which
the
14
entire
signal
gives
a
15
recognizable...
entire
system
gives
a
16
recognizable
signal.
Another
17
definition
is
the
lowest
acceptable
18
calibration
point,
and
a
third
19
definition
is
10
times
the
standard
20
deviation
of
the
MDL
replicates.
We
21
think
these
three
definitions
are
kind
22
of
difficult,
because
they
would
all
23
tend
to
lead
you
to
a
different
24
number,
and
they
are
supposed
to
be
25
defining
the
same
thing.
39
1
We
also
think
that
if
the
MDL
2
were
targeted
at
LD
and
it
were
3
determined
with
full
consideration
of
4
the
routine
variability
and
blank
5
bias,
then
two
or
three
times
the
MDL
6
would
be
a
reasonable
quantitation
7
estimate.
Again,
if
a
quantitation
8
limit
is
going
to
be
set
for
a
9
nationwide
method
to
be
proposed
at
10
Part
136,
then
a
multi­
lab
study
11
should
be
performed
to
define
the
12
precision
and
accuracy
at
the
ML.

13
We
don't
think
that
ten
times
14
the
standard
deviation
should
be
used,

15
because
that
could
be
below
the
level
16
at
which
there
is
confidence
17
distinguishing
a
result
from
a
blank,

18
as
we
have
seen
in
some
of
the
EPA
19
6000
data.
We
shouldn't
use
a
20
calibration
point
alone,
although
a
21
calibration
point
should
be
included
22
at
the
ML,
and
that
is
because
of
23
weaknesses
in
the
criteria
for
an
24
acceptable
calibration
which
is
25
typically
based
on
the
correlation
40
1
coefficient.
Because
of
weaknesses
in
2
correlation
coefficient
concept,
you
3
can
essentially
have
your
low
4
calibration
point
wandering
around
all
5
over
the
place
and
still
get
a
very
6
good
correlation
coefficient.
You
can
7
certainly
get
.999
correlation
8
coefficient
even
if
you
get
no
result
9
at
all
for
your
low
calibration
point.

10
So,
to
summarize
our
position,

11
we
believe
that
routine
lab
12
variability
must
be
included
in
any
13
detection
limit
procedure.
A
sample
14
coming
into
a
lab
may
be
analyzed
15
randomly
on
different
qualified
16
instruments
by
different
analysts,

17
different
batches,
and
different
18
calibrations.
If
you
don't
include
19
those
sources
of
variability
in
your
20
detection
limit
estimate,
then
there
21
is
no
hope
of
getting
an
accurate
22
assessment
of
the
detection
limit.

23
We
think
the
detection
limit
24
should
target
Curry's
limit
of
25
detection,
but
we
would
avoid
41
1
excessive
false
negatives
at
a
true
2
concentration
of
LD
by
reporting
3
results
down
to
the
calculated
value
4
for
LC,
the
critical
value.
And
we
5
believe
that
there
is
not
a
single
6
procedure
that
will
provide
a
good
7
detection
limit
estimate
for
both
8
unfiltered
tests
such
as
ICP
and
9
highly
filtered
tests
such
as
GC/
MS,

10
but
we
are
proposing
two
11
straightforward
methods
for
determining
12
detection
limits
for
all
types
of
13
tests.

14
Further
information
is
available
15
on
our
website
that
we
have
16
established
where
you
can
also
visit
17
and
make
comments
that
would
be
18
directed
back
to
the
ACIL
Technical
19
Committee
on
our
procedure.
That
is
20
at
acil­
mdl.
org,
or
you
can
call
or
21
email
me
directly.

22
Thank
you.

23
MR.
TELLIARD:
Thank
24
you,
Richard.
Our
next
speaker
is
25
Nancy
Grams
from
Advanced
Earth
42
1
Technologies.
2
MS.
GRAMS:
I
will
go
3
ahead
and
start.
I
am
here
today
4
representing
the
task
group
of
ASTM.

5
It
develops...
can
you
hear
me?

6
SPEAKER:
We
can't
hear
7
you.

8
MS.
GRAMS:
All
right.

9
There
is
a
handout
that
is
going
10
around.
Naomi
passed
a
few
of
them
11
out.
Unfortunately,
I
didn't
make
12
enough
copies.
I
didn't
expect
to
see
13
so
many
faces
here
today
of
so
many
14
people
so
interested
in
the
MDL.

15
Hopefully,
most
of
you
can
share
them.

16
Many
of
you
who
are
members
of
my
17
task
group
have
seen
these
materials
18
and
don't
really
need
a
copy.

19
Additionally,
my
comments
today,

20
my
written
comments
as
well
as
my
oral
21
comments,
will
be
a
part
of
the
22
record,
making
them
a
part
of
the
23
record,
and
those
materials
will
be
in
24
the
written
materials.
So,
if
you
25
don't
get
them
today,
you
can
go
to
43
1
the
docket
and
find
them.

2
Also,
ACIL
has
made
an
offer
to
3
allow
people
to
post
information
to
4
their
website
relevant
to
these
5
comments,
and
I
will
try
to
make
sure
6
that
the
materials
get
posted.

7
So,
to
repeat,
I
am
here
today
8
representing
the
task
group
that
9
developed
the
IDE
and
the
IQE.

10
In
the
Office
of
Water's
MDL
11
assessment
effort,
of
all
the
myriad
12
of
detection
and
quantitation
terms
13
that
Bill
mentioned
in
his
opening
14
remarks,
the
IDE
and
the
IQE
were
two
15
concepts
that
were
carried
through
the
16
entire
assessment
process
and
were
17
critically
evaluated
at
the
very
end
18
against
the
MDL
and
the
ML
using
the
19
six
criteria
that
Bill
outlined
20
earlier.

21
I
fear
that
for
many
of
the
22
people
here
today
who
are
not
regular
23
participants
in
ASTM,
you
never
heard
24
of
the
IDE
and
the
IQE
before,
so
25
what
I
wanted
to
do
today
was
to
try
44
1
to
acquaint
you
folks
both
with
ASTM,

2
with
the
history
of
the
standards,
the
3
IDE
and
the
IQE,
and
to
introduce
you
4
just
a
little
bit
to
the
standards.

5
I
am
probably
not
going
to
have
a
6
lot
of
time
to
do
that,
but
I
did
put
7
some
materials
in
the
written
comments
8
that
give
you
a
little
bit
more
depth
9
about
the
IDE
and
the
IQE
in
terms
of
10
the
technical
content,
and
I
put
in
11
some
references
to
some
documents
that
12
I
think
are
also
very
valuable
in
13
understanding
the
technical
side
of
14
these
issues.

15
So,
after
that,
I
guess
my
16
other
purpose
here
today
is
to
try
to
17
persuade
you
of
the
merits
of
the
IDE
18
and
the
IQE
and
of
the
merits
of
19
EPA's
Office
of
Water
adopting
these
20
procedures.

21
I
just
want
to
mention
in
22
passing
the
Technology
Transfer
Act
in
23
case
many
of
you
have
not
heard
of
24
this.
This
Act
requires
Federal
25
agencies
such
as
the
EPA
to
first
45
1
consider
the
use
of
public
consensus
2
processes
and
standards
before
3
unilaterally,
and
at
taxpayer
expense,

4
going
it
alone.
There
are
many
5
examples
of
Government
agencies
working
6
with
ASTM,
in
particular,
to
develop
7
standards
and
the
agencies
adopting
8
them.

9
And
I
guess
one
of
the
first
10
things...
the
first
point
I
would
like
11
to
make
is
that
I
think
that
we
all
12
should
consider
the
Technology
Transfer
13
Act
as
a
positive
thing
for
us
as
14
taxpayers,
and
I
believe
the
ASTM
15
process
has
a
number
of
merits
that
16
unilateral
development
by
the
EPA
does
17
not
have
for
standards,
and
I
just
18
want
to
start
by
giving
you
some
19
background
to
give
you
my
reasoning
20
for
this.

21
The
American
Society
for
22
Testing
Materials
has
recently
been
23
renamed
ASTM
International,
and
this
24
organization
has
been
around
a
long
25
time.
There
are
30,000
volunteers.
I
46
1
think
a
lot
of
people
don't
recognize
2
how
big
an
organization
and
how
3
important
an
organization
ASTM
is.

4
It
has
130
technical
5
committees.
We
have
face­
to­
face
6
meetings.
Unlike
a
lot
of
standards
7
development
organizations,
we
actually
8
get
together,
and
we
fight
it
out.

9
Don't
we,
Bill?
There
is
an
extensive
10
tiered
balloting
system
so
that
there
11
is
really
protection
in
the
balloting
12
system,
and
there
is
strict
13
adjudication
of
negative
votes.
The
14
negative
voter
in
ASTM
has
a
very
15
strong
position.

16
And,
as
I
said,
one
of
the
17
merits
of
the
ASTM
standards
is
that
18
they
meet
the
criteria
for
use
under
19
the
Technology
Transfer
Act.

20
I
want
to
talk
just
a
little
21
bit
about
the
committee
itself.
There
22
are
a
lot
of
different
committees,
but
23
the
ASTM
committee
that
I
participate
24
in
primarily
is
D19,
and
this
25
committee
talks
about
water.
We
do
47
1
everything
about
water.
We
measure
2
water,
we
measure
the
stuff
in
water,

3
we
do
all
sorts
of
stuff
with
water.

4
This
committee
was
established
in
5
1932,
and
I
want
you
to
think
about
6
this
for
a
minute.
This
committee
is
7
70
years
old.
Long
before
anybody
was
8
even
thinking
about
an
EPA,
this
9
committee
was
developing
water
10
standards.

11
We
have
300
current
standards,

12
and
one
of
the
things
about
ASTM
is
13
that
all
of
these
standards
have
to
be
14
reassessed
at
least
every
five
years.

15
So,
any
standards
that
are
outdated
16
are
removed
if
they
are
not
being
used
17
anymore,
and
we
update
the
information
18
that
is
in
them
so
they
are
not
19
stagnant.

20
We
have
400
active
members
on
21
the
committee,
and
in
ACM,
you
have
to
22
be
active.
If
you
don't
vote,
you
23
are
thrown
out.

24
The
committee
is
diverse
in
25
representation.
Agencies
are
48
1
represented
there.
For
many,
many
2
years,
EMSL­
Cincinnati,
when
they
were
3
doing
methods
development
for
EPA,
was
4
at
the
committee,
and
more
recently,

5
Bill
Telliard's
group
at
the
Office
of
6
Water
has
been
participating
as
has
7
Bill
himself.

8
There
are
many
different
9
persons
from
industry
and
consultants
10
to
industry
who
participate
in
the
11
committee,
and
this
includes
the
12
litigants
in
the
matter
of
the
MDL.

13
We
also
have
good
representation
by
14
instrument
manufacturers
and
by
15
laboratories.
So,
we
are
really
16
diverse
and
representative
in
the
17
environmental
industry.

18
I
just
want
to
point
out
that
19
ASTM
D19,
both
within
ASTM
and
within
20
our
industry,
was
the
leader
in
21
requiring
interlaboratory
method
22
validation.
Now,
there
is
23
interlaboratory,
and
there
is
24
intralaboratory,
and
it
is
going
to
be
25
very
important
for
everyone
today
to
49
1
keep
those
two
concepts
separate.

2
This
is
among
laboratories.

3
And
it
was
also
very
early
on
4
in
providing
precision
and
bias
5
statements
within
all
of
its
methods.
6
If
you
go
to
the
ASTM
methods
under
7
D19,
you
will
see
that
every
method
8
there
has
a
precision
and
bias
9
statement.
Many
of
them
are
in
10
individual
matrices
of
water
such
as
11
groundwater
and
waste
water,
et
12
cetera,
et
cetera.

13
We
have
a
standard
that
we
use
14
in
order
to
determine
the
precision
15
and
bias
and
to
make
these
precision
16
and
bias
statements.
This
standard
17
was
first
approved
in
1969,
and
I
want
18
you
to
think
about
this
as
well.

19
Maybe
EPA
was
a
twinkle
in
Uncle
Sam's
20
eye
at
the
time.

21
The
committee
recognized
the
22
importance
and
the
responsibility,
as
23
method
developers
themselves,
for
24
validating
the
methods
they
put
out
25
there.
The
standard
D2777
has
often
50
1
been
revised.
It
is
revised
to,
you
2
know,
record
the
accumulated
wisdom,

3
and
it
is
revised
based
on
the
state
4
of
the
science.
There
are
many
things
5
that
have
changed
over
the
years,

6
statistics,
environmental
chemistry,

7
lots
of
changes.

8
It
specifies
that
for
every
9
method
that
our
committee
produces,

10
testing
method,
we
will
do
an
11
interlaboratory
method
validation
12
study.
This
is
what
is
commonly
13
called
in
the
committee
a
round
robin
14
study.
It
is
multi­
lab,
multi­

15
concentration.

16
D2777
defines
the
working
range
17
of
the
method.
In
our
methods
at
our
18
committee,
when
you
write
the
method,

19
you
have
to
state
what
the
working
20
range
of
the
method
is,
and
you
cannot
21
extrapolate
outside
of
your
22
interlaboratory
study.

23
Precision
and
bias
statements
24
are
generally
equations
that
are
25
developed
by
matrix
within
the
water
51
1
matrices.
The
studies,
the
analyses
2
of
the
study
data,
and
the
precision
3
and
bias
statements
are
all
reviewed
4
and
must
be
approved
within
the
5
committee,
and
the
information
is
also
6
used
to
establish
ongoing
CQ
7
parameters.

8
What
the
committee
was
really
9
thinking
about
in
doing
and
generating
10
these
IDE
and
IQE
standards
was
to
11
extend
the
range
of
the
validation,

12
and
I
think
you
need
to
think
about
13
this
for
a
minute,
because
it
may
not
14
be
totally
intuitive.
This
industry
15
is
going
to
be
moving
toward
16
measurement
uncertainty
and
reporting
17
measurement
uncertainty.
All
the
IDE
18
and
IQE
do
for
our
committee
is
extend
19
from
the
general
working
range
to
20
probe
down
and
try
to
figure
out,
on
21
an
interlaboratory
basis
for
a
method,

22
where
the
limits
of
detection
and
23
where
the
limits
of
quantitation
24
really
are.

25
In
discussing
the
IDE
and
IQE,
52
1
I
think
it
is
important
for
me
to
2
talk
about
the
history
at
D19
of
3
detection
and
quantitation,
about
the
4
task
group,
and
the
participation
of
5
EPA
and
the
litigants.
In
the
late
6
1980s,
committee
D19
attempted
to
7
develop
a
single
standard,
one
8
standard,
to
address
all
the
issues
of
9
detection
and
quantitation,
both
single
10
lab
and
multiple
lab,
intralab
and
11
interlab.

12
After
many
drafts
and
rounds
of
13
balloting,
the
committee
was
totally
14
frustrated.
We
couldn't
get
anywhere.

15
In
1990,
I
volunteered
to
head
16
an
effort
to
try
again.
I
set
out
to
17
bring
interested
parties
together
and
18
to
foster
communication
and
to
clarify
19
both
the
science
and
the
issues.

20
Instead
of
just
trying
to
draft
a
21
standard
which
hadn't
worked
before...

22
BEGIN
DISC
2,
TRACK
23
01...
groups.
We
do
meet
twice
yearly.

24
As
I
said,
ASTM
does
go
face­
to­
face.

25
We
had
joint
meetings
both
with
53
1
the
quality
group
which
is
the
group
2
in
which
I
participate,
that
3
subcommittee,
and
organic
and
inorganic
4
subcommittees.
We
had
40­
plus
members
5
who
attended
each
and
every
meeting.

6
At
many
of
our
meetings,
we
had
80­

7
plus
people.

8
My
task
group
consists
of
200
9
people.
By
the
time,
which
was
ten
10
years,
we
got
the
IDE
and
the
IQE
11
standards
done,
we
had
invested
12
thousands
of
hours,
and
these
are
all
13
highly
qualified
professionals
who
had
14
been
involved.
It
is
an
enormous
15
investment
both
by
agencies
and
by
16
private
individuals
and
corporations.

17
Now,
this
is
the
next
point
I
18
would
like
to
make.
It
is
extremely
19
unlikely
that,
in
our
lifetimes,
this
20
level
of
effort
will
ever
be
repeated
21
again.

22
And
what
are
our
work
products?

23
As
I
mentioned
before,
rather
than
24
just
writing
a
standard,
I
think
our
25
most
important
work
product
was
54
1
communication,
and
it
continues
to
be.

2
In
our
meetings,
we
have
people
from
3
USGS,
we
have
people
from
the
Office
4
of
Water,
we
have
people
from
other
5
Federal
agencies,
we
have
States,
and
6
we
have
people
from
the
industry
side,

7
and
we
all
get
together,
and
we
8
communicate.

9
My
task
group
did
another
thing
10
that
was
quite
different.
We
did
a
11
lot
of
education.
We
had
a
number
of
12
tutorials,
because
we
found
that
13
people
really
didn't
understand
some
14
of
the
underlying
concepts
associated
15
with
detection
and
quantitation
and
16
the
needs,
especially
statistical
17
needs,
in
developing
an
appropriate
18
procedure.

19
We
also
developed
software.

20
You
know,
one
of
the
criticisms
of
the
21
more
complex
procedures
is
that
people
22
can't
do
them
themselves,
and
even
I
23
can
use
the
software.
It
is
very
24
simple.

25
I
just
want
to
note
for
anybody
55
1
who
is
seriously
interested
in
2
detection
or
quantitation
and/
or
coming
3
to
better
understand
the
ASTM,
in
4
June,
we
are
meeting
in
Louisville,

5
and
the
task
group
meets
on
Tuesday,

6
and
then
we
have
a
half­
day
seminar
on
7
Wednesday
where
we
are
really
going
to
8
go
into
a
lot
more
depth,
and
Bill
9
and
his
folks
at
the
Office
of
Water
10
are
going
to
be
there,
and
we
are
11
going
to
have
a
good
time.

12
So,
now
I
need
to
look
at
how
13
much
time
I
have,
and
I
am
running
a
14
little
behind,
so
I
am
just
going
to
15
kind
of
rush
ahead
here.

16
As
I
mentioned
before,
ASTM
has
17
this
D2777
standard,
and
we
actually
18
piggybacked
the
IDE
and
the
IQE
on
it,

19
and
I
kind
of
mentioned
that
before.

20
Again,
as
I
said,
in
the
written
21
materials,
there
is
a
lot
more
depth
22
of
information
on
what
the
interlab
23
detection
estimate
is,
so
I
am
just
24
going
to
run
through
this.

25
It
is
based
on
Curry's
theory.

56
1
Rather
than
using
a
single
2
concentration,
though,
it
uses
a
3
multiple
concentration
study
design.

4
We
determine
the
critical
level
and
a
5
detection
level
which
is
the
IDE.
We
6
regress
the
measured
concentration
7
versus
true
concentration.
That
is
8
what
is
called
the
calibration
design.

9
We
use
true
interlaboratory
data.
We
10
don't
just
pool
single
operator
data.

11
And
we
use
weighted
least
squares
to
12
take
care
of
the
difference
in
the
13
magnitudes,
the
concentration
issues.

14
Here
is
one
of
the
things
we
15
had
a
lot
of
discussion
on
and
a
16
tutorial
on,
this
thing
called
17
confidence
bounds.
You
know,
I
18
haven't
used
the
word
confidence
19
limit.
Modeled
around
the
regression.

20
So,
you
are
just
kind
of
figuring
out,

21
you
know,
where
most
of
your
data
is
22
going
to
be,
whether
it
is,
you
know,

23
the
99
percent
control
limit
or
the
24
95,
and
we
use
both,
one
to
control
25
false
positives,
the
other,
negatives.

57
1
We
use
statistical
tolerance
2
intervals,
and
in
my
next
slide,
I
am
3
going
to
start
talking
about
this
a
4
little
bit.
There
are
other
5
intervals,
and
EPA
has
used
confidence
6
intervals.
In
the
standard
in
the
7
software,
we
have
used
a
tiered
8
approach
to
modeling
standard
9
deviation.
If
I
have
time,
I
will
go
10
into
that
just
a
little
bit.

11
Those
of
you
who
have
been
12
doing
MDLs
have
thought
you
have
been
13
doing
them
to
say
something
about
the
14
next
or
any
future
detection
decision
15
you
may
make,
but
when
you
are
doing
16
an
MDL,
that
is
not
actually
what
you
17
are
doing.
When
you
are
using
a
18
confidence
interval,
you
are
developing
19
an
understanding
of
the
specified
20
confidence
you
have
in
the
mean
of
the
21
population.
So,
if
you
went
back
and,

22
in
every
sample
you
ever
analyzed,
you
23
analyzed
it
seven
times
and
took
the
24
mean,
then
your
MDL
might
mean
25
something.

58
1
Another
interval
that
is
2
frequently
used
is
the
prediction
3
interval,
and
this
is
getting
there,

4
because
it
says
with
a
specified
5
degree
of
confidence,
the
next
6
measurement
that
you
will
measure
will
7
be
within
a
calculated
interval,
and
8
many
people
use
the
prediction
9
interval.

10
We
use
the
tolerance
interval,

11
and,
you
know,
there
was
a
big
debate,

12
and
we
finally
voted
on
it,
and
we
13
decided
to
use
the
tolerance
interval.

14
It
was
sort
of
the
most
scientifically
15
sound
approach
from
the
majority
of
16
the
task
group's
point
of
view.
We
17
used
it
to
quantify
the
uncertainty
in
18
a
proportion
of
future
measurements.
19
The
interlaboratory
quantitation
20
estimate
was
our
second
standard.

21
This
was
finished
in
2000.
Same
22
calibration
regression,
same
modeling,

23
and
I
will
just
talk
about
the
24
modeling
for
a
second
here.

25
Our
standard...
and
we
debated
59
1
this
at
length
as
well...
requires
that
2
you
go
from
a
very
simple
model
for
3
standard
deviation
to
a
more
complex
4
on.
So,
we
start
with
assuming
that
5
the
standard
deviation
is
constant,

6
then
we
assume
the
standard
deviation
7
is
linear,
then
we
assume
that
it
is
8
hybrid
which
is
the
same
as
the
Rocke
9
and
Lorizano
model,
for
any
of
you
who
10
are
interested,
and
only
then,
if
none
11
of
those
fit
in
ascending
order,
then
12
we
allow
the
use
of
exponential.
And
13
these
are
all
in
the
software.
The
14
software
does
them
all.

15
Based
on
a
quantification
16
defined
as
known
and
control
17
precision,
this
is
another
one
you
18
have
heard
today
in
the
prior
talks,

19
that
you
can
take
your
detection
limit
20
and
multiply
it
by
a
factor
and
get
21
your
quantitation
limit.
We
spent
22
many
an
hour
discussing
this
in
our
23
task
group,
and
the
basic
consensus
24
was
that
quantitation
is
a
totally
25
different
concept,
and
while
you
may
60
1
get
by
with
making
some
simplifying
2
assumptions
to
use
a
factor,
that
it
3
is
really
not
the
scientific
way
to
go
4
about
it,
and
the
scientific
approach
5
to
quantitation
is
that
you
must
6
control
precision.

7
In
the
references
I
give
you,

8
David
Coleman
has
an
article,
and
it
9
relates
detection
and
quantitation
to
10
significant
digits
and
to
relative
11
percent
difference.
I
strongly
12
recommend
that
anybody
who
is
13
seriously
interested
in
this
topic
14
read
David's
article.

15
We
use
10
percent
RSD
as
our
16
default
in
our
software,
because
that
17
is
pretty
well
known
throughout
the
18
sciences,
but
the
software
allows
19
other
DQOs.

20
I
will
just
summarize
what
I
21
have
said
so
far.
The
Office
of
22
Water's
extensive
evaluation
identified
23
ASTM
IDE
and
IQE
as
the
only
fully
24
developed,
strongly
founded
procedures
25
worthy
of
detailed
assessment.
I
have
61
1
sought
today
to
give
you
an
2
understanding
of
the
stature
of
ASTM
3
and
of
D19
and
of
the
level
playing
4
field
of
consensus
established
at
5
ASTM,
and
I
have
tried
to
give
you
6
some
understanding
of
the
standards
7
that
have
been
developed
in
our
task
8
group
and
to
fill
you
in
on
the
9
history
of
the
development
of
these
10
standards
along
with
the
participation
11
of
EPA,
others,
and
the
litigants
in
12
our
efforts.

13
I
have
tried
to
portray
the
14
level
of
effort
that
went
into
the
15
development
of
these
standards,
the
16
significance
of
these
standards
in
17
that
it
is
very
unlikely
that
a
group
18
will
be
brought
together
to
do
this
19
ever
again.
I
have
tried
to
show
you
20
that
it
was
developed
by
a
large,

21
diverse
group
of
credentialed
22
professionals
with
a
long­
term
vision
23
for
our
committee
and
with
a
24
commitment
to
high­
quality
water
25
testing
validation.

62
1
Our
committee
is
working
to
2
utilize
the
IDE
and
the
IQE
and
the
3
software
itself,
and
we
are
committed
4
to
improving
these
standards
and
our
5
software
over
time.
I
also
hoped
to
6
show
today
that
the
IDE
and
the
IQE
7
are
the
only
viable
candidates
under
8
the
Technology
Transfer
Act
for
EPA's
9
Office
of
Water
adoption.

10
Most
importantly,
I
have
tried
11
to
highlight
the
actual
issues
that
12
are
here
in
the
litigation
today,

13
interlab/
intralab,
intralab/
interlab,

14
because
it
is
the
litigants
in
this
15
matter
who
are
concerned
about
16
potential
interlaboratory
uses
of
the
17
MDL
procedure.
Most
everybody
else
18
who
is
going
to
comment
on
this
is
19
going
to
be
interested
in
20
intralaboratory.

21
The
Office
of
Water
has
to
deal
22
with
both
realities
and
has
proposed
23
to
use
a
single
procedure
for
both.

24
Under
the
Technology
Transfer
Act,
the
25
office
could,
instead,
choose
to
adopt
63
1
the
IDE
and
the
IQE
specifically
and
2
just
for
the
interlaboratory
purpose.

3
This
way,
they
would
issue
high
4
quality
validation
of
methods,
they
5
would
satisfy
the
litigants
in
the
6
matter,
they
would
produce
detection
7
and
quantitation
estimates
with
a
sound
8
technical
basis,
and
it
would
allow
9
EPA
to
harness
the
ASTM
D19
process,

10
its
resources,
and
its
volunteer
11
laboratories.
It
would
also
facilitate
12
maintenance
and
update
of
the
13
procedures,
as
ASTM
does
update
their
14
procedures
which
the
CFR
discourages.

15
I
believe
it
is
extremely
16
important
that
when
you,
you
yourselves
17
here
today,
develop
your
comments
that
18
you
recognize
and
differentiate
and
19
critically
evaluate
the
interlaboratory
20
uses
of
the
MDL.
Should
the
MDL
21
procedure,
as
it
is
now
written,
be
22
used
by
the
Office
of
Water
to
23
establish
minimum
sensitivity
24
at
an
interlaboratory
level
when
it
25
approves
methods,
or
should
the
Office
64
1
of
Water
use
the
Technology
Transfer
2
Act
and
the
IDE
and
the
IQE?

3
Is
the
new
MDL
procedure
4
detailed
enough?
Does
it
describe
5
exactly
how
EPA
will
do
its
national
6
level
studies
to
determine
the
7
sensitivity
of
methods
at
a
national
8
level?
This
is
not
the
9
intralaboratory
stuff
that
Richard
was
10
talking
about.
It
is
the
1
percent
of
11
MDL
studies
that
are
ever
done
which
12
are
the
published
MDLs
that
are
13
published
with
the
methods.

14
Is
the
MDL
procedure
the
right
15
balance
of
science
and
defensibility
16
against
practicality?
Is
the
proposed
17
procedure
detailed
enough
to
ensure
18
controlled
use?
Will
the
MDL
19
procedure,
using
the
new
MDL
procedure,

20
be
a
good
estimate?
Is
it
really
a
21
good
enough
estimate?

22
Will
it
be
a
sound
basis...
and
23
here
is
the
real
crux...
for
setting
24
compliance
limits
where
water­
quality
25
based
effluent
limits
are
lower
than
65
1
the
capabilities
of
the
method?
2
In
closing
and
as
you
might
3
guess,
I
would
like
to
challenge
EPA
4
to
consider
the
IDE
and
the
IQE,
and
I
5
would
like
to
challenge
them
as
to
6
their
assumptions
that
were
made
about
7
their
appropriate
criteria.
In
the
8
materials
I
passed
out,
I
have
9
provided...
and
also
in
the
written
10
docket
will
be...
an
alternative
11
assessment.
I
used
their
criteria.
I
12
assess
their
criteria,
and
I
assess
13
the
IDE,
the
IQE,
and
I
assess
the
MDL
14
against
those
criteria.

15
As
time
is
limited,
I
wanted
to
16
just
focus
real
quickly
on
two
17
criteria.
The
Office
of
Water,
in
18
criteria
3
and
6,
makes
the
assumption
19
that
it
is
critically
important
that
20
one
have
only
one
procedure,
one
MDL
21
procedure.
And
as
I
said,
we
divided
22
it
into
four
when
the
majority
of
the
23
task
group
made
their
decisions
about
24
how
to
deal
with
this
issue.

25
I
think
these
assumptions
that
66
1
they
made,
3
and
6,
are
inappropriate.

2
I
think
that
whether
or
not
you
need
3
one
procedure
or
whether
you
need
more
4
than
one
procedure
should
have
been
5
assessed
itself.

6
And,
at
the
very
end
of
the
7
materials
I
passed
out,
I
have
8
provided
an
alternative
list
of
9
criteria
that
I
think
should
have
been
10
assessed,
and
I
ask
you
to
look
at
11
that
list
and
to
look
at
the
varied
12
uses
of
the
MDL,
including
the
13
interlaboratory
uses,
and,
for
14
yourself,
make
a
determination
about
15
whether
or
not
one
MDL
procedure
is
16
adequate.

17
Thank
you
very
much
for
your
18
consideration,
and
I
want
to
take
the
19
time
to
thank
all
the
various
members
20
of
my
task
group
for
the
hours
and
21
hours
of
efforts
they
have
put
in
over
22
the
years.

23
MR.
TELLIARD:
Our
next
24
speaker
is
Naomi
Goodman
from
the
25
Electric
Power
Research
Institute.
67
1
MS.
GOODMAN:
Nancy
said
2
a
lot
of
what
I
had
intended
to
say,

3
so
I
will
spare
you.

4
I
represent
EPRI,
and
just
to
5
give
you
a
little
history,
EPRI
began
6
considering
issues
of
detection
and
7
quantitation
in,
as
far
as
I
can
8
determine,
about
1984.
That
is
19
9
years.
In
that
time,
they
have
10
conducted
at
least
five
or
six
major
11
round
robin
studies
to
evaluate
low­

12
level
detection
and
quantitation
of
13
various
analytical
methods.

14
They
have
been
involved
in
15
efforts
to
develop
alternative
16
detection
and
quantitation
criteria,

17
starting
with
the
alternate
minimum
18
level,
continuing
with
the
involvement
19
in
the
IDE
and
IQE
development
20
process.
I
am
currently
a
member
of
21
Nancy
Grams'
task
group,
and
I
have
22
been
for
the
last
three
or
four
years.

23
So,
I
don't
want
to
take
any
24
more
of
your
time,
tack
that
on
top
of
25
those
19
years,
to
restate
things
that
68
1
have
been
said
many
times.
The
2
deficiencies
of
the
MDL
and
ML
have
3
been
covered
by
the
previous
speakers,

4
and
I
am
sure
the
following
speakers
5
will
also
address
them.

6
I
just
want
to
make
a
couple
of
7
comments
about
EPRI's
perspective
on
8
these
issues.

9
My
primary
concern
about
the
10
current
proposed
procedure
is
that
it
11
does
not
specify
the
methods
that
are
12
going
to
be
used
for
interlaboratory
13
validation
nor
say
what
is
going
to
be
14
done
with
the
data
that
are
produced
15
as
a
result
of
that
study.
Although
16
the
Office
of
Water
has
recognized
17
that
method
development
on
a
nationwide
18
level
does
require
multi­
laboratory
19
validation,
they
do
not
go
into
any
20
details.

21
In
previous
interlab
studies,

22
the
procedure
had
been
to
take
the
23
maximum
MDL
produced
by
any
of
the
24
participating
laboratories
that
passed
25
project
criteria
and
to
use
that
as
69
1
the
MDL
upon
which
the
ML
would
then
2
be
based.
This
was
the
procedure
that
3
was
used
in
the
1631
study
and
the
4
draft
1638
study
which
EPRI
5
participated
in.
Financially,
we
6
collaborated
with
EPA,
and
we
also
did
7
our
own
independent
data
analysis.

8
Since
that
subset
of
9
laboratories
is
fairly
random,
you
try
10
to
pick
laboratories
that
are
well
11
performing,
but
another
set
of
12
laboratories
would
produce
an
entirely
13
different
set
of
MDLs,
that
maximum
14
MDL
value
is
totally
arbitrary
and
15
would
be
different
if
you
picked
16
different
laboratories.
This
is
a
17
very
weak
effort
on
the
part
of
the
18
Office
of
Water
to
ensure
that
19
laboratories
do
not
fall
out
of
20
compliance
merely
because
they
are
on
21
the
wrong
side
of
the
distribution
22
curve.
There
has
to
be
some
more
23
scientific
way
to
go
about
it.

24
However,
my
primary
concern,
as
25
Nancy
Grams
has
already
addressed,
is
70
1
the
use
of
the
MDL
and
ML
for
setting
2
prescriptive
criteria
that
can
be
3
applied,
rightly
or
wrongly,
by
4
jurisdictions
to
ensure
compliance
with
5
the
permits
and
other
regulatory
6
limits.
ASTM
developed
the
IDE
and
7
IQE
to
address
this
particular
usage
8
and
no
other,
and
in
their
technical
9
assessment
document,
EPA
has
made
some
10
attempts
to
cram
the
standards
intro
11
an
application
for
which
they
were
12
never
intended
which
is
single
lab
13
application.

14
In
their
criteria
number
3,
the
15
Office
of
Water
states,
the
approach
16
should
be
supported
by
a
practical
and
17
affordable
procedure
that
a
single
18
laboratory
can
use
to
evaluate
method
19
performance.
This
is
a
limitation
20
that
the
Office
of
Water
set
on
21
itself.
It
is
not
required
by
the
22
settlement
agreement.

23
My
viewpoint
is
that
detection
24
and
quantitation
criteria
that
are
used
25
on
a
nationwide
basis
must
be...
excuse
71
1
me...
that
are
used
on
a
single
lab
2
basis
must
be
obtainable
at
a
low
cost
3
in
instrument
time
and
chemist
time.

4
However,
this
is
not
necessarily
the
5
case
for
nationwide
usage.

6
A
full
range
validation
study
7
to
support
a
procedure
such
as
the
8
IDE/
IQE
is
affordable
for
a
procedure
9
that
will
only
be
completed
once
for
a
10
single
method
or
method
revision.
We
11
don't
expect
that
to
happen
very
12
often,
and
we
expect
it
to
be
paid
13
for,
hopefully,
by
EPA
in
collaboration
14
with
interested
parties
who
may
include
15
the
laboratory
community,
may
include
16
industry
groups,
may
include
vendors.

17
There
has
certainly
been
no
lack
of
18
parties
in
the
past
who
have
been
19
willing
to
step
up
and
support
these
20
kinds
of
studies
where
their
own
21
interests
are
involved.

22
EPRI
collaborated
with
the
23
Office
of
Water
on
two
such
studies,

24
on
Method
1631
and
the
draft
Method
25
1638
which
was
nine
metals
by
ICP/
MS.

72
1
We
funded
addition
of
blank
pairs
and
2
low­
level
concentration
pairs
that
3
would
allow
calculation
of
an
IDE
or
4
IQE,
and
we
conducted
an
independent
5
analysis
of
the
data.

6
Our
findings
were
that
these
7
data
produce
reasonable
estimates
of
8
the
interlab
detection
and
quantitation
9
limits.
Unlike
the
proposed
MDL/
ML
10
procedure,
the
ASTM
approach
can
be
11
used
to
determine
when
a
method
is
not
12
capable
of
achieving
data
quality
13
objectives.

14
In
the
1638
study,
we
15
determined
the
overall
interlab
16
variability
at
the
ML
determined
by
17
the
Office
of
Water
for
reagent
water.

18
Out
of
nine
metals,
only
one
had
an
19
RSD
of
less
than
10
percent.
The
20
other
eight
had
RSDs
ranging
from
19
21
to
90
percent.

22
For
single
lab
variability,
if
23
you
want
to
look
at
that,
five
of
the
24
nine
elements
had
RSDs
greater
than
20
25
percent
at
the
Office
of
Water's
ML.
73
1
I
suggest
to
the
Office
of
2
Water
they
should
broaden
their
3
assessment
of
detection
and
4
quantitation
criteria
to
allow
5
alternative
procedures
such
as
the
6
IDE/
IQE.
Although
the
proposed
7
Appendix
B
offers
the
possibility
of
8
using
such
procedures,
it
then
9
nullifies
the
concession
by
stating,

10
providing
the
resulting
detection
11
limits
meets
the
sensitivity
needs
for
12
the
specific
application,
that
is
to
13
say,
if
a
method
is
sensitive
enough
14
that
the
quantitation
limit
is
well
15
below
the
regulatory
limit,
it
is
16
acceptable
to
consider
all
sources
of
17
variability,
but
it
is
precisely
those
18
parameters
with
detection
and
19
quantitation
limits
close
to
regulatory
20
criteria
that
have
the
greatest
need
21
to
consider
interlab
variability.

22
Thank
you.

23
MR.
TELLIARD:
Thanks,

24
Naomi.
Our
next
speaker
is
Larry
25
LaFleur.
74
1
MR.
CHRISTMAN:

2
Substituting
for
Larry
LaFleur...

3
MR.
TELLIARD:
We
4
switched?

5
MR.
CHRISTMAN:
We
6
switched
again.

7
MR.
TELLIARD:
Okay,
Jim
8
Christman.

9
MR.
CHRISTMAN:
Thanks,

10
Bill.

11
Good
morning.
I
am
going
to
12
make
just
a
few
brief
remarks
on
13
behalf
of
the
Inter­
Industry
Analytical
14
Group
or
IIAG,
as
we
call
it,
which
is
15
a
group
that,
in
turn,
represents
the
16
views
of
many
NPDES
permittees
who
17
depend
on
these
analytical
procedures
18
to
prove
compliance
with
their
permit
19
limits.

20
As
you
know,
as
you
well
know,

21
the
Office
of
Water's
proposals
for
22
23
detection
and
quantitation
limits
were
24
published
in
March
of
this
year
in
the
25
Federal
Register.
The
comment
period
75
1
is
now
open,
and
comments
are
due
July
2
10th.
The
IIAG
feels
like
this
is
a
3
really
important
rulemaking
for
some
4
reasons
that
I
will
go
into
directly.

5
Way
back
when
the
world
were
6
young,
we
were
all
young
and
foolish
7
and
before
we
grew
embittered
by
the
8
world
and
had
our
dreams
dashed
9
against
the
rocks
of
experience,
when
10
the
first
fish
with
lungs
were
11
crawling
out
of
the
primordial
12
ooze...
this
was
back
in
March
of
13
1994...
there
was
a
draft
guidance
14
document
from
EPA,
never
made
final,

15
discussing
use
of
detection
and
16
quantitation
limits
in
the
permitting
17
process.
That
was,
I
guess,

18
controversial,
but
some
of
the
ideas
19
in
that
paper
are
still
showing
up
or
20
reflected,
I
think,
in
the
March
21
Federal
Register
notice.

22
After
that
guidance
document
in
23
March
of
'
94
came
out,
the
IIAG
formed
24
itself.
Even
back
then,
the
folks
in
25
this
group
recognized
that
EPA
was
76
1
going
to
keep
setting
water
quality
2
criteria
lower
and
lower
and
lower.

3
Mercury
limits,
for
example,

4
are
very
low,
based
on
wildlife
5
criteria.
At
the
same
time,
we
knew
6
the
Agency
would
be
needing
analytical
7
methods
that
would
be
capable
of
8
measuring
down
at
these
lower
and
9
lower
levels,
and
the
IIAG,
as
a
10
group,
felt
that
there
is
an
11
obligation
here
to
make
sure
that
12
these
new
methods
are
as
reliable
as
13
they
can
be
made.

14
So,
the
IIAG
tried
to
work
15
along
with
EPA
through
the
years,
and
16
while
this
work
was
going
on,
EPA
came
17
out
with
Method
1631
for
mercury
and
18
came
out
with
1668
for
PCBs
which
have
19
very
low
levels
of
detection
and
20
quantitation.
And
as
you
have
heard
21
several
times,
there
was
some
22
litigation
on
that
and
a
settlement
23
agreement,
and
in
the
agreement,
EPA
24
undertook
to
do
a
broad
reassessment
25
of
detection
and
quantitation.
It
77
1
arose
out
of
the
mercury
litigation,

2
but
it
wasn't
just
directed
at
3
mercury.
It
was
a
more
general
view
4
of
the
subject
as
a
whole.

5
The
IIAG
continued,
you
know,

6
working
with
EPA.
We
were
very
7
hopeful
that
that
reassessment
would
8
produce
concepts
of
quantitation
and
9
detection
that
would
allow
us
to
agree
10
on
what
they
should
be,
that
would
use
11
sound
science,
and,
most
of
all,
would
12
avoid
legal
disputes
in
the
future,

13
because
like
everybody
else,
we
want
14
to
avoid
litigation
as
much
as
we
can,

15
and
we
want
to
avoid
disputes.
So,
we
16
had
hoped...
we
were
very
hopeful
about
17
this
reassessment
and
the
rulemaking
18
that
has
come
out
of
it.

19
Now,
when
we
looked
at
the
20
proposal
in
the
Federal
Register,
on
21
the
other
hand,
we
were
concerned
that
22
maybe
this
was
a
continuation
of
the
23
old
way
of
doing
things
with
minor
24
modifications
that
don't
solve
the
25
problems
that
we
are
concerned
about.
78
1
One
of
the
concerns
that
the
2
permittees
have,
the
regulated
3
community,
is
that...
remember
that
that
4
March,
'
94
guidance
document
suggested
5
that
when
a
permit
limit
is
set
below
6
the
level
that
the
method
can
reliably
7
measure,
you
can
use
the
ML
as
a
8
measure
of
quantitation
limit
and
as
a
9
compliance
limit.
So,
it
is
this
10
unique
feature
of
the
ML
which
can
be
11
used
as
a
compliance
limit
when
permit
12
limits
are
set
very
low
that
is
of
13
much
concern
to
the
members
of
the
14
IIAG,
and
it
is
one
of
the
reasons
we
15
are
very
interested
and
very
concerned
16
about
this
rulemaking.

17
The
problem
is
if
the
ML
is
not
18
got
right
at
the
front
end
with
good
19
science,
the
ML
is
set
too
low,
then
20
there
may
be
violations
of
permits
21
that
are
not
true
violations
but
22
merely
the
result
of
variability
23
or
noise.
And
everybody
wants
to
24
avoid
permit
violations
for
all
kinds
25
of
reasons.
There
are
penalties
79
1
attached,
but,
you
know,
it
is
hard
on
2
the
reputations
of
the
companies
that
3
have
the
permits,
and
many
companies
4
have
imposed
upon
themselves
zero
5
violations
policies,
and
if
you
get
a
6
violation,
even
if
it
is
a
supposed
7
violation
or
an
apparent
violations,
it
8
violates
these
policies
and
causes
lots
9
of
trouble.

10
So,
we
want
to
get
the
11
detection
and
quantitation
levels
right
12
right
at
the
front.

13
We
think
for
that
reason
alone,

14
this
rulemaking
should
be
important
not
15
only
to
the
people
who
have
permits
16
but
to
the
States,
the
regulators
who
17
write
permits,
the
regulators
who
18
enforce
permits,
and
everybody
who
has
19
those
sorts
of
interests
in
this
20
subject.

21
Well,
moreover,
we
think
there
22
is
a
legal
principle,
and
I
will
just
23
touch
briefly
on
it.
The
legal
24
process
goes
into
the
due
process
of
25
law
which
is
really
just
fairness.
It
80
1
is
written
into
the
Federal
2
Constitution
and
probably
into
the
3
constitution
of
every
State
in
the
4
country,
and
the
way
the
courts
have
5
expressed
this
legal
principle
is
every
6
citizen
and
every
NPDES
permitted
is
7
entitled
to
fair
notice,
fair
notice
8
of
the
conduct
which
will
make
him
9
break
the
law,
conduct
which
the
10
Government
deems
to
be
illegal.

11
A
further
principle
is
that
if
12
you
are
being
penalized
for
illegal
13
conduct
in
an
area
where
you
can't
14
measure
it
reliably,
that
probably
15
violates
this
fundamental
principle.

16
One
court
in
the
District
of
Columbia
17
said,
in
a
very
short
passage,
that
18
well,
it
may
be
okay
to
set
permit
19
limits
below
where
you
can
reliably
20
measure
them,
but
you
may
not
be
able
21
to...
may
not
have
the
legal
power
to
22
enforce
them
down
at
those
levels,
and
23
these
legal
principles
are
one
reason
24
that
we
see
the
importance
of
this
25
rulemaking,
because...
END
OF
TRACK
01
81
1
BEGIN
DISC
2,
TRACK
02:

2
...
proposed
MDL
and
ML
procedure.

3
They
are
three
in
number.
There
are
4
others,
but
let
me
just
give
you
three
5
important
ones.

6
One,
which
has
been
mentioned
7
several
times,
is
the
use
of
8
interlaboratory
variability.
We
think
9
incorporating
interlaboratory
10
variability
is
very
important,
because
11
many
regulatory
decisions
using
data
12
rely
on
data
from
many
labs.
So,
the
13
variability
between
and
among
different
14
labs
is
going
to
be
very
important.

15
Second
is
a
concern
for
16
accuracy.
If
the
quantitation
level
17
and,
therefore,
the
compliance
level
in
18
some
permits
is
based
on
mere
19
repetition
of
measurements
and
how
20
close
together
they
are
rather
than,

21
in
addition,
how
close
they
are
to
a
22
true
number,
a
concern
for
accuracy,

23
then
there
will
be
a
problem
and
24
vulnerability
to
legal
challenge
there.

25
Finally,
we
feel
like
the
MDL
82
1
is
being
asked
to
carry
too
much
2
weight
on
its
shoulders.
The
MDL
is
3
used
for
a
variety
of
ways
in
the
4
regulatory
process,
as
you
know.
It
5
is
used
as
a
figure
of
merit
when
a
6
method
is
published
or
promulgated,

7
used
as
a
QA/
QC
tool
for
individual
8
laboratories,
used
as
a
basis
for
the
9
ML,
as
a
basis
for
the
compliance
10
level
for
some
permits,
and
in
11
addition
to
all
of
those
uses
that
the
12
MDL
has,
EPA
has
had
to
try
to
come
13
up
with
a
procedure
that
is
relatively
14
cheap
and
simple
so
that
the
labs
can
15
use
it
without
undue
difficulty.

16
Well,
those
are
good,
you
know,

17
good
motives,
but
we
kind
of
feel
that
18
the
MDL
just
can't
bear
all
that
19
burden
by
itself,
and
it
would
be
20
better
to
split
up
some
of
the
21
functions
that
the
MDL
is
expected
to
22
serve
and
pick
a
tool
that
is
more
23
precisely
tailored
using
sound
science
24
for
the
individual
uses
of
it.

25
To
that
end...
and
I
now
come
to
83
1
my
close...
the
IIAG,
although
we
really
2
like
the
ASTM
International
proposals
3
you
just
heard
about,
we
also
came
up
4
with
a
sort
of
another
alternative
5
which
we
proposed
to
EPA
and
they
were
6
kind
enough
to
mention
in
the
Federal
7
Register
notice
and
put
in
the
docket.

8
So,
you
can
get
a
copy
of
the
IIAG
9
proposal
which
is
two
parts
either
10
from
me
if
you
will
see
me
after
the
11
meeting,
or
I
am
sure
you
can
get
it
12
from
EPA
as
well,
and
EPA
has
invited
13
comment
on
it.
We
hope
you
will
look
14
at
that
and,
if
you
like
it,
comment
15
to
EPA
before
July
or
by
July
10th.

16
The
two­
part
IIAG
alternate
17
proposal
which
you
are
going
to
hear
18
more
about
from
Larry
LaFleur
in
a
few
19
minutes
has
a
sensitivity
test
and
a
20
full
range
of
validation.
Again,
we
21
split
up
the
functions.

22
The
sensitivity
test
is
23
designed
to
be
used
by
labs,
for
24
example,
as
a
proficiency
test
upon
25
startup.
It
does
not
rely
on
84
1
interlaboratory
data,
and
the
labs
can
2
do
it
themselves,
and
it
is
fairly
3
simple
and
easy
to
do.

4
The
second
part
of
the
proposal
5
is
the
full
range
validation
which
6
does
collect
interlaboratory
data,
but
7
the
individual
labs
would
not
have
to
8
bear
that
burden.
We
envision
that
9
the
interlaboratory
data
would
be
10
collected
by
the
Government,
by
method
11
developers,
or
maybe
even
by
groups
of
12
volunteers
who
have
an
interest
in
13
getting
a
method
approved.
14
So,
you
can
look
at
the
15
detailed...
the
writings
we
have
put
16
into
the
docket,
and
in
order
to
give
17
you
a
few
more...
a
little
bit
more
18
information
about
this
IIAG
proposal,
I
19
would
like
to
present
Larry
LaFleur
20
who
is
with
the
National
Council
for
21
Air
and
Steam
Improvement,
and
he
is
22
going
to
tell
you
about
both
parts
of
23
the
IIAG
proposal,
both
the
sensitivity
24
test
and
the
full
range
validation.

25
Larry?

85
1
MR.
LAFLEUR:
Thanks,

2
Jim.

3
As
Jim
said,
I
am
going
to
try
4
and
provide
you
a
little
more
5
clarification
as
to
some
of
the
6
details
in
the
IIAG
sensitivity
test
7
and
then
go
on
and
talk
about
the
full
8
range
validation.

9
What
we
did
was
we
started
out
10
looking
at
some
of
the
multiple
uses
11
the
EPA
has
for
the
MDL,
and,

12
specifically,
we
were
looking
at
their
13
QA/
QC
uses.
Those
we
identified
were
14
initial
proficiency
demonstrations,

15
basically,
a
test
that
is
required
if
16
an
analyst
makes
a
significant
change
17
in
the
method,
and
the
third
one
we
18
found
was,
you
know,
documentation
for
19
method
modifications
that
are
allowed
20
under
the
flexibility
in
Part
136
21
methods.

22
So,
we
asked
ourselves,
if
we
23
are
just
focusing
on
this
one
end
use,

24
what
is
the
MDL?
From
the
analyst's
25
point
of
view,
it
is
really
just
low­

86
1
level
precision,
and
that
is
all
we
2
are
really
doing
in
the
MDL
study
from
3
the
context
of
QA/
QC,
challenging
the
4
lab's
low­
level
precision.
So,
if
you
5
focus
on
just
that
one
end
use,
you
6
will
get
a
better
idea
of
what
we
are
7
shooting
for
when
we
talk
about
the
8
sensitivity
test.
It
is
targeted
just
9
for
the
initial
startup
criteria,
just
10
for,
you
know,
documentation
of
changes
11
from
changes
in
equipment,
different
12
analysts,
as
EPA
currently
interprets
13
that
in
their
methods,
and
then,

14
finally,
as
a
verification
of
15
sensitivity
when
method
modifications
16
are
implemented.

17
What
we
are
not
trying
to
do
is
18
add
a
new
element
or
a
new
layer
of
19
QA/
QC.
We
are
trying
to
just
identify
20
the
current
uses
the
EPA
has
21
for
the
MDL
in
the
QA/
QC
context
and
22
see
if
there
is
a
more
efficient
way
23
to
do
it.
So,
this
is
not
a
new
24
requirement.
We
don't
propose
this
25
for
continued
use.
It
is
a
substitute
87
1
for
the
use
of
the
MDL.

2
One
easy
way
to
think
of
it
3
conceptually
is
it
is
really
very
4
similar
to
the
initial
precision
and
5
accuracy
test
except
we
are
using
it
6
to
test
low­
level
precision.
So,
the
7
spike
level
will
be
very
low,
but
we
8
are
not
imposing
a
recovery
criterion.

9
The
MDL
doesn't
do
that.
We
are
not
10
adding
to
that
test.
So,
it
is
really
11
just
a
test
of
low­
level
precision
12
similar
to
your
initial
precision
and
13
accuracy
test.

14
So,
how
would
this
work?
Well,

15
when
EPA
is
developing
a
method,
they
16
would
select
the
spiking
levels
and
17
the
number
of
replicates.
If
they
18
modeled
the
sensitivity
test
after
the
19
IPR
test,
they
would
use
four
20
replicates.
We
wouldn't
envision
them
21
using
any
more
than
seven,
because
22
that
is
what
they
are
currently
using
23
in
the
context
of
the
MDL.

24
The
idea
is
to
keep
this
simple
25
and
functional
and
focused
on
the
88
1
objective.
So,
we
would
think
that
2
the
spike
level
would
be
just
a
simple
3
dilution
of
the
lowest
calibration
4
standard.
That
might
be
half,
or
it
5
might
be
a
third,
but
the
idea
is
to
6
keep
it
simple.
We
don't
want
to
7
challenge
the
lab
to
make
out
8
detailed,
complicated
cocktails
in
9
order
to
preform
this
test.

10
During
method
development
and
11
validation,
EPA
would
generate
QC
12
performance
specifications
for
the
13
method.
It
seems
like
the
logical
14
thing
to
do
would
be
use
the
model
15
they
use
now
for
developing
the
16
precision
specifications
in
the
IPR.

17
They
have
a
certain
experimental
18
procedure
and
statistical
procedures
19
that
are
outlined
in
their
guidelines
20
for
alternative
test
procedures
at
21
tire
3,
national
approval.

22
Once
they
have
done
all
that
23
work,
then
we
would
envision
the
24
sensitivity
test
being
incorporated
25
into
the
analytical
protocol.
The
89
1
spike
levels
would
be
designated.
The
2
number
of
replicates
would
be
3
designated.
The
QC
spec
would
be
4
designated,
and
that
would
be
first
5
proposed
and
then,
finally,
promulgated
6
as
part
of
the
method.

7
How
would
it
work
from
the
8
laboratory's
perspective?
Well,
it
is
9
fairly
simple.
Basically,
the
lab
10
would
prepare
the
required
number
of
11
replicates,
and
this
would
be
spiked
12
reagent
water.
They
would
spike
it
at
13
the
level
designated
in
the
method.

14
They
would
analyze
the
samples
through
15
the
entire
analytical
protocol
so
it
16
is
similar
in
that
it
tests
the
entire
17
analytical
procedure
and
not
just
an
18
element
of
calibration.
They
would
19
just
simply
calculate
the
standard
20
deviation,
and
in
this
sense,
compare
21
it
against
the
QC
specification.
They
22
wouldn't
multiply
it
by
the
Student's
23
t
value
to
calculate
an
MDL.
They
24
would
just
compare
it
against
the
QC
25
spec.
90
1
If
you
are
below
that
QC
spec,

2
you
have
passed,
and
you
move
on
and
3
start
running
samples.

4
What
are
the
advantages?
Well,

5
the
primary
advantage
is
in
order
to
6
meet
the
spiking
requirements
in
the
7
current
CFR
Part
136
Appendix
B
8
procedure,
for
a
multi­
analyte
method,

9
many
laboratories
are
required
to
10
perform
three,
even
four
sets
of
seven
11
replicates.
So,
you
are
talking
21
or
12
28
analyses.
That
seems
like
an
awful
13
lot
of
work
just
to
demonstrate
a
14
certain
level
of
proficiency
at
low
15
levels.

16
So,
the
primary
advantage
of
17
the
sensitivity
test
is
it
sets
a
18
certain
number...
and,
again,
we
think
19
that
is
between
4
and
7...
number
of
20
replicates
at
a
prescribed
level.

21
Yet,
at
the
same
time,
it
will
provide
22
a
credible
challenge
of
the
23
laboratory's
performance
at
low
24
concentrations.

25
Other
advantages
we
think
would
91
1
be
that
all
labs
would
be
performing
2
the
tests
at
the
same
level.

3
Currently,
each
lab
picks
their
own
4
MDL
spike
level,
so
you
really
can't
5
compare
one
MDL
in
one
lab
versus
6
another,
because
you
don't
know
what
7
the
spike
levels
were.

8
There
would
be
a
very
specific
9
pass­
fail
criterion
which
is
explicitly
10
stated
in
the
method.
Thus,
all
the
11
labs
would
be
evaluated
against
that
12
same
criterion.

13
Again,
since
we
focused
14
specifically
on
the
QA/
QC
aspects
of
15
the
use
of
the
MDL,
it
will
be
a
far
16
more
efficient
test
for
the
17
laboratories.

18
So,
in
summary,
by
focusing
on
19
very
specific
objectives,
we
can
20
develop
a
much
more
efficient
tool
21
than
the
current
use
of
the
MDL.
Yet,

22
at
the
same
time,
we
will
still
be
23
providing
a
credible
challenge
of
24
laboratories'
performance
at
low
25
concentrations.
The
test
is
as
simple
92
1
as
we
can
possibly
make
it,
consistent
2
with
the
goal
of
providing
a
3
meaningful
challenge
to
the
labs.

4
Now,
there
has
been
some
5
confusion
about
what
we
intended,
so
I
6
want
to
tell
you
what
we
don't
think
7
the
sensitivity
test
would
be.
It
is
8
not
a
detection
limit.
It
is
really
9
just
a
QA/
QC
tool
that
would
be
used
10
in
very
limited
aspects
for
proficiency
11
testing.

12
It
is
not
a
substitute
for
a
13
reporting
threshold.
You
would
have
14
to
look
elsewhere
to
find
that.

15
Again,
there
are
different
objectives
16
and
goals
and
quality
objectives
that
17
you
would
want
to
establish
for
that
18
kind
of
a
reporting
threshold.

19
Finally,
there
was
no
intent
20
here
for
this
to
be
utilized
as
a
21
quantification
level.

22
Again,
the
second
part,
this
23
presentation
was
actually
prepared
by
24
John
Phillips.
He
was
unable
to
25
attend
today,
so
he
asked
me
if
I
93
1
would
try
to
work
my
way
through
it,

2
and
I
will
do
my
best.
This
covers
3
the
second
part
of
the
IIAG
proposal,

4
that
of
the
full
range
validation.

5
The
basic
assumptions
I
think
6
you
have
heard
a
little
bit
of
7
earlier.
Basically,
the
IIAG
believes
8
that
any
data
used
for
regulatory
9
compliance
purposes
must
fall
within
10
the
methods
quantification
range.

11
Secondly,
we
believe
quantification
is
12
defined
by
both
acceptable
precision
13
and
bias.

14
So,
when
we
look
at
the
MDL
15
when
it
is
derived
as
a
multiple
of
16
the
ML,
we
don't
believe
it
is
a
17
quantification
level.
First
of
all,

18
it
does
not
have
known
precision.

19
Secondly,
it
doesn't
even
consider
20
accuracy
or
bias.

21
As
we
envision
it,
the
full
22
range
validation
would
be
an
23
interlaboratory
study
using
a
minimum
24
of
seven
laboratories.
Nancy
mentioned
25
one
of
many
standardized
protocols
for
94
1
conducting
these
kinds
of
studies,
the
2
D2777
process,
but
there
are
others.

3
It
would
develop
precision
and
bias
4
through
the
entire
working
range
of
5
the
method.
The
data
could
be
used
to
6
establish
method
performance
criteria.

7
It
could
also
be
used,
if
it
was
a
8
correctly
defined
study,
to
establish
9
the
LC
and
LD
that
Richard
spoke
about
10
earlier
as
well
as
LQ,
and
it
has
some
11
utility
in
the
context
for
reporting
12
measurement
uncertainties.

13
We
believe
that
the
full
range
14
validation
should
be
used
for
every
15
method
used
for
regulatory
compliance
16
purposes
or
any
new
use
of
a
method
17
for
regulatory
compliance.
This
would
18
include,
of
course,
in
the
context
of
19
this
meeting,
the
methods
that
are
20
promulgated
at
40
CFR
136.

21
At
a
minimum,
we
would
like
to
22
see
it
applied
to
all
newly
23
promulgated
methods.

24
Now,
we
see
that
the
full
range
25
validation
would
only
be
carried
out
95
1
by
a
method
developer,
by
EPA,
or
by
2
consortia
of
laboratories
and/
or
3
industries,
and
I
will
give
you
an
4
example
of
the
latter
a
little
later
5
in
the
talk.

6
The
key
thing
here
is
that
any
7
individual
laboratory
would
not
be
8
expected
to
repeat
this
work.
This
is
9
really
just
for
method
development
and
10
promulgation.

11
Why
do
we
think
we
need
12
interlaboratory
data?
We
think
we
13
need
to
characterize
the
performance
of
14
the
method
under
the
same
conditions
15
that
it
will
be
used
and
practiced
for
16
regulatory
compliance
testing
purposes,

17
and
that
includes
accounting
for
all
18
sources
of
variability.
If
you
19
include
20
interlaboratory
data,
you
really
21
capture,
in
some
respects,
all
the
22
other
sources
of
variability
which
23
would
include
batch
to
batch,

24
instrument
to
instrument,
analyst
to
25
analyst.
So,
it
is
sort
of
an
all­

96
1
encompassing
way
to
account
for
2
uncertainty
and
variability.

3
Well,
we
are
all
aware
of
the
4
various
different
sources
of
5
variability
and
error
that
are
in
6
methods,
calibration,
cross
sample
7
contamination,
carryover,
individual
8
interpretations
of
how
the
method
ought
9
to
be
performed,
differences
in
10
techniques
or
skills
by
a
given
11
analyst.
Variation
in
reagents
is
12
becoming
more
important,
because
13
reagents
contamination
is
starting
to
14
become
an
issue
when
we
go
to
very,

15
very
low
trace
levels.
Finally,
of
16
course,
the
accuracy
of
the
standards
17
that
we
use
for
calibration.

18
Another
reason
why
full
range
19
validation
is
important
is
there
is
a
20
variety
of
different
uses
that
States,

21
permittees,
and
the
Office
of
Water
22
use
Part
136
methods
for.
These
23
include
effluent
guideline
24
developments,
NPDES
permit
application
25
procedures
that
the
States
might
use
97
1
then
for
reasonable
potential
2
determinations,
receiving
water
3
surveys,
TMDLs,
and,
of
course,
NPDES
4
permit
compliance
testing.

5
We
feel
that
many
of
these
6
different
uses
have
different
7
measurement
quality
objectives,
so
each
8
of
them
should
be
assessed...
or
the
9
method
should
be
assessed
against
those
10
measurement
quality
objectives.

11
To
illustrate
this,
we
are
12
going
to
use
a
few
examples
and
then
13
sort
of
walk
through
how
the
full
14
range
validation
would
work.

15
The
Office
of
Water's
16
definition
for
quantification,
for
17
compliance
purposes,
is
essentially
18
based
upon
a
goal
of
10
percent
RSD,

19
but
they
have
no
criteria
at
present
20
for
accuracy.
So,
the
purpose
of
21
this...
we
are
assuming
that
the
Agency
22
would
establish
a
recovery
requirement,

23
say,
in
the
range
of
50
to
100
percent
24
which
most
of
the
EPA
methods
are
25
capable
of
producing.
98
1
For
our
second
example,
let's
2
assume
that
a
State
wants
to
survey
a
3
river
looking
for
hot
spots,
and
they
4
want
to
use
State
certified
labs
for
5
this.
They
might
even
write
this
in
6
as
a
special
permit
condition
for
a
7
permitted.
They
may
define
their
8
measurement
quality
objective
as
plus
9
or
minus
30
percent
RSD
with
a
10
recovery
of
70
to
130
percent,
and
in
11
this
case,
they
want
to
control
both
12
false
positives
but
also
have
no
more
13
than
5
percent
false
negatives.

14
Our
third
example
is
a
client
15
who
may
want
to
submit
samples
to
any
16
one
of
three
or
more
qualified
labs
to
17
perform
some
sort
of
in­
process
18
technology
evaluation,
process
19
optimization,
perhaps
in
preparation
to
20
coming
into
compliance
with
their
NPDES
21
permit.
Here,
the
engineers
think
22
they
might
need
25
percent
RSD,

23
accuracy
of
80
to
120
percent,
and,

24
again,
the
same
1
percent
false
25
positives
and
5
percent
false
negative
99
1
rate.
2
So,
how
can
EPA,
the
State,
or
3
the
lab
or
the
permittee
decide
4
whether
a
method
will
be
suitable
for
5
meeting
these
measurement
quality
6
objectives?
What
we
would
like
to
7
show
you
is
how
we
think
the
full
8
range
method
validation
would
work
to
9
accomplish
that
end.

10
I
am
going
to
use
that
using
11
some
data
that
was
generated
in
an
12
interlaboratory
validation
study
on
13
PCBs
using
Method
608
using
capillary
14
electron
capture
analysis.
This
work
15
was
undertaken
by
the
Michigan
16
Manufacturers
Association
which
is,
I
17
think,
a
good
illustration
of
how
we
18
envision
interested
parties
willing
to
19
undertake
this
kind
of
work
and
then
20
provide
the
kind
of
interlaboratory
21
data
necessary.

22
The
study
included
11
23
laboratories,
16
different
24
concentrations,
looked
for
aroclor
1016
25
and
1260.
They
conducted
the
study
100
1
over
a
20­
week
period
to
account
for
2
routine
measurement
variability,
and
3
overall,
they
generated
19,800
data
4
points.

5
Now,
this
is
a
plot
of
the
6
actual
concentrations
reported
by
the
7
laboratories
versus
the
trues
8
concentrations.
This
gives
you
an
9
idea
of
the
number
of
different
spike
10
levels
that
were
used,
the
spacing
of
11
those
spike
levels,
and
we
have
got
a
12
regression
analysis
using
weighted
13
least
squares
and
ordinary
least
14
squares.

15
You
can't
really
see
the
16
details
in
this,
but
if
you
plot
17
recovery
versus
true
concentration,
you
18
can
see
that
the
study
revealed
that
19
there
was
a
severe
bias
of
the
20
recovery
in
the
low
concentration
21
range,
and
at
the
high
range,
they
22
observed
a
low
bias.

23
Well,
in
evaluating
the
three
24
examples
we
have,
we
can
look
at
these
25
data
to
see
where
a
method
would
give
101
1
us
acceptable
recoveries.
In
the
2
first
example,
we
assumed
that
EPA
3
would
use
a
50
to
150
percent
recovery
4
criterion.
That
criterion
could
be
5
met
using
these
data
or
should
be
met
6
using
this
method
from
0.02
ppb
up
to
7
1.2
ppb,
in
this
case,
the
highest
8
level
studied
in
the
experimental
9
design.

10
In
our
second
example,
the
11
State
river
survey,
we
said
we
12
presumed
that
they
would
want
70
to
13
130
percent.
Those
criteria
could
be
14
met
between
0.3
ppb
and
0.35
ppb.

15
The
third
in­
process
survey,

16
the
criteria
could
be
met
between
0.04
17
and
0.18.
And
you
can
go
on.
I
18
mean,
this
is
just
an
illustration
of
19
how
this
kind
of
data
can
be
used
to
20
examine
different
measurement
quality
21
objectives
in
the
context
of
recovery.

22
Next,
we
look
at
precision.

23
Here
is
the
raw
data
plotted,

24
interlaboratory
standard
deviation
25
versus
true
concentration.
Again,
I
102
1
am
going
to
look
at
a
different
plot
2
that
is
a
little
easier
to
see.
This
3
is
relative
standard
deviation
versus
4
true
concentration.
We
stated
our
5
criterion
on
the
basis
of
relative
6
standard
deviation.

7
They
looked
at
four
different
8
models,
the
models
that
Nancy
mentioned
9
earlier
that
are
in
the
10
IQE...
IDE...
excuse
me...
and
in
this
11
case,
they
found
the
hybrid
model
and
12
the
SL
model...
I
can't
remember
which
13
one
that
is...
but
they
found
those
two
14
matched
the
data
best.

15
So,
again,
going
to
our
16
examples,
the
first
one,
EPA's
goal
of
17
10
percent
RSD
for
their
methods
for
18
NPDES
permit,
we
find
that
the
method
19
never
did
achieve,
in
the
entire
range
20
of
the
method,
a
10
percent
RSD.

21
If
you
look
at
the
22
State's...
our
example
2,
a
State
river
23
survey,
the
measurement
quality
24
objective
was
30
percent.
That
can
be
25
achieved
at
any
concentration
above
103
1
0.08.
2
The
third
example
was
the
in­

3
process
survey.
We
said
plus
or
minus
4
25
percent
RSD,
and
that
can
be
5
achieved
at
any
concentration
above
6
0.1.

7
So,
now
we
have
examined
using
8
the
full
range
validation
recovery
and
9
precision.
The
next
is
the
false
10
positive/
false
negative
rate.
This
11
particular
study
was,
in
fact,
designed
12
using
the
ASTM
IDE
procedure.
So,
we
13
can
calculate
an
LD
for
a
1
percent
14
false
positive
rate
and
a
5
percent
15
false
negative
rate,
and
this
study
16
showed
that
that
could
be
achieved
at
17
levels
above
0.082.

18
So,
in
conclusion,
what
we
have
19
tried
to
do
is
illustrate
how
full
20
range
validation
can
be
used
to
21
provide
the
necessary
information
for
22
all
data
users
of
Part
136
methods
for
23
a
variety
of
the
different
uses
to
24
which
those
methods
are
put
to
meet
25
their
measurement
objectives.
In
the
104
1
first
example,
we
found
that
the
2
methodology
would
not
be
suitable
for
3
regulatory
compliance
testing,
because
4
it
was
never
able
to
achieve
the
5
target
10
percent
standard
deviation.

6
In
the
second
example,
we
found
7
the
working
range
in
the
low
end
was
8
bounded
at
0.082
by
control
of
false
9
positives
and
false
negatives,
while
10
the
upper
end
was
the
0.35
was
11
controlled
by
acceptable
recovery.

12
In
our
third
example,
we
showed
13
that
the
low
end
of
the
workable
range
14
for
that
purpose
was
0.1
ppb
driven
by
15
the
lowest
level
at
which
you
could
16
achieve
acceptable
precision
up
to
the
17
0.18
ppb
upper
limit
which
was
again
18
controlled
by
acceptable
recovery.

19
So,
in
general,
that
is
how
we
20
envision
a
full
range
validation
study
21
being
conducted
and
how
it
might
be
22
used.

23
Thank
you,
Bill.

24
MR.
TELLIARD:
Thank
25
you.
Our
next
speaker
is
Ed
Askew,

105
1
Davenport
Water
Pollution
Control
2
Laboratory.

3
MR.
ASKEW:
Good
4
morning.
I
will
make
it
brief,
short,

5
and
to
the
point,
and
you
all
can
go
6
to
lunch
or
whatever
you
want
to
do,

7
Bill.

8
I
am
here...
basically,
I
run
a
9
wastewater
treatment
laboratory.
I
am
10
also
a
pretreatment
coordinator
for
the
11
City
of
Davenport.
In
that
city
of
12
125,000
people,
we
have
over
60
13
categorical
and
significant
industries,

14
including
the
largest
aluminum
rolling
15
mill
in
the
world,
discharging
to
our
16
POTW.

17
I
am
also
the
Part
4000
18
coordinator
for
Standard
Methods
for
19
the
examination
of
water
and
20
wastewater,
the
oldest
and
most
senior
21
voluntary
consensus
body
for
water
and
22
wastewater.
I
also
contribute...
I
23
mean,
I
vote,
read
and
vote,
for
D19
24
ASTM.

25
But
what
I
am
here
for
today
is
106
1
specifically...
and
I
will
put
my
hat
2
on
for
this...
the
hat
I
wear
all
the
3
time
is
I
am
a
chemist,
and
I
work
4
the
bench.
I
manage
the
laboratory.

5
I
apply
the
analytical
results
to
6
enforcement
under
the
pretreatment
7
program.
So,
this
MDL
document
is
of
8
importance
to
me
and,
actually,
to
my
9
people
in
the
laboratory.

10
To
start
it
out,
I
am
in
11
support
of
the
proposed
rule.
There
12
can
be
modifications
to
it.
There
can
13
be
improvements
to
it,
but
I
think
I
14
will
be
key
on
that
is
I
am
in
15
support
of
it.

16
I
have
two
procedural
concerns
17
with
this.
Now,
I
am
going
to
use
a
18
term
called
classical
analysis.
As
19
Part
4000
coordinator,
we
do
20
distillations,
colorimetric
analysis.

21
Thank
God
I
don't
do
BOD.
I
do
it
on
22
a
regular
basis,
but
I
am
not
on
that
23
committee.
So,
that
is
classical,

24
titrametric,
gravimetric,
colorimetric.

25
Instrumentation,
if
I
use
that,
107
1
I
am
applying
that
to
GC,
GC/
mass
2
spec,
HPCL,
and
such.

3
So,
first,
I
have
two
4
procedural
concerns,
and
they
are
very,

5
very
small.
The
first
is
there
is
in
6
the
technical
support
document
for
the
7
assessment
of
detection
and
8
quantitation
approaches
which
I
read
9
3.3.1,
sources
of
variability,
page
10
3.19,
quote,
however,
it
has
been
11
EPA's
experience
that
concern
over
12
matrix
effects
may
be
somewhat
13
overblown.

14
This
statement
which
may
be
15
true
for
instrumental
analysis
such
as
16
GC
and
GC/
mass
spec
is
not
equally
17
true
for
classical
analysis
such
as
18
ammonia,
cyanide,
phenols,
oil
and
19
grease,
et
cetera
when
applied
to
20
direct
industrial
discharges
to
the
21
collection
system
of
a
POTW,
something
22
that
has
not
really
been
pointed
out
23
here,
but
if
an
industry
meets
or
is
24
below
the
categorical
limits
or
the
25
local
limits,
they
do
not
have
to
108
1
retreat
their
discharge
before
it
is
2
discharged
to
the
POTW,
publicly
owned
3
treatment
words,
basically,
the
4
municipal
or
the
districts
that
collect
5
this
wastewater
and
treat
it.

6
Therefore,
matrix
interferences
7
are
key
when
you
apply
40
CFR
136
8
methods
and
the
method
detection
limit
9
to
those.
So,
that
is
something
I
am
10
just
pointing
out,
that
this
needs
to
11
be
added
in,
considered,
whatever.

12
And
a
case
in
point
near
and
13
dear
to
my
heart...
and
I
am
sure
Bill
14
is
going
to
just
cringe...
is
the
15
abysmal
failure
of
the
oil
and
grease
16
1664
method
standard
of
stearic
17
acid/
hexadecane
to
work
in
industrial
18
wastewater
from
industries
that
have
19
high
concentrations
of
surfactants
such
20
as
commercial
laundries
that
wash
rags
21
used
in...
those
little
red
rags
you
22
see
that
the
mechanics
wipe
their
23
hands
off
with
and
such.
We
have
one
24
industry
or
one
industrial
laundry
in
25
our
POTW
collection
system
that
bulk
109
1
washes
those
rags.
Trying
to
get
1664
2
with
stearic
acid
and
hexadecane
to
3
work
in
that
just
doesn't
work.
There
4
are
just
too
many
surfactants
in
5
there.

6
Also,
we
have
industries
that
7
produce
a
lot
of
fatty
acids,
grease,

8
trying
to
get
hexadecane
and
stearic
9
acid
to
work
in
that.
So,
you
know,

10
that
is
the
type
of
example
where
a
11
matrix
needs
to
be
considered
in
12
method
detection
limits.

13
The
other
thing
is
the
overall
14
method
detection
limits
for
15
instrumental
analysis
are
more
clearly
16
defined
by
the
analytical
process
than
17
those
for
classical
analysis.
This
is
18
clearly
illustrated
by
the
comparison
19
of
gas
chromatography
to
a
gravimetric
20
analysis
such
as
total
suspended
solids
21
or
residue
non­
filterable.

22
A
gas
chromatography
can
have
23
increasing
sample
volume
or
mass
loaded
24
on
the
column.
So,
you
can
increase
25
the
sample
volume.
You
can
use
a
110
1
Goerstel,
a
concentrator,
purge
trap,
2
something
like
that.
You
are
limited
3
by
the
diameter
of
the
column,
the
4
thickness
of
the
film
on
the
5
capillary,
the
stationary
phase,
and
6
by,
basically,
detector
overload
seen
7
on
some,
say,
mass
selective
detectors
8
and
mass
spectrometer.
You
can
only
9
put
so
much
in
it,
and
after
that
10
point,
it
doesn't
matter
if
you
can
11
concentrate
it
more
than
that.
Your
12
detector
can
only
take
so
much.

13
But
let's
take
total
suspended
14
solids.
According
to
Standard
Method
15
2540D,
20th
edition,
you
must
be
able
16
to
recover
at
least
2.5
mg
of
solid.

17
It
also
mentions
later
on
in
the
18
method
that
you
must
filter
up
to...
it
19
is
allowing
you
up
to
1
liter
to
20
filter.
In
some
cases,
you
can't
get
21
2.5
mg
out
of
a
liter.
So,
what
has
22
happened
in
a
lot
of
State,
they
have
23
put
the
emphasis
on
mass
collected
and
24
not
the
amount
of
volume
filtered.

25
So,
they
allow
you
to
filter
more
than
111
1
a
liter.
Collect
2
liters,
collect
3
2
liters,
collect
10
liters,
we
still
3
want
2.5
mg.

4
So,
basically,
for
the
5
classical
analytical
methods,
as
you
6
develop
the
method
detection
limits,

7
part
of
the
process
will
be
8
determining
what
are
you
going
to
9
measure.
Volume?
Mass?
Unlike
your
10
GC
mass
specs
where
you
have
an
11
electrical
signal,
in
some
cases,
you
12
are
actually
talking
about
unique
13
physical
measurements.

14
And
let's
not
even
bring
in
15
temperature,
pH,
BOD.
All
of
those
16
are
applied
to
every
NPDES
permit
17
holder.

18
So,
like
anything
else,
if
you
19
complain,
you
also
have
to
have
some
20
solutions.
So,
here
is
my
solution
to
21
it
or
proposed
corrective
action.

22
I
think
the
EPA,
along
with
the
23
voluntary
consensus
standard
bodies
24
responsible
for
the
analytical
methods
25
in
40
CFR
136...
and
we
are
talking
112
1
about
ASTM,
Standard
Methods,
USGS,

2
AOAC...
should
develop
a
guidance
3
manual.
Now,
it
has
been
pointed
out
4
to
me
everybody
is
short
of
staff.

5
But
for
the
voluntary
consensus
boards,

6
they
already
have
people
in
place
that
7
have
helped
develop
these
methods.
I
8
know
for
Standard
Methods,
there
are
9
joint
task
groups
that
actually
develop
10
these
methods.
Those
would
be
key
11
people
to
be
brought
into
a
guidance
12
manual
development,
and
it
13
specifically,
then,
can
outline
how
do
14
you
develop
the
method
detection
limit
15
for
this
method.

16
Now,
we
have
already
seen
a
lot
17
of
work
done
by
ASTM
that
should
not
18
be
put
to
waste.
I
think
in
a
19
consensus
building,
we
can
develop
a
20
guidance
manual
that
can
address
these.

21
So,
my
suggestion
is,

22
basically,
completion
of
this
meeting
23
today
and
in
the
process
during
the
24
public
comment,
the
EPA
should
contact
25
the
voluntary
consensus
standards
113
1
bodies
such
as
Standard
Methods,
ASTM,
2
USGS,
AOAC,
et
cetera...
that
includes
3
the
Lab
Practices
Committee
for
the
4
Water
Environment
Federation
and
any
5
other
voluntary
group...
and
start
6
developing
the
guidance
manual
to
7
actually
implement
whether
you
call
it
8
method
detection
limit,
quantitation
9
limit,
anything
else,
to
provide
the
10
working
manual
for
the
laboratorians
or
11
the
chemists,
the
analysts,
to
actually
12
put
some
of
this
into
place.

13
So,
in
conclusion,
the
need
of
14
the
proposed
rule
establishing
15
procedures
for
detection
and
16
quantitation
is
manifest
by
the
spate
17
of
lawsuits
by
special
interest
groups
18
bent
on
preventing
the
further
19
implementation
of
the
Clean
Water
Act.

20
I
clearly
believe
this.
I
see
this
21
every
day.
Not
to
knock
anybody
in
22
here,
but
lawyers
and
business
leaders
23
consistently
use
the
lack
of
consensus
24
of
the
scientific
community
on
25
detection
limits
to
negate
good
114
1
analytical,
scientific
results.
I
have
2
seen
this
on
a
daily
basis.

3
Though
there
will
never
be
a
4
unified
field
theory...
nods
to
Albert
5
Einstein...
for
detection
limits,
never
6
will
be,
the
proposed
EPA
rule
is
an
7
excellent
starting
point,
and
the
8
authorship
of
the
guidance
manual
will
9
only
help
clarify
the
majority
of
10
concerns
from
the
scientific
community.

11
Thank
you.

12
MR.
TELLIARD:
Thank
13
you.

14
We
have
on
our
calendar
here
a
15
one­
hour
break
where
you
can
have
your
16
caviar
and
champagne
for
lunch.
For
17
those
of
you
who
can't
do
that,
there
18
is
a
restaurant
upstairs,
there
is
a
19
deli
downstairs,
and
a
McDonald's
20
across
the
street.
We
didn't
want
to
21
give
you
more
than
an
hour,
because
we
22
thought
you
might
enjoy
yourselves.

23
So,
we
will
see
you
back
here
24
at
1:
00
o'clock.

25
I
would
like
to
thank
this
115
1
morning's
speakers
very
much.
Thank
2
you.

3
(
WHEREUPON,
a
luncheon
recess
was
4
taken.)

5
MR.
TELLIARD:
The
first
6
speaker
this
afternoon
is
Patty
Lee
7
from
the
Hampton
Roads
Sanitary
8
District.

9
MS.
LEE:
Thank
you
all
10
very
much
for
coming
back
after
lunch.

11
I
know
that
could
have
been
a
12
challenge.

13
Jim
Pletl
and
myself
are
going
14
to
be
commenting
on
our
issues
that
we
15
have
with
the
proposed
MDL
procedure,

16
and
we
would
like
to
thank
EPA
for
17
opening
this
forum
up
for
us
to
give
18
our
comments.
We
are
also
19
representing,
in
addition
to
HRSD,
we
20
are
representing
the
Virginia
21
Association
of
Municipal
Wastewater
22
Agencies,
and
we
are
also
going
to
be
23
providing
comments
to
AMSA
on
detection
24
and
quantitation.

25
Our
general
comments...
we
would
116
1
like
to
start
with
that
and
then
2
address
each
one
of
the
six
criteria
3
that
EPA
used
for
the
development
of
4
the
technical
support
document.
First
5
of
all,
we
would
like
to
say
that
we
6
agree
with
ACIL's
proposal.
We
7
support
it
in
concept.
There
are
a
8
few
things
that
we
might
would
like
to
9
have
changed
or
whatnot,
but
we
will
10
provide
that
in
the
comments
to
EPA
11
later
on.

12
The
ACIL
proposal
is
user
13
friendly
and
has
minimal
costs
14
associated
with
it,
because
you
are
15
basically
using
QA/
QC
criteria
that
you
16
already
have
with
each
method
batch
or
17
each
analytical
batch
anyway.

18
Therefore,
it
makes
it
very
affordable
19
and
something
that
your
bench
analysts
20
can
use
without
a
lot
of
problems.

21
ACIL's
approach,
as
we
heard
22
earlier,
does
take
into
account
false
23
negatives,
false
positives,
and,
most
24
importantly,
it
does
correct
for
bias,

25
because
that
is
definitely
a
problem
117
1
with
some
of
the
methods
that
we
have.

2
Like
several
other
people
in
3
the
room,
I
am
also
a
member
of
the
4
ASTM
committee
that
developed
the
IDE
5
and
the
IQE,
and
we
looked
at
that
6
concept
as
well,
and
although
we
feel
7
that
it
has
merit,
it
probably
is
more
8
appropriate.
The
most
appropriate
use
9
would
be
for
interlaboratory
data.

10
You
have
to
have
a
lot
of
data
points,

11
and
the
data
points
need
to
be
in
a
12
specific
range.
Then,
you
need
13
someone
that
has
a
good
deal
of
14
statistical
knowledge
to
actually
set
15
up
and
run
the
study.

16
So,
because
it
is
so
onerous
17
and
it
is
so
large,
we
feel
that
18
although
it
is
very
well
suited
for
19
interlaboratory
MDL
development
such
as
20
that
associated
with
method
validation,

21
we
feel
that
it
is
a
little
too
22
onerous
for
a
single
laboratory
to
23
have
to
do
that.

24
So
far
as
the
first
criterion
25
that
EPA
used
which
was
that
detection
118
1
and
quantitation
approaches
should
be
2
scientifically
valid,
our
first
comment
3
is
they
must
be
scientifically
valid.

4
People
are
signing
off
on
DMRs,
so
we
5
have
to
have
good
science
behind
these
6
limits
that
we
are
getting.

7
EPA's
proposal
is
also
still
8
concentration
dependent
in
that
it
is
9
based
on
the
EDL.
If
you
select
the
10
wrong
concentration,
you
could
end
up
11
with
falsely
low
or
falsely
high
MDLs.

12
It
still
does
not
really
encompass
13
anything
for
bias.

14
Because
of
this,
because
it
is
15
so
concentration
dependent,
and
because
16
it
doesn't
correct
for
bias,
we
feel
17
that
we
really
can't
meet
the
18
condition
of
99
percent
confidence
in
19
the
analyte
being
present.
Therefore,

20
we
feel
that
it
fails
condition
3
of
21
criterion
1.

22
Criterion
2
states
that
the
23
approach
should
address
demonstrated
24
expectations
of
the
laboratory
and
25
method
performance,
including
routine
119
1
variability.
Although
the
ML
proposal
2
or
concept
that
they
have
on
the
books
3
now
does
address,
you
know,
rounding
4
and
things
like
that,
it
really
5
doesn't
address
are
we
really
at,
can
6
we
really
even
see
ten
standard
7
deviations
of
the
MDL,
because
that
is
8
currently
what
EPA's
ML
is,
that
is,

9
ten
standard
deviations
of
the
MDL.

10
It
works
really
good
for
some
methods,

11
but
there
are
a
lot
of
methods
out
12
there
where
ten
standard
deviations
is
13
just
inadequate,
and
there
is
no
14
guidance
on
what
to
do
if
that
happens
15
to
be
the
case.

16
Another
issue
is
that
the
long­

17
term
variability
is
not
addressed
by
18
EPA's
procedure.
Standard
deviation,

19
over
time,
does
increase,
and
that
20
will
ultimately
affect
the
laboratory's
21
performance
of
this.

22
Again,
the
ML
is
based
on
a
23
multiplication
factor,
and
it
really
is
24
not
suitable
for
one
size
fits
all.

25
Really,
we
need
some
guidance
on
if
120
1
they
decide
to
stay
with
this
concept,
2
what
do
we
do
when
ten
standard
3
deviations
is
just
simply
inadequate?

4
We
also
feel
that
the
values,

5
the
MDL
and
ML
values,
that
you
obtain
6
should
be
something
that
can
be
7
routinely
achieved
by
most
competent
8
laboratories,
and
the
value
shouldn't
9
change.
Based
on
what
happens
now
is
10
you
develop
an
MDL
one
day,
and
it
is
11
basically
good
for
that
one
day.
It
12
may
only
be
good
for
one
calibration
13
in
some
cases.
So,
we
really
need
14
something
when
we
get
a
number
it
is
a
15
firm
number,
not
something
that
is
16
floating.

17
For
criterion
3,
the
approach
18
should
be
supported
by
a
practical
and
19
affordable
procedure
that
a
single
20
laboratory
can
use
to
evaluate
method
21
performance.
We
agree
with
this.

22
Bench
analysts
need
to
be
able
to
23
understand
and
to
use
these
types
of
24
concepts.

25
ACIL's
concept
really
is
cost
121
1
effective,
and
it
is
easy
to
use,

2
because
you
are
already
using
data
3
points
that
you
are
going
to
generate
4
in
your
analytical
batches
anyway.
5
Another
problem
with
evaluating
6
method
performance
is
the
lack
of
any
7
mention
of
a
guidance
on
how
to
8
determine
significant
figures
in
the
9
MDL
and
the
ML.
We
are
currently
10
seeing
regulatory
requirements
that
11
specify
the
number
of
significant
12
figures
that
must
be
reported,
and,
in
13
a
lot
of
cases,
it
exceeds
the
14
capability
of
the
method.

15
So,
that
is
a
problem
when
we
16
are
reporting
data
that
is
not
17
certain,
and
we
have
people
having
to
18
sign
off
on
DMRs.
So,
it
is
very
19
important
that
we
have
certain
data,

20
we
have
an
approach
or
a
guide
on
how
21
to
determine
the
significant
figures
22
and
how
to,
you
know,
appropriately
23
round
the
data,
because
we
really
need
24
consistency
in
this
area
for
data
sets
25
to
be
comparable.

122
1
Criterion
4
states
the
2
detection
level
approach
should
3
identify
the
signal
or
estimated
4
concentration
at
which
there
is
99
5
percent
confidence
that
the
substance
6
is
actually
present
when
the
analytical
7
method
is
performed
by
experienced
8
staff
in
a
well
operated
laboratory.

9
Because
the
EPA's
MDL
procedure
is
10
concentration
dependent,
it
does
not
11
correct
for
bias,
and
that,
in
a
lot
12
of
cases,
could
make
it
absolutely
13
impossible
to
consistently
recognize
14
the
analyte
as
being
present
at
a
99
15
percent
level
of
confidence.
So,
we
16
really,
again,
we
need
a
number
that
17
is
going
to
be
18
firm.

19
Criterion
5
states
that
the
20
quantitation
limit
approach
should
21
identify
the
concentration
that
gives
a
22
recognizable
signal
that
is
consistent
23
with
the
capabilities
of
the
method
24
when
the
method
is
performed
by
25
experienced
staff
in
a
well
operated
123
1
laboratory.
Because
the
ML
is
a
2
straight
multiplication
factor
of
your
3
MDL
and
you
are
trying
to
achieve
ten
4
standard
deviations
of
your
MDL,
you
5
can't
always
produce
a
recognizable
6
signal
at
ten
standard
deviations
of
7
your
ML,
and
there
is
no
guidance
on
8
how
to
handle
this
situation
if
your
9
method
or
your
capabilities
simply
10
can't
meet
that.
11
So,
we
would
like
to
see
12
something
in
the
regulations
stating
13
that
this
is
how
it
works.

14
Significant
figures,
again,
we
have
to
15
report
self
to
fulfill
regulatory
16
requirements
that
really
does
exceed
17
the
capabilities
of
the
method.

18
Unfortunately,
that
is
another
19
instance
of
the
regulations
ruling
the
20
science
instead
of
the
other
way
21
around.
We
really
need
to
base
this
22
stuff
on
good
science
and
have
the
23
regulations
follow
the
good
science.

24
The
rounding
rules
that
were
in
25
the
ML
procedure
don't
always
work,

124
1
because,
in
some
instances,
when
you
2
have
to
round
down,
you
are
violating
3
some
of
the
basic
precepts
of
the
4
quantitation
level.
What
is
happening
5
is,
when
you
are
rounding
down,
you
6
are
going
into
an
area
of
higher
7
uncertainty,
because
you
are
8
approaching
your
MDL.

9
Then,
that
ultimately
will
mean
10
that
you
can't
always
and
consistently
11
meet
a
99
percent
confidence
level
12
that
your
analyte
is
there,
let
alone
13
quantify
it,
and
this
is
a
real
issue,
14
because
we
need
to
have
data
that
the
15
laboratory
and
the
data
user
feel
16
confident
in.

17
It
is
very
difficult,
as
a
lab
18
person,
to
know
I
am
reporting
more
19
significant
figures
that
I
really
20
should
be,
than
I
really
can
see,
but
21
we
have
the
regulatory
requirements,

22
and
it
is
also
very
difficult
when
you
23
stop
and
think
maybe
I
really
don't
24
have
a
good
MDL.
Maybe
it
truly
was
25
a
snapshot
of
the
one
day
that
I
did
125
1
it,
and
because
of
the
way
the
ML
is
2
structured,
if
your
MDL
is
flawed,

3
then
your
ML
is
going
to
be
flawed
as
4
well.

5
So,
what
we
would
like
to
see
6
is
that
to
come
up
with
a
procedure
in
7
which
the
data
user
and
the
laboratory
8
have
a
great
deal
of
confidence
in
the
9
MDL
and
ML,
because
from
a
regulatory
10
perspective,
it
is
very
important
when
11
you
sign
off
on
your
DMR
that
you
feel
12
really
good
and
you
have
very
sound
13
data.

14
With
that,
I
would
like
to
turn
15
this
over
to
my
colleague,
Jim
Pletl,

16
for
further
comments.
17
MR.
TELLIARD:
Thanks,

18
Patty.

19
MR.
PLETL:
Can
20
everybody
hear
me
okay?
Patty
just
21
went
through
five
of
the
six
criteria
22
that
EPA
used
in
trying
to
select
an
23
approach
for
calculating
MDLs
and
MLs,

24
and
I
am
going
to
focus
on
the
sixth
25
criterion
which
is
really
the
126
1
regulatory
implications
of
the
2
proposal.
To
refresh
your
memory,

3
criterion
number
6
is
the
applicability
4
of
the
approach
to
decisions
under
the
5
Clean
Water
Act
and
support
for
State
6
and
local
implementation
of
7
measurements.

8
EPA
states
that
the
MDL/
ML
9
approach
proposed
has
been
successfully
10
applied
and
successfully
used
across
11
the
country,
but
EPA
did
not
define
12
the
term
successful
in
its
support
13
document.
It
appears
that
their
14
definition
is
based
on
the
fact
that
15
this
approach
has
been
used
in
various
16
regulatory
programs
over
a
period
of
17
time.

18
Success
is
usually
based
on
19
attainment
of
a
goal
or
goals
rather
20
than
how
long
or
how
hard
you
have
21
been
trying
to
meet
that
goal.
In
22
this
particular
case,
a
successful
23
approach
would
meet
the
needs
of
all
24
of
its
users
while
minimizing
decision
25
error
rates,
but
it
is
not
evident
127
1
that
that
evaluation
has
actually
taken
2
place.

3
For
example,
successful
4
application
of
an
MDL
approach
might
5
include
MDLs
and
MLs
where
the
limits
6
are
low
enough
to
ensure
protection
of
7
the
environment
but
to
allow
the
8
reporting
of
data
with
adequate
9
certainty
that
the
data
generated
was
10
actually
reliable.

11
So,
what
I
am
going
to
do
is
I
12
am
going
to
go
over
a
number
of
issues
13
that
were
either
touched
on
in
the
14
technical
support
document
that
relate
15
to
regulatory
issues
and
talk
a
little
16
bit
about
those
implications
and
why
17
we
believe,
at
least,
that
we
18
really
don't
have
evidence
or,
at
19
least,
obvious
evidence
that
the
20
approach
that
we
are
going
to
be
using
21
or
the
one
we
have
been
using
really
22
characterizes
success.
23
EPA
has
chosen
not
to
include
24
establishment
of
compliance
thresholds
25
using
MDLs
or
MLs
as
a
criterion
for
128
1
selecting
an
approach
for
developing
2
detection
and
quantitation
limits,
but
3
this
may
result
in
selecting
an
4
approach
that
does
not
meet
the
needs
5
of
all
of
its
users.
Data
collected
6
to
meet
NPDES
permits
are
often
7
interpreted
relative
to
detection
and
8
quantitation
limits,
to
determine
9
reasonable
potential
to
exceed
water
10
quality
standards,
limit
magnitude,
how
11
stringent
your
limit
actually
is,
and
12
whether
you
can
comply
with
that
13
limit.

14
The
proposed
detection
and
15
quantitation
limits
which
do
not
16
reasonably
address
all
factors
17
influencing
them
will
lead
to
either
18
false
positive
or
false
negative
errors
19
in
their
procedures,
and
the
procedures
20
I
am
referring
to
are
these
RP,
the
21
reasonable
potential,
and
the
limit
22
calculations.
Therefore,
the
certainty
23
and
reliability
of
detection
or
24
quantitation
limits
will
directly
25
determine
whether
the
discharger
129
1
receives
a
legally
enforceable
permit
2
limit,
the
magnitude
of
that
limit,

3
and
whether
a
discharger
can
comply
4
with
that
limit.

5
It
is
imperative
that
the
6
selected
approach
provide
detection
and
7
quantitation
limits
that
minimize
the
8
error
rates
of
these
procedures.
Our
9
review
indicates
that
the
proposal
made
10
by
EPA
does
not
adequately
address
all
11
those
factors
affecting
the
reliability
12
of
the
detection
and
quantitation
13
limits
and
that
the
uncertainties
of
14
regulatory
decisions
using
the
EPA
15
approach
are
and
will
be
unacceptable.

16
So,
in
our
view,
we
have
two
17
choices.
We
can
either
eliminate
the
18
use
of
the
MDL
and
ML
as
a
regulatory
19
decision
point
and
use
it
only
as
a
20
measure
of
lab
performance,
or
we
can
21
continue
to
use
MDLs
in
regulatory
22
decisions,
but
we
have
to
adjust
the
23
way
we
calculate
them
to
provide
the
24
certainty
required
to
make
regulatory
25
decisions.
It
is
one
or
the
other.

130
1
EPA
makes
quite
clear
in
the
2
support
document
that
it
is
focused
on
3
driving
MDLs
lower
to
make
sure
that
4
we
protect
the
environment,
and,

5
apparently,
from
the
way
it
is
6
currently
calculated,
they
believe
that
7
the
approach
adequately
addresses
false
8
positives
when
they
use
the
99
percent
9
confidence
level.
However,
there
is
10
no
quantitative
evaluation
showing
that
11
the
uncertainty
associated
with
false
12
negatives
is
balanced
by
that
for
13
false
positives.

14
In
fact,
the
decision
error
15
rates,
both
analytically
and
in
the
16
regulatory
program,
should
have
been
17
used
as
a
criterion
for
selecting
an
18
approach.
This
evaluation
is
necessary
19
to
conclude
that
the
proposed
approach
20
is
defensible
and
will
truly
lead
to
a
21
successful
program.

22
Another
issue
I
want
to
talk
23
about
is
censoring
which
was
briefly
24
addressed
in
the
support
document.

25
The
proposal
does
not
address
how
data
131
1
should
be
censored
relative
to
the
2
detection
and
quantitation
limits.

3
Yet,
as
I
just
talked
about,
these
4
limits
have
a
great
deal
of
impact
on
5
permit
issues.

6
Although
it
does
not
believe,

7
EPA
does
not
believe,
that
censoring
8
is
an
issue
when
selecting
an
MDL
or
9
ML
approach,
EPA
states
in
its
support
10
document
that
it
is
acceptable
to
use
11
the
worst
case
approach
of
reporting
12
data
less
than
MDL
or
ML
as
equal
to
13
these
values.
So,
even
though
they
14
don't
think
censoring
should
be
15
addressed,
they
voice
their
opinion
as
16
to
how
it
should
be
addressed.

17
The
detection
and
quantitation
18
limits
themselves
and
how
data
is
19
interpreted
relative
to
these
limits
20
are
critical
components
of
the
NPDES
21
program.
Without
instruction
in
the
22
regulation
or
guidance
on
how
to
23
interpret
data
less
than
detection
and
24
quantitation,
there
will
be
and
25
currently
is
great
variability
among
132
1
the
States
and
EPA
regions
on
whether
2
a
permittee
should
get
a
limit,
the
3
magnitude
of
that
limit,
and
compliance
4
with
the
limit.

5
EPA
acknowledges
this
openly
in
6
their
support
document.
EPA
also
7
states
that
censoring
is
an
issue
no
8
matter
which
approach
you
choose.
We
9
agree
and
believe
that
this
is
10
justification
in
itself
that
censoring
11
should
be
addressed
in
the
regulation.

12
Given
the
current
proposal,

13
dischargers
characterized
by
the
same
14
data
will
be
held
to
very
different
15
NPDES
requirements,
depending
on
their
16
location
in
the
country
rather
than
17
the
quality
of
that
discharge.
This
18
approach
does
not
adequately
balance
19
the
risk
of
decision
errors
and
20
unacceptably
displaces
regulatory
21
decisions
from
the
sample
quality
which
22
is
where
it
should
be
to
the
23
geographic
location
of
the
lab
or
the
24
regulatory
agency
that
is
making
the
25
decisions.

133
1
Additionally,
lack
of
regulation
2
on
this
topic
will
lead
to
data
sets
3
that
cannot
be
pooled
between
4
regulatory
agencies.
If
you
have
5
different
censoring
approaches,
you
are
6
going
to
have
a
mishmash
of
all
kinds
7
of
data
out
there,
and
this
is
going
8
to
hinder
any
efforts...
we
heard
Jim
9
Hanlon
talk
earlier
in
the
week
about
10
watershed
management...
this
certainly
11
is
going
to
hinder
that
effort.

12
Patty
talked
a
little
bit
about
13
the
issue
of
DMR
certification.
If
14
permittees
are
unsure
as
to
whether
15
data
were
truly
detectable
or
16
quantifiable,
based
on
the
approach
17
used
to
calculate
these
reporting
18
limits,
or
how
data
is
censored,
how
19
can
permittees
certify
the
data
as
20
true
and
accurate?
Once
data
is
21
signed,
it
is
no
longer
open
to
22
interpretation
or
challenge
by
a
23
permitted.
Therefore,
it
is
critical
24
that
permittees
be
very
sure
that
data
25
deemed
detectable
of
quantifiable
134
1
actually
meet
these
definitions
with
a
2
known
degree
of
uncertainty.

3
EPA
has
stated
in
the
WET
4
methods
rule...
and
WET
stands
for
Whole
5
Effluent
Toxicity,
and
these
were
6
methods
that
were
recently
finalized
7
back
in
November
of
2002...
that
the
8
detection
limit
concept
does
not
apply
9
to
WET
methods,
and
such
limits
cannot
10
be
calculated
for
WET
methods.

11
WET
methods
have
analytical
12
limitations
just
like
chemistry.
EPA
13
has
attempted
to
address
these
14
limitations
using
its
PMSD
approach,

15
and
those
of
you
who
were
here
earlier
16
in
the
week
heard
EPA
talk
about
that.

17
Although
the
intent
of
this
approach
18
is
to
address,
in
part,
false
19
positives,
it
does
not
provide
the
20
protection
against
false
positives
that
21
is
necessary
in
a
regulatory
program.

22
AMSA,
Association
of
23
Metropolitan
Sewage
Agencies,
has
met
24
with
EPA
and
provided
information
25
allowing
the
development
of
WET
135
1
censoring
points
comparable
to
EPA's
2
MDL,
but
it
is
too
early
to
say
where
3
that
is
going
to
go.
The
regulation
4
must
provide
an
approach
for
5
calculating
detection
limits
for
the
6
WET
methods
if
it
continues
to
hold
7
that
the
WET
methods
must
be
used
to
8
assess
compliance
with
water
quality
9
standards
and
form
the
basis
for
10
legally
enforceable
NPDES
limits.

11
EPA
plans
on
developing
MDLs
12
and
MLs
for
all
new
methods
that
it
13
develops,
but
it
also
needs
to
think
14
about
the
ones
that
it
has
already
15
promulgated
and
to
go
back
and
make
16
sure
that
we
have
those
spaces
filled
17
in.

18
In
its
proposal,
EPA
has
not
19
included
a
procedure
for
calculating
20
MDLs
and
MLs
for
methods
rather
than
21
MDLs
and
MLs
for
laboratories.
EPA
22
states
that
labs
must
achieve
23
acceptable
levels
of
sensitivity,
but
24
it
does
not
provide
a
procedure
for
25
defining
what
is
acceptable.

136
1
Standards
must
be
set
so
that
2
using
environmental
data
to
implement
3
regulatory
programs
like
the
NPDES
4
permit
program
will
know
the
5
limitations
of
those
methods
and
6
accommodate
those
limitations.
Failure
7
to
do
this
leaves
regulators
and
8
permittees
not
knowing
whether
specific
9
methods
are
capable
of
reliably
10
supporting
various
regulatory
decisions
11
and
actions.
This
regulation
must
12
include
methods
for
calculating
MDLs
13
and
MLs
which
most
competent
labs
can
14
meet
before
it
is
finalized
and
15
promulgated.

16
Methods
developed
by
consensus
17
based
organizations
must
also
provide
18
the
data
that
will
meet
this
new
19
requirement
before
EPA
can
adopt
them
20
for
Clean
Water
Act
use.
Clearly,

21
there
will
always
be
inconsistency
in
22
lab
performance
and
data
uncertainty
23
without
such
a
standard.

24
There
are
some
regulatory
25
programs
other
than
those
for
the
137
1
Clean
Water
Act
that
reference
the
MDL
2
procedure
proposed
by
EPA.
Yet,
this
3
current
rule,
as
I
understand
it,
is
4
restricted
to
Clean
Water
Act
programs.

5
For
consistency
in
decision
making
6
across
programs,
it
would
seem
7
desirable
to
have
one
approach
for
all
8
programs.

9
It
is
also
desirable
for
labs
10
conducting
analyses
for
multiple
11
programs
to
use
the
same
procedure
for
12
each
program.
Otherwise,
you
have
13
multiple
efforts
going
on
within
a
14
single
laboratory.

15
EPA
does
not
provide
sufficient
16
reasons
for
why
standardized
procedures
17
for
all
programs
using
the
MDL
and
ML
18
approach
is
not
being
proposed,
and
we
19
do
not
see
a
scientific
or
technical
20
reason
why
standardization
across
21
programs
cannot
occur.
We
ask
that
22
this
issue
be
revisited
and
reasons
be
23
better
communicated.

24
Finally,
EPA
states
in
its
25
support
document
that
all
new
methods
138
1
developed
by
EPA
for
promulgation
in
2
40
CFR
should
use
standardized
3
procedures
for
calculating
MDLs
and
4
MLs,
but
it
will
not
support
this
5
standard
for
other
organizations
6
developing
methods.
EPA
provides
four
7
reasons
for
this.

8
First,
EPA
states
that
no
9
single
method
will
meet
the
needs
of
10
all
organizations
and
applications.

11
Yet,
it
is
not
clear
that
an
12
identification
and
an
evaluation
of
13
these
needs
has
been
conducted.

14
We
would
hold
that
not
only
the
15
needs
of
organizations
developing
the
16
method
should
be
addressed
but
all
17
users
of
data,
including
the
18
permittees.

19
The
Office
of
Water
is
20
concerned
that
a
new
approach
would
21
expend
resources
in
other
programs
in
22
order
to
use
the
new
mandatory
23
approach.
This
may
be
true,
but
if
24
the
approach
is
better
technically,
and
25
it
meets
all
the
users'
needs,
139
1
basically,
we
are
inferring
that
we
2
shouldn't
go
there
even
if
we
come
up
3
with
a
better
method.

4
The
third,
EPA
believes
that
5
fewer
methods
would
be
promulgated
6
under
40
CFR
if
other
organizations
do
7
not
adopt
the
new
mandatory
approach
8
to
calculating
MDLs
and
MLs.
Well,

9
quite
frankly,
promulgation
of
new
10
methods
is
currently
at
a
standstill,

11
and
it
can't
get
much
worse.

12
Finally,
EPA
holds
that
the
new
13
approach
could
be
burdensome,
but
it
14
didn't
acknowledge
when
it
did
its
15
analysis
that
this
should
be
precluded
16
by
its
criterion
number
3
which
says
17
that
it
has
to
be
practical
and
18
affordable.

19
So,
EPA's
reasons
for
not
20
requiring
one
approach
for
calculating
21
MDLs
and
MLs
is
not
persuasive,
and
22
failure
to
do
so
will
simply
23
perpetuate
inconsistency
in
the
24
reliability
of
the
data.
In
summary,

25
we
believe
that
use
of
the
MDL
and
ML
140
1
concepts
in
regulatory
programs
demands
2
that
the
promulgated
approach
3
adequately
address
the
issues
4
associated
with
implementation
of
these
5
concepts
in
these
programs.

6
Thanks.

7
MR.
TELLIARD:
Thanks,

8
Jim.
Appreciate
it.

9
I
have
an
announcement.
We
are
10
asking
if
anyone
has
picked
up
a
11
leather
daytimer
belonging
to
Barbara
12
Burn...
pardon?
Beard.
That
is
how
13
you
spell
Beard?
Okay.
And
if
so,

14
would
you
please
return
it
to
Jen
who
15
is
at
the
registration
table?
So,
we
16
are
looking
for
a
leather
daytimer.

17
Thank
you.

18
Our
next
speaker
is
Scott
19
Houston.

20
MR.
HOATSON:
Hoatson.

21
MR.
TELLIARD:
Hoatson.

22
MR.
HOATSON:
Hello.

23
From
our
standpoint...
I
am
from
Delmar
24
Analytical,
and
I
have
been
in
the
25
commercial
lab
business
for
a
lot
of
141
1
years,
going
on
25,
I
think,
atthis
2
point
in
time.

3
A
lot
of
the
MDL
issues
that
we
4
have
or
that
I
see
are
not
necessarily
5
from
a
procedure
itself
but
on
how
it
6
is
used,
and
if
we
don't
change
how
it
7
is
used,
then
we
need
to
change
the
8
procedure.
So,
right
now,
up
here,
if
9
data
reported
below
the
level
of
10
calibration
wasn't
used
for
compliance
11
purposes,
we
probably
wouldn't
even
be
12
here,
because
this
is
where
the
13
biggest
argument
is.

14
The
MDL
is
used
currently
to
15
legitimize
reporting
of
data
below
16
levels
of
calibration
or
reasonable
17
quantitation.
Clients
and
regulators
18
treat
these
values
as
real
numbers.

19
This
is
a
little
example
here
20
of
we
have
lab
A
which
has
got
an
MDL
21
of
2
ppb
and
reports
a
value
of
2.6,

22
and
lab
B
has
got
an
MDL
of
4
and
23
reports
a
value
of
ND.
The
client
of
24
lab
A
may
have
to
do
more
testing
and
25
be
out
of
compliance,
whereas
the
one
142
1
for
lab
B,
since
they
have
a
not
2
detect,
is
just
find
and
dandy.
They
3
don't
do
anything.

4
So,
MDLs
should
be
stated
in
5
such
a
way
as
not
to
be
used
for
6
regulatory
compliance.

7
I
realize
that
we
don't
really
8
want...
or
the
MDL
procedure
is
for
40
9
CFR
136.
In
reality,
everybody
uses
10
it,
drinking
water,
hazardous
waste,

11
and
we
have
got
to
realize
that
12
whatever
is
finalized
is
going
to
be
13
used
throughout
the
industry.

14
Definitions
of
the
MDL
must
take
into
15
account
other
regulatory
uses
besides
16
Clean
Water
Act,
as
we
just
stated
a
17
few
minutes
ago.

18
Now,
whether
or
not
the
MDL
19
should
be
based
on
a
specific
program
20
or
not
is
another
item
that
I
kind
of
21
question,
because
it
states
currently
22
in
the
procedure
that
spike
levels
may
23
be
set
based
on
regulatory
objectives
24
rather
than
the
actual
method
ability.

25
This
gives
you
a
lot
of
variety
of
143
1
different
spike
levels
which,
in
turn,

2
give
you
different
MDLs.

3
The
ACIL
approach
is
based
on
4
method
rather
than
variable
based
on
5
different
programs,
and
I
agree
with
6
that.
Programs
should
set
minimum
7
requirements
that
must
be
achieved
and
8
all
labs
report
to
the
minimum
program
9
level
rather
than
using
the
MDL
and
10
use
the
MDL
to
show
that
it
is
11
achievable
but
not
on
an
ongoing
12
basis.

13
The
definition
of
an
MDL,
it
is
14
an
estimate
of
the
measured
15
concentration
in
which
there
is
a
99
16
percent
confidence
level
that
any
given
17
analyte
is
present
in
its
sample
18
matrix.
The
MDL
is
the
concentration
19
at
which
decision
is
made...
these
are
20
right
out
of
the
40
CFR
definition.

21
What
I
wanted
to
kind
of
focus
22
on
is
that
word
estimate.
Regardless
23
of
how
many
iterations,
how
many
times
24
we
run
samples,
it
is
always
going
to
25
be
an
estimated
value.
Whatever
144
1
manipulations
we
go
through,
they
are
2
all
estimated
values.

3
MDLs
apply
only
to
the
4
conditions
at
the
point
in
time
when
5
the
MDL
study
was
done,
and
there
is
6
no
discussion
regarding
that.

7
Daily
variations
in
instrument
8
conditions
and
sample
matrices
affect
9
MDLs
over
time.
So,
since
we
are
10
reporting
down
to
the
MDL,
just
doing
11
an
MDL
just
doesn't
quite
cut
it,
and
12
I
will
get
into
that
in
a
minute.

13
Can
we
simplify
the
process?
I
14
may
have
a
Polyanna
kind
of
point
of
15
view
on
it,
but
we
are
discussing,

16
since
we
are
talking
about
estimated
17
values,
let's
see
if
we
can
simplify
18
it
rather
than
make
it
more
19
complicated.

20
Can
we
simplify
the
process
and
21
still
achieve
the
desired
end
results?

22
I
think
we
can.

23
As
was
discussed
during
the
24
ACIL
presentation,
there
are
a
lot
of
25
differences...
actually
it
was
brought
145
1
up
in
another
one...
there
are
large
2
differences
in
basic
principles
between
3
the
different
types
of
analyses,
atomic
4
spectroscopy,
chromatography,

5
gravimetric
analysis,
filtered
versus
6
unfiltered.
At
least,
the
ACIL
7
approach
does
take
into
consideration
8
these
variations,
and
I
think
it
is
a
9
good
approach,
rather
than
trying
to
10
take
one
approach
for
it.

11
And
we
need
to
be
able
to
12
account
for
bias,
and
bias
shows
up
13
mostly,
that
I
have
seen,
in
methods
14
that
require
extractions.
They
tend
15
to
be
prone
to
bias.
There
is
usually
16
some
kind
of
loss.
Benzadine
is
my
17
favorite
case.
That
compound
doesn't
18
recover
worth
beans.

19
But
if
you
have
got
a
low
bias,

20
you
are
going
to
get
an
artificially
21
low
method
detection
level.
Actually,

22
I
think
our,
quote,
MDL
for
benzadine
23
calculates
out
to
be
something
we
will
24
never
see,
something
less
than
10
ppb
25
or
something
on
that
order,
and
it
146
1
doesn't
matter
what
you
spike
it
at.

2
You
will
never
get
it.

3
An
MDL
model
must
acknowledge
4
the
issues
related
to
bias.
The
ACIL
5
proposal
is
a
good
effort
towards
this
6
end.

7
The
number
of
analytes
in
a
8
method,
single
analyte
versus
multi
9
analyte,
and
the
number
of
instruments
10
used
can
have
a
bearing,
a
major
11
bearing,
on
the
practicality
and
12
economic
feasibility
of
the
approach.

13
As
was
discussed
earlier,
the
number
14
of
spike
levels
that
a
lab
might
have
15
to
do
to
achieve
the
MDLs
according
to
16
the
current
procedure
can
be
quite
17
excessive,
and
since
we
are
doing
this
18
on
several
instruments
besides
19
different
spike
levels,
this
number
20
just
gets
multiplied.

21
The
current
procedure
in
the
22
Federal
Register
is
to
estimate
the
23
detection
limit
and
we
are
going
to
24
take
concentration
approximately
equal
25
to
three
times
the
standard
deviation
147
1
of
the
replicate
measurements
of
the
2
blank,
or
we
take
the
lowest
3
concentration
that
can
be
detected
by
4
analyzing
samples
containing
5
successively
lower
concentrations.

6
Like
I
said
before
about
this
being
an
7
estimate,
expanding
on
these
items
8
would
make
for
a
very
usable
MDL,
and
9
as
was
discussed
earlier
also
in
the
10
ACIL
presentation,
these
two
premises
11
are
the
whole
basis
for
their
12
approach.

13
How
often
should
MDLs
be
14
performed?
There
is
no
discussion
in
15
there
whatsoever
about
that,
and
the
16
interpretation
on
how
often
they
need
17
to
be
analyzed
and
verified
is
wide
18
open
for
interpretation.

19
Since
MDLs
are
based
on
20
conditions
at
a
single
point
in
time,

21
since
we
are
reporting
down
to
them
22
and
actually
trying
to
use
those
23
numbers,
they
should
be
verified
at
a
24
regular
frequence
in
some
way,
shape,

25
or
form.
Conditions
change.

148
1
Right
now,
we
have
got
a
lot
of
2
clients
and
regulators
that
have
taken
3
the
extreme
approach,
and
they
want
4
the
full
MDL
studies
on
all
5
instrumentation
annually.
For
our
6
laboratory,
actually,
the
combination
7
of
our
three
laboratories,
and
since
8
this
crosses
all
programs
and
not
just
9
the
Clean
Water
Act,
this
costs
us
10
between
$
400,000
and
$
500,000
a
year
11
to
do
all
these
MDL
studies.
And
we
12
need
to
simplify
this.
I
mean,
we
13
need
to
verify
it,
but
we
have
got
to
14
simplify
it.

15
So,
the
current
procedure
is
16
estimate
the
lowest
detectable
amount,

17
and
then
we
are
going
to
seven
times
18
the
spike
replicates
at
one
to
five
19
times
that
estimated
MDL,
and
then
the
20
MDL
should
be
between
0.2
and
1
times
21
the
spike
amount
which
gets
us
back
to
22
the
first
number
in
the
first
place,

23
the
estimate.
So,
why
go
through
all
24
these
actions
just
to
get
us
back
to
25
the
same
number
we
started
with?
149
1
Calculated
MDLs
based
on
2
statistical
models
are
theoretical
only
3
and
have
the
potential
for
being
below
4
the
actual
detection.
I
constantly,

5
when
we
see
papers
being
given
on
new
6
methods,
new
procedures,
they
will
show
7
an
MDL
slide
up
there,
and
in
a
lot
8
of
cases,
if
you
look
at
the
9
calibration
curve,
the
intercepts
of
10
those
curves
don't
match
up
with
what
11
they
are
saying
the
MDLs
are.

12
So,
it
is
just
a
theoretical
13
number
that
doesn't
have
a
whole
lot
14
of
meaning,
but
it
is
well
documented.

15
So,
data
reported
at
or
above
the
16
calculated
MDLs
are
of
unknown,

17
although
well
documented,
quality.

18
By
basing
the
MDLs
on
a
19
function
of
the
mean
blank
value
plus
20
three
standard
deviations
or
actually
21
detected
spike
amounts,
we
can
have
an
22
MDL
that
could
be
supported
by
real
23
data
rather
than
an
MDL
that
is
24
theorized.
Effective,
flexible,

25
relatively
simple,
and
affordable.

150
1
And
thank
you
very
much.
2
MR.
TELLIARD:
Thanks,

3
Scott.
Our
next
speaker
is
Paul
4
Georgiou.

5
MR.
GEORGIOU:
I
am
Paul
6
Georgiou.
I
am
here
from
Bran
Luba
7
U.
S.
A.
I
am
working
on
about
four
8
weaknesses
that
I
have.
Usually,
I
9
improvise
speeches,
and
here
I
am
with
10
a
documented
speech.
So,
if
it
11
doesn't
go
that
smoothly,
you
will
12
know
why.

13
I
also
work
for
a
Fortune
500
14
company,
so
that
means
I
am
doing
the
15
work
of
five
people,
so
I
didn't
have
16
enough
time
to
prepare
for
this.
I
17
found
out
about
this
at
pit
time,
and
18
like
I
said,
I
have
been
working
19
through
all
the
documentation
since
20
then.

21
Also,
I
like
walking
around,
so
22
I
am
going
to
be
sitting
here
standing
23
in
one
place,
and
I
usually
take
24
questions
as
I
am
giving
a
speech,

25
because,
normally,
I
am
a
trainer.

151
1
So,
this
is
not
what
I
am
normally
2
used
to.

3
But
I
am
Paul
Georgiou,
a
4
senior
applications
chemist
engineer
5
from
Bran
Luba
U.
S.
A.
This
is
the
6
company
that
was
formerly
known
as
7
Technicon
Industrial
Systems.
We
8
probably
have
more
EPA
approved
methods
9
than
just
about
anybody
other
than
ICP
10
and
things
like
that.
We
are
a
part
11
of
SPX
Process
Equipment.
If
you
want
12
to
send
me
anything,
the
name
you
have
13
on
there
is
14
paul.
georgiou@
bran­
luba.
spx.
com.

15
I
have
13
years'
experience
16
running
inorganic
chemistries
for
the
17
EPA,
Waste
Management,
and
Bran
Luba.

18
This
kind
of
gives
me
a
unique
19
understanding
of
the
way
all
three
20
work,
because
I
have
worked
for
the
21
EPA,
and
I
have
worked
for
industry,

22
and
I
have
worked
for
an
instrument
23
manufacturer.
In
all
three
positions,

24
I
have
run
hundreds
of
MDLs.

25
One
statement
of
that
is
when
I
152
1
worked
for
Waste
Management,
we
were
2
doing
some
of
the
COP
protocols.
We
3
would
do
three
MDLs
on
all
the
ICP
4
methods
for
three
days
in
a
row.
So,

5
just
alone
there,
we
were
running,
I
6
don't
know,
200,
300
every
6
months.

7
So,
I
have
a
lot
of
experience
running
8
the
MDLs.

9
I
am
currently
an
applications
10
chemist
for
the
autoanalyzer
II,

11
autoanalyzer
III,
and
other
12
colorimetric
systems.
Additionally,
I
13
have
responsibility
for
the
regulatory
14
compliance
of
North
America.

15
Once
I
heard
about
EPA
changing
16
its
MDL
definitions,
I
was
quite
17
interested
in
the
discussions.
As
I
18
told
you,
I
have
done
hundreds
of
19
these
MDLs.

20
I
recall
slaving
over
an
AA2
21
with
exacting
standards
trying
to
get
22
an
acceptable
MDL
for
my
supervisors.

23
I
have
not
let
my
former
animosity
24
change
my
opinion
of
what
has
been
25
going
on
here.

153
1
I
started
looking
through
the
2
reams
of
paper.
I
went
on
the
3
Internet,
did
all
the
research
in
4
this,
and,
actually,
I
have
a
good
5
quantity
of
comments
which
I
won't
go
6
into,
but
I
will
give
the
information
7
to
the
EPA,
but
I
have
since
turned
8
into
a
support.
I
will
give
you
the
9
three
reasons
why
in
just
a
minute,

10
but
I
concur
with
the
peer
reviewers.
11
Each
one
of
these
MDLs,
there
12
are
strengths
and
weaknesses
to
each
13
one
of
them,
but
the
fact
is
that
14
most
industry
has
been
using
this
for
15
probably
20
or
30
years.
There
are
16
weaknesses.
For
instance,
if
you
talk
17
about
a
linear
curve,
as
the
gentleman
18
just
said
before,
you
know,
where
it
19
crosses
the
axis,
y=
mx+
b,
if
you
have
20
a
negative
slope,
your
blanks
will
be
21
negative
which
is
also
I
made
a
22
statement
of
that
later
in
my
talk.

23
I
kind
of
relate
the
MDL
24
discussion
to
who
created
the
largest
25
tax
increase
in
history,
which
154
1
President.
Well,
according
to
Bill
2
Clinton,
it
was
George
Bush
41.
Then,

3
according
to
who
Bob
Dole,
it
was
4
actually
Bill
Clinton.

5
As
a
scientist,
I
always
6
wondered,
how
can
two
completely
7
opposite
statements
be
true?
So,
I
8
did
some
investigation.
What
I
found
9
out
was
that
Bill
Clinton
had
the
10
largest
tax
increase
in
straight
11
dollars,
but
corrected
for
inflation,

12
George
Bush
41
did.
So,
we
have
two
13
completely
opposite
statements
which
14
are
both
technically
true.

15
I
kind
of
see
the
MDL
16
discussion
the
same
way.
Once
again,

17
industry...
like
I
said,
there
were
some
18
great
talks
here.
Each
one
of
them
I
19
have
seen,
you
know,
as
it
crosses
20
through
zero.
I
have
done
statistical
21
analysis
across
the
entire
curves
using
22
zero.
I
was
thinking
that
if
you
did
23
the
entire
curve,
you
would
get
a
24
higher
MDL
than
just
using
someplace
25
around
the
detection
limit.

155
1
Well,
as
I
looked
at
each
of
2
the
different
methods,
each
of
the
3
different
methods
was
different.
Some
4
had
a
higher
MDL;
some
had
a
lower
5
MDL.
So,
I
think
it
is
all
kind
of
a
6
discussion.
How
you
treat
your
data
7
is
essentially
how
you
are
going
to
8
come
up
with
your
MDL.

9
But
what
I
do
like
about
the
10
current
MDL
is
it
is
universal.
I
11
said
no
matter
how
much
money
your
12
laboratory
has
or
how
little
money
13
your
laboratory
has,
you
are
all
able
14
to
do
it.
And
just
as
a
gentleman
15
said
this
morning,
I
never
thought
16
about
the
titrametric
methods.
You
17
know,
I
am
doing
an
automated
method,

18
but
it
is
true
it
is
a
lot
more
19
difficult
to
run
those
methods
than
it
20
is
to
run
an
automated
system.

21
So,
it
is
all
dependent
on,
you
22
know,
what
you
are
doing
with
your
23
system.

24
Second
of
all,
with
some
of
the
25
changes
that
are
made,
as
you
are
able
156
1
to
correct
for
blanks
or
force
the
2
curve
through
zero,
as
long
as
it
is
3
stated
in
your
method,
you
are
able
to
4
use
that.
That
will
take
into
account
5
that.

6
We
have
seawater
methods
which
7
have
high
part
per
trillion
detection
8
methods.
Then,
we
go
out
to
a
9
laboratory,
and
they
start
using
it,

10
but
this
is
mainly
used
in
seawater,

11
so
they
want
to
see
10
ppb
accurately,

12
and
then,
when
they
start
running
it,

13
they
see
12
ppb,
but
they
know,

14
because
they
have
been
running
the
15
same
samples
over
and
over
and
over
16
again,
that
it
is
10
ppb.

17
So,
it
just
so
happened
that
18
there
was
an
offset
in
the
blank
of
2
19
ppb.
Due
to
the
regulations
right
20
now,
they
don't
really
state
whether
21
you
can
correct
for
that
blank
or
22
force
it
through
zero.
So,
that
is
23
another
way
the
MDLs
can
be
changed.

24
Also,
using
the
ML
which
is
ten
25
times
the
standard
deviation,
it
gives
157
1
a
better
argument
as
far
as
what
is
a
2
real
concentration.
Once
you
take
3
into
account,
you
know,
the
blank
and
4
where
all
that
stuff
is,
once
you
take
5
the
standard
deviation
of
that
number,

6
you
multiply
it
by
ten,
and
you
get
a
7
much
more
believable
number.

8
One
of
my
favorite
examples
9
inside
my
industry,
there
is
the
10
phenol
colorimetric
method.
It
states
11
that
you
can
get
a...
certain
regulatory
12
agencies
state
you
can
get
a
1
ppb
13
detection
limit.
Well,
me
14
who
is
in
this
industry
and
sells
the
15
instruments
say
I
have
never
reliably
16
seen
that.
So,
like
I
said,
our
17
instruments
can
probably
do
2
or
3
18
ppb,
but
still,
in
certain
regulations,

19
that
is
what
it
states,
but
that
was
20
because
it
was
a
regulatory
21
concentration,
not
whether
or
not
the
22
instrument
could
do
it.
23
So,
when
I
sell
instruments
to
24
get
the
1
ppb
detection
limit,

25
usually,
I
said
well,
it
probably
158
1
can't
be
done,
but
you
can
probably
2
get
2
or
3.
So,
otherwise,
we
haven't
3
sold
too
many
of
those
instruments
4
lately.

5
I
have
one
example.
Like
I
6
said,
I
have
put
all
this
7
documentation.
It
will
probably
be
8
whatever
they
put
on
the
Internet
in
9
the
future,
but
I
have
an
MDL
study
of
10
what
I
did
at
a
customer's
site.

11
I
went
out
to
a
customer's
12
site.
Oh,
this
instrument
is
a
piece
13
of
garbage,
and
they
couldn't
get
14
their
detection
limit.
They
were
15
looking
for
a
deviation
in
their
blank
16
of
plus
or
minus
10
ppb.
Through
17
experience,
what
I
have
found
out
is
18
usually...
which
was
stated
also
in
one
19
of
the
presentations
earlier
20
today...
that
a
deviation
in
the
blank
21
is
usually
higher
than
your
detection
22
limit.
I
just
found
that
through
23
experience.

24
But
what
I
did
find
is
I
was
25
running
these
samples.
I
got
a
159
1
0.0174,
0.0169,
0.0173,
roughly,
around
2
20
ppb.
I
got
a
standard
deviation
of
3
0.3
ppb
which
was
within
their,
you
4
know,
limits.
I
said,
to
be
quite
5
honest
with
you,
this
is
the
best
run
6
I
have
ever
seen
in
my
life.
I
got
a
7
1
ppb
detection
limit
with
a
1
ppm
8
high
standard.

9
I
have
this
thing
what
I
call
10
Georgiou's
thesis
which
customers
ask
11
what
detection
limit
can
you
get.

12
Inside
the
colorimetric
industry,
if
13
you
are
using
a
1
ppm
high
standard,

14
chances
are
you
should
probably
get
a
15
10
ppb
detection
limit.
On
our
AA3
16
system,
we
are
actually
able
to
do
17
half
of
that,
but,
you
know,
using
18
that
statement,
I
was
running
a
1
ppm
19
high
standard
on
total
phos.
So,
it
20
is
even
digested.
I
got
a
1
ppb
21
detection
limit,
and,
eventually,
they
22
came
back
to
me
and
said
well,
this
is
23
no
good,
the
QC
is
out.

24
The
concentrations
I
was
25
seeing,
the
0.0174,
was
actually
only
160
1
an
85.9
percent
recovery.
Problem
2
number
one.
They
were
doing
drinking
3
water,
so
it
is
supposed
to
be
plus
or
4
minus
10
percent.
So,
they
basically
5
threw
the
data
out
because
of
that.

6
Second
of
all,
the
7
concentration
of
the
MDL
was
too
high.

8
If
you
look
at
the
way
it
results
now,

9
you
are
supposed
to
use
three
to
five
10
times
the
MDL.
So,
with
a
1
ppb
11
detection
limit,
I
used
a
20
ppb.
So,

12
once
again,
I
had
to
throw
this
13
data
out,
because
the
concentration
was
14
too
high.

15
And,
last,
as
I
stated
before,

16
the
blank
results
were
deviating
more
17
than
the
actual
replicates.

18
I
think
most
of
this
can
be
19
taken
into
account
with
the
new
20
changes
that
are
being
recommended
by
21
the
EPA.
Like
I
said,
if
you
can
use
22
zero
forcing
or
zero
correction
or
23
things
like
that,
I
think
that
will
24
have
a
larger
effect
on
what
results
25
you
get
rather
than
what
is
actually
161
1
calculated.

2
Thank
you.

3
MR.
TELLIARD:
Thanks,

4
Paul.

5
The
Metropolitan
Water
6
Reclamation
District
of
Greater
Chicago
7
will
not
be
presenting
this
afternoon,

8
so
that
brings
to
a
close
this
public
9
meeting.

10
The
closing
date
for
comments
11
is
July
10th.
Please
keep
those
12
letters
and
cards
coming
in.
Any
13
checks
you
want
to
send
for
gifts,

14
because
you
really
like
it,
will
be
15
appreciated,
because
our
budget
is
16
really
low
this
year.

17
So,
thanks
for
coming
today.

18
The
information
in
question
is
up
on
19
our
website
if
you
want
to
take
a
look
20
at
it.
If
you
have
any
questions
on
21
getting
your
comments
in
or
anything
22
else,
feel
free
to
give
us
a
call.

23
Thanks
for
your
attention.
I
24
want
to
thank
the
speakers
for
taking
25
the
time
to
give
us
their
comments.
162
1
Have
a
good
day.

2
(
WHEREUPON,
the
Proceedings
were
3
concluded.)

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
163
1
CAPTION
2
.

3
The
Public
Meeting
in
the
4
matter,
on
the
date,
and
at
the
time
5
and
place
set
out
on
the
title
page
6
hereof.

7
.

8
It
was
requested
that
the
9
Meeting
be
taken
by
the
reporter
and
10
that
the
same
be
reduced
to
11
typewritten
form.

12
13
14
15
16
17
18
19
20
21
22
23
24
