Optimization
of
the
Sliced
Testis
Steroidogenesis
Assay
Draft
Part
I
Letter
Report
Research
Triangle
Institute
Laboratory
of
Reproductive
and
Endocrine
Toxicology
Chemistry
and
Life
Sciences
Group
Center
for
Life
Sciences
and
Toxicology
Post
Office
Box
12194
Research
Triangle
Park,
NC
27709­
2194
Battelle
Memorial
Institute
505
King
Avenue
Columbus,
OH
43201­
2693
EPA
Contract
No.:
68­
W­
01­
023
(
Battelle
Prime
Contractor,
WA
2­
27)
Contract
No.:
08055.001.020
Study
Code:
Rt02­
OPST
Master
Protocol
No.:
RTI­
870
TABLE
OF
CONTENTS
Page
1.0
Introduction
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
1.1
Background
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
1.2
Objectives
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
Figure
1.
Sliced
Testis
Steroidogenesis
Assay
Experimental
Design
Organizational
Diagram
.
.
6
2.0
Materials
and
Chemicals
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
2.1
Reagents
and
Solutions
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
2.2
Standards
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
2.2.1
Testosterone
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
2.2.2
Human
Chorionic
Gonadotropin
(
hCG)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
8
2.2.3
LDH
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
8
3.0
Method
for
Prototypical
Assay
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
8
Figure
2.
Technical
Flow
Illustration
of
the
Sliced
Testis
Steroidogenesis
Assay
.
.
.
.
.
.
.
.
.
.
.
.
9
4.0
Methods
for
Phase
I
­
Preliminary
Experimental
Phase
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
10
4.1
Testosterone
Radioimmunoassay
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
10
4.2
Lactate
Dehydrogenase
Spectrophotometric
Assay
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
4.3
Phase
I
Optimization
Experiments
for
Media
Type,
Gaseous
Atmosphere,
Rat
Age,
and
Storage
Container
Type
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12
4.3.1
Media
Type
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12
4.3.2
Gaseous
Atmosphere
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12
4.3.3
Rat
Age
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12
4.3.4
Storage
Container
Type
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
4.4
Phase
I
Experimental
Design
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
4.4.1
Factorial
Design
Experiments
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
Table
1.
Summary
of
Experimental
Factors
for
Phase
I
Optimization
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
Table
2.
Factorial
Test
Conditions
for
Phase
I
Optimization
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
14
4.4.2
Phase
I
Single
Factor
Experimental
Design
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
4.5
Phase
I
Data
Evaluation
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
5.0
Results
and
Statistical
Analyses
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
5.1
Testosterone
Ria
Verification
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
Table
3.
Values
of
Standards
on
Testosterone
RIA
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
16
Table
4.
Testosterone
RIA
Intra­
assay
CV
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
17
Table
5.
Testosterone
RIA
Percent
Recovery
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
17
Table
6.
Testosterone
RA
Parallelism
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
17
5.2
LDH
Verification
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
18
TABLE
OF
CONTENTS
Page
Table
7
Factor
levles
in
the
Phase
I
Assay
Optimization
Experiment
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
22
Table
8.
Data
Listing
for
Samples
Without
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
23
Table
9.
Data
Listing
for
Samples
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
24
Table
10.
Summary
of
Data
 
Original
Scale,
Without
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
26
Table
11.
Summary
of
Data
­
Log
Scale,
Without
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
Table
12.
Summary
of
Data
­
Original
Scale,
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
28
Table
13.
Summary
of
Data
­
Log
Scale,
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
29
Table
14.
Summary
of
Statistical
Analysis
of
Baseline
Data
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
30
Table
15.
Summary
of
ANOCOVA
Results
for
Individual­
Hour
Original­
Scale
Models
.
.
.
.
.
.
.
31
Table
16.
Adjusted
Mean
Levels
Based
on
Original­
Scale
Models
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
32
Table
17.
Summary
of
ANOCOVA
Results
for
Individual­
Hour
Log­
Scale
Models
.
.
.
.
.
.
.
.
.
.
33
Table
18.
Adjusted
Mean
Levels
Based
on
Log­
Scale
Models
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
34
Table
19.
Least
Squares
Means
for
Reduced
Log­
Scale
Model:
Without
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
35
Table
20.
Least
Squares
Means
for
Reduced
Original­
Scale
Model:
Without
hCG
.
.
.
.
.
.
.
.
.
35
Table
21.
Least
Squares
Means
for
Reduced
Log­
Scale
Model:
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
36
Table
22.
Least
Squares
Means
for
Reduced
Original­
scale
Model:
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
36
Table
23.
Analysis
of
Differences
in
Levels
for
With
and
Without
hCG
Stimulation:

Original­
Scale
Models
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
37
Table
24.
Adjusted
Mean
Differences
(
With­
Without
hCG),
Based
on
Original­
Scale
Models
.
38
Table
25.
Analysis
of
Differences
in
Levels
for
With
and
Without
hCG
Stimulation:
39
Log­
Scale
Models
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
39
Table
26.
Adjusted
Mean
Differences
(
With­
Without
hCG),
Based
on
Log­
Scale
Models
.
.
.
.
40
Table
27.
Phase
IB
Data:
Samples
Without
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
43
Table
28.
Phase
IB
Data:
Samples
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
44
Table
29.
Summary
of
Data
­
Original
Scale,
Without
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
45
Table
30.
Summary
of
Data
­
Log
Scale,
Without
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
46
Table
31.
Summary
of
Data
­
Original
Scale,
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
47
Table
32.
Summary
of
Data
­
Log
Scale,
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
48
Table
33.
Summary
of
Results
by
Sample
Condition
­
Original
Scale,
Without
hCG
.
.
.
.
.
.
.
.
49
Table
34.
Summary
of
Results
by
Sample
Condition
­
Log
Scale,
Without
hCG
.
.
.
.
.
.
.
.
.
.
.
.
49
Table
35.
Summary
of
Results
by
Sample
Condition
­
Original
Scale,
With
hCG
.
.
.
.
.
.
.
.
.
.
.
50
Table
36.
Summary
of
Results
by
Sample
Condition
­
Log
Scale,
With
hCG
.
.
.
.
.
.
.
.
.
.
.
.
.
.
50
6.0
Discussion
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
51
7.0
Conclusions
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
51
8.0
References
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
51
Appendix
1.
LDH
Validation
and
Verification
with
Media
199
without
Phenol
Red
.
.
.
.
.
.
.
.
.
.
52
4
1.0
INTRODUCTION
1.1
Background
In
1996,
the
Food
Quality
Protection
Act
(
FQPA)
amendments
were
enacted
by
Congress
to
authorize
the
Environmental
Protection
Agency
(
EPA)
to
implement
an
Endocrine
Disruptor
Screening
Program
(
EDSP)
on
pesticides
and
other
substances
found
in
food
or
water
sources
for
endocrine
effects
in
humans
(
FQPA,
1996).
In
this
program,
comprehensive
toxicological
and
ecotoxicological
screens
and
tests
are
being
developed
for
identifying
and
characterizing
the
endocrine
effects
of
various
environmental
contaminants,
industrial
substances,
and
pesticides.
A
two­
tiered
approach
will
be
utilized.
Tier
1
employs
a
combination
of
in
vivo
and
in
vitro
screens,
and
Tier
2
involves
in
vivo
testing
methods
using
two­
generation
reproductive
studies.
A
steroidogenesis
assay
is
proposed
as
one
of
the
Tier
1
screening
battery
assays.

A
detailed
review
paper
(
DRP)
about
steroidogenesis
was
prepared.
The
DRP
(
1)
summarized
the
state
of
the
science
of
the
in
vivo,
ex
vivo,
and
in
vitro
methodologies
available
for
measuring
gonadal
steroidogenesis;
(
2)
for
each
methodology,
presented
a
review
of
the
individual
assays
and
representative
data
generated
by
investigators
who
used
the
assay
to
evaluate
a
substance
for
steroidogenic­
altering
activity;
(
3)
provided
an
evaluation
of
the
various
methodologies
and
the
assays
as
tools
for
screening
substances
with
suspected
steroidogenic
activity;
(
4)
recommended
a
particular
screening
method
and
assay
as
a
screening
tool;
and
(
5)
described
the
strengths,
weaknesses,
and
implications
for
further
research
associated
with
the
recommended
screening
assay.

The
in
vitro
sliced
testis
steroidogenesis
assay
was
selected
as
the
most
promising
screening
tool
for
identifying
substances
with
steroidogenic­
altering
activity.
The
sliced
testis
assay
was
recommended
because
it
can
be
conducted
at
a
minimal
cost,
quickly,
and
simply
with
standard
laboratory
equipment
and
basic
laboratory
training;
the
preparation
is
stable
and
the
parenchyma
remains
viable
over
a
sufficient
time
period
to
measure
changes
in
end­
product
hormone
production;
the
assay
is
relatively
sensitive
and
specific;
the
assay
uses
parenchyma
that
maintains
the
cytoarchitecture
of
the
organ;
the
assay
uses
a
reduced
number
of
animals
(
up
to
quartered
testis
slices);
the
assay
should
be
relatively
easy
to
standardize
(
by
optimization);
and
the
assay
has
a
well­
defined
endpoint
in
testosterone
and,
if
desired,
can
be
modified
to
include
additional
intermediate
hormonal
endpoints.

Although
a
promising
tool,
the
sliced
testis
assay
remains
to
be
fully
tested
as
an
assay
that
can
meet
all
the
demands
of
an
endocrine
disruptor
screening
tool.
Concerns
raised
by
the
EPA
and
Endocrine
Disruptor
Methods
Validation
Subcommittee
(
EDMVS)
suggested
that
experiments
be
conducted
to
ensure
the
optimization
of
the
assay
prior
to
more
rigorous
pre­
validation
and
validation
testing.
The
most
notable
concerns
were
associated
with
1)
various
incubation
variables,
2)
variables
that
5
affect
optimal
human
chorionic
gonadotrophin
(
hCG)
stimulation,
3)
characterization
of
the
parenchymal
post­
slicing
equilibration
time,
and
4)
parenchymal
viability.
In
addition
to
these
most
notable
concerns,

other
factors
that
could
potentially
affect
the
optimal
performance
of
this
assay
were
identified.
The
objective
of
this
optimization
was
to
describe
in
detail
the
experiments
designed
to
provide
data
for
setting
in
place
the
procedures
and
parameters
that
will
optimize
the
performance
of
this
assay.

1.2
Objectives
The
study
plan
for
testing
the
factors
described
in
the
previous
section
involved
two
phases
and
utilized
single
factor
and
factorial
experimental
designs.
A
diagram
of
the
experimental
design
for
this
study
is
illustrated
in
Figure
1.
6








	


















































































The
study
was
divided
into
Phases
I
and
II.

In
Phase
I,
the
Preliminary
Experimental
Phase,
the
analytical
assays
planned
for
use
were
verified,
storage
containers
and
lengths
of
storage
were
selected
and
three
factors
that
may
affect
the
7
performance
of
the
assay
were
tested.
The
reasoning
for
including
these
three
factors
in
the
preliminary
phase
was
to
establish
early
whether
a
given
level
of
each
factor
was
going
to
affect
assay
performance.

Although
any
factor
listed
in
the
study
plan
could
have
been
rationalized
to
fit
such
a
criteria,
inclusion
in
the
preliminary
phase
also
required
that
the
factor
be
unlikely
to
have
an
interaction,
or
at
best
a
minimal
interaction,
with
another
experimental
factor.
Although
subjective,
these
factors
were
believed
to
best
fit
these
criteria.
Furthermore,
it
was
believed
essential
to
establish
the
optimal
level
for
each
of
these
factors
before
proceeding
with
the
factorial
experiments
since
an
effect
of
one
of
these
would
require
additional
verification
experiments
after
sensitivity
analysis.
Finally,
by
establishing
the
media
type
early
on
in
the
experiment,
the
analytical
assay
verification
testing
(
Phase
I
)
and
Optimization
of
Sample
Testing
(
Phase
II)
could
be
initiated
earlier
in
the
study
milestone
schedule.

2.0
MATERIALS
AND
CHEMICALS
2.1
Reagents
andSolutions
No
test
substances
were
used
in
this
study.
The
chemicals
used
in
this
study
were
used
to
prepare
reagents
and
solutions
for
the
assay.
All
reagents
and
solutions
had
appropriate
information
documented,
which
included
the
identity,
concentration,
storage
requirements,
and
expiration
date.

Reagents
and
solutions
were
prepared
according
to
Standard
Operating
Procedures.

2.2
Standards
Verification
of
the
analytical
assays
required
testosterone
for
the
radioimmunoassay
(
RIA)

method
and
LDH
for
the
spectrophotometric
assay.
In
addition,
hCG
was
used
as
a
stimulant
of
the
sliced
testis
bioassay.
These
substances
were
considered
standards.

2.2.1
Testosterone
Chemical
Name:
Testosterone
CAS
No.:
58220
Molecular
Weight:
288.4
Solubility:
Clear
colorless
to
very
faint
yellow
solution
at
100
mg/
mL
in
chloroform
Supplier:
Sigma­
Aldrich
Chemical
Company
Lot
No.:
6384­
70­
8
Purity:
NLT
98%

Storage
Conditions:
2
Year
shelf
life
A
safety
protocol
exists
for
the
use
of
the
radioactive
form
of
testosterone.
8
2.2.2
Human
Chorionic
Gonadotropin
(
hCG)

Chemical
Name:
hCG
CAS
No.:
9002­
61­
3
Molecular
Weight:
36,700
Solubility:
H
2
O
Supplier:
Calbiochem
Lot
No.:
B11174,
B51120
Purity:
approx.
30%
hCG
by
weight
Storage
Conditions:
Freezer
(­
20
°
C).
Following
reconstitution,
aliquot
and
freeze
(­
20
°
C).
Stable
for
2
years
as
supplied.

2.2.3
LDH
Chemical
Name:
Lactate
Dehydrogenase
CAS
No.:
EC
1.1.127
Source:
Rabbit
muscle
Solubility:

Supplier:
Sigma­
Aldrich
Chemical
Company
Lot
No.:
99H7480
Purity:
NLT
%

Storage
Conditions:
2­
8
°
C
Year
shelf
life
3.0
METHOD
FOR
PROTOTYPICAL
ASSAY
The
prototypical
assay
describes
the
sliced
testis
assay
using
the
conditions
that
are
believed
to
be
the
starting
conditions
of
the
assay.
This
section
does
not
describe
any
experiments
to
be
conducted;

rather,
it
describes
the
settings
of
all
factors,
except
for
the
one
that
is
being
tested
in
order
to
perform
the
optimization
experiments
described
in
the
following
sections.
A
run
is
defined
as
a
single
sample
vessel
with
assay
components.

The
sliced
testis
assay
prototype
uses
a
15
week
old
male
Sprague­
Dawley
rat,
which
is
euthanized
and
its
testes
removed.
The
testes
are
decapsulated,
weighed,
and
placed
in
cold
(
4
°
C)

DPBS.
The
media
is
medium­
199
(
Gibco)
that
has
added
0.71
g
sodium
bicarbonate,
2.1
g
HEPES,
1.0
g/
L
BSA,
and
0.025
g/
L
soybean
trypsin
inhibitor,
and
is
adjusted
to
a
pH
of
7.4.
The
time
from
removal
to
the
time
of
slicing
is
held
to
under
1
hour.
Each
testis
is
sliced
along
the
longitudinal
axis
into
approximately
4
slices.
Each
slice
is
placed
in
a
20
mL
borosilicate
scintillation
vial
(
loosely
capped)
that
contains
5
mL
of
media
(
Figure
2).
The
vials
containing
the
testicular
sections
and
media
are
incubated
at
9
Figure
2.
Technical
Flow
Illustration
of
the
sliced
testis
steroidogenesis
assay
34
°
C
on
a
shaker
(
at
135
rpm)
in
5
percent
CO
2/
95
percent
air.
After
the
first
period
of
incubation,
the
media
is
removed
and
discarded.
Fresh
media
(
5
mL)
is
added
to
the
vial
and
an
aliquot
of
media
(
0.5
mL)
is
collected.
The
sample
is
centrifuged
and
the
supernatant
transferred
to
a
labeled
vial
and
stored
at
approximately
­
70
°
C
in
a
siliconized
plastic
container
for
no
more
than
one
month.
This
sample
is
the
baseline
sample.
Next,
one
half
of
the
vials
are
challenged
with
a
stimulant,
e.
g.,
hCG,
and
the
other
half
are
not.
The
final
hCG
concentration
is
0.1
IU/
mL.
Additional
media
samples
are
collected
from
the
vials
at
1,
2,
3,
and
4
hours
post­
challenge.
These
media
samples
are
also
stored
frozen
for
later
analysis.

Samples
are
analyzed
for
testosterone
using
an
RIA
method.
All
samples
for
a
given
day's
set
of
runs
should
be
analyzed
in
the
same
testosterone
RIA.
10
4.0
METHODS
FOR
PHASE
I
­
PRELIMINARY
EXPERIMENTAL
PHASE
Phase
I
is
comprised
of
the
verification
experiments
for
the
two
analytical
assay
methods
(
testosterone
RIA
and
LDH
spectrophotometry),
the
determination
of
the
preferred
storage
container
and
storage
length,
and
the
optimization
experiments
for
the
three
factors
to
be
tested
(
media
type,
incubation
atmosphere,
and
animal
age).
The
testosterone
RIA
method
was
verified
prior
to
conducting
any
optimization
assay
since
it
is
was
needed
as
an
endpoint
to
determine
optimization
of
the
sliced
testes
assay.
The
storage
container
determination
was
also
performed
before
any
further
optimization
assays
were
run.
The
determination
was
made
using
standards
of
testosterone
in
the
prototypical
media,

modified
M­
199
without
phenol
red.

4.1
Testosterone
Radioimmunoassay
The
objective
of
this
experiment
was
to
verify
that
testosterone
can
be
measured
in
the
sliced
testis
assay
media.
A
RIA
commercial
kit
(
Diagnostic
Products
Corporation,
Los
Angeles,
CA),
that
utilizes
125I­
testosterone
and
a
testosterone­
specific
antibody
affixed
to
polypropylene
culture
tubes,
was
used
to
measure
testosterone.
The
assay
was
verified
in
all
three
of
the
potential
assay
media.

Testosterone
(
Sigma,
St.
Louis,
MO;
T­
1500)
was
used
for
preparing
the
standard
curve
and
was
stored
desiccated
at
room
temperature
in
the
RTI
vault.
A
standard
was
prepared
in
ethanol
(
0.1
mg/
ml).

Up
to
an
8­
point
standard
curve
but
not
less
than
a
4­
point
standard
curve
was
prepared
using
standards
with
concentrations
of
0.07,
0.16,
0.41,
1.02
,
2.56,
6.4,
16,
and
40
ng/
ml
in
PBS­
Gel
Buffer
(
0.1
M
phosphate
buffered
saline
with
0.1%
(
w/
v)
sodium
azide
and
0.1%
(
w/
v)
gelatin,
pH
7.4).
In
addition,

procedural
controls
were
included
in
each
run.
The
standard
curve
points
and
the
procedural
controls
were
prepared
in
quadruplicate;
the
bioassay
unknowns
and
the
internal
standards
(
see
below)
were
prepared
in
duplicate.
The
volume
of
all
standards
and
controls
(
including
bioassay
unknowns)
were
adjusted
to
50

L
by
adding
the
PBS­
Gel
Buffer.
Next,
1
ml
125I­
testosterone
was
added
to
each
antibody­
coated
tube
and
mixed
(
Vortex).
The
tubes
were
incubated
in
a
37

C
water
bath
for
three
hours,

during
which
time
testosterone,
whether
labeled
or
unlabeled,
competed
for
testosterone
specific
antibody
binding
sites.
At
the
end
of
the
incubation
period,
the
free
(
unbound)
testosterone,
in
the
supernatant
fluid
of
all
tubes
was
aspirated
and
tubes
were
wiped
clean
of
fluid.
The
bound
testosterone
was
counted
in
a
gamma
counter
for
1
minute.
The
concentration
of
testosterone
was
estimated
against
the
standard
curve.
Values
were
reported
as
a
mean
concentration
(
ng/
mL)
of
duplicate
analyses,
unless
only
a
single
value
was
available.
Verification
of
the
testosterone
assay
involved
preparation
of
internal
standards
(
at
least
three)
using
spiked
media
with
concentrations
ranging
from
12.5
to
500
ng/
mL.
Each
concentration
was
run
at
each
of
three
volumes
­
10,
25,
and
50
µ
L,
to
check
for
parallelism,
and
each
sample
was
adjusted
to
50
µ
L
by
adding
the
PBS­
Gel
Buffer.
The
low
and
high
standards
were
analyzed
in
at
least
duplicate.
Verification
was
based
on
results
determined
for
accuracy,
precision,
specificity,
and
linearity.

Accuracy
is
expressed
as
the
relative
error,
which
was
determined
by
comparing
the
measured
to
the
target
concentration.
Relative
errors
within
15
percent
were
acceptable.
Precision
is
expressed
as
the
11
relative
standard
deviation
(
RSD)
or
coefficient
of
variance
(
CV),
which
was
determined
by
calculating
the
mean
and
standard
deviation
(
sd)
of
the
low
and
high
standards.
A
RSD
or
CV
within
15
percent
was
acceptable.
The
sensitivity
was
acceptable
if
the
means
of
the
blanks
and
low
standards
were
significantly
different
at
the
5
percent
significance
level.
Linear
determinations
of
the
standard
curve
line
were
made
and
a
correlation
coefficient
(
r)
calculated.
An
r
of
0.90
or
greater
was
considered
acceptable.

Inter­
and
intra­
assay
variability
was
determined.
The
intra­
assay
variability
was
determined
from
the
precision
results
calculated
from
the
results
obtained
by
measuring
the
low
and
high
standards
in
triplicate.
The
inter­
assay
variability
was
determined
by
repeated
analyses
of
the
standards
by
generating
a
standard
curve
on
three
different
days.

4.2
Lactate
Dehydrogenase
Spectrophotometric
Assay
The
objective
of
this
experiment
was
to
verify
that
LDH
can
be
measured
in
the
sliced
testis
assay
media.
The
LDH
assay
measures
the
rate
at
which
NADH
is
formed
when
NAD
is
reduced
when
it
catalyzes
the
oxidation
of
lactate
to
pyruvate.
NADH
is
measured
at
340nm
using
a
kineticspectrophotometric
method.
The
assay
and
samples
are
temperature
sensitive.
The
samples
should
not
be
refrigerated
or
frozen.
The
assay
has
been
characterized
for
assay
conditions
at
37oC.
LDH
activity
is
expressed
in
U/
L.

Qualification
of
the
assay
using
Media
199
without
phenol
red
consisted
of
determination
of
sensitivity
and
dilutional
linearity.
These
were
the
only
qualification
tests
specific
to
media
use.

Intraassay
imprecision
and
inter­
assay
imprecision
were
based
on
serum
based
controls
normally
used
for
instrument
monitoring
of
quality
control.
Accuracy
was
based
on
the
linear
regression
of
a
media
sample
spiked
with
purified
LDH
and
diluted
with
media.

Intra­
assay
imprecision
was
measured
by
assaying
ten
quality
control
samples
within
a
single
run.

The
mean
standard
deviation
(
SD)
and
coefficient
of
variation
(
CV)
was
calculated.
The
CV
should
be
less
than
10%.
Inter­
assay
imprecision
was
measured
by
assaying
ten
quality
control
samples
over
ten
separate
runs.
The
mean
standard
deviation
(
SD)
and
coefficient
of
variation
(
CV)
was
calculated.
The
CV
should
be
less
than
10%.

Sensitivity
of
the
assay
or
limit
of
detection
was
determined
by
analysis
of
twenty
samples
of
media
with
no
LDH
present
in
the
media.
A
low
level
of
control
was
also
analyzed
to
determine
the
lowest
level
of
detection.
Two
standard
deviations
from
the
mean
of
the
media
activity
was
used
as
the
limit
of
detection.

Dilutional
linearity
was
used
to
determine
accuracy
and
linearity
limits.
A
sample
of
media
was
spiked
with
purified
LDH
(
Sigma
#
L1254).
This
was
diluted
with
Media
199
without
phenol
red
to
within
12
the
linear
limits
of
the
assay.
A
series
of
dilutions
by
serial
dilution
of
the
spiked
sample
was
conducted
in
triplicate.
The
linear
regression
of
the
line
compared
to
the
theoretical
activity
should
be
between
0.990
and
1.100.

All
calculations
were
performed
using
EP
Evaluator,
release
3.0
statistical
analysis
software
from
David
Rhoads
Associates,
Inc.,
Kennett
Square,
PA.

4.3
Phase
1
Optimization
Experiments
for
Media
Type,
Gaseous
Atmosphere,
Rat
Age,
and
Storage
Container
Type
4.3.1
Media
Type
The
objective
was
to
determine
the
effect
of
different
types
of
media
(
with
specified
components)
on
testosterone
production
using
the
sliced
testis
assay.
The
prototypical
assay
conditions
were
used
except
with
regard
to
the
types
of
media.

The
media
types
tested
were:

RPMI­
1640
media
(
without
phenol
red),
10%
FCS,
50
ug/
mL
soybean
trypsin
inhibitor
Medium­
199
(
Gibco),
0.71
g
Na
bicarbonate,
2.1
g
HEPES,
1.0
g/
L
BSA,
0.025
g/
L
soybean
trypsin
inhibitor,
adjusted
to
pH
7.4
Eagles
MEM
4.3.2
Gaseous
Atmosphere
The
objective
was
to
determine
the
effect
of
different
types
of
gaseous
atmospheres
on
testosterone
production
using
the
sliced
testis
assay.
The
prototypical
assay
conditions
were
used
except
with
regard
to
the
gaseous
atmosphere.

The
atmospheres
tested
were:

5%
CO
2/
95%
air
5%
CO
2/
95%
O
2
air
(
three
gases).

4.3.3
Rat
Age
The
objective
will
be
to
determine
the
effect
of
age
on
testosterone
production
using
the
sliced
testis
assay.
The
prototypical
assay
conditions
will
be
used
except
with
regard
to
the
age
of
the
rat
used
to
obtain
the
testes.
13
The
ages
to
be
tested
are:

11
weeks
of
age
15
weeks
of
age
22
weeks
of
age.

4.3.4
Storage
Container
Type
The
objective
was
to
determine
the
effect
of
storage
type
containers
on
the
stability
of
testosterone
in
media.
The
types
of
containers
tested
were:
siliconized
plastic
and
non­
siliconized
plastic.

4.4
Phase
1
Experimental
Design
4.4.1
Factorial
Design
Experiments
These
experiments
were
conducted
as
a
33
full
factorial
design
with
one
replicate
per
condition.

The
experimental
factor
levels
are
summarized
in
Table
2.
The
factorial
test
conditions
are
displayed
in
Table
3.
The
27
factor
level
combinations
were
run
in
random
order.
Each
combination
was
run
with
and
without
hCG
stimulation,
for
a
total
of
54
test
runs.
For
each
test
run
responses
(
ng
T/
ml)
were
determined
at
1,
2,
3,
and
4
hours
after
media
refreshment.

Table
1.
Summary
of
Experimental
Factors
for
Phase
1
Optimization
Factor
Identification
Units
Experimental
Levels
Coded
Experimental
Levels
1
2
3
1
2
3
Media
Type
NA
RPMI­
1640
medium­
199
Eagles­
MEM
­
1
0
+
1
Gaseous
Atmosphere
NA
5%
CO
2
/
95%
air
5%
CO
2
/
95%
O
2
air
­
1
0
+
1
Rat
Age
wks
11
15
22
­
1
0
+
1
NA
=
not
applicable.
14
Table
2.
Factorial
Test
Conditions
for
Phase
1
Optimization
Experiment
Media
Type
Gaseous
Atmosphere
Rat
Age
­
1
­
1
­
1
­
1
­
1
0
­
1
­
1
+
1
­
1
0
­
1
­
1
0
0
­
1
0
+
1
­
1
+
1
­
1
­
1
+
1
0
­
1
+
1
+
1
0
­
1
­
1
0
­
1
0
0
­
1
+
1
0
0
­
1
0
0
0
0
0
+
1
0
+
1
­
1
0
+
1
0
0
+
1
+
1
+
1
­
1
­
1
+
1
­
1
0
+
1
­
1
+
1
+
1
0
­
1
+
1
0
0
+
1
0
+
1
+
1
+
1
­
1
+
1
+
1
0
+
1
+
1
+
1
15
4.4.2
Phase
1
Single
Factor
Experimental
Design
The
objective
of
this
set
of
experiments
was
to
determine
the
effect
of
storage
container
type
on
the
stability
of
testosterone
in
the
media.
This
experimental
series
did
not
use
the
sliced
testis
assay.
Stability
was
assessed
as
a
function
of
sample
handling
factors.
The
incubation
medium
that
will
be
used
will
be
determined
in
the
factorial
experiments
(
Phase
I).
To
conduct
the
stability
experiments,
a
known
amount
of
testosterone
was
added
to
the
media
to
achieve
a
specified
target
concentration.
The
target
concentration
was
determined
from
the
media
type
experiments
in
Phase
I
and
was
in
the
range
of
the
lowest
testosterone
concentrations
measured
in
the
sliced
testis
assay.
Using
this
target
concentration,
the
measured
concentration
was
compared
to
the
target
concentration.
The
stability
was
evaluated
based
on
the
difference
between
the
measured
and
target
concentrations.
If
no
statistical
difference
existed
at
the
5
percent
level,
then
the
sample
was
determined
to
be
stable
under
the
conditions
tested.

4.5
PHASE
I
DATA
EVALUATION
Upon
completion
of
the
Phase
I
optimization
experiments,
the
results
were
reviewed
for
a
possible
change
in
the
prototype
conditions
with
regard
to
the
four
factors
tested.
A
decision
was
made
as
to
whether
the
Phase
II
optimization
experiments
would
be
conducted
using
the
original
prototype
or
modifications
made
to
the
media
type,
atmosphere,
age
of
rat
and/
or
storage
container
type
used
for
the
remaining
experiments.
The
endpoint
was
the
amount
of
testosterone
released
into
the
media
during
the
sliced
testis
assay.
The
test
conditions
resulting
in
the
highest
values
for
testosterone
were
used
in
Phase
II.

5.0
RESULTS
AND
STATISTICAL
ANALYSES
5.1
Testosterone
RIA
Verification
The
Testosterone
RIA
kit
from
Diagnostic
Products
was
used
to
verify
that
the
M­
199
media
without
phenol
red
could
be
used
for
the
sliced
testis
assay
and
provide
accurate
results
for
the
RIA
assay.
16
Table
3.
Values
of
Standards
on
Testosterone
RIA
Value
of
Standard
Factor
Reading
8
ng/
ml
1
9.33
8
ng/
ml
1
8.70
8
ng/
ml
1
9.19
8
ng/
ml
1
8.86
8
ng/
ml
1
9.35
8
ng/
ml
2
9.63
8
ng/
ml
2
10.16
8
ng/
ml
2
10.54
8
ng/
ml
2
11.08
8
ng/
ml
2
11.86
8
ng/
ml
5
10.96
8
ng/
ml
5
11.60
8
ng/
ml
5
12.41
8
ng/
ml
5
11.56
8
ng/
ml
5
13.03
2
ng/
ml
1
2.56
2
ng/
ml
1
2.77
2
ng/
ml
1
2.56
2
ng/
ml
1
2.59
2
ng/
ml
1
2.46
0.5
ng/
ml
1
0.63
0.5ng/
ml
1
0.80
0.5ng/
ml
1
0.76
0.5ng/
ml
1
0.74
0.5ng/
ml
1
0.71
Unspiked
Media
Below
Detection
Limits
of
0.04ng/
mL
17
Table
4.
Testosterone
RIA
Intra­
assay
CV
Number
50
µ
l
25
µ
l
10
µ
l
Unspiked
M199
2
Blanks
+
8
ng/
ml
10
5.24%
8.64%
7.68%

+
2
ng/
ml
10
6.09%

+
0.5
ng/
ml
10
13.34%

Table
5.
Testosterone
RIA
Percent
Recovery
50
µ
l
25
µ
l
10
µ
l
+
8
ng/
ml
113.8
133.3
149.0
+
2
ng/
ml
129.5
0.5
ng/
ml
146.5
Table
6.
Testosterone
RIA
Parallelism
50
µ
l
25
µ
l
10
µ
l
+
8
ng/
ml
9.10
ng/
ml
10.66
ng/
ml
11.92
ng/
ml
The
Index
between
50
and
25
10
µ
l
was
117.1%,
between
25
and
10
µ
l
was
111.8%
and
between
50
and
10
µ
l
was
130.99%.
18
5.2
LDH
Verification
See
Appendix
A
Statistical
Analysis
of
the
Phase
I
Assay
Optimization
Experiment
Objectives
The
assay
optimization
experiment
involves
three
factors
(
media
type,
atmosphere,
and
rat
age)

which
are
run
in
a
33
factorial
arrangement;
each
of
these
27
trials
is
run
with
and
without
hCG
stimulation
and
repeated
measurements
are
taken
at
baseline
(
time
0)
and
at
1,
2,
3,
and
4
hours
after
baseline.
The
conditions
are
identified
in
Table
7.
Objectives
of
the
experiment
are:

1.
To
determine
the
set
of
conditions
that
yields
the
highest
estimated
testosterone
level,

and
2.
To
determine
the
set
of
conditions
that
yields
the
largest
with­
versus­
without­
hCG
difference
in
testosterone
levels.

Data
Two
basic
SAS
data
sets
were
constructed
from
the
raw
data
and
two
fundamental
types
of
dependent
variables
were
used
in
the
analyses
of
each
type:

Date
Set
1:
Cases
without
hCG
stimulation
Dependent
variables:
testosterone
concentrations
Dependent
variables:
(
natural)
logarithm
of
testosterone
concentrations
Data
Set
2:
Cases
with
hCG
stimulation
Dependent
variables:
testosterone
concentrations
Dependent
variables:
(
natural)
logarithm
of
testosterone
concentrations
Each
data
set
can
be
viewed
as
consisting
of
27
observations
(
rows).
Each
observation
includes
dependent
variable
values
for
4
time
points
and
a
corresponding
baseline
level.
Each
observation
also
includes
data
identifying
the
levels
of
the
pertinent
factors.
Data
are
listed
in
Table
8
for
the
unchallenged
samples,
and
in
Table
9,
for
the
challenged
cases.

Statistical
Analysis
Methods
Objective
1.
Several
statistical
analysis
methods
were
used
to
address
the
first
objective.

Analysis
of
variance
(
ANOVA)
and
analysis
of
covariance
(
ANOCOVA)
were
used
to
analyze
the
data
for
each
individual
time
point
(
including
the
baseline)
and
a
mixed­
model
ANOCOVA
method
was
used
to
jointly
analyze
the
data
(
across
time
points
1
through
4).
The
ANOCOVA
models
utilized
the
baseline
level
(
or
log­
level)
as
a
covariate.
For
each
type
of
analysis,
all
main
effects
and
two­
factor
interactions
19
(
2fi)
of
the
three
factors
were
initially
included
in
the
models.
Tests
for
interactions
were
conducted
and
where
they
were
not
detected
as
statistically
significant
(
p=
0.05),
a
reduced
model
was
employed
that
retained
the
main
effects,
the
baseline
covariate
(
where
applicable),
and
only
those
2fi
deemed
to
have
significant
effects.
Additional
details
are
provided
in
the
Results
section.

Objective
2.
For
each
of
the
27
trials,
differences
between
the
with­
hCG
and
the
without­
hCG
testosterone
levels
were
computed
for
each
hour
(
including
baseline).
These
differences
were
computed
on
both
the
original
and
log
scales.
Analysis
of
variance
(
ANOVA)
was
used
to
analyze
these
differences
for
each
individual
time
point
(
including
the
baseline).
For
each
model
and
type
of
data,
all
main
effects
and
2fi
of
the
three
factors
were
initially
included
in
the
models.
Tests
for
interactions
were
conducted
and
where
they
were
not
detected
as
statistically
significant
(
p=
0.05),
a
reduced
model
was
employed
that
retained
the
main
effects
and
only
those
2fi
deemed
to
have
significant
effects.
Additional
details
are
provided
in
the
Results
section.

Results
Overall
Characterization
of
the
Data.
Table
10
provides
summary
statistics
characterizing
the
testosterone
levels
in
the
non­
hCG­
stimulated
data
set.
This
summary
ignores
the
particular
experimental
factors.
The
top
portion
of
the
table
gives,
by
hour,
the
sample
size
(
n),
the
mean,
standard
deviation,

sum,
minimum,
and
maximum.
These
variables
are
denoted
as
yJ,
where
J
denotes
the
hour
and
takes
on
values
of
0,
1,
2,
3,
and
4.
The
lower
portion
of
the
table
gives
the
correlations
between
the
hourly
data.
The
following
trends
are
apparent:

°
the
means
continue
to
increase
over
time
°
the
standard
deviations
also
increase
over
time
(
i.
e.,
as
the
mean
level
gets
larger)

°
the
correlations
are
generally
high,
as
tend
to
be
largest
for
adjacent
hours.

Table
11
provides
a
similar
summary
for
the
log­
scaled
data;
these
variables
are
denoted
as
lyJ,

where
J
denotes
the
hour.
Similar
trends
for
the
means
and
correlations
are
evident,
but
the
standard
deviations
tend
to
be
fairly
stable
across
the
various
time
points.

Tables
12
and
13
furnish
comparable
information
for
the
hCG­
stimulated
samples.
Similar
trends
are
evident
for
these
data.
Mean
levels
tend
to
be
much
higher
than
for
the
non­
stimulated
samples.

Analysis
of
Baseline
Data.
Since
we
intend
to
adjust
for
baseline
(
time
0)
levels
for
subsequent
analyses
of
the
hourly
(
non­
baseline)
data,
it
is
important
to
understand
how
the
experimental
factors
affect
the
baseline
levels.
For
instance,
if
one
of
the
factors
does
impact
the
baseline
levels,
then
adjusting
for
baseline
levels
in
those
subsequent
analyses
may
obscure
the
effect
of
the
experimental
factor.
Table14
presents
the
results
that
summarize
the
ANOVA
results
for
the
baseline
data.
Initially,
we
fit
an
ANOVA
model
that
included
all
mean
effects
(
denoted
as
z1,
z2,
and
z3)
and
all
two­
factor
20
interactions
(
denoted
as
z1*
z2,
z1*
z3,
and
z2*
z3).
We
examined
the
statistical
significance
of
each
of
the
interactions
and
reduced
the
model
to
contain
only
main
effects
and
the
pertinent
2fi.
For
three
of
the
four
cases
considered,
only
the
main
effects
were
retained;
for
the
log­
scale,
without
hCG
case,
two
of
the
interaction
effects
were
deemed
significant.
Among
the
three
experimental
factors,
it
is
clear
that
the
rat
age
(
z3)
has
the
most
pronounced
effect
on
the
baseline
levels
(
estimated
testosterone
levels
are
0.54,

0.58,
and
0.17
for
11­
week,
15­
week,
and
22­
week
old
animals).
For
the
non­
hCG­
stimulated
case,

media
type
also
impacted
the
baseline
levels
with
the
highest
level
occurring
for
the
z1=
0
case
(
0.54
versus
0.38
for
the
other
two
media).
The
lower
portion
of
Table
14
furnishes
statistics
characterizing
the
model
fit:

R2
=
the
proportion
of
variability
accounted
for
by
the
model,

RMSE
=
root
mean
squared
error
=
the
square
root
of
the
residual
variance,

C.
V.
=
the
coefficient
of
variation
=
the
RMSE
divided
by
the
mean
testosterone
level
(
times
100%).

Analyses
Directed
at
Objective
1.
Two
fundamental
types
of
statistical
analysis
were
used
to
address
objective
1
(
assessing
the
effects
of
the
experimental
factors
on
the
testosterone
levels)
 
separate
analyses
for
each
hour
and
a
combined
mixed­
model
approach.

Individual­
hour
analyses.
These
analyses
involved:

°
fitting
the
testosterone
data
for
a
given
hour
as
a
function
of
the
experimental
factors,

their
two­
factor
interactions,
and
the
baseline
level
°
examining
the
significance
of
the
two­
factor
interactions
(
2fi)

°
choosing
(
and
fitting)
a
reduced
model
form
by
eliminating
any
2fi
that
was
not
statistically
significant
in
any
of
the
four
hourly
models.

Results
for
the
original­
scale
data
are
summarized
in
Tables
15
and
16.
Table
15
provides
an
indication
of
which
effects
were
retained
in
the
reduced
model
and
which
of
those
terms
were
statistically
significant.
The
lower
portion
of
the
table
gives
statistics
characterizing
the
fit
of
the
models.

Table
15
indicates
that
mean
concentration
levels
increase
with
time.
The
models
for
the
without
and
with­
hCG
stimulation
data
are
somewhat
different,
but
both
indicate
statistical
significance
for
z2
(
atmosphere
type)
and
z3
(
rat
age).
For
the
without­
hCG
case,
the
model
also
shows
a
significant
effect
of
the
covariate
and
of
the
z1*
z2
(
media
type
by
atmosphere
type)
interaction.
For
the
with­
hCG
case,
the
covariate
was
not
statistically
significant
and
there
was
some
(
weak)
indication
of
a
z1*
z3
(
media
type
by
rat
age)
interaction.
The
RMSE
values
tend
to
increase
with
increasing
concentration
levels
(
i.
e.,
with
time).
The
C.
V.
s,
on
the
other
hand,
tend
to
be
fairly
stable,
suggesting
that
a
log­
tranform
of
the
concentrations
should
result
in
data
with
approximately
homogeneous
variances
over
the
various
time
points.
21
Adjusted
means
based
on
the
models
that
were
indicated
in
Table
15
are
presented
in
Table
16.

The
means
are
those
that
are
estimated
to
occur
for
a
given
level
of
a
factor
(
or
given
combination
of
factors)
when
other
effects
in
the
model
(
e.
g.,
the
baseline
level
covariate)
are
fixed
at
their
mean
values.

The
three
first
columns
of
the
table
identify
the
factor
levels
(
see
Table
7).
Within
each
set
of
the
levels
(
e.
g.,
the
three
rows
with
z1=
­
1,
0,
and
+
1),
the
estimated
adjusted
mean
that
is
largest
is
highlighted.

Asterisks
beside
the
other
non­
highlighted
means
indicate
if
that
particular
mean
is
deemed
to
be
statistically
significant
from
the
one
that
is
highlighted.
For
the
non­
stimulated
data,
for
instance,
the
table
indicates
no
significant
difference
among
the
z1
levels,
although
the
zero
level
is
consistently
estimated
to
be
the
largest.
For
the
atmosphere
and
rat
age
factors,
the
zero
levels
generally
have
the
highest
estimated
mean
testosterone
concentrations
and
the
other
levels
typically
have
significantly
lower
means.

An
exception
is
the
rat
age
(
z3)
factor
for
the
hCG­
stimulated
case,
where
the
11­
week
old
and
the
15­

week
old
rats
had
similar
adjusted
mean
levels.
Even
when
interactions
are
considered
(
lower
portion
of
Table
16),
the
zero
levels
of
all
three
factors
are
either
estimated
to
have
the
highest
adjusted
means
or
to
have
adjusted
means
that
are
not
significantly
different
from
the
factor
combination
having
the
highest
estimated
mean
level.

Tables
17
and
18
show
results
for
the
log­
transformed
data.
These
tables
are
analogous
to
Tables
15
and
16,
respectively.
The
models
are
somewhat
different
than
those
indicated
in
Table
15.
On
the
log
scale,
the
rat
age
factor
does
not
appear
to
be
as
prominent.
Also,
the
Eagle­
MEM
(
z1=
1)
medium
yields
the
highest
mean
levels,
although
the
199
medium
(
z1=
0)
levels
are
not
significantly
smaller.
The
air
atmosphere
consistently
produces
lower
mean
levels.
As
with
the
original­
scale
data,
the
zero
levels
of
all
three
factors
are
either
estimated
to
have
the
highest
adjusted
means
or
to
have
adjusted
means
that
are
not
significantly
different
from
the
factor
combination
having
the
highest
estimated
mean
level.

Mixed­
model
analyses.
For
each
of
two
data
sets
and
two
types
of
dependent
variables
indicated
above
(
see
"
Data"),
these
analyses
involved
several
steps.
First,
we
employed
a
mixed
model
that
included
°
the
main
effects
of
the
experimental
factors
°
two­
factor
interactions,

°
the
baseline
testosterone
level
°
a
linear
and
a
quadratic
time
component
°
cross
products
of
the
linear
and
quadratic
time
components
with
the
main
effects
°
cross
products
of
the
linear
and
quadratic
time
components
with
the
2fi.

For
this
"
full"
model,
we
utilized
the
SAS
PROC
MIXED
procedure
to
determine
a
relevant
covariance
structure
for
the
data
set;
in
particular,
we
examined
10
different
possible
covariance
structures,
using
maximum
likelihood
estimation,
and
selected
one
that
appeared
to
be
optimal
or
near
optimal.
Using
that
structure,
we
estimated
fixed
effects
for
all
of
the
above
model
terms.
We
then
22
reduced
the
model
by
eliminating
non­
significant
higher­
order
terms.
We
then
re­
examined
the
covariance
structure
(
again
selecting
from
among
10
possible
structures)
for
this
reduced
model.
Using
the
selected
structure,
we
estimated
the
fixed
effects
in
the
reduced
model
(
using
restricted
maximum
likelihood
estimation)
and
computed
adjusted
means
for
the
various
factor
levels,
along
with
approximate
95%
confidence
intervals
for
the
means.

The
adjusted
means
for
the
various
cases
and
factors
are
given
in
Tables
19
through
22.
These
are
denoted
as
EST.,
where
J
=
1,
2,
3,
or
4
denotes
hour.
Approximate
95%
confidence
limits
are
given
in
the
right
portion
of
the
table.
The
lower
and
upper
limits
are
denoted
as
LOWJ
and
HIJ,
respectively,

where
J
=
1,
2,
3,
or
4
denotes
hour.
Unlike
the
individual­
hour
analyses,
these
estimates
(
and
the
interval
estimates)
rely
on
data
from
all
four
hours
and
also
reflect
a
smoothing
over
time
(
due
to
the
assumed
quadratic
time
dependence).

Analyses
Directed
at
Objective
2.
Differences
in
hCG­
stimulated
and
non­
stimulated
testosterone
levels
were
computed
for
the
27
trials
on
an
hour­
by­
hour
basis.
These
differences
were
then
analyzed,
by
hour,
using
an
initial
ANOVA
model
that
included
all
main
effects
and
2fi.
A
reduced
ANOVA
model
was
then
selected
by
eliminating
those
2fi
that
were
not
statistically
significant.
The
original­
scale
models
are
summarized
in
Table
23;
no
2fi
were
deemed
necessary,
so
that
the
model
only
includes
the
main
effects.
Both
atmosphere
type
and
rat
age
were
judged
to
have
impact
on
the
testosterone
concentration
levels.
Table
24
shows
the
adjusted
means
derived
form
the
model.
The
estimated
adjusted
mean
difference
that
is
largest
is
highlighted.
Asterisks
beside
the
other
nonhighlighted
mean
differences
indicate
if
that
particular
difference
is
deemed
to
be
statistically
significant
from
the
one
that
is
highlighted.
The
media
type
appears
to
have
little
effect,
but
rat
age
and
atmosphere
type
are
significant
factors
affecting
the
with­
minus­
without­
hCG
differences.

Tables
25
and
26
show
comparable
results
for
the
differences
of
the
log­
transformed
data.
In
this
case,
rat
age
appears
less
important
and
atmosphere
type
is
the
most
dominant
factor.

Table
7.
Factor
Levels
in
the
Phase
I
Assay
Optimization
Experiment
Factor
Identification
Units
Factor
Name
Experimental
Levels
Coded
Experimental
Levels
1
2
3
1
2
3
Media
Type*
Z1
RPMI­
1640
medium­
199
Eagles­
MEM
­
1
0
+
1
Atmosphere*
Z2
5%
CO
2
/
95%
air
5%
CO
2
/
95%
O
2
air
­
1
0
+
1
Rat
Age*
wks
Z3
11
15
22
­
1
0
+
1
*
Treated
as
a
3­
level
discrete
factor.
23
Table
8.
Data
Listing
for
Samples
Without
hCG
Animal
Ear
Tag
Set
Number
whole
testis
weight
g
testis
section
g
z1
z2
z3
y0
=
Testos.
Conc.
Baseline
y1
=
Testos.
Conc.
Hour
1
y2
=
Testos.
Conc.
Hour
2
y3
=
Testos.
Conc.
Hour
3
y4
=
Testos.
Conc.
Hour
4
3
303
A
RIGHT­
1.5531
0.2331
­
1
­
1
­
1
0.54
3.23
5.21
6.19
6.11
5
308
B
RIGHT­
1.8189
0.2519
­
1
­
1
0
0.45
4.32
5.02
6.81
8.43
16
317
C
LEFT­
2.0203
0.2626
­
1
­
1
1
0.13
1.17
1.96
2.54
2.94
3
303
D
LEFT­
1.5632
0.2688
­
1
0
­
1
0.43
3.60
5.14
6.48
7.11
5
308
E
LEFT­
1.7979
0.2750
­
1
0
0
0.36
5.56
7.85
8.30
10.48
19
316
F
RIGHT­
2.0218
0.2524
­
1
0
1
0.18
2.10
3.77
5.11
6.26
2
304
G
RIGHT­
1.7860
0.2692
­
1
1
­
1
0.58
3.73
4.99
5.72
6.06
7
315
H
RIGHT­
1.7847
0.2449
­
1
1
0
0.57
5.17
7.01
8.39
9.17
16
317
I
RIGHT­
1.9804
0.2603
­
1
1
1
0.17
1.70
2.67
3.64
4.79
3
303
J
RIGHT­
1.5531
0.2742
0
­
1
­
1
0.63
5.35
6.27
7.52
8.64
5
308
K
RIGHT­
1.8189
0.2349
0
­
1
0
0.81
6.60
8.54
10.28
10.97
16
317
L
LEFT­
2.0203
0.2502
0
­
1
1
0.15
1.10
2.04
2.93
4.15
2
304
M
RIGHT­
1.7860
0.2668
0
0
­
1
0.86
6.60
9.81
11.35
13.48
5
308
N
LEFT­
1.7979
0.2504
0
0
0
0.49
5.58
8.02
10.00
11.23
19
316
O
RIGHT­
2.0218
0.2715
0
0
1
0.34
3.50
6.56
8.68
9.65
2
304
P
LEFT­
1.7809
0.2577
0
1
­
1
0.67
3.11
4.85
5.23
5.58
7
315
Q
RIGHT­
1.7847
0.2419
0
1
0
0.78
5.23
6.54
7.29
7.85
16
317
R
RIGHT­
1.9804
0.2563
0
1
1
0.13
1.31
1.82
2.46
2.83
3
303
S
LEFT­
1.5632
0.2651
1
­
1
­
1
0.21
3.98
6.10
7.37
10.16
5
308
T
RIGHT­
1.8189
0.2371
1
­
1
0
0.40
5.33
8.05
9.07
12.25
16
317
U
LEFT­
2.0203
0.2586
1
­
1
1
0.08
1.01
1.45
1.97
2.33
2
304
V
RIGHT­
1.7860
0.2741
1
0
­
1
0.48
3.44
5.53
6.98
8.12
5
308
W
LEFT­
1.7979
0.2340
1
0
0
0.69
4.91
7.36
10.05
11.72
19
316
X
RIGHT­
2.0218
0.2686
1
0
1
0.24
2.31
3.98
5.61
6.55
2
304
Y
LEFT­
1.7809
0.2681
1
1
­
1
0.45
4.44
5.49
6.71
6.86
7
315
Z
RIGHT­
1.7847
0.2560
1
1
0
0.74
5.34
6.15
7.87
8.43
16
317
AA
RIGHT­
1.9804
0.2408
1
1
1
0.09
0.84
1.18
1.73
1.74
24
Table
9.
Data
Listing
for
Samples
With
hCG
Animal
Ear
Tag
Set
Number
Whole
testis
weight
g
Testis
section
g
z1
z2z3
yc0
=
Testos.
Conc.
Baseline
yc1
=
Testos.
Conc.
Hour
1
yc2
=
Testos.
Conc.
Hour
2
yc3
=
Testos.
Conc.
Hour
3
yc4
=
Testos.
Conc.
Hour
4
3
303
AC
RIGHT­

1.5531
0.2422
­
1
­
1
­
1
0.54
6.56
16.16
28.30
42.09
5
308
BC
RIGHT­

1.8189
0.2690
­
1
­
1
0
1.11
5.96
9.69
15.19
23.41
16
317
CC
LEFT­

2.0203
0.2680
­
1
­
1
1
0.16
2.03
4.88
8.87
15.93
3
303
DC
LEFT­

1.5632
0.2374
­
1
0
­
1
0.73
5.91
20.96
41.76
51.74
5
308
EC
LEFT­

1.7979
0.2398
­
1
0
0
0.43
4.39
10.84
22.92
33.45
19
316
FC
RIGHT­

2.0218
0.2515
­
1
0
1
0.26
3.80
9.94
18.38
24.77
2
304
GC
RIGHT­

1.7860
0.2494
­
1
1
­
1
1.29
4.71
7.14
10.26
13.39
7
315
HC
RIGHT­

1.7847
0.2401
­
1
1
0
0.62
6.35
8.09
11.90
14.64
16
317
IC
RIGHT­

1.9804
0.2642
­
1
1
1
0.10
1.30
2.74
3.63
5.13
3
303
JC
RIGHT­

1.5531
0.2613
0
­
1
­
1
0.72
8.26
18.51
31.30
39.20
5
308
KC
RIGHT­

1.8189
0.2386
0
­
1
0
0.47
7.84
15.75
28.83
43.60
16
317
LC
LEFT­

2.0203
0.2672
0
­
1
1
0.11
2.19
6.70
11.23
18.38
2
304
MC
RIGHT­

1.7860
0.2631
0
0
­
1
0.86
9.73
31.45
55.27
77.68
5
308
NC
LEFT­

1.7979
0.2461
0
0
0
0.89
7.33
18.38
33.23
59.89
19
316
OC
RIGHT­

2.0218
0.2516
0
0
1
0.42
4.15
10.45
18.93
27.32
2
304
PC
LEFT­

1.7809
0.2461
0
1
­
1
0.61
5.57
9.65
12.75
15.73
(
continued)
25
Table
9.
Data
Listing
for
Samples
With
hCG
(
continued)

Animal
Ear
Tag
Set
Number
Whole
testis
weight
g
Testis
section
g
z1
z2
z3
yc0
=
Testos.
Conc.
Baseline
yc1
=
Testos.
Conc.
Hour
1
yc2
=
Testos.
Conc.
Hour
2
yc3
=
Testos.
Conc.
Hour
3
yc4
=
Testos.
Conc.
Hour
4
7
315
QC
RIGHT­
1.7847
0.2618
0
1
0
1.38
7.03
9.69
12.35
13.47
16
317
RC
RIGHT­
1.9804
0.2607
0
1
1
0.15
1.03
2.14
2.92
4.01
3
303
SC
LEFT­
1.5632
0.2375
1
­
1
­
1
0.34
6.43
18.23
26.69
38.80
5
308
TC
RIGHT­
1.8189
0.2668
1
­
1
0
0.65
8.79
18.07
29.25
46.72
16
317
UC
LEFT­
2.0203
0.2683
1
­
1
1
0.11
1.81
4.44
7.96
12.42
2
304
VC
RIGHT­
1.7860
0.2593
1
0
­
1
0.66
5.91
15.48
24.92
32.07
5
308
WC
LEFT­
1.7979
0.2515
1
0
0
0.90
9.38
25.90
45.21
67.45
19
316
XC
RIGHT­
2.0218
0.2595
1
0
1
0.18
3.98
10.13
18.33
27.01
2
304
YC
LEFT­
1.7809
0.2591
1
1
­
1
0.43
5.17
10.59
16.60
25.20
7
315
ZC
RIGHT­
1.7847
0.2684
1
1
0
0.52
7.09
12.48
20.01
27.96
16
317
AAC
RIGHT­
1.9804
0.2459
1
1
1
0.12
2.44
5.53
9.16
13.76
26
Table
10.
Summary
of
Data
­­
Original
Scale,
Without
hCG
Simple
Statistics
Variable
N
Mean
Std
Dev
Sum
Minimum
Maximum
Label
y0
27
0.43148
0.24012
11.65000
0.08000
0.86000
y0
=_
Testos._
Conc._
Baseline
y1
27
3.72444
1.77928
100.56000
0.84000
6.60000
y1
=_
Testos._
Conc._
Hour
1
y2
27
5.30963
2.33770
143.36000
1.18000
9.81000
y2
=_
Testos._
Conc._
Hour
2
y3
27
6.52889
2.67986
176.28000
1.73000
11.35000
y3
=_
Testos._
Conc._
Hour
3
y4
27
7.55148
3.16645
203.89000
1.74000
13.48000
y4
=_
Testos._
Conc._
Hour
4
Pearson
Correlation
Coefficients,
N
=
27
Prob
>
|
r|
under
H0:
Rho=
0
y0
y1
y2
y3
y4
y0
y0
=_
Testos._
Conc._
Baseline
1.00000
0.83359
<.
0001
0.77555
<.
0001
0.75779
<.
0001
0.64894
0.0003
y1
y1
=_
Testos._
Conc._
Hour
1
0.83359
<.
0001
1.00000
0.96006
<.
0001
0.93735
<.
0001
0.89924
<.
0001
y2
y2
=_
Testos._
Conc._
Hour
2
0.77555
<.
0001
0.96006
<.
0001
1.00000
0.98329
<.
0001
0.96280
<.
0001
y3
y3
=_
Testos._
Conc._
Hour
3
0.75779
<.
0001
0.93735
<.
0001
0.98329
<.
0001
1.00000
0.97499
<.
0001
y4
y4
=_
Testos._
Conc._
Hour
4
0.64894
0.0003
0.89924
<.
0001
0.96280
<.
0001
0.97499
<.
0001
1.00000
27
Table
11.
Summary
of
Data
­­
Log
Scale,
Without
hCG
Simple
Statistics
Variable
N
Mean
Std
Dev
Sum
Minimum
Maximum
ly0
27
­
1.04291
0.70964
­
28.15853
­
2.52573
­
0.15082
ly1
27
1.15995
0.62490
31.31878
­
0.17435
1.88707
ly2
27
1.53822
0.57738
41.53197
0.16551
2.28340
ly3
27
1.76591
0.52363
47.67970
0.54812
2.42922
ly4
27
1.90916
0.52876
51.54720
0.55389
2.60121
Pearson
Correlation
Coefficients,
N
=
27
Prob
>
|
r|
under
H0:
Rho=
0
ly0
ly1
ly2
ly3
ly4
ly0
1.00000
0.90646
<.
0001
0.88113
<.
0001
0.86395
<.
0001
0.79479
<.
0001
ly1
0.90646
<.
0001
1.00000
0.97732
<.
0001
0.96441
<.
0001
0.93255
<.
0001
ly2
0.88113
<.
0001
0.97732
<.
0001
1.00000
0.99193
<.
0001
0.97266
<.
0001
ly3
0.86395
<.
0001
0.96441
<.
0001
0.99193
<.
0001
1.00000
0.98411
<.
0001
ly4
0.79479
<.
0001
0.93255
<.
0001
0.97266
<.
0001
0.98411
<.
0001
1.00000
28
Table
12.
Summary
of
Data
­­
Original
Scale,
With
hCG
Simple
Statistics
Variable
N
Mean
Std
Dev
Sum
Minimum
Maximum
Label
yc0
27
0.54667
0.36060
14.76000
0.10000
1.38000
yc0
=_
Testos._
Conc._
Baseline
yc1
27
5.37556
2.48467
145.14000
1.03000
9.73000
yc1
=_
Testos._
Conc._
Hour
1
yc2
27
12.37074
7.01613
334.01000
2.14000
31.45000
yc2
=_
Testos._
Conc._
Hour
2
yc3
27
20.96852
12.82440
566.15000
2.92000
55.27000
yc3
=_
Testos._
Conc._
Hour
3
yc4
27
30.19333
18.74161
815.22000
4.01000
77.68000
yc4
=_
Testos._
Conc._
Hour
4
Pearson
Correlation
Coefficients,
N
=
27
Prob
>
|
r|
under
H0:
Rho=
0
yc0
yc1
yc2
yc3
yc4
yc0
yc0
=_
Testos._
Conc._
Baseline
1.00000
0.66941
0.0001
0.44159
0.0211
0.38125
0.0497
0.34634
0.0768
yc1
yc1
=_
Testos._
Conc._
Hour
1
0.66941
0.0001
1.00000
0.86213
<.
0001
0.80157
<.
0001
0.78474
<.
0001
yc2
yc2
=_
Testos._
Conc._
Hour
2
0.44159
0.0211
0.86213
<.
0001
1.00000
0.98464
<.
0001
0.96305
<.
0001
yc3
yc3
=_
Testos._
Conc._
Hour
3
0.38125
0.0497
0.80157
<.
0001
0.98464
<.
0001
1.00000
0.98016
<.
0001
yc4
yc4
=_
Testos._
Conc._
Hour
4
0.34634
0.0768
0.78474
<.
0001
0.96305
<.
0001
0.98016
<.
0001
1.00000
29
Table
13.
Summary
of
Data
­­
Log
Scale,
With
hCG
Simple
Statistics
Variable
N
Mean
Std
Dev
Sum
Minimum
Maximum
lyc0
27
­
0.86822
0.80586
­
23.44183
­
2.30259
0.32208
lyc1
27
1.53650
0.61019
41.48556
0.02956
2.27521
lyc2
27
2.34072
0.64378
63.19938
0.76081
3.44840
lyc3
27
2.83766
0.70489
76.61684
1.07158
4.01223
lyc4
27
3.19518
0.71702
86.26981
1.38879
4.35260
Pearson
Correlation
Coefficients,
N
=
27
Prob
>
|
r|
under
H0:
Rho=
0
lyc0
lyc1
lyc2
lyc3
lyc4
lyc0
1.00000
0.84699
<.
0001
0.68581
<.
0001
0.61367
0.0007
0.54138
0.0035
lyc1
0.84699
<.
0001
1.00000
0.92139
<.
0001
0.87024
<.
0001
0.82124
<.
0001
lyc2
0.68581
<.
0001
0.92139
<.
0001
1.00000
0.98599
<.
0001
0.95954
<.
0001
lyc3
0.61367
0.0007
0.87024
<.
0001
0.98599
<.
0001
1.00000
0.98703
<.
0001
lyc4
0.54138
0.0035
0.82124
<.
0001
0.95954
<.
0001
0.98703
<.
0001
1.00000
30
Table
14.
Summary
of
Statistical
Analysis
of
Baseline
Data
Original­
Scale
Models
Log­
Scale
Models
Without
hCG
Stimulation
With
hCG
Stimulation
Without
hCG
Stimulation
With
hCG
Stimulation
Dependent
Variable:
Y0
Y0
log(
Y0)
log(
Y0)

Dependent
Variable
Mean:
0.431
0.547
­
1.043
0.868
Significance
of
Model
Terms:
z1
XX
XXX
z2
XX
z3
XXX
XXX
XXX
XXX
z1*
z2
na
na
XX
na
z1*
z3
na
na
na
na
z2*
z3
na
na
XX
na
R2
0.766
0.628
0.950
0.822
RMSE
0.132
0.251
0.233
0.388
C.
V.
30.7
45.9
­­
­­
X
=
statistically
significant
effect
at
0.10
level
of
significance.
XX
=
statistically
significant
effect
at
0.05
level
of
significance.
XXX
=
statistically
significant
effect
at
0.01
level
of
significance.
na
=
not
applicable
(
effect
not
included
in
the
model)
31
Table
15.
Summary
of
ANOCOVA
Results
for
Individual­
Hour
Original­
Scale
Models
Without
hCG
Stimulation
With
hCG
Stimulation
Dependent
Variable:
Y1
Y2
Y3
Y4
Y1
Y2
Y3
Y4
Dependent
Variable
Mean:
3.72
5.31
6.53
7.55
5.38
12.37
20.97
30.19
Significance
of
Model
Terms:
Y0
XX
XX
XX
z1
XX
X
z2
X
XXX
XXX
XXX
XX
XXX
XXX
XXX
z3
XXX
XX
XXX
XX
XXX
XXX
XXX
XXX
z1*
z2
XX
XX
XX
X
na
na
na
na
z1*
z3
na
na
na
na
X
X
R2
0.931
0.913
0.917
0.877
0.904
0.862
0.865
0.847
RMSE
0.62
0.91
1.02
1.46
1.01
3.43
6.21
9.65
C.
V.
16.6
17.1
15.5
19.3
18.9
27.7
29.6
32.0
X
=
statistically
significant
effect
at
0.10
level
of
significance.

XX
=
statistically
significant
effect
at
0.05
level
of
significance.

XXX
=
statistically
significant
effect
at
0.01
level
of
significance.

na
=
not
applicable
(
effect
not
included
in
the
model)
32
Table
16.
Adjusted
Mean
Levels
Based
on
Original­
Scale
Models
Level
of
Independent
Variable
Mean
Levels
of
Dependent
Variables:
Without
hCG
Stimulation
Mean
Levels
of
Dependent
Variables:
With
hCG
Stimulation
z1
z2
z3
Y1
Mean
Y2
Mean
Y3
Mean
Y4
Mean
Y1
Mean
Y2
Mean
Y3
Mean
Y4
Mean
­
1
3.55
5.05
6.14
6.99
4.52*
10.08*
18.11
25.30
0
3.95
5.64
6.83
7.92
5.83
13.70
23.41
34.01
+
1
3.67
5.24
6.62
7.75
5.77
13.34
21.38
31.26
­
1
3.72
5.16*
6.31**
7.50*
5.62
12.43*
20.40**
30.39**
0
4.12
6.37
7.97
9.33
6.02
17.10
31.25
45.05
+
1
3.34*
4.40**
5.30**
5.82**
4.49**
7.59**
11.25**
15.14**

­
1
3.86**
5.53*
6.59*
7.67*
6.34
16.58
28.33
38.71
0
4.89
6.58
7.99
9.56
6.91
14.50
25.61
38.99
+
1
2.43**
3.83**
5.01**
5.43**
2.88**
6.03**
8.97**
12.88**

­
1
­
1
3.07**
4.28**
5.44**
6.01**
­
1
0
4.06
6.00
7.11*
8.30
­
1
+
1
3.51*
4.86**
5.88**
6.65**
0
­
1
4.07
5.24**
6.48**
7.60*
0
0
4.85
7.63
9.43
11.03
0
+
1
2.94**
4.04**
4.57**
5.11**
+
1
­
1
4.02
5.96
7.02*
8.89
+
1
0
3.44*
5.48*
7.38*
8.67
+
1
+
1
3.55*
4.29**
5.46**
5.69**

­
1
­
1
5.43**
15.00
28.51
38.78
­
1
0
5.40**
9.68**
17.65**
25.55**
­
1
+
1
2.73**
5.55**
8.18**
11.58**
0
­
1
7.68
20.02
34.14
46.02
0
0
7.05
14.90
26.88
42.62
0
+
1
2.76**
6.17**
9.21**
13.40**
+
1
­
1
5.90*
14.71
22.34*
31.33*
+
1
0
8.28
18.93
32.30
48.80
+
1
+
1
3.13**
6.37**
9.50**
13.67**
Shaded
cell
indicates
highest
mean
estimated
level.
*
Indicates
that
the
mean
level
is
significantly
lower
than
the
cell
with
the
maximum
estimated
level,
p=
0.05.
**
Indicates
that
the
mean
level
is
significantly
lower
than
the
cell
with
the
maximum
estimated
level,
p=
0.01.
33
Table
17.
Summary
of
ANOCOVA
Results
for
Individual­
Hour
Log­
Scale
Models
Without
hCG
Stimulation
With
hCG
Stimulation
Dependent
Variable:
log(
Y1)
log(
Y2)
log(
Y3)
log(
Y4)
log(
Y1)
log(
Y2)
log(
Y3)
log(
Y4)

Dependent
Variable
Mean:
1.160
1.538
1.766
1.909
1.537
2.341
2.838
3.195
Significance
of
Model
Terms:
log(
Y0)
XXX
XXX
XXX
XXX
XX
z1
X
XX
X
X
z2
X
XXX
XXX
XXX
XX
XXX
XXX
XXX
z3
X
X
X
z1*
z2
XX
XX
XX
XX
X
X
z1*
z3
X
X
XX
na
na
na
na
R2
0.963
0.964
0.962
0.951
0.894
0.900
0.917
0.924
RMSE
0.185
0.167
0.156
0.181
0.261
0.268
0.268
0.260
X
=
statistically
significant
effect
at
0.10
level
of
significance.

XX
=
statistically
significant
effect
at
0.05
level
of
significance.

XXX
=
statistically
significant
effect
at
0.01
level
of
significance.

na
=
not
applicable
(
effect
not
included
in
the
model)
34
Table
18.
Adjusted
Mean
Levels
Based
on
Log­
Scale
Models
Level
of
Independent
Variable
Mean
Levels
of
Dependent
Variables:
Without
hCG
Stimulation
Mean
Levels
of
Dependent
Variables:
With
hCG
Stimulation
z1
z2
z3
log(
Y1)
Mean
log(
Y2)
Mean
log(
Y3)
Mean
log(
Y4)
Mean
log(
Y1)
Mean
log(
Y2)
Mean
log(
Y3)
Mean
log(
Y4)
Mean
­
1
1.162
1.551
1.764
1.903
1.381*
2.147*
2.669*
3.021*
0
1.085
1.456
1.679
1.834
1.527
2.362
2.836
3.190
+
1
1.232
1.608
1.855
1.990
1.702
2.513
3.008
3.374
­
1
1.208
1.591
1.810
1.98
1.628
2.431
2.949*
3.356*
0
1.240
1.690
1.935
2.10
1.644
2.696
3.305
3.687
+
1
1.031*
1.334**
1.552**
1.64**
1.337*
1.895**
2.259**
2.542**

­
1
1.096
1.442
1.640
1.780
1.651
2.610
3.098
3.444
0
1.279
1.550
1.774
1.956
1.698
2.463
2.982
3.408
+
1
1.105
1.623
1.884
1.991
1.261
1.949*
2.433*
2.734*

­
1
­
1
1.021*
1.406**
1.645**
1.752**
1.419
2.192*
2.730*
3.211*
­
1
0
1.371
1.805
1.993
2.157
1.515
2.566
3.251
3.552
­
1
+
1
1.094*
1.440*
1.654*
1.801*
1.208*
1.682**
2.026**
2.302**
0
­
1
1.066*
1.403*
1.659*
1.857*
1.755
2.582
3.124
3.477
0
0
1.292
1.740
1.981
2.144
1.675
2.782
3.379
3.864
0
+
1
0.898**
1.225**
1.396**
1.501**
1.151*
1.723**
2.005**
2.229**
+
1
­
1
1.539
1.963
2.127
2.341
1.711
2.520
2.991
3.381
+
1
0
1.057*
1.524*
1.833
2.001
1.743
2.740
3.285
3.645
+
1
+
1
1.102*
1.338**
1.606**
1.628**
1.651
2.281
2.747*
3.097**

­
1
­
1
0.946*
1.307*
1.512*
1.587**
­
1
0
1.401
1.658
1.855
2.053
­
1
+
1
1.139
1.686
1.925
2.070
0
­
1
0.982*
1.288*
1.470*
1.650*
0
0
1.213
1.470
1.685
1.821
0
+
1
1.061
1.609
1.882
2.031
+
1
­
1
1.359
1.730
1.939
2.104
+
1
0
1.221
1.521
1.781
1.996
+
1
+
1
1.117
1.572
1.846
1.871
Shaded
cell
indicates
highest
mean
estimated
level.

*
Indicates
that
the
mean
level
is
significantly
lower
than
the
cell
with
the
maximum
estimated
level,
p=
0.05.
**
Indicates
that
the
mean
level
is
significantly
lower
than
the
cell
with
the
maximum
estimated
level,
p=
0.01.
35
Table
19.
Least
Squares
Means
for
Reduced
Log­
scale
Model:
Without
hCG
Mean
covariate
value:
ly0=­
1.043
x1
x2
x3
EST1
EST2
EST3
EST4
LOW
1
HI1
LOW
2
HI2
LOW
3
HI3
LOW
4
HI4
­
1
_
_
1.163
1.527
1.773
1.901
1.059
1.267
1.424
1.629
1.670
1.875
1.797
2.005
0
_
_
1.116
1.471
1.708
1.828
0.942
1.290
1.298
1.644
1.536
1.881
1.655
2.002
1
_
_
1.211
1.587
1.846
1.988
1.064
1.357
1.442
1.733
1.701
1.992
1.842
2.134
_
­
1
_
1.184
1.563
1.824
1.968
1.039
1.330
1.418
1.708
1.680
1.969
1.823
2.114
_
0
_
1.278
1.672
1.949
2.109
1.146
1.409
1.542
1.802
1.819
2.080
1.978
2.240
_
1
_
1.028
1.350
1.554
1.640
0.928
1.128
1.251
1.448
1.455
1.652
1.540
1.741
_
_
­
1
1.137
1.472
1.683
1.771
0.901
1.374
1.236
1.708
1.448
1.919
1.534
2.007
_
_
0
1.327
1.606
1.813
1.949
1.031
1.622
1.311
1.901
1.518
2.108
1.654
2.245
_
_
1
1.026
1.507
1.831
1.998
0.523
1.529
1.004
2.010
1.328
2.334
1.494
2.501
Reduced
model
was
fit
using
compound
symmetry
covariance
structure.
Effects
retained
in
the
reduced
model
were
the
following
(
tL
denote
the
linear
time
effect,
tQ
denotes
the
quadratic):
ly0,
z1,
z2,
z3,
z1*
z2,
z1*
z3,
tL,
tL*
z1,
tL*
z2,
tL*
z3,
tL*
z1*
z3,
tQ,
tQ*
z3.

Table
20.
Least
Squares
Means
for
Reduced
Original­
scale
Model:
Without
hCG
Mean
covariate
value:
y0=
0.43
x1
x2
x3
EST
1
EST
2
EST
3
EST
4
LO
W1
HI1
LO
W2
HI2
LO
W3
HI3
LO
W4
HI4
­
1
_
_
3.70
5.11
6.24
7.09
2.99
4.40
4.41
5.80
5.55
6.94
6.39
7.79
0
_
_
3.74
5.35
6.67
7.72
2.82
4.66
4.43
6.27
5.75
7.59
6.79
8.64
1
_
_
3.76
5.40
6.75
7.82
3.05
4.48
4.69
6.10
6.04
7.46
7.11
8.53
_
­
1
_
3.78
5.30
6.54
7.50
3.07
4.48
4.60
6.00
5.84
7.24
6.80
8.21
_
0
_
4.17
6.18
7.91
9.36
3.54
4.81
5.56
6.81
7.29
8.54
8.73
9.99
_
1
_
3.25
4.37
5.21
5.77
2.59
3.90
3.72
5.01
4.56
5.85
5.12
6.42
_
_
­
1
3.65
5.20
6.47
7.46
2.74
4.57
4.29
6.12
5.56
7.38
6.54
8.38
_
_
0
4.45
6.30
7.87
9.15
3.29
5.62
5.14
7.46
6.70
9.03
7.99
10.32
_
_
1
3.09
4.35
5.32
6.02
1.31
4.87
2.57
6.13
3.55
7.10
4.24
7.80
Reduced
model
was
fit
using
compound
symmetry
covariance
structure.
Effects
retained
in
the
reduced
model
were
the
following
(
tL
denote
the
linear
time
effect,
tQ
denotes
the
quadratic):
y0,
z1,
z2,
z3,
z1*
z2,
z1*
z3,
z2*
z3,
tL,
tL*
z1,
tL*
z2,
tL*
z3,
tL*
z1*
z2,
tL*
z1*
z3,
tL*
z2*
z3,
tQ.
36
Table
21.
Least
Squares
Means
for
Reduced
Log­
scale
Model:
with
hCG
Mean
covariate
value:
lyc0=­
0.868
x1
x2
x3
EST1
EST2
EST3
EST4
LOW
1
HI1
LOW
2
HI2
LOW
3
HI3
LOW
4
HI4
­
1
_
_
1.388
2.153
2.694
3.012
1.199
1.576
1.970
2.335
2.512
2.876
2.823
3.200
0
_
_
1.554
2.314
2.852
3.166
1.360
1.748
2.126
2.503
2.664
3.040
2.972
3.360
1
_
_
1.668
2.471
3.051
3.408
1.469
1.867
2.278
2.665
2.858
3.244
3.209
3.607
_
­
1
_
1.598
2.400
2.996
3.385
1.402
1.795
2.206
2.595
2.801
3.190
3.188
3.581
_
0
_
1.682
2.662
3.319
3.651
1.481
1.884
2.463
2.862
3.119
3.518
3.449
3.852
_
1
_
1.329
1.876
2.283
2.550
1.141
1.518
1.690
2.062
2.097
2.469
2.361
2.738
_
_
­
1
1.726
2.560
3.109
3.372
1.489
1.963
2.325
2.795
2.873
3.344
3.135
3.609
_
_
0
1.793
2.443
2.951
3.317
1.530
2.056
2.182
2.704
2.690
3.212
3.054
3.580
_
_
1
1.091
1.935
2.537
2.896
0.713
1.470
1.558
2.313
2.159
2.914
2.518
3.275
Reduced
model
was
fit
using
first­
order
autoregressive
covariance
structure.
Effects
retained
in
the
reduced
model
were
the
following
(
tL
denote
the
linear
time
effect,
tQ
denotes
the
quadratic):
lyc0,
z1,
z2,
z3,
z1*
z2,
tL,
tL*
z1,
tL*
z2,
tL*
z3,
tL*
z1*
z2,
tQ,
tQ*
z2,
tQ*
z3.

Table
22.
Least
Squares
Means
for
Reduced
Original­
scale
Model:
with
hCG
Mean
covariate
value:
yc0=
0.55
x1
x2
x3
EST
1
EST2
EST3
EST4
LO
W1
HI1
LOW
2
HI2
LOW
3
HI3
LOW
4
HI4
­
1
_
_
5.11
12.20
20.55
30.17
4.43
5.79
10.49
13.90
17.76
23.34
26.03
34.31
0
_
_
5.52
12.60
20.96
30.57
4.83
6.21
10.90
14.31
18.17
23.75
26.43
34.71
1
_
_
5.59
12.68
21.03
30.65
4.89
6.29
10.97
14.39
18.24
23.82
26.51
34.79
_
­
1
_
5.68
12.69
20.97
30.52
4.71
6.64
9.80
15.59
16.09
25.85
23.51
37.52
_
0
_
6.14
17.64
30.40
44.43
5.18
7.09
14.75
20.53
25.52
35.28
37.43
51.44
_
1
_
4.40
7.15
11.16
16.44
3.45
5.36
4.26
10.04
6.28
16.04
9.43
23.45
_
_
­
1
6.32
16.54
27.02
37.76
5.36
7.29
13.66
19.42
22.24
31.81
30.62
44.90
_
_
0
6.78
14.01
23.91
36.48
5.79
7.78
11.12
16.90
19.12
28.70
29.34
43.63
_
_
1
3.11
6.92
11.60
17.15
2.04
4.17
4.01
9.84
6.80
16.41
9.99
24.30
Reduced
model
was
fit
using
first­
order
autoregressive
covariance
structure
with
heterogeneous
variances.
Effects
retained
in
the
reduced
model
were
the
following
(
tL
denote
the
linear
time
effect,
tQ
denotes
the
quadratic):
y0,
z1,
z2,
z3,
tL,
tL*
z2,
tL*
z3,
tQ,
tQ*
z3.
37
Table
23.
Analysis
of
Differences
in
Levels
for
With
and
Without
hCG
Stimulation:
Original­
Scale
Models
Dependent
Variable:
diff(
Y0)
diff(
Y1)
diff(
Y2)
diff(
Y3)
diff(
Y4)

Dependent
Variable
Mean:
0.12
1.65
7.06
14.44
22.64
Significance
of
Model
Terms:
z1
z2
XXX
XXX
XXX
z3
XX
XXX
XXX
XXX
R2
0.175
0.457
0.687
0.722
0.712
RMSE
0.25
1.05
3,49
6,57
10.04
X
=
statistically
significant
effect
at
0.10
level
of
significance.
XX
=
statistically
significant
effect
at
0.05
level
of
significance.
XXX
=
statistically
significant
effect
at
0.01
level
of
significance.
38
Table
24.
Adjusted
Mean
Differences
(
With­
Without
hCG),
Based
on
Original­
Scale
Models
Level
of
Independent
Variable
Mean
Levels
of
Dependent
Variables
z1
z2
z3
Mean
diff
(
Y0)
Mean
diff
(
Y1)
Mean
diff
(
Y2)
Mean
diff
(
Y3)
Mean
diff
(
Y4)
­
1
0.20
1.16
5.20
12.00
18.13
0
0.08
1.64
7.59
15.67
24.99
+
1
0.06
2.16
8.40
15.64
24.80
­
1
0.09
1.98
7.53
14.77*
23.84*
0
0.14
1.89
10.61
22.93
35.20
+
1
0.12
1.09
3.04**
5.62**
8.89**

­
1
0.15
2.31
10.53
20.48
29.31
0
0.19
1.79
7.15
15.65
26.67
+
1
0.01
0.85**
3.50**
7.19**
11.94**
Shaded
cell
indicates
highest
mean
estimated
level.
*
Indicates
that
the
mean
level
is
significantly
lower
than
the
cell
with
the
maximum
estimated
level,
p=
0.05.
**
Indicates
that
the
mean
level
is
significantly
lower
than
the
cell
with
the
maximum
estimated
level,
p=
0.01.
39
Table
25.
Analysis
of
Differences
in
Levels
for
With
and
Without
hCG
Stimulation:
Log­
Scale
Models
Dependent
Variable:
diff
(
log(
Y0))
diff
(
log(
Y1))
diff
(
log(
Y2))
diff
(
log(
Y3))
diff
(
log(
Y4))

Dependent
Variable
Mean:
0.175
0.377
0.802
1.072
1.286
Significance
of
Model
Terms:
z1
X
z2
XX
XXX
XXX
z3
R2
0.142
0.305
0.528
0.585
0.503
RMSE
0.393
0.289
0.313
0.335
0.405
X
=
statistically
significant
effect
at
0.10
level
of
significance.
XX
=
statistically
significant
effect
at
0.05
level
of
significance.
XXX
=
statistically
significant
effect
at
0.01
level
of
significance.
40
Table
26.
Adjusted
Mean
Differences
(
With­
Without
hCG),
Based
on
Log­
Scale
Models
Level
of
Independent
Variable
Mean
Levels
of
Dependent
Variables
z1
z2
z3
Mean
diff
log(
Y0)
Mean
diff
log(
Y1)
Mean
diff
log(
Y2)
Mean
diff
log(
Y3)
Mean
diff
log(
Y4)
­
1
0.282
0.289
0.659**
0.963
1.166
0
0.079
0.307
0.732
0.995
1.197
+
1
0.164
0.533
1.017
1.278
1.496
­
1
0.186
0.493
0.956
1.248
1.486
0
0.242
0.380
0.931
1.298
1.503
+
1
0.096
0.257
0.520**
0.670**
0.869**

­
1
0.236
0.446
0.957
1.259
1.447
0
0.243
0.274
0.637*
0.948
1.170
+
1
0.045
0.410
0.814
1.008
1.241
Shaded
cell
indicates
highest
mean
estimated
level.
*
Indicates
that
the
mean
level
is
significantly
lower
than
the
cell
with
the
maximum
estimated
level,
p=
0.05.
**
Indicates
that
the
mean
level
is
significantly
lower
than
the
cell
with
the
maximum
estimated
level,
p=
0.01.

Statistical
Analysis
of
the
Phase
I
B
Assay
Optimization
Experiment
Objectives
Since
the
gaseous
atmosphere
of
5%
CO
2
/
95%
O
2
was
optimal
and
it
was
thought
that
most
laboratories
would
not
have
incubators
to
accommodate
this
mixture,
a
comparison
was
made
between
incubated
samples
and
those
in
media
that
had
be
gassed
with
the
mixture.

The
Phase
IB
assay
optimization
experiment
involved
assessing
the
effect
of
a
single
experimental
factor
 
using
gassed
or
incubated
samples.
All
other
factors
were
held
fixed.
Ten
trials
of
each
condition
were
run
both
with
and
without
hCG
stimulation.
For
each
trial,
repeated
measurements
are
taken
at
baseline
(
time
0)
and
at
1,

2,
3,
and
4
hours
after
baseline.
Objective
of
the
experiment
was
to
assess
whether
the
treatments
differed
in
terms
of
the
resultant
testosterone
levels.

Data
Two
basic
SAS
data
sets
were
constructed
from
the
raw
data
and
two
fundamental
types
of
dependent
variables
were
used
in
the
analyses
of
each
type:
41
Date
Set
1:
Cases
without
hCG
stimulation
Dependent
variables:
testosterone
concentrations
Dependent
variables:
(
natural)
logarithm
of
testosterone
concentrations
Data
Set
2:
Cases
with
hCG
stimulation
Dependent
variables:
testosterone
concentrations
Dependent
variables:
(
natural)
logarithm
of
testosterone
concentrations
Each
data
set
can
be
viewed
as
consisting
of
20
observations
(
rows).
Each
observation
includes
dependent
variable
values
for
4
time
points
and
a
corresponding
baseline
level.
Each
observation
also
includes
data
identifying
the
levels
of
the
pertinent
factors.
Data
are
listed
in
Table
27
for
the
unchallenged
samples,
and
in
Table
28,
for
the
challenged
cases.

Statistical
Analysis
Methods
Analysis
of
variance
(
ANOVA)
were
used
to
analyze
the
data
for
each
individual
time
point
(
including
the
baseline).
ANOVAs
were
performed
for
both
original­
scale
data
and
log­
scaled
data
(
natural
logarithm).
Analysis
of
covariance
(
ANOCOVA)
models
utilizing
the
baseline
level
(
or
log­
level)
as
a
covariate
were
also
employed.

Results
Overall
Characterization
of
the
Data.
Table
29
provides
summary
statistics
characterizing
the
testosterone
levels
in
the
non­
hCG­
stimulated
data
set.
This
summary
ignores
the
experimental
factor.
The
top
portion
of
the
table
gives,
by
hour,
the
sample
size
(
n),
the
mean,
standard
deviation,
sum,
minimum,
and
maximum.
These
variables
are
denoted
as
yJ,
where
J
denotes
the
hour
and
takes
on
values
of
0,
1,
2,
3,
and
4.

The
lower
portion
of
the
table
gives
the
correlations
between
the
hourly
data.
The
following
trends
are
apparent:

°
the
means
continue
to
increase
over
time
°
the
standard
deviations
also
increase
over
time
(
i.
e.,
as
the
mean
level
gets
larger)

°
the
correlations
are
generally
high,
and
tend
to
be
largest
for
adjacent
hours.

Table
30
provides
a
similar
summary
for
the
log­
scaled
data;
these
variables
are
denoted
as
lyJ,
where
J
denotes
the
hour.
Similar
trends
for
the
means
and
correlations
are
evident,
but
the
standard
deviations
tend
to
be
fairly
stable
across
the
various
time
points.

Tables
31
and
32
furnish
comparable
information
for
the
hCG­
stimulated
samples.
Similar
trends
are
evident
for
these
data.
Mean
levels
tend
to
be
much
higher
than
for
the
non­
stimulated
samples.
42
Analysis
of
Variance
and
Covariance.
Means,
by
hour
and
sample
condition,
are
presented
in
Table
33
for
the
non­
hCG­
stimulated
original­
scale
data.
The
table
gives
the
number
of
observations,
the
approximate
95%
confidence
limits
for
the
mean,
the
estimated
mean
and
standard
deviation.
The
last
column
gives
the
mean
(
for
hours
1,
2,
3,
and
4)
adjusted
for
the
baseline
level,
as
determined
from
the
ANOCOVA.
Tables
34,
35,
and
36
give
corresponding
results
for
the
log­
scaled
data
and
the
hCG­
stimulated
cases.
The
table
below
summarizes
the
findings
detailed
in
Tables
33
through
36.
For
the
most
part,
the
ANOVAs
and
ANOCOVAs
of
the
hourly
data
did
not
detect
significant
differences
between
the
testosterone
levels
of
the
gassed
and
incubated
samples.
If
the
0.05
significance
level
is
used
to
judge
statistical
significance,
then
only
the
baseline
case
of
Table
33
yielded
a
significant
difference.

Source
Type
Data
Analyzed
ANOVA
Results
ANOCOVA
Results
Table
33
Original
scale,
non­
hCG­
stimulated
Baseline
gassed
samples
have
lower
testosterone
mean
(
p=
0.05);
all
other
hours
not
significantly
different
at
0.10
level.
Hour­
1
and
hour­
4
gassed
samples
have
higher
testosterone
means
(
p=
0.08
and
0.07,
respectively);
other
hours
not
significantly
different
at
0.10
level.

Table
34
Log
scale,
non­
hCG­
stimulated
Baseline
gassed
samples
have
lower
testosterone
mean
(
p=
0.06);
all
other
hours
not
significantly
different
at
0.10
level.
No
significant
differences
at
0.10
level.

Table
35
Original
scale,
hCG­
stimulated
No
significant
differences
at
0.10
level.
No
significant
differences
at
0.10
level.

Table
36
Log
scale,
hCG­
stimulated
No
significant
differences
at
0.10
level.
No
significant
differences
at
0.10
level.
43
Table
27.
Phase
IB
Data:
Samples
Without
hCG
Ear
Tag
Body
Weight
g
Testis
wt
g
Section
wt
g
Run
Number
x0
y0
=
Testos.
Conc.
Baseline
y1
=
Testos.
Conc.
Hour
1
y2
=
Testos.
Conc.
Hour
2
y3
=
Testos.
Conc.
Hour
3
y4
=
Testos.
Conc.
Hour
4
320
361.74
RIGHT
­
1.5641
0.2413
1
1
0.35
3.49
5.46
6.66
8.05
321
361.74
LEFT
­
1.5303
0.2657
2
1
0.38
3.92
4.95
6.88
8.27
321
365.84
RIGHT
­
1.8176
0.2623
3
1
0.36
4.66
6.52
7.88
8.53
321
365.84
LEFT
­
1.8273
0.2463
4
1
0.34
3.94
4.94
6.30
7.71
321
365.84
LEFT
­
1.8273
0.2389
5
1
0.32
4.53
6.36
7.56
7.73
322
372.13
RIGHT
­
1.5835
0.2341
6
1
0.14
3.30
4.34
5.65
6.40
322
372.13
LEFT
­
1.5263
0.2436
7
1
0.18
1.67
2.47
3.16
3.63
323
365.46
RIGHT
­
1.7040
0.2722
8
1
0.45
3.87
5.41
6.06
7.90
324
356.94
RIGHT
­
1.4908
0.2626
9
1
0.42
5.32
7.49
9.71
11.01
323
365.46
LEFT
­
1.6760
0.2711
10
1
0.35
5.32
7.07
8.83
10.17
320
361.74
RIGHT
­
1.5641
0.2713
11
2
0.49
5.06
7.37
9.13
10.97
320
361.74
LEFT
­
1.5303
0.2543
12
2
0.38
2.94
4.36
5.44
6.45
321
365.84
RIGHT
­
1.8176
0.2610
13
2
0.38
4.76
6.78
8.51
9.21
321
365.84
LEFT
­
1.8273
0.2408
14
2
0.36
3.64
5.37
6.34
6.87
322
372.13
RIGHT
­
1.5835
0.2472
15
2
0.33
3.38
4.87
5.53
6.41
322
372.13
LEFT
­
1.5263
0.2499
16
2
0.46
4.45
5.56
6.85
8.15
323
365.46
RIGHT
­
1.7040
0.2474
17
2
0.40
4.26
6.30
7.99
9.13
323
365.46
LEFT
­
1.6760
0.2630
18
2
0.39
3.56
5.63
6.78
7.71
324
356.94
RIGHT
­
1.4908
0.2365
19
2
0.37
2.62
3.69
4.82
5.84
324
356.94
LEFT
­
1.4136
0.2684
20
2
0.49
4.56
5.85
6.55
9.02
x0
=
sample
condition:
1
=
gassed,
2
=
incubated
44
Table
28.
Phase
IB
Data:
Samples
With
hCG
Ear
Tag
Body
Weight
g
Testis
wt
g
Section
wt
g
Run
Number
x0
y0
=
Testos.
Conc.
Baseline
y1
=
Testos.
Conc.
Hour
1
y2
=
Testos.
Conc.
Hour
2
y3
=
Testos.
Conc.
Hour
3
y4
=
Testos.
Conc.
Hour
4
320
361.74
RIGHT
­
1.5641
0.2387
1
1
0.36
9.56
20.60
32.84
41.13
320
361.74
LEFT
­
1.5303
0.2727
2
1
0.52
6.31
13.49
22.25
29.13
321
365.84
RIGHT
­
1.8176
0.2518
3
1
0.40
5.12
13.29
18.32
25.33
321
365.84
LEFT
­
1.8273
0.2748
4
1
0.48
7.57
14.92
23.79
30.83
321
365.84
LEFT
­
1.8273
0.2681
5
1
0.43
4.84
9.66
13.47
18.59
322
372.13
RIGHT
­
1.5835
0.2429
6
1
0.28
8.73
18.74
30.71
45.84
322
372.13
LEFT
­
1.5263
0.2523
7
1
0.23
4.18
9.38
12.68
18.38
322
365.46
RIGHT
­
1.7040
0.2550
8
1
0.42
5.95
12.41
21.00
30.28
324
356.94
RIGHT
­
1.4908
0.2489
9
1
0.12
5.09
11.99
17.25
28.43
323
365.46
LEFT
­
1.6760
0.2513
10
1
0.21
3.40
8.04
13.78
19.14
320
361.74
RIGHT
­
1.5641
0.2451
11
2
0.34
6.71
17.75
29.37
39.76
320
361.74
LEFT
­
1.5303
0.2409
12
2
0.38
5.07
12.01
17.63
23.46
321
365.84
RIGHT
­
1.8176
0.2471
13
2
0.24
7.90
17.49
27.99
39.46
321
365.84
LEFT
­
1.8273
0.2695
14
2
0.41
7.52
19.79
30.29
43.69
322
372.13
RIGHT
­
1.5835
0.2452
15
2
0.37
4.24
10.43
14.70
19.41
322
372.13
LEFT
­
1.5263
0.2635
16
2
0.31
5.06
11.70
19.98
29.90
323
365.46
RIGHT
­
1.7040
0.2615
17
2
0.44
5.71
13.87
21.35
30.16
323
365.46
LEFT
­
1.6760
0.2328
18
2
0.72
6.93
21.00
31.05
40.41
324
356.94
RIGHT
­
1.4908
0.2632
19
2
0.81
5.73
13.45
23.01
34.28
324
356.94
LEFT
­
1.4136
0.2491
20
2
0.31
6.14
16.68
26.65
33.43
x0
=
sample
condition:
1
=
gassed,
2
=
incubated
45
Table
29.
Summary
of
Data
­­
Original
Scale,
Without
hCG
Simple
Statistics
Variable
N
Mean
Std
Dev
Sum
Minimum
Maximum
Label
Y0
20
0.36700
0.08652
7.34000
0.14000
0.49000
y0
=_
Testos._
Conc._
Baseline
Y1
20
3.96250
0.92905
79.25000
1.67000
5.32000
y1
=_
Testos._
Conc._
Hour
1
Y2
20
5.53950
1.26145
110.79000
2.47000
7.49000
y2
=_
Testos._
Conc._
Hour
2
Y3
20
6.83150
1.57510
136.63000
3.16000
9.71000
y3
=_
Testos._
Conc._
Hour
3
Y4
20
7.95800
1.76957
159.16000
3.63000
11.01000
y4
=_
Testos._
Conc._
Hour
4
Pearson
Correlation
Coefficients,
N
=
20Prob
>
|
r|
under
H0:
Rho=
0
Y0
Y1
Y2
Y3
Y4
Y0y0
=_
Testos._
Conc._
Baseline
1.00000
0.56839
0.0089
0.57435
0.0081
0.51050
0.0215
0.65502
0.0017
Y1y1
=_
Testos._
Conc._
Hour
1
0.56839
0.0089
1.00000
0.95710
<.
0001
0.94357
<.
0001
0.94257
<.
0001
Y2y2
=_
Testos._
Conc._
Hour
2
0.57435
0.0081
0.95710
<.
0001
1.00000
0.97439
<.
0001
0.94352
<.
0001
Y3y3
=_
Testos._
Conc._
Hour
3
0.51050
0.0215
0.94357
<.
0001
0.97439
<.
0001
1.00000
0.95462
<.
0001
Y4y4
=_
Testos._
Conc._
Hour
4
0.65502
0.0017
0.94257
<.
0001
0.94352
<.
0001
0.95462
<.
0001
1.00000
46
Table
30.
Summary
of
Data
­­
Log
Scale,
Without
hCG
Simple
Statistics
Variable
N
Mean
Std
Dev
Sum
Minimum
Maximum
ly0
20
­
1.03875
0.30294
­
20.77491
­
1.96611
­
0.71335
ly1
20
1.34511
0.27439
26.90217
0.51282
1.67147
ly2
20
1.68265
0.26153
33.65299
0.90422
2.01357
ly3
20
1.89300
0.25530
37.86004
1.15057
2.27316
ly4
20
2.04705
0.25043
40.94100
1.28923
2.39880
Pearson
Correlation
Coefficients,
N
=
20Prob
>
|
r|
under
H0:
Rho=
0
ly0
ly1
ly2
ly3
ly4
ly0
1.00000
0.58422
0.0068
0.60978
0.0043
0.55761
0.0106
0.659980.0015
ly1
0.58422
0.0068
1.00000
0.96741
<.
0001
0.95797
<.
0001
0.95831<.
0001
ly2
0.60978
0.0043
0.96741
<.
0001
1.00000
0.97796
<.
0001
0.95576<.
0001
ly3
0.55761
0.0106
0.95797
<.
0001
0.97796
<.
0001
1.00000
0.96540<.
0001
ly4
0.65998
0.0015
0.95831
<.
0001
0.95576
<.
0001
0.96540
<.
0001
1.00000
47
Table
31.
Summary
of
Data
­­
Original
Scale,
With
hCG
Simple
Statistics
Variable
N
Mean
Std
Dev
Sum
Minimum
Maximum
Label
YC0
20
0.38900
0.16189
7.78000
0.12000
0.81000
yc0
=_
Testos._
Conc._
Baseline
YC1
20
6.08800
1.58725
121.76000
3.40000
9.56000
yc1
=_
Testos._
Conc._
Hour
1
YC2
20
14.33450
3.87932
286.69000
8.04000
21.00000
yc2
=_
Testos._
Conc._
Hour
2
YC3
20
22.40550
6.48074
448.11000
12.68000
32.84000
yc3
=_
Testos._
Conc._
Hour
3
YC4
20
31.05200
8.66685
621.04000
18.38000
45.84000
yc4
=_
Testos._
Conc._
Hour
4
Pearson
Correlation
Coefficients,
N
=
20Prob
>
|
r|
under
H0:
Rho=
0
YC0
YC1
YC2
YC3
YC4
YC0yc0
=_
Testos._
Conc._
Baseline
1.00000
0.14785
0.5339
0.24875
0.2903
0.25182
0.2841
0.19761
0.4037
YC1yc1
=_
Testos._
Conc._
Hour
1
0.14785
0.5339
1.00000
0.88953
<.
0001
0.90763
<.
0001
0.87874
<.
0001
YC2yc2
=_
Testos._
Conc._
Hour
2
0.24875
0.2903
0.88953
<.
0001
1.00000
0.97501
<.
0001
0.93343
<.
0001
YC3yc3
=_
Testos._
Conc._
Hour
3
0.25182
0.2841
0.90763
<.
0001
0.97501
<.
0001
1.00000
0.96812
<.
0001
YC4yc4
=_
Testos._
Conc._
Hour
4
0.19761
0.4037
0.87874
<.
0001
0.93343
<.
0001
0.96812
<.
0001
1.00000
48
Table
32.
Summary
of
Data
 
Log
Scale,
With
hCG
Simple
Statistics
Variable
N
Mean
Std
Dev
Sum
Minimum
Maximum
lyc0
20
­
1.02614
0.42653
­
20.52279
­
2.12026
­
0.21072
lyc1
20
1.77407
0.26184
35.48142
1.22378
2.25759
lyc2
20
2.62709
0.27611
52.54184
2.08443
3.04452
lyc3
20
3.06710
0.30327
61.34207
2.54003
3.49165
lyc4
20
3.39608
0.29444
67.92156
2.91126
3.82516
Pearson
Correlation
Coefficients,
N
=
20Prob
>
|
r|
under
H0:
Rho=
0
lyc0
lyc1
lyc2
lyc3
lyc4
lyc0
1.00000
0.25901
0.2702
0.29545
0.2060
0.30624
0.1891
0.211980.
3696
lyc1
0.25901
0.2702
1.00000
0.91838
<.
0001
0.91443
<.
0001
0.89007<.
0001
lyc2
0.29545
0.2060
0.91838
<.
0001
1.00000
0.96817
<.
0001
0.93546<.
0001
lyc3
0.30624
0.1891
0.91443
<.
0001
0.96817
<.
0001
1.00000
0.97426<.
0001
lyc4
0.21198
0.3696
0.89007
<.
0001
0.93546
<.
0001
0.97426
<.
0001
1.00000
49
Table
33.
Summary
of
Results
by
Sample
Condition
­­
Original
Scale,
Without
hCG
x0
N
Ob
s
Variable
Lower
95%
CL
for
Mean
Upper
95%
CL
for
Mean
Mean
Std
Dev
ANOCOVA
Adjusted
Mean
1
10
Y0
Y1
Y2
Y3
Y4
0.26
3.23
4.45
5.57
6.51
0.40
4.77
6.55
8.17
9.37
0.
33
4.00
5.50
6.87
7.94
0.10
1.08
1.47
1.82
2.00
4.30
5.89
7.32
8.57
2
10
Y0
Y1
Y2
Y3
Y4
0.37
3.34
4.79
5.80
6.82
0.44
4.51
6.36
7.79
9.13
0.41
3.92
5.58
6.79
7.98
0.06
0.81
1.10
1.39
1.62
3.62
5.19
6.34
7.34
x0
=
sample
condition:
1
=
gassed,
2
=
incubated
Bolded
entries
are
statistically
significant
at
the
0.10
level.
Bolded
and
underlined
entries
are
statistically
significant
at
the
0.05
level.

Table
34.
Summary
of
Results
by
Sample
Condition
­­
Log
Scale,
Without
hCG
x0
N
Ob
s
Variable
Lower
95%
CL
for
Mean
Upper
95%
CL
for
Mean
Mean
Std
Dev
ANOCOVA
Adjusted
Mean
1
10
ly0
ly1
ly2
ly3
ly4
­
1.433
1.105
1.437
1.667
1.820
­
0.898
1.583
1.893
2.110
2.252
­
1.165
1.344
1.665
1.889
2.036
0.374
0.334
0.318
0.309
0.302
1.426
1.743
1.961
2.118
2
10
ly0
ly1
ly2
ly3
ly4
­
1.008
1.190
1.553
1.751
1.914
­
0.816
1.502
1.848
2.044
2.202
­
0.912
1.346
1.700
1.897
2.058
0.134
0.218
0.206
0.205
0.202
1.264
1.623
1.825
1.976
x0
=
sample
condition:
1
=
gassed,
2
=
incubated
Bolded
entries
are
statistically
significant
at
the
0.10
level.
Bolded
and
underlined
entries
are
statistically
significant
at
the
0.05
level.
50
Table
35.
Summary
of
Results
by
Sample
Condition
­­
Original
Scale,
With
hCG
x0
N
Ob
s
Variable
Lower
95%
CL
for
Mean
Upper
95%
CL
for
Mean
Mean
Std
Dev
ANOCOVA
Adjusted
Mean
1
10
YC0
YC1
YC2
YC3
YC4
0.25
4.65
10.39
15.60
22.10
0.44
7.50
16.12
25.62
35.31
0.35
6.08
13.25
20.61
28.71
0.13
1.99
4.00
7.00
9.23
6.14
13.44
20.94
29.02
2
10
YC0
YC1
YC2
YC3
YC4
0.30
5.27
12.82
20.13
27.80
0.57
6.93
18.01
28.28
38.99
0.43
6.10
15.42
24.20
33.40
0.18
1.16
3.62
5.69
7.82
6.03
15.22
23.87
33.09
x0
=
sample
condition:
1
=
gassed,
2
=
incubated
Bolded
entries
are
statistically
significant
at
the
0.10
level.
Bolded
and
underlined
entries
are
statistically
significant
at
the
0.05
level.

Table
36.
Summary
of
Results
by
Sample
Condition
 
Log
Scale,
With
hCG
x0
N
Ob
s
Variable
Lower
95%
CL
for
Mean
Upper
95%
CL
for
Mean
Mean
Std
Dev
ANOCOVA
Adjusted
Mean
1
10
lyc0
lyc1
lyc2
lyc3
lyc4
­
1.474
1.524
2.331
2.737
3.085
­
0.818
1.989
2.757
3.214
3.538
­
1.146
1.757
2.544
2.975
3.312
0.458
0.325
0.297
0.334
0.317
1.776
2.562
2.995
3.323
2
10
lyc0
lyc1
lyc2
lyc3
lyc4
­
1.176
1.652
2.539
2.978
3.295
­
0.637
1.931
2.881
3.340
3.665
­
0.906
1.792
2.710
3.159
3.480
0.377
0.195
0.239
0.253
0.259
1.772
2.693
3.139
3.469
x0
=
sample
condition:
1
=
gassed,
2
=
incubated
Bolded
entries
are
statistically
significant
at
the
0.10
level.
Bolded
and
underlined
entries
are
statistically
significant
at
the
0.05
level.
Page
51
of
52
6.0
DISCUSSION
The
optimization
of
the
sliced
testis
assay
is
necessary
in
order
to
proceed
to
the
pre­
validation
and
validation
stages
of
the
testing
of
the
assay
for
use
in
the
Tier
I
tests
for
screening
of
substances
for
potential
as
endocrine
disruptors.
The
Phase
I
studies
have
contributed
to
the
initial
portion
of
this
optimization.
These
factors
will
be
used
for
the
rest
of
the
optimization
in
Part
II.
Without
these
initial
studies
we
would
not
have
used
the
best
gaseous
atmosphere
for
optimal
testosterone
concentrations
from
the
testicular
tissues.

7.0
CONCLUSIONS
The
testosterone
RIA
and
the
LDH
assay
can
both
be
verified
with
M­
199
without
phenol
red.
Both
were
validated
and
show
the
characteristics
necessary
for
use
to
optimize
the
sliced
testis
assay.

There
were
certain
factors
in
the
initial
Phase
I
experiments
that
definitely
were
not
beneficial
to
use
in
the
assay,
for
instance,
22
week
old
rats
do
not
show
the
responsiveness
in
their
testicular
tissue
that
is
necessary
for
an
optimal
assay.
The
air
atmosphere
was
also
not
a
favorable
condition
for
the
assay.
The
prototypical
assay
media,
Media
199
without
phenol
red,
was
equal
to
any
of
the
others
tested.
Statistical
analysis
was
necessary
to
show
that
the
atmosphere
of
5%
CO
2/
95%
O
2
was
optimal
and
that
rats
of
the
11­
15
week
range
could
be
used
for
the
assay.
From
these
conclusions
,
we
were
ready
to
advance
to
the
Phase
II
experiments.

Media
199
without
phenol
red
will
be
used
after
it
is
gassed
with
the
5%
CO
2/
95%
O
2
mixture
and
pH
adjusted
to
7.4
for
testicular
tissues
from
11­
15
week
old
rats
for
the
Phase
II
experimental
studies.

8.0
REFERENCES
EP
Evaluator,
release
3.0
statistical
analysis
software
from
David
Rhoads
Associates,
Inc.,
Kennett
Square,
PA.

FQPA
(
1996).
Food
Quality
Protection
Act
of
1996,
U.
S.
Public
Law
104­
170,
21
U.
S.
C.

46a(
p),
Section
408(
p),
110
STAT.
1489,
August
3,
1996.
Page
52
of
52
APPENDIX
1
LHD
Validation
and
Verification
with
Media
199
without
Phenol
Red
