Meta-analysis of yield response of hybrid field corn to foliar fungicides in the U.S. Corn Belt


Paul, P.A.; Madden, L.V.; Bradley, C.A.; Robertson, A.E.; Munkvold, G.P.; Shaner, G.; Wise, K.A.; Malvick, D.K.; Allen, T.W.; Grybauskas, A.; Vincelli, P.; Esker, P.

Phytopathology 101(9): 1122-1132

2012


The use of foliar fungicides on field corn has increased greatly over the past 5 years in the United States in an attempt to increase yields, despite limited evidence that use of the fungicides is consistently profitable. To assess the value of using fungicides in grain corn production, random-effects meta-analyses were performed on results from foliar fungicide experiments conducted during 2002 to 2009 in 14 states across the United States to determine the mean yield response to the fungicides azoxystrobin, pyraclostrobin, propiconazole + trifloxystrobin, and propiconazole + azoxystrobin. For all fungicides, the yield difference between treated and nontreated plots was highly variable among studies. All four fungicides resulted in a significant mean yield increase relative to the nontreated plots (P < 0.05). Mean yield difference was highest for propiconazole + trifloxystrobin (390 kg/ha), followed by propiconazole + azoxystrobin (331 kg/ha) and pyraclostrobin (256 kg/ha), and lowest for azoxystrobin (230 kg/ha). Baseline yield (mean yield in the nontreated plots) had a significant effect on yield for propiconazole + azoxystrobin (P < 0.05), whereas baseline foliar disease severity (mean severity in the nontreated plots) significantly affected the yield response to pyraclostrobin, propiconazole + trifloxystrobin, and propiconazole + azoxystrobin but not to azoxystrobin. Mean yield difference was generally higher in the lowest yield and higher disease severity categories than in the highest yield and lower disease categories. The probability of failing to recover the fungicide application cost (p(loss)) also was estimated for a range of grain corn prices and application costs. At the 10-year average corn grain price of $0.12/kg ($2.97/bushel) and application costs of $40 to 95/ha, p(loss) for disease severity <5% was 0.55 to 0.98 for pyraclostrobin, 0.62 to 0.93 for propiconazole + trifloxystrobin, 0.58 to 0.89 for propiconazole + azoxystrobin, and 0.91 to 0.99 for azoxystrobin. When disease severity was >5%, the corresponding probabilities were 0.36 to 95, 0.25 to 0.69, 0.25 to 0.64, and 0.37 to 0.98 for the four fungicides. In conclusion, the high p(loss) values found in most scenarios suggest that the use of these foliar fungicides is unlikely to be profitable when foliar disease severity is low and yield expectation is high.

Ecology
and
Epidemiology
Meta
-Analysis
of
Yield
Response
of
Hybrid
Field
Corn
to
Foliar
Fungicides
in
the
U.S.
Corn
Belt
P.
A.
Paul,
L.
V.
Madden,
C.
A.
Bradley,
A.
E.
Robertson,
G.
P.
Munkvold,
G.
Shaner,
K.
A.
Wise,
D.
K.
Malvick,
T.
W.
Allen,
A.
Grybauskas,
P.
Vincelli,
and
P.
Esker
First
and
second
authors:
Department
of
Plant
Pathology,
The
Ohio
State
University,
Ohio
Agricultural
Research
and
Development
Center,
Wooster
44691;
third
author:
Department
of
Crop
Sciences,
University
of
Illinois,
Urbana
61801;
fourth
and
fifth
authors:
Department
of
Plant
Pathology,
Iowa
State
University,
Ames
50011;
sixth
and
seventh
authors:
Department
of
Botany
and
Plant
Pathology,
Purdue
University,
West
Lafayette,
IN
47907;
eighth
author:
Department
of
Plant
Pathology,
University
of
Minnesota,
St.
Paul
55108;
ninth
author:
Mississippi
State
University,
Delta
Research
and
Extension
Center,
Stoneville
38776;
tenth
author:
Department
of
Plant
Science
and
Landscape
Architecture,
University
of
Maryland,
College
Park
20742;
eleventh
author:
Department
of
Plant
Pathology,
University
of
Kentucky,
Lexington
40546;
and
twelfth
author:
Department
of
Plant
Pathology,
University
of
Wisconsin,
Madison
53706.
Accepted
for
publication
29
April
2011.
ABSTRACT
Paul,
P.
A.,
Madden,
L.
V.,
Bradley,
C.
A.,
Robertson,
A.
E.,
Munkvold,
G.
P.,
Shaner,
G.,
Wise,
K.
A.,
Malvick,
D.
K.,
Allen,
T.
W.,
Grybauskas,
A.,
Vincelli,
P.,
and
Esker,
P.
2011.
Meta
-analysis
of
yield
response
of
hybrid
fi
eld
corn
to
foliar
fungicides
in
the
U.S.
Corn
Belt.
Phyto-
pathology
101:1122-1132.
The
use
of
foliar
fungicides
on
field
corn
has
increased
greatly
over
the
past
5
years
in
the
United
States
in
an
attempt
to
increase
yields,
despite
limited
evidence
that
use
of
the
fungicides
is
consistently
profitable.
To
assess
the
value
of
using
fungicides
in
grain
corn
production,
random
-
effects
meta
-analyses
were
performed
on
results
from
foliar
fungicide
experiments
conducted
during
2002
to
2009
in
14
states
across
the
United
States
to
determine
the
mean
yield response
to
the
fungicides
azoxy-
strobin,
pyraclostrobin,
propiconazole
+
trifloxystrobin,
and
propicona-
zole
+
azoxystrobin.
For
all
fungicides,
the
yield
difference
between
treated
and
nontreated
plots
was
highly
variable
among
studies.
All
four
fungicides
resulted
in
a
significant
mean
yield
increase
relative
to
the
nontreated
plots
(P
<
0.05).
Mean
yield
difference
was
highest
for
propiconazole
+
trifloxystrobin
(390
kg/ha),
followed
by
propiconazole
+
azoxystrobin
(331
kg/ha)
and
pyraclostrobin
(256
kg/ha),
and
lowest
for
azoxystrobin
(230
kg/ha).
Baseline
yield
(mean
yield
in
the
nontreated
plots)
had
a
significant
effect
on
yield
for
propiconazole
+
azoxystrobin
(P
<
0.05),
whereas
baseline
foliar
disease
severity
(mean
severity
in
the
nontreated
plots)
significantly
affected
the
yield response
to
pyra-
clostrobin,
propiconazole
+
trifloxystrobin,
and
propiconazole
+
azoxy-
strobin
but
not
to
azoxystrobin.
Mean
yield
difference
was
generally
higher
in
the
lowest
yield
and
higher
disease
severity
categories
than
in
the
highest
yield
and
lower
disease
categories.
The
probability
of
failing
to
recover
the
fungicide
application
cost
(m
oss
)
also
was
estimated
for
a
range
of
grain
corn
prices
and
application
costs.
At
the
10
-year
average
corn
grain
price
of
$0.12/kg
($2.97/bushel)
and
application
costs
of
$40
to
95/ha,
m
oss
for
disease
severity
<5%
was
0.55
to
0.98
for
pyraclo-
strobin,
0.62
to
0.93
for
propiconazole
+
trifloxystrobin,
0.58
to
0.89
for
propiconazole
+
azoxystrobin,
and
0.91
to
0.99
for
azoxystrobin.
When
disease
severity
was
>5%,
the
corresponding
probabilities
were
0.36
to
95,
0.25
to
0.69,
0.25
to
0.64,
and
0.37
to
0.98
for
the
four
fungicides.
In
conclusion,
the
high
m
oss
values
found
in
most
scenarios
suggest
that
the
use
of
these
foliar
fungicides
is
unlikely
to
be
profitable
when
foliar
disease
severity
is
low
and
yield
expectation
is
high.
Additional
keywords:
quinone
outside
inhibitor
fungicides,
research
syn-
thesis,
risk
analysis.
The
use
of
foliar
fungicides
is
often
profitable
in
seed
corn
(Zea
mays
L.)
production
(38,67,68)
but
has
been
much
less
profitable
in
grain
corn
production
because
of
the
substantially
lower
value
of
grain
compared
with
seed.
Profitable
fungicide
use
in
corn
grown
for
grain
is
influenced
by
grain
price
and
application
costs
and
is
strongly
dependent
on
the
yield
potential
and
disease
-sus-
ceptibility
or
resistance
of
the
hybrid
planted,
and
foliar
disease
intensity
throughout
the
growing
season
(38).
In
the
U.S.
Corn
Belt,
several
foliar
diseases
are
of
concern,
depending
on
the
pro-
duction
region,
but
gray
leaf
spot
(GLS),
caused
by
Cercospora
zeae-maydis
Tehon
&
E.
Y.
Daniels,
has
been
the
disease
of
greatest
concern
since
first
becoming
a
problem
in
the
1980s
and
1990s
(31-33).
The
elevation
of
GLS
from
a
disease
of
secondary
im-
portance
to
a
major
problem
throughout
the
eastern
United
States
and
the
Midwest
paralleled
the
adoption
of
reduced
tillage
(31,33).
Disease
severity
tends
to
be
higher
in
areas
where
susceptible
hybrids
are
planted
in
no
-till,
continuous
-corn
fields
(10,33,49,
Corresponding
author:
P.
A.
Paul;
E-mail:
pau1.661@osu.edu
doi:10.1094/
PHYTO-03-11-0091
©
2011
The
American
Phytopathological
Society
66).
However,
the
presence
of
corn
residue
on
the
soil
surface,
although
increasing
the
risk
of
foliar
disease caused
by
necrotro-
phic
fungi,
does
not
always
lead
to
severe
disease.
Unless
weather
conditions
are
favorable
for
infection
(3,53,63)
and
sporulation
(46),
GLS
may
not
reach
yield
-limiting
levels
(e.g.,
leaves
above
the
ear
become
severely
blighted)
(1),
even
if
residue
is
present.
As
a
result,
foliar
fungicide
applications
for
management
of
GLS
(or
other
residue
-borne
diseases)
may
not
be
warranted,
even
when
crop
production
practices
favor
GLS
or
other
diseases.
Since
2006,
there
has
been
an
increased
interest
in
foliar
fungicide
application
in
corn
(and
other
field
crops)
in
the
United
States
for
reasons
other
than
simply
disease
control
(37).
Claims
of
substantial
yield
increase
in
hybrid
corn
in
response to
foliar
fungicides,
even
in
the
absence
of
foliar
disease
symptoms,
have
led
to
fungicide
applications
on
several
million
hectares,
with
costs
in
the
millions
of
dollars
across
the
U.S.
Corn
Belt
(37).
Modern
hybrids
with
high
yield
potential
and
new
fungicide
active
ingredients
with
effects
on
crop
physiology
have
been
given
as
possible
motivations
for
increased
fungicide
application
in
field
corn
production
(37).
In
particular,
based
on
bioassays
and
studies
conducted
under
controlled
conditions,
quinone
outside
inhibitor
(QoI)
fungicides
have
been
shown
to
induce
1122
PHYTOPATHOLOGY
physiological
and
developmental
changes
in
plants,
including
retardation
of
senescence
due
to
reduced
oxidative
stress
(72),
in-
creased
photosynthetic
capacity,
transient
inhibition
of
respiration,
inhibition
of
ethylene
biosynthesis
(15),
and
reduction
of
stomatal
aperture
and
water
loss
through
transpiration
(14,39).
These
changes
are
believed
to
translate
into
greater
stress
tolerance
and
higher
yields
in
QoI-treated
crops
under
field
conditions,
and
have
prompted
recent
additions
to
the
label
of
one
of
the
most
widely
used
QoI
fungicide
products,
23.6%
pyraclostrobin
(Headline;
BASF
Corporation
Agricultural
Products,
Research
Triangle
Park,
NC).
In
2009,
the
U.S.
Environmental
Protection
Agency
granted
a
supplemental
label
for
the
use
of
Headline
for
disease
control
and
its
"plant
health"
benefit,
which
may
lead
to
more
widespread
use
of
fungicides
without
regard
to
disease
risk.
Claims
of
substantial
yield
increases
in
response to
QoI-based
fungicides
have
not
always
been
substantiated
by
adequate
data
analysis
and
research
synthesis.
For
instance,
conclusions
regard-
ing
the
overall
yield
benefit
and
economics
of
fungicide
use
in
corn
have
been
based
largely
on
tests
of
treatment
significance
from
individual
trials,
a
tally
of
the
number
of
trials
with
significant
results
(vote
counting),
and
simple
unweighted
arith-
metic
mean
yield
difference
between
treatments
across
multiple
trials
(27,37).
There
are
several
reasons
why
these
approaches
may
not
necessarily
be
appropriate
for
the
synthesis
of
this
type
of
data
(4,18).
Simple
means
of
effect
sizes
across
studies
give
the
same
weight
to
studies
with
high
variability
(low
precision)
and
to
those
with
low
variability
(high
precision)
and,
hence,
do
not
account
for
inherent
differences
among
trials
(including
between
-
study
variability
and
the
presence
of
study
-specific
conditions
as
well
as
unequal
within
-study
variability)
that
likely
influence
the
magnitude
and
precision
of
the
estimated
overall
fungicide
effect.
Madden
and
Paul
(36)
presented
a
discussion
of
the
fallacy
of
research
synthesis
based
on
vote
counting
and
provided
several
justifications
why
a
quantitative
method
known
as
meta
-analysis
is
much
more
appropriate.
In
brief,
the
statistical
power
of
the
vote
-counting
method
is
generally
very
low,
perhaps
lower
than
that
of
the
majority
of
the
individual
studies,
and
may
actually
decrease
as
the
number
of
studies
increase.
The
statistical
power
of
meta
-analysis,
on
the
other
hand,
is
higher
than
that
of
indi-
vidual
analyses
and
vote
counting
and,
therefore,
this
approach
is
less
likely
to
lead
to
erroneous
conclusions
regarding
treatment
effects
(4,21,36).
Meta
-analysis,
a
quantitative
synthesis
of
research
findings
from
multiple
individual
trials
(4,18,34,44),
provides
an
alterna-
tive
to
narrative
review
or
qualitative
research
synthesis.
In
meta
-
analysis,
an
overall
effect
size
is
estimated,
in
which
each
individual
effect
size
is
given
a
weight
that
is
an
inverse
function
of
(i)
the
within
-study
variance
and,
for
a
random
-effects
analysis,
(ii)
the
between
-study
variance.
An
effect
size
is
any
statistic
(mean
yield
difference
in
this
case)
that
can
be
used
to
evaluate
the
overall
effect
of
some
treatment
or
the
strength
of
a
relation-
ship
between
variables
(4,17,34,36).
Both
within-
and
between
-
study
variability
are
considered
when
estimating
the
overall
mean
effect
size
through
an
iterative
weighing
algorithm.
The
objective
of
this
study
was
to
use
meta
-analysis
to
determine
(i)
the
overall
mean
effect
of
QoI-based
fungicide
treatments
on
grain
yield
of
hybrid
field
corn
and
the
corresponding
confidence
interval
(CI)
and
between
-study
variability,
(ii)
the
influence
of
categorical
levels
of
yield
and
foliar
disease
severity
in
the
nontreated
plot
(baseline)
on
mean
effect
size,
and
(iii)
the
probability that
there
is
an
economic
benefit
to
applying
a
fungicide
in
a
randomly
selected
study
or
field
under
a
range
of
scenarios
of
grain
market
prices
and
application
costs.
MATERIALS
AND
METHODS
Data
set.
The
data
used
in
this
investigation
were
obtained
from
fungicide
research
and
on
-farm
studies
conducted
by
uni-
versity
researchers
(the
co-authors
of
this
article)
and
from
summaries
of
foliar
fungicide
studies
published
in
Fungicide
and
Nematicide
Tests
(F&N
Tests)
and
Plant
Disease
Management
Reports
(PDMR)
(The
American
Phytopathological
Society,
St.
Paul,
MN).
For
the
purpose
of
this
research,
a
study
is
defined
as
an
entry
(a
row
in
the
data
matrix)
with
a
unique
combination
of
year,
location,
hybrid,
and
corresponding
yield
and
disease
(when
available)
data.
In
order
for
a
study
to
be
considered
for
inclusion
in
the
analysis,
it
had
to
be
replicated,
with
a
random
assignment
of
treatments
consisting
of
a
single
application
of
at
least
one
of
the
following
fungicides:
23.6%
pyraclostrobin
(Headline;
BASF
Corporation
Agricultural
Products),
11.4%
propiconazole
+
11.4%
trifloxystrobin
(Stratego;
Bayer
CropScience,
Research
Triangle
Park,
NC),
7%
azoxystrobin
+
11.7%
propiconazole
(Quilt;
Syn-
genta
Crop
Protection
Inc.,
Greensboro,
NC),
or
22.9%
azoxy-
strobin
(Quadris;
Syngenta
Crop
Protection
Inc.)
applied
at
label
-
recommended
rates
(438
ml/ha
for
Headline,
730
ml/ha
for
Stratego,
1,023
ml/ha
for
Quilt,
and
658
ml/ha
for
Quadris)
be-
tween
the
VT
(tassel emergence)
and
R1
(silk
emergence)
growth
stages
(52).
Studies
also
had
to
have
some
measure
of
yield
(volume
or
weight
per
unit
area)
for
the
nontreated
comparison
plots
and
at
least
one
of
the
fungicide
treatments
and
some
measure
of
variability
of
the
yield
response,
such
as
the
least
sig-
nificant
difference
or
coefficient
of
variation.
Inclusion
of
infor-
mation
on
disease
severity
was
not
a
criterion
for
selection
because
this
research
synthesis
encompasses situations
with
or
without
visible
foliar
disease.
To
be
included,
studies
had
to
be
published
by
February
2011
and
report
on
results
from
2009
or
earlier.
Unpublished
summaries
and
raw
data
were
gathered
from
a
total
of
187
studies
conducted
by
the
co-authors
of
this
article.
Most
of
the
studies
were
conducted
with
fungicide
treatments
and
a
nontreated
check
in
a randomized
complete
block
design;
however,
some
studies
(25%)
had
fungicide
treatment
and
hybrid
in
a
split
-plot
arrangement,
with
hybrid
as
the
whole
-plot
factor.
For
the
latter
group
of
studies,
each
hybrid
was
treated
as
a
separate
observation
in
the
meta
-analysis,
because
hybrids
were
in
separate
replicated
plots
with
fungicide
treatments
randomized
within
each
hybrid.
The
number
of
replicate
blocks
ranged
from
two
to
six,
and
experimental
units
were
7.62
to
several
hundred
meters
long
x
3.05
to
several
meters
wide.
The
specific
fungicide
treatments
varied
among
studies
but,
in
all
of
the
187
selected
trials,
one
or
more
of
the
treatments
satisfied
the
aforementioned
criteria.
In
most
cases,
a
nonionic
surfactant
was
added
with
the
fungicide
mixture
at
a
rate
of
0.25
to
1.0%
(vol/vol).
Treatments
were
either
air
applied
(in
13
studies)
or
ground
applied
using
tractor
-mounted
sprayers,
specialized
high
-clearance
sprayers,
or
hand-held
booms,
at
46.67
to
187.08
liters/ha.
Grain
was
har-
vested
using
a
research
-plot
or
commercial
combine
harvester
and
yield
was
adjusted
to
15.5%
moisture.
Yield
was
converted
from
the
original
unit
of
bushels
per
acre
to
metric
tons
per
hectare,
based
on
a
test
weight
of
25
kg/bushel.
Using
corn
and
azoxystrobin,
pyraclostrobin,
or
trifloxystrobin
as
keywords,
the
PDMR
search
engine
was
used
to
screen
for
trials
that met
the
aforementioned
criteria.
In
general,
for
any
given
issue
of
PDMR,
each
published
summary
consisted
of
a
table
with
mean
grain
yield
for
fungicide
-treated
and
nontreated
plots
from
a
single
trial.
However,
in
some
cases,
the
published
summary
consisted
of
data
from
separate
trials
(with
the
same
or
similar
sets
of
treatments)
conducted
at
multiple
locations
or
using
different
hybrids
at
the
same
location.
Each
trial
was
con-
sidered
an
individual
study
for
the
analysis.
Data
were
collected
from
a
total
of
25
PDMR
studies
for
the
period
2002
to
2009
(6-
8,22-25,28-30,58-62,69-71).
Eight
additional
studies
were
found
in
PDMR,
however,
in
which
four
of
these
were
not
considered
because
they
did
not
meet
the
aforementioned
study
selection
criteria.
The
other
four
PDMR
studies
(41,47,48,56)
were
already
included
in
the
dataset
as
raw
data
collected
from
the
co-authors.
Vol.
101,
No.
9,
2011
1123
In
total,
212
studies
(187
as
raw
data
from
experiments
and
25
as
summaries
from
PDMR)
were
compiled
from
14
states
for
the
period
2002
to
2009:
86
from
Ohio,
33
from
Iowa,
23
from
Illinois,
24
from
Minnesota,
8
from
Maryland,
and
38
from
Virginia,
Mississippi,
Kansas,
Wisconsin,
Missouri,
Kentucky,
Indiana,
North
Dakota,
and
Nebraska.
Among
the
fungicide
treatments
evaluated
in
this
study,
pyraclostrobin
was
tested
in
172
studies
(67
from
Ohio,
31
from
Iowa,
17
from
Illinois,
19
from
Minnesota,
and
38
from
the
other
10
states),
propiconazole
+
trifloxystrobin
in
72
studies
(24
from
Ohio,
10
from
Iowa,
23
from
Illinois,
and
15
from
8
other
states),
propiconazole
+
azoxystrobin
in
61
studies
(17
from
Illinois
and
44
from
12
other
states),
and
azoxystrobin
in
25
studies
(16
from
Illinois,
4
from
Iowa,
3
from
Indiana,
and
2
from
Ohio).
Meta
-analysis
of
fungicide
effect
on
grain
yield.
For
each
study,
the
difference
in
mean
yield
between
treated
and
nontreated
plots
for
each
fungicide
tested
was
used
as
the
effect
size
to
determine
the
overall
mean
yield
response
(mean
effect
size)
of
h1brid
corn
to
QoI-based
fungicide
treatments.
For
instance,
if
Check
is
the
mean
yield
for
the
nontreated
plot
in
a
study
and
XTreated
the
mean
yield
for
a
treatment
in
a
study
(where
"Treated"
corresponds
to
either
pyraclostrobin,
propiconazole
+
trifloxystrobin,
propiconazole
+
azoxystrobin,
or
azoxystrobin),
then
the
effect
size
(D,
unstandardized
mean
difference)
was
computed
simply
as
D
-
=
7
Treated
7Check
(4,34,36).
D
is
an
intui
tive
and
informative
summary
of
fungicide
effect
on
grain
yield
and,
as
such,
an
appropriate
effect
size
for
the
questions
being
asked
in
this
investigation
(36).
Separate
random
-effects
meta
-analyses
were
performed
as
described
by
Paul
et
al.
(42)
and
Madden
and
Paul
(36)
for
each
fungicide
to
estimate
the
overall
mean
effect
size
(D)
and
to
determine
the
variability
in
effect
size
among
studies.
Models
were
fitted
to
the
data
in
PROC
MIXED
of
SAS
using
maximum
likelihood
as
described
elsewhere
(36,40,55,64).
In
the
analysis,
each
study
was
given
an
initial
weight
that
was
inversely
proportional
to
the
sampling
variance
(within
-study
variance)
of
the
mean
difference,
computed
as
s,
2
=
(2
x
V)I
r
,
where
the
i
sub-
script
refers
to
the
ith
study
and
r
the
number
of
replicates
within
a
study.
V
is
the
mean
square
error
(residual
variance)
from
an
analysis
of
variance
(ANOVA)
of
the
effect
of
fungicide
treatment
on
yield.
For
studies
for
which
the
original
data
were
available
(fungicide
research
and
on
-farm
strip
trials),
the
residual
variance
was
obtained
directly
from
a
preliminary
ANOVA
of
the
raw
data
in
PROC
MIXED
(35)
of
SAS
(SAS
Institute
Inc.,
Cary,
NC).
For
studies
from
F&N
Tests
and
PDMR
for
which
only
the
means
and
the
least
significant
difference
(LSD)
were
available,
V
was
calcu-
lated
from
the
presented
LSD
as
described
previously
(36,42).
In
the
maximum
likelihood
approach,
the
initial weights
are
updated
in
an
iterative
fashion
by
inclusion
of
the
between
-study
variance.
The
random
effects
model
can
be
written
as
D,
N(1,t,s
+6
2
)
,
where
D,
is
the
mean
difference
(effect
size)
in
yield
between
the
treatment
and
the
nontreated
for
the
ith
study
(i
=
1,...,k),
—A/(0)
indicates
a
normal
distribution,
n
is
the
expected
value
of
D
(mean
effect
size),
6
2
is
the
between
-study
variance,
and
s
1
2
is
the
within
-study
variance
for
the
ith
study.
Study
was
considered
a
random
effect
in
each
analysis,
and
models
were
fitted
using
D,
as
the
response
variable
in
the
model
statement.
The
within
-study
variances
were
incorporated
(as
fixed
weights,
1/s,
2
)
into
the
model
fitting
procedure
using
the
weight
statement
in
PROC
MIXED.
The
estimate
of
n
equals
D
,
and
separate
values
TPYRA
TPROP+TRIF
TPROP+AZOX
and
D
Az°x
)
were
computed
for
each
fungicide
in
separate
analyses.
Standard
normal
test
statistics
(Z)
(11,19)
were
used
to
determine
whether
r )
PYRA
TPROP+TRIF
TPROP+AZOX
and
D
Az°x
were
significantly
different
from
zero.
Standard
errors
of
the
effect
sizes
were
determined
by
PROC
MIXED
based
on
mixed
model
theory
(35)
and
used
to
determine
the
95%
CIs
for mean
effect
sizes,
as
described
elsewhere
(4,34).
Yield
response
to
fungicides
as
influenced
by
baseline
yield
and
foliar
disease
severity.
Two
of
the
most
common
expla-
nations
for
differences
in
yield
response to
fungicide
treatments
among
trials
are
the
difference
in
baseline
yield
(YLD_BASE)
among
hybrids
used
in
the
trials
and
difference
in
disease
intensity.
YLD_BASE
differences
could
be
a
function
of
several
factors
acting
separately
or
in
combination
to
affect
yield,
including
the
yield
potential
of
the
hybrid,
soil
and
weather
conditions,
and
stresses
caused
by
pests
and
diseases.
For
the
population
of
studies
included
in
this
investigation,
mean
yield
and
foliar
disease
severity
(when
reported)
in
the
nontreated
check
were
used
as
measures
of
YLD_BASE
and
baseline
disease
severity
(DIS_BASE),
respectively,
in
the
study.
YLD_BASE
served
a
surrogate
for
possible
biotic
and
abiotic
factors
affecting
yield
in
a
given
study.
This
is
similar
to
the
approach
taken
in
medical
statistics,
where
the
response
variable
in
the
control
is
used
to
represent
baseline
risk
of
disease,
disorder,
or
clinical
condition
(2,9).
Based
on
the
range
(difference
between
minimum
and
maxi-
mum
yields)
and
distribution
of
the
yield
data
for
the
nontreated
plots,
studies
were
grouped
into
different
categories,
using
histograms
as
a
guide
for
defining
cutoffs
(natural
breaks
in
the
distribution).
For
pyraclostrobin
and
propiconazole
+
trifloxy-
strobin,
the
categories
were
(i)
YLD_BASE1
<
9.1
metric
tons
per
hectare
(MT/ha)
0145
bushels/acre),
(ii)
YLD_BASE2
of
9.1
to
11.9
MT/ha
(145
and
190
bushels/acre),
and
(iii)
YLD_BASE2
11.9
MT/ha.
There
were
57,
61,
and
54
studies
with
pyra-
clostrobin
in
categories
i,
ii,
and
iii,
respectively,
and
18, 17,
and
37
studies
with
propiconazole
+
trifloxystrobin
in
the
three
categories.
For
propiconazole
+
azoxystrobin
and
azoxystrobin,
the
categories
were
(i)
YLD_BASE1
<
11.9
MT/ha
and
(ii)
YLD_BASE2
11.9
MT/ha.
There
were
28
and
33
studies
with
propiconazole
+
azoxystrobin
in
the
first
and
second
categories,
respectively,
and
12
and
13
with
azoxystrobin
in
the
two
categories.
For
each
fungicide,
if
foliar
disease
severity
(percent
diseased
leaf
area)
was
reported,
studies
were
grouped
into
two
categories
based
on
ear
leaf
severity
in
the
nontreated
plot
between
the
R4
and
R6
growth
stages
(52):
(i)
DIS_BASE1
<
5%
and
(ii)
DIS_
BASE2
5%.
The
5%
cutoff
was
based
on
a
preliminary
sum-
mary
of
a
subset
of
the
data
(data
collected
in
2008)
which
showed
a
clear
difference
in
yield
response
between
the
two
categories
(G.
Shaner,
unpublished).
The
diseases
reported
were
GLS
(C.
zeae-maydis),
northern
corn
leaf
blight
(Exserohilum
turcicum),
and
common
rust
(Puccinia
sorghi).
GLS
was
the
disease
most
frequently
reported,
either
as
the
only
disease
or
in
combination
with
one
or
both
of
the
other
two
diseases
as
total
diseased
ear
leaf
area.
A
third
category,
DIS_BASE3,
was
created
for
studies
without
reported
foliar
disease
severity.
These
studies
may
represent
situations
when
visible
disease
symptoms
were
not
observed
(and
not
mentioned)
or
when
disease
severity
assess-
ments
were
not
conducted
or
reported.
Thus,
it
would
likely
represent
a
diverse
set
of
conditions.
Studies
in
which
disease
intensity
was
reposted
as
area
under
the
disease
progress
curve
or
severity
on
an
ordinal
rating
scale
were
included
in
the
third
category,
unless
the
severity
category
could
have
been
deduced
from
the
reports.
For
instance,
in
one
study
(61),
GLS
severity
was
reported
on
a
1
-to
-5
scale,
with
1
=
a
trace
number
of
lesions
on
leaves
below
the
ear,
none
on
leaves
above;
2
=
many
lesions
on
leaves
below
the
ear,
trace
above;
3
=
severe
lesion
develop-
ment
on
leaves
below
the
ear,
all
leaves
above
with
lesions;
4
=
all
leaves
with
severe
lesion
development,
but
green
tissue
still
visible;
and
5
=
all
leaves
dry
and
dead.
The
ordinal
scores
were
accompanied
by
corresponding
whole
-plant
severity
scores
(0
to
100%),
with
a
3
on
the
ordinal
scale
corresponding
to
35%
whole
-plant
severity,
4
on
the
ordinal
scale
to
,-
-t,'75%
whole
-plant
severity,
and
5
to
>96%
whole
-plant
severity.
These
severity
scores,
on
both
scales,
corresponded
to
>5%
GLS
severity
on
the
1124
PHYTOPATHOLOGY
ear
leaf
(E.
Stromberg,
personal
communication).
Foliar
disease
severity
data
were
available
for
101
studies
with
pyraclostrobin
(65
in
the
first
category
and
36
in
the
second),
59
with
propi-
conazole
+
trifloxystrobin
(28
in
the
first
category
and
31
in
the
second),
37
with
propiconazole
+
azoxystrobin
(15
and
22
in
categories
i
and
ii,
respectively),
and
20
with
azoxystrobin
(7
and
14
in
categories
i
and
ii,
respectively).
For
this
part
of
the
study,
the
meta
-analytical
models
were
expanded
to
evaluate
the
effect
of
YLD_BASE
and
DIS_BASE,
categorical
moderator
variables
(4,34),
on
the
effect
size
(D).
This
is
a
common
analytical
approach
in
meta
-analysis
when
hetero-
geneity
of
the
effect
sizes
between
studies
is
verified
(42,43).
Because
the
moderator
variables
were
treated
as
fixed
effects
in
the
analysis,
the
overall
model
was
a
mixed
effect
model.
Sepa-
rate
models
were
fitted
for
each
fungicide
and
each
moderator
variable
in
PROC
MIXED
to
determine
whether
YLD_BASE
and
DIS_BASE
significantly
affected
D
and
to
estimate
separate
D
values
for
each
level
of
each
moderator
variable.
Models
were
fitted
using
the
same
within
-study
(sampling)
variances
as
above
in
the
analysis
without
the
moderator
variable.
With
the
moderator
variable,
the
mixed
-effect
model
can
be
written
as
D,
+
(75
2
)
,
where
6,
is
the
effect
of
the
moderator
variable
for
the
ith
study
on
the
mean
effect
size.
All
other
terms
are
as
described
previously.
In
this
model,
the
mean
effect
size
is
not
a
constant
across
all
studies
but
a
sum
of
an
overall
mean
(n)
and
the
effect
of
the
moderator
variable
(s,).
To
determine
whether
the
moderator
variable
had
a
significant
effect
on
the
mean
effect
sizes,
x
2
tests
(11,19,34,42)
were
used,
and
linear
contrasts
were
used
to
estimate
the
mean
effect
sizes
and
their
standard
errors
and
CIs
for
each
level
of
the
moderator
variable.
Study
heterogeneity.
A
likelihood
ratio
statistic
(LRS)
was
used
to
test
whether
the
between
-study
variance
was
>0
(36).
For
each
fungicide,
the
meta
-analytical
model
was
refitted
without
the
random
effect
of
study
(a
fixed
-effect
model)
and
the
difference
in
—2
times
the
log
-likelihood
between
the
fixed-
and
random
-effects
model
fits
(the
LRS)
was determined.
Under
the
null
hypothesis
of
cr
2
=
0,
LRS
has
a
distribution
that
is
a
mixture
of
x
2
distri-
butions
with
0
and
1
degrees
of
freedom
(35,36).
In
addition,
the
R
2
statistic
(this
is
not
the
coefficient
of
determination)
of
Higgins
and
Thompson
(20)
was
calculated.
This
is
a
statistic
used
to
determine
the
impact
of
the
between
-study
variability
on
the
effect
sizes.
R
2
>
1.5
indicates
considerable
heterogeneity
and
the
need
to
account
for
between
-study
variability
in
the
analysis.
Prediction
and
risk
analysis.
Although
the
overall
mean
yield
difference
(
D
)
can
be
used
to
determine
average
cost
and
bene-
fits
of
fungicide
application
in
the
long
run,
D
alone
cannot
re-
veal
the
chance
of
a
given
yield
response
in
any
single
field
or
study.
It
is
important
for
corn
producers
and
researchers
to
know
what
can
be
expected
of
a
certain
treatment
or
crop
management
practice
in
a
future
trial
or
growing
season.
It
would
also
be
of
interest
to
estimate
the
risk
of
losing
money
(because
of
increased
production
cost)
if
a
certain
production
practice
is
used
when
it
is
not
warranted.
One
can
estimate
the
probability
of
a
certain
effect
-size
outcome
based
on
results
from
previous
studies
in
a
meta
-analysis
(36).
For
instance,
for
each
of
the
fungicides
evalu-
ated
in
this
investigation,
the
mean
effect
size
(
D
)
and
estimated
between
-study
variance
(
es
2
)
from
the
meta
-analyses
can
be
used
to
estimate
the
probability
of
the
yield
response to
fungicide
in
a
new
randomly
selected
study
—done
in
the
same
manner
as
the
studies
considered
in
this
analysis
—being
lower
(or
higher)
than
some
constant
(C).
This
probability
is
estimated
as
p
=
O[(C
D
)I&
],
where
6(.)
is
the
cumulative
standard
-normal
function
and
&
is
the
estimated
between
-study
standard
deviation
(36,43,
64).
In
particular,
one
can
estimate
the
probability
that
the
yield
response
(effect
size
in
a
future
study)
is
lower
than
that
necessary
to
offset
the
cost
of
the
fungicide
treatment
(product
plus
application
costs).
This
probability
is
the
risk
of
failing
to
recover
the
cost
of
applying
the
fungicide.
Assuming
grain
prices
of
$79
to
276
U.S./MT
(
-
-t$2
to
7
U.S./
bushel)
and
fungicide
application
costs
of
$40
to
95
U.S./ha
(
-
-t$16
to
40
U.S./acre),
the
minimum
yield
increase
in
response to
fungicide
treatment
necessary
to
break
even
was
estimated
for
each
combination
of
grain
price
and
application
cost.
D
and
&
2
from
the
meta
-analyses,
with
disease
severity
as
moderator
vari-
able,
were
then
used
to
estimate
the
probability
of
not
achieving
the
minimum
breakeven
yield
increase
in
a
new
study,
for
each
selected
grain
price
—application
cost
combination,
fungicide,
and
baseline
disease
class.
For
the
purpose
of
this
investigation,
this
probability
was
called
m
oss
.
For
instance,
assuming
all
other
crop
production
costs
remain
constant,
if
fungicide
application
cost
is
$62
U.S./ha
and
grain
market
price
is
$197
U.S./MT,
it
would
take
a
yield
increase
of
C
=
314
kg/ha
to
offset
the
cost
of
fungicide
application.
Hence,
the
probability
of
D
<
314
kg/ha
can
be
estimated
as
1,
10
,,,
=
4(314
D)
I
&]
.
RESULTS
Yield
response
and
meta
-analysis.
Mean
grain
yield
varied
among
studies
and
among
fungicide
treatments.
In
general,
mean
yield
was
higher
in
plots
treated
with
fungicides
than
in
the
non
-
treated
plots
(Fig.
1).
For
all
fungicides,
the
difference
in
mean
yield
between
treated
and
nontreated
plots
(
D
=XTreated
X
Check)
varied
among
studies.
A
subset
of
the
studies
had
a
negative
yield
response,
meaning
that
the
nontreated
plots
had
higher
mean
yields
than
the
fungicide
-treated
plots
(Fig.
2).
This
occurred
in
26
to
48%
of
the
studies,
depending
on
the
fungicide.
D
was
—1,940
to
2,211
kg/ha
for
pyraclostrobin,
—1,793
to
2,999
kg/ha
for
propiconazole
+
trifloxystrobin,
—2,368
to
3,034
kg/ha
for
propiconazole
+
azoxystrobin,
and
—886
to
1,821
kg/ha
for
azoxy-
strobin.
As
indicated
by
the
vertical
standard
error
bars
in
Figure
2,
a
measure
of
the
within
-study
variability
of
the
effect
size,
the
precision
with
which
D
was
estimated,
varied
substantially
among
studies
for
all
of
the
tested
fungicides.
Based
on
the
standard
normal test
statistics
from
the
meta
-
analyses,
the
overall
mean
yield
difference
(the
effect
size,
D
)
was
positive
and
significantly
different
from
zero
for
all
of
the
tested
fungicides
(Table
1).
D
was
highest
for
propiconazole
+
trifloxystrobin
followed
by
propiconazole
+
azoxystrobin
and
pyraclostrobin,
and
lowest
for
azoxystrobin.
The
width
of
the
95%
CI
around
D
was
narrowest
for
pyraclostrobin
and
in-
creased
with
decreasing
sample
size
(number
of
studies),
being
Yield
(1000
kg/ha)
20
18
-
16
14 -
12
-
10
-
-
6
-
4
-
2
-
0
S
0
06
'
-
Ok
\
1.
01-
c
4
,0
9
'`
ck
cp?
P
'
?
Fig.
1.
Box
plots
summarizing
the
distribution
of
grain
yield
of
hybrid
corn
treated
with
the
fungicides pyraclostrobin
(PYRA)
propiconazole
+
trifloxy-
strobin
(PROP+TRIF),
propiconazole
+
azoxystrobin
(PROP+AZOX),
and
azoxystrobin
(AZOX)
and
the
nontreated
plot.
Solid
and
dashed
lines
within
each
box
represent
median
and
mean,
respectively.
Top
and
bottom
lines
of
the
boxes
represent
the
75th
and
25th
percentiles
of
the
data,
respectively.
Vertical
bars
extending
beyond
the
boxes
represent
the
10th
and
90th
percen-
tiles,
whereas
circles
indicate
outliers.
Vol.
101,
No.
9,
2011
1125
the
widest
for
azoxystrobin.
Hence,
the
precision
with
which
the
effect
size
was
estimated
was
partly
affected
by
the
number
of
studies
in
the
analysis.
For
instance,
there
were
147
more
studies
with
pyraclostrobin
than
with
azoxystrobin,
with
the
width
of
the
95%
CI
being
280
kg/ha
(4.46
bushels/acre)
narrower
for
the
former
than
the
latter
fungicide.
Effect
of
moderator
variables
and
between
-study
vari-
ability.
Based
on
the
x2
test
statistics
from
the
analyses,
the
effect
of
YLD_BASE
on
the
effect
size
was
statistically
significant
for
propiconazole
+
azoxystrobin
(P
=
0.03),
marginally
significant
for
propiconazole
+
trifloxystrobin
(P
=
0.07),
but not
significant
for
azoxystrobin
(P
=
0.103)
and
pyraclostrobin
(P
=
0.805).
Studies
in
the
lowest
YLD_BASE
category
consistently
had
higher
mean
yield
responses
to
the
fungicides
than
those
in
the
highest
yield
category
(Fig.
3).
For
propiconazole
+
trifloxy-
Yield
difference
(1,000
kglha
2
-4
6
4
2
0
-2
4
strobin,
T
P
"
±TRIF
was
significantly
(P
=
0.02)
higher
for
YLD_BASE1
(YLD_BASE
<
9.1
MT/ha)
than
YLD_BASE3
(YLD_BASE
11.9
MT/ha).
However,
the
differences
between
YLD_BASE1
versus
YLD_BASE2
and
YLD_BASE2
versus
YLD_BASE3
were
not
statistically
significant.
Similarly,
for
propiconazole
+
azoxystrobin,
YLD_BASE1 (YLD_BASE
<
11.9
MT/ha)
had
a
significantly
(P
=
0.03)
higher
7
PR°P±Az°x
(Table
2)
than
YLD_BASE2
(YLD_BASE
11.9
MT/ha).
The
width
of
the
95%
CI
around
7
(Fig.
3)
varied
from
one
YLD_BASE
cate-
gory
to
another,
tending
to
be
narrower
for
categories
with
the
larger
sample
sizes
(Fig.
3).
The
effect
of
DIS_BASE
on
the
mean
effect
size
was
statisti-
cally
significant
for
pyraclostrobin
(P
=
0.008),
propiconazole
+
trifloxystrobin
(P
=
0.030),
and
propiconazole
+
azoxystrobin
(P
=
0.012)
but not
significant
for
azoxystrobin
(P
=
0.239).
For
A
K=
172
PYRA
'If
_1
IA
144
Ii
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100105110115120125130135140145150155160165170
B
K=72
PROP+TRIF
C
K=61
I
PROP+AZOX
_
4.144"
.
D
K=25
AZOX
flfili
rri
n
ri
1
li
l I
0
1.J
U
L.'
0
10
20
30
40
50
60
70
0
5
10
15
20
25
30
35
40
45
50
55
60
0
2
4
6
8
10
12
14
16
18
20
22
24
26
Study
number
(sorted)
Fig.
2.
Mean
yield
difference
between
fungicide
treatments
and
the
nontreated,
sorted
from
the
lowest
to
the
highest,
for
the
fungicides
A,
pyraclostrobin,
(PYRA);
B,
propiconazole
+
trifloxystrobin
(PROP+TRIF);
C,
propiconazole
+
azoxystrobin
(PROP+AZOX);
and
D,
azoxystrobin
(AZOX).
Each
bar
represents
the
yield
difference
averaged
across
two
to
six
replicates
and
the
vertical
lines
extending
from
each
bar
are
standard
errors.
K
is
the
number
of
studies
in
which
the
mean
yield
difference
was
determined
for
each
fungicide.
Studies
came
from
a
total
of
18
articles
published
in
Plant
Disease
Management
Reports
(with
some
articles
reporting
on
multiple
studies)
and
187
trials
conducted
by
the
co-authors
of
this
article.
TABLE
1.
Effect
sizes
and
corresponding
statistics
for
the
effect
of
fungicides
on
yield
of
hybrid
field
corn
Fungicide'
k
b
Statistics`
se(D)
CI
L
CI
u
PYRA
172
255.91
(4.08)
36.91
183.05
328.76
6.93
<0.001
PROP+TRIF
72
390.39
(6.22)
83.02
224.86
555.91
4.70
<0.001
PROP+AZOX
61
331.19
(5.27)
91.39
148.38
514.01
3.62
0.001
AZOX
25
229.75
(3.67)
103.11
16.95
442.56
1.64
0.036
a
Active
ingredients:
PYRA
=
pyraclostrobin,
PROP
=
propiconazole,
TRIF
=
trifloxystrobin,
and
AZOX
=
azoxystrobin.
b
Total
number
of
studies
used
in
each
analysis
(based
on
the
number
of
studies
in
which
the
indicated
variable
was
measured
for
the
specific
fungicide
treatment
and
the
control).
The
dataset
consisted
of
25
studies
from
articles
published
in
Plant
Disease
Management
Reports
(with
some
articles
dealing
with
multiple
studies)
and
187
from
experiments
conducted
by
the
co-authors
of
this
article.
D
=
effect
size
as
mean
yield
difference
(kg/ha)
for
each
treatment
relative
to
the
nontreated,
se(
=
standard
error
of
D
,
CI
L
and
C/
u
=
lower
and
upper
limits
of
the
95%
confidence
interval
around
D,
Z
=
(standard
normal)
statistic
from
the
meta
-analysis,
and
P
=
probability
value
(significance
level)
for
the
effect
of
treatment
on
the
effect
size.
Numbers
in
parentheses
indicated
mean
yield
converted
to
bushels/acre.
1126
PHYTOPATHOLOGY
the
first
three
fungicides,
7
was
significantly
higher
in
DIS_BASE2
(foliar
disease
severity
5%)
than
in
one
or
both
of
the
other
two
categories,
DIS_BASE1
(severity
<
5%)
and
DIS_BASE3
(trials
without
reported
foliar
disease
severity).
However,
7
was
not
significantly
different
between
DIS_BASE1
and
DIS_BASE3,
and
the
difference
between
DIS_BASE2
and
DIS_BASE1
was
not
statistically
significant
for
each
fungicide
(Table
2).
Depending
on
the
fungicide,
studies
in
the
high
DIS_BASE2
category
had
a
114
to
400
kg/ha
higher
mean
yield
Grain
yield
baseline
Overall
-
YLD_BASE1
-
YLD_BASE2
-
YLD_BASE3
-
A
(K
=
172)
(K
=
57)
(K=
61)
(K
=
54)
1-0—I
PYRA
Overall
-
YLD_BASE1
-
YLD_BASE2
-
YLD_BASE3
-
0
C
(K
=
72)
PROP+TRIF
(K=
18)
(K
=
17)
(K=
37)
E
Overall
-
YLD_BASE1
-
YLD_BASE2
-
PROP+AZOX
(K=
61)
(K
=
28)
(K
=
33)
Overall
-
YLD_BASE1
-
YLD_BASE2
G
(K
=
25)
(K=
12)
C
(K
=
13)
0
AZOX
than
those
in
the
DIS_BASE1
category
(Table
2;
Fig.
3).
For
both
DIS_BASE
categories,
the
width
of
the
95%
CI
around
the
effect
size
was
wider
for
azoxystrobin
than
for
the
other
fungicides
(Fig.
3H).
With
<15
studies
for
some
moderator
-variable
categories
for
azoxystrobin,
estimated
effect
sizes
for
each
category
would
be
imprecise
for
this
fungicide
(4,34).
The
results
are
still
shown
in
the
figure,
however,
for
comparisons
with
the
other
treatments.
The
effect
of
DIS_BASE3
(no
reported
foliar
disease)
on
the
effect
sizes
was
ambiguous,
as
anticipated.
This
category
likely
Foliar
disease
baseline
Overall
-
DIS_BASE1
-
DISBASE2
DIS_BASE3
B
(K
=
172)
(K
=
65)
(K
=
36)
I
0-
(K
=
67)
I
--4i
D
Overall
-
DIS_BASE1
-
DIS_BASE2
-
DIS_BASE3
(K
=
72)
(K
=
28)
(K
=
12)
0
(K=
31)
F
Overall
-
DIS_BASE1
-
DIS_BASE2
-
DIS_BASE3
(K
=
61)
(K=
15)
0
(K
=
20)
(K
=
22)
I
0
Overall
-
DIS_BASE1
DIS_BASE2
-
DIS_BASE3
H
(K
=
26)
(K
=
7)
1
_
0
_
4
(K
=
14)
(K=
5)
-200
0
200
400
600
800
1000 1200
-400
-200
0
200
400
600
800
1000
Mean
yield
difference
(kg/ha)
Fig.
3.
Mean
yield
differences
between
fungicide
-treated
and
nontreated
plots
and
their
95%
confidence
intervals
(horizontal
bar)
for
all
studies
(Overall)
and
for
studies
categorized
according
to
baseline
yield
(YLD_BASE,
mean
grain
yield
in
the
nontreated
plot;
left
column)
and
baseline
foliar
disease
severity
(DIS_BASE,
mean
disease
severity
on
the
ear
leaf
in
the
nontreated,
between
the
dough
and
dent
growth
stages)
for
the
fungicides
A
and
B,
pyraclostrobin,
(PYRA);
C
and
D,
propiconazole
+
trifloxystrobin
(PROP+TRIF);
E
and
F,
propiconazole
+
azoxystrobin
(PROP+AZOX);
and
G
and
H,
azoxystrobin
(AZOX).
For
PYRA
and
PROP+TRIF,
YLD_BASE1
=
YLD_BASE
<
9.1
MT/ha,
YLD_BASE2
=
YLD_BASE
of
9.1-11.9
MT/ha,
and
YLD_BASE3
=
YLD_BASE
11.9
MT/ha.
For
PROP+AZOX
and
AZOX,
YLD_BASE1
=
YLD_BASE
<
11.9
MT/ha
and
YLD_BASE2
=
YLD_BASE
11.9
MT/ha.
DIS_BASE1
=
DIS_BASE
<
5%,
DIS_BASE2
=
DIS_BASE
5%
severity,
and
DIS_BASE3
=
studies
without
reported
disease
severity.
Confidence
intervals
show
precision
of
the
means
but
do
not
directly
indicate
significant
differences.
See
Table
2
for
contrasts
of
means.
K
is
the
total
number
of
studies
(Overall)
and
the
number
of
studies
in
each
category.
The
dataset
consisted
of
25
studies
from
articles
published
in
Plant
Disease
Management
Reports
(with
some
articles
dealing
with
multiple
studies)
and
187
from
experiments
conducted
by
the
co-authors
of
this
article.
Vol.
101,
No.
9,
2011
1127
includes
a
diverse
collection
of
unreported
disease
levels,
for
different
diseases,
including
those
with
no
actual
foliar
disease
being
present
(or
found).
The
mean
effect
size
for
this
category
was
typically
the
lowest
numerically
of
the
three
DIS_BASE
categories
but
the
CIs
were
also
quite
wide
for
some
treatments.
Based
on
likelihood
ratio
test,
the
estimated
between
-study
variance
(
6
2
)
was
significantly
different
from
zero
(P
<
0.05)
for
models
fitted
with
and
without
moderator
variables
for
three
of
the
four
tested
fungicides
(Table
3).
This
test
indicated
that
the
random
-effects
meta
-analysis
model
was
appropriate,
fitting
the
data
significantly
better
than
a
fixed
-effects
model
(model
with
zero
between
-study
variances
or
covariances)
(P
<
0.05)
for
pyra-
clostrobin,
propiconazole
+
trifloxystrobin,
and
propiconazole
+
azoxystrobin
but
not
azoxystrobin.
The
R
2
statistics
(20)
were
>1.5
for
the
first
three
mean
effect
sizes
in
Table
3, indicating
that
there
was
considerable
heterogeneity
among
the
studies
with
pyraclostrobin,
propiconazole
+
trifloxystrobin,
and
propicona-
zole
+
azoxystrobin
and,
as
such,
the
between
-study
variability
needed
to
be
accounted
for
in
the
analyses.
In
all
cases,
accounting
for
the
effects
of
YLD_BASE
and
DIS_BASE
on
the
effect
size
led
to
a
reduction
in
6
2
.
In
general,
DIS_BASE
explained
a
greater
proportion
of
the
between
study
variability
than
did
YLD_BASE
(Table
3).
Prediction
and
risk
analysis.
Based
on
estimated
break-even
grain
yield
for
a
range
of
fungicide
application
cost
-grain
price
combinations,
and
the
statistics
(
D
and
&
2
)
from
the
meta
-
analyses
reported
here,
the
probability
of
the
expected
yield
response
in
a
new
randomly
selected
trial
being
insufficient
to
offset
the
cost
of
fungicide
application
(p
i
)
was
estimated.
For
almost
all
of
the
grain
price
-application
cost
combinations
evaluated
(85%),
pionswas
>0.5
when
foliar
disease
severity
in
the
nontreated
plot
(DIS_BASE)
was
<5%
(Fig.
4A,
C,
E,
and
G).
On
the
other
hand,
pionswas
>0.5
for
only
33%
of
the
grain
price
-
application
cost
combinations,
when
disease
severity
was
5%
(Fig.
4B,
D,
F,
and
H).
As
required
based
on
the
probability
formula,
for
all
four
fungicides
and
both
DIS_BASE
categories,
the
pionsvalues
increased
with
increasing
fungicide
application
cost
and,
at
any
given
application
cost,
decreased
with
increasing
TABLE
2.
Differences
between
mean
effect
sizes
(
D
piff
)
and
corresponding
x2
statistics
and
probability
values
(P)
for
comparisons
between
categories
of
moderator
variables
for
evaluating
the
effects
of
foliar
fungicides
on
grain
yield
of
hybrid
corn,
based
on
random
-effects
meta
-analyses
Fungicide'
Comparisonb
D
DS
x
2
P
PYRA
YLD_BASE1
vs.
YLD_BASE2
-8.79
0.01
0.921
YLD_BASE1
vs.
YLD_BASE3
-56.52
0.37
0.545
YLD_BASE2
vs.
YLD_BASE3
-47.73
0.29
0.591
DIS_BASE1
vs.
DIS_BASE2
-113.67
1.16
0.281
DIS_BASE1
vs.
DIS_BASE3
156.37
3.48
0.063
DIS_BASE2
vs.
DIS_BASE3
270.67
8.62
0.003
DIS_BASE2
vs.
DIS_BASE1&3
192.17
4.58
0.032
PROP+TRIF
YLD_BASE1
vs.
YLD_BASE2
357.96
2.33
0.127
YLD_BASE1
vs.
YLD_BASE3
450.90
5.21
0.023
YLD_BASE2
vs.
YLD_BASE3
92.32
0.21
0.649
DIS_BASE1
vs.
DIS_BASE2
-400.04
5.10
0.024
DIS_BASE1
vs.
DIS_BASE3
58.40
0.06
0.803
DIS_BASE2
vs.
DIS_BASE3
458.44
4.25
0.040
DIS_BASE2
vs.
DIS_BASE1&3
429.55
6.87
0.009
PROP+AZOX
YLD_BASE1
vs.
YLD_BASE2
395.64
4.91
0.027
DIS_BASE1
vs.
DIS_BASE2
-387.48
2.58
0.108
DIS_BASE1
vs.
DIS_BASE3
201.59
0.76
0.382
DIS_BASE2
vs.
DIS_BASE3
588.44
8.78
0.003
DIS_BASE2
vs.
DIS_BASE1&3
487.96
6.70
0.010
a
Active
ingredients:
PYRA
=
pyraclostrobin,
PROP
=
propiconazole,
TRIF
=
trifloxystrobin,
and
AZOX
=
azoxystrobin.
b
Comparisons
between
categories
of
baseline
yield
(YLD_BASE,
mean
grain
yield
in
the
nontreated
plot)
and
baseline
foliar
disease
severity
(DIS_BASE,
mean
disease
severity
on
the
ear
leaf
in
the
nontreated,
between
the
R4
and
R6
growth
stages).
For
PYRA
and
PROP+TRIF,
YLD_BASE1
=
YLD_BASE
<
9.1
MT/ha,
YLD_BASE2
=
YLD_BASE
of
9.1
to
11.9
MT/ha,
and
YLD_BASE3
=
YLD_BASE
11.9
MT/ha.
For
PROP+AZOX,
YLD_BASE1
=
YLD_BASE
<
11.9
MT/ha
and
YLD_BASE2
=
YLD_BASE
11.9
MT/ha.
DIS_BASE1
=
DIS_BASE
<
5%,
DIS_BASE2
=
DIS_BASE
5%
severity,
and
DIS_BASE3
=
studies
without
reported
disease
severity.
c
D
=
effect
size
as
mean
yield
difference
(kg/ha)
for
each
treatment
relative
to
the
nontreated
plot.
TABLE
3.
Estimated
between
-study
variance
and
corresponding
statistics
from
random
-effects
meta
-analysis
of
the
effect
of
fungicides
on
yield
of
hybrid
field
corn
Statisticb
Overall
With
baseline
yield
With
baseline
disease
Fungicide'
C5^
2
LRS
P
R
2
2
C5
P
Percent`
&
2
P
Percent`
PYRA
75.21
47.10
<0.001
6.00
73.96
<0.001
1.66
56.86
<0.001
24.40
PROP+TRIF
211.48
18.00
<0.001
2.89
192.34
<0.001
9.05
167.35
0.002
20.87
PROP+AZOX
273.17
134.80
<0.001
55.07
242.92
<0.001
11.07
206.43
<0.001
24.43
AZOX
80.76
0.90
0.171
1.36
32.98
0.376
59.16
40.56
0.327
49.78
a
Active
ingredients:
PYRA
=
pyraclostrobin,
PROP
=
propiconazole,
TRIF
=
trifloxystrobin,
and
AZOX
=
azoxystrobin.
b
&2
=
estimated
between
-study
variance,
P
=
probability
value
(significance
level)
for
testing
the
equality
of
&
2
to
zero
for
models
fitted
without
moderator
variable
(Overall)
and
with
baseline
yield
or
foliar
disease
severity
as
a
categorical
moderator
variable,
LRS
=
likelihood
ratio
statistic,
and
R
2
of
Higgins
and
Thompson
(20)
(see
text
for
explanation).
For
ease
of
presentation,
&
2
was
divided
by
1,000.
c
Percentage
of
the
between
-study
variability
explained
by
baseline
yield
(mean
yield
in
the
nontreated
plot)
and
baseline
foliar
disease
severity
(mean
disease
severity
on
the
ear
leaf
in
the
nontreated
at
R4
-R6).
1128
PHYTOPATHOLOGY
grain
prices
(Fig
4).
For
instance,
when
severity
was
<5%
and
grain
price
was
$0.16/kg,
p
i
„,
increased
from
0.41
to
0.90
for
pyraclostrobin;
0.54
to
0.84
for
propiconazole
+
trifloxystrobin;
0.50
to
0.78
for
propiconazole
+
azoxystrobin;
and
0.83
to
0.99
for
azoxystrobin
as
application
cost
increased
from
$40.00
to
96.00/ha.
The
corresponding
increases
when
disease
severity
was
5%
were
from
0.24 to
0.78,
0.19
to
0.50,
0.20
to
0.47,
and
0.23
to
0.85
for
pyraclostrobin,
propiconazole
+
trifloxystrobin,
propiconazole
+
azoxystrobin,
and
azoxystrobin,
respectively.
At
an
application
cost
of
$65/ha,
p
i
„,
increased
from
0.39
to
0.99
when
disease
severity
was
<5%
and
from
0.18
to
0.98
when
severity
was
5%
as
grain
price
decreased
from
$0.28
to
0.08/kg.
For
205
of
the
384
(
-
=.53%)
grain
price
-application
cost
-fungi-
cide
scenarios
that
were
considered,
there
was
a
>70%
chance
of
Probability
of
not
offsetting
fungicide
cost
Foliar
disease
severity
<
5%
1.00
-
0.75
0.50
025
-0-
0-
0-
--
C
r-c
'
,0"
r
--
1.-
--
v
es
-
_
_
_
-0
-
•-•"'
Y
' '
--
...4
..,......N
.----
-.4.-
-
"
6
`
.
Cr
2----.--.1
tr.
_Cr-
••0-
a
- -
ii
,...•
-
-
4
,-.Y
.-
7
-
ir
•F•
p
,0
-•d
-
-
E
.
-M.
Az
r
---
...-te
,...••••
-4:16:9
MO&
kg
MANI*
fr
-
1..a•rt
-o-c
-CI
-0-
5.0.12.rkg
($3.00.rbu}
a
-
---
____
____.-
9
-f4.1•L•
9
4$4.Pax••)--
A
PYRA
0.00
1.00
0.20(kg
(53.Carbu).
-
•N-
-
Sa.24.1.3
($6.Carbu)
-
$0.2arkg
•($7.61.11bu).
0.75
0.50
0.25
-0-
0
- °
..--•••
-
•"
- t
..
_a--
A-
-
a
t
irp
t
4.
0 .
-A
:LL.Z.
--CP
-
CI
PROP+TRIF
0-00
1.00
-•
-0-C---<)
0.75
-
....
-or
o-
-
"
°
...
al
43-
-
O.
50
E
PROP+AZOX
0.00
-
. . . . . . . . . .
1.00
0.75
a-AI:fin
-
E
-
---
-
G
AZOX
0.00
-
not
seeing
a
return
on
investment
when
disease
severity
was
<5%
(p
i
„,
0
.70).
For
the
low
DIS_BASE
category,
there
was
>75%
chance
of
not
obtaining
a
yield
increase
high
enough
to
offset
the
cost
of
applying
pyraclostrobin
or
azoxystrobin
(p
i
„,
0.75)
if
grain
price
was
$0.08/kg
and
fungicide
application
costs
were
>$40/ha
(Fig.
4).
For
the
high
DIS_BASE
category
and
the
two
fungicides
with
the
highest
D
values,
propiconazole
+
trifloxy-
strobin
and
propiconazole
+
azoxystrobin,
the
chance
of
losing
money
on
fungicide
investment
was
<25%
(p
i
„,
<
0.25)
when
grain
prices
were
>$0.16/kg
and
application
costs
were
<$65/ha.
The
chance
of
at
least
recovering
the
cost
of
fungicide
appli-
cation
through
increased
grain
yield
was
greatest
at
the
highest
grain
price
and
lowest
fungicide
application
cost
considered
when
disease
severity
was
being
88%
for
pyraclostrobin
Foliar
disease
severity
>5%
Ce
Z
4
4
:r
Tr"
V
--
V
---
T"'
V
----
A
---6
.'-'
.„,r
..„.
'
-
'
4'
y....y.--
r
te
-
.47.76'-'4"
-
7.7.
-
..6---4"
.4/
4,
..e
ip---
-
'
,.--ar-
-
.rr•
0
.
-0
,
4
2
,
--r--E7.2,..-.4.-
..
„-w.R'_g_c...-Q--0
__
--
C.-
-
6
.
&S
-0--0-E1
B
:Asfr4!1•
117•Z
ir'+'&:71-":Pi
c'"
-cr-c)
cf
-D.
Cr
°
-
••10r
.
-
- - -
s
:t
o.
40
44
48
52
56
60
64
68
72
76
80
84
88
92
96
40
44
48
52
56
60
64
68
72
76
80
84
88
92
96
t
.'>\1005).
0
,5‘
40P
N
9\5'
6
%.0-10
(;)
,
),-0"
V
1.515
,
Za0.
1
9!°
oi%1
6
\r,
0
5
0‘.
5
.1
1
\•,%!
3\
1
/
4
(i•
‘'?
Fungicide
application
cost
[$/ha(S/Acren
Fig.
4.
Probability
of
not
offsetting
fungicide
application
cost
(product
plus
application)
for
a
range
of
application
costs
and
grain
market
prices
for
the
fungicides
A
and
B,
pyraclostrobin,
(PYRA);
C
and
D,
propiconazole
+
trifloxystrobin
(PROP+TRIF);
E
and
F,
propiconazole
+
azoxystrobin
(PROP+AZOX);
and
G
and
H,
azoxystrobin
(AZOX),
based
on
estimates
of
the
mean
yield
difference
(
D
)
and
the
between
-study
variance
(
6
2
)
from
meta
-analyses
of
fungicide
effects
on
grain
yield,
for
studies
with
baseline
foliar
disease
severity
on
the
ear
leaf
in
the
nontreated
plot
A,
C,
E,
and
G,
<5%
and
B,
D,
F,
and
H,
Vol.
101,
No.
9,
2011
1129
(i.e.,
1
—pions
0.88),
87%
for
propiconazole
+
trifloxystrobin,
86%
for
propiconazole
+
azoxystrobin,
and
90%
for
azoxystrobin.
Similar
results
for
p
i
„,
were
found
when
YLD_BASE
was
used
as
the
moderator
variable
(data
not
shown).
In
brief,
p
i
„,,
was
the
greater
for
the
highest
YLD_BASE
categories
than
for
the
lowest
YLD_BASE categories.
DISCUSSION
Since
2006,
there
has
been
a
substantial
increase
in
the
appli-
cation
of
foliar
fungicides,
especially
those
belonging
to
the
QoI
group
of
compounds,
on
hybrid
field
corn
in
the
U.S.
Corn Belt.
This
increase
can
at
least
be
partially
attributed
to
claims
of
"plant
health",
"plant
performance",
yield,
and
physiological
benefits
associated
with
fungicide
applications
(37).
Many
plant
pathol-
ogists
have
questioned
the
magnitude
of
yield
benefits
associated
with
these
physiological
effects
(37).
To
address
these
questions,
several
studies
on
corn
have
been
conducted
by
both
university
-
and
industry
-based
researchers.
Although
the
summaries
of
yield
responses
to
fungicides
presented
by
the
different
research
groups
show
similar
trends
(with
graphs
similar
to
those
shown
in
Figure
2),
there
have
been
different
interpretations
of
the
findings
and,
consequently,
conclusions
regarding
the
value
and
economics
of
using
foliar
fungicides
in
hybrid
corn
when
foliar
diseases
are
absent
or
at
nominal
levels
(37).
Most
of
the
previous
conclusions
regarding
the
effects
of
fungi-
cides
on
hybrid
corn
yield were
based
on
simple
arithmetic
means
or
a
tally
of
studies
with
significant
results
(vote
counting)
(27,37).
Using
data
from
fungicide
trials
conducted
across
the
U.S.
Corn
Belt
during
2002
to
2009,
results
presented
here
constitute
the
first
comprehensive
quantitative
synthesis
of
hybrid
corn
yield
response
across
a
wide
range
of
environments
to
four
of
the
most
widely
used
foliar
fungicides.
Through
random
-
effects
meta
-analyses,
the
overall
mean
yield
difference
(the
effect
size,
D
)
between
treated
and
nontreated
plots
was
determined
for
pyraclostrobin,
propiconazole
+
trifloxystrobin,
propiconazole
+
azoxystrobin,
and
azoxystrobin,
all
applied
at
label
-recommended
rates
at
either
tasseling
or
silking.
For
all
four
products,
there
was
a
significant
increase
in
yield
relative
to
the
nontreated
plots.
For
three
of
the
fungicides
for
which
the
yield
increase
was
signifi-
cant
(pyraclostrobin,
propiconazole
+
trifloxystrobin,
and
propi-
conazole
+
azoxystrobin),
the
magnitude
of
the
response
was
affected
by
baseline
foliar
disease
severity.
D
was
generally
greater
for
the
higher
disease
severity
category
than
the
lower
severity
category.
In
addition,
YLD_BASE
affected
the
response
to
propiconazole
+
trifloxystrobin
and
propiconazole
+
azoxy-
strobin,
with
D
being
higher
in
the
lowest
YLD_BASE
category
than
in
the
highest
category.
The
observed
influence
of
YLD_BASE
on
the
effect
sizes
and,
consequently,
the
probability
of
profitable
fungicide
use
is
some-
what
contrary
to
that
reported
in
a
similar
investigation
conducted
by
Munkvold
et
al.
(38)
based
on
a
much
smaller
number
of
studies.
These
authors
reported
that
the
probability
of
profitable
fungicide
use
in
field
corn
was
lower
in
fields
with
low
yield
potential
and
suggested
that
such
fields
should
not
be
considered
for
fungicide
application.
There
are
several
possible
explanations
for
significantly
higher
effect
sizes
in
the
low
YLD_BASE
cate-
gories
than
in
the
higher
categories.
These
include
not
only
the
inherent
genetic
yield
potential
of
the
hybrid
but
also
environ-
mental
factors,
yield
-impacting
stresses
such
as
pest
and
diseases,
hybrid
resistance,
and
complex
interactions
involving
these
fac-
tors.
It
is
quite
possible
for
the
profitability
of
fungicide
use
to
be
low
in
fields
with
high
yield
potential,
especially
if
the
hybrid
is
resistant
or
tolerant
to
foliar
diseases.
Munkvold
et
al.
(38)
also
suggested
that
fungicide
use
was
less
likely
to
be
profitable
when
resistant
hybrids
were
planted,
because
resistant
hybrids
were
observed
to
be
less
responsive
to
fungicide
treatments
than
susceptible
hybrids.
In
general,
YLD_BASE
would
be
low
when
foliar
disease
severity
is
high
(66,68)
and
fungicide
effects
on
foliar
diseases
could
translate
into
a
yield
response
(58-62,66).
The
foliar
disease
effect
was
supported
by
results
from
our
analysis
of
the
effect
of
DIS_BASE
on
the
mean
yield
response,
which
showed
that
D
was
higher
in
trials
with
disease
severity
5%
than
in
trials
with
severity
<5%.
As
expected,
results
are
less
clear
when
foliar
disease
severity
was
not
reported,
although
the
trend
for
this
category
was
for
a
smaller
mean
effect
size
compared
with
the
situation
with
severity
5%.
Our
findings
are
comparable
with
those
from
other
studies,
which
showed
that
mean
yield
response
to
fungicides
was
higher
for
trials
in
which
hybrids
with
fair
to
poor
resistance
to
GLS
were
planted
than
in
those
planted
with
GLS-resistant
hybrids
(37,38),
and
higher
for
trials
in
which
corn
was
planted
after
corn
than
those
in
which
corn
was
rotated
with
soybean
(37).
This
is
also
consistent
with
the
findings
from
studies
on
hybrid
corn
grain
yield
response to
foliar
diseases
and
defoliation
(1,26,66),
the
effects
of
genetic
resistance
on
disease
and
yield
responses
(12,27,38),
and
the
value
of
using
fungicides
to
minimize
yield
loss
in
corn
when
foliar
disease
levels
are
above
critical
thresholds
(1,38,54).
In
addition
to
estimating
the
expected
effect
size
and
evaluating
the
influence
of
the
moderator
variables
on
the
effect
sizes,
meta
-
analysis
was
used
here
to
make
projections
regarding
the
prob-
ability
of
future
outcomes
and
to
assess
the
economic
value
of
using
a
fungicide
in
hybrid
corn
when
disease
severity
is
<5%
and
5%
(or
yield
was
high
or
low).
This
is
of
particular
importance
because,
from
a
farmer's
perspective,
it
is
not
enough
to
know
that
a
fungicide
(or
any
other
production
practice,
for
that
matter)
may
lead
to
a
positive
mean
yield
response
in
the
long
run.
Knowing
whether
the
average
increase
is
large
enough
to
offset
the
cost
of
applying
the
fungicide
(and
under
what
conditions
this
occurs)
and
the
variability
of
the
response
are
also
important.
At
average
grain
prices
and
fungicide
application
costs
over
the
study
period,
the
mean
yield
increases
of
64
to
306
kg/ha
in
trials
where
disease
severity
was
<5%,
depending
on
the
fungicide,
generally
were
insufficient
to
offset
the
cost
of
fungicide
appli-
cation.
The
probability
of
not
recovering
the
cost
of
applying
a
fungicide
(p
i
„,,)
at
low
disease
severity
was
>0.70
for
a
wide
range
of
application
costs
and
grain
prices,
and
>0.5
for
almost
every
scenario
considered.
For
all
tested
fungicides,
p
i
„,
decreased
with
increasing
grain
price
and
decreasing application
cost
but,
even
at
the
highest
grain
price
($0.28/kg
[$7/bushel])
and
lowest
fungi-
cide
application
cost
($40/ha
[$16.20/acre]),
there
was
a
25,
44,
41,
and
65%
chance
of
the
yield
increase
being
insufficient
to
cover
the
cost
of
applying
pyraclostrobin,
propiconazole
+
trifloxystrobin,
propiconazole
+
azoxystrobin,
and
azoxystrobin,
respectively,
when
disease
severity
was
low.
Based
on
data
from
the
United
States
Department
of
Agricul-
ture,
National Agricultural
Statistics
Service,
grain
corn
prices
have
not
reached
$0.28/kg
($7.00/bushel)
in
the
last
decade
(1999
to
2009)
but
have
been
$0.06
to
0.23/kg
($1.53
to
5.80/bushel),
with
an
average
of
$0.12/kg
($2.97/bushel).
At
this
10
-year
average
grain
price
and
for
application
costs
of
$40
to
95/ha,
the
chance
of
failing
to
recoup
expenses
when
disease
severity
was
<5%
was
55
to
98%
for
pyraclostrobin,
62
to
93%
for
propicona-
zole
+
trifloxystrobin,
57
to
88%
for
propiconazole
+
azoxystro-
bin,
and
91
to
99%
for
azoxystrobin.
However,
p
i
„,
was
signifi-
cantly
lower
when
foliar
disease
severity
was
5%
on
the
ear
leaf
and,
consequently,
the
probability
of
at
least
recovering
the
cost
of
fungicide
application
was
higher
in
trials
with
>5%
severity,
especially
at
low
application
costs
and
high
grain
prices.
When
baseline
severity
was
>5%,
p
i
,„
values
at
the
10
-year
average
grain
price
were
<0.50
(0.28
to
0.48)
for
propiconazole
+
trifloxy-
strobin
and
propiconazole
+
azoxystrobin,
at
fungicide
applica-
tion
costs
of
<$73/ha.
These
results,
based
on
meta
-analysis,
are
comparable
with
those
reported
by
Munkvold
et
al
(38),
based
on
Bayesian
inference
methods
(5).
They
estimated
profit
probabili-
1130
PHYTOPATHOLOGY
ties
of
0.02
to
0.98
(corresponding
to
m
os
,
of
0.98
to
0.02)
for
a
single
application
of
propiconazole,
based
on
grain
prices
of
$0.79
to
0.118/kg
and
an
expected
net
return
of
$25/ha.
In
that
study,
the
highest
profit
probabilities
(and,
consequently,
the
lowest
m
oss
values)
occurred
when
susceptible
hybrid
were
planted
and
GLS
severity
on
the
ear
leaf
was
>5%
(22
to
73%).
Accounting
for
YLD_BASE
and
DIS_BASE
reduced
6
2
but
a
substantial
portion
of
the
between
-study
variability
remained
un-
explained.
These
large
6
2
values
suggest
that
other
factors
(un-
recorded
or
not
reported)
contributed
to
the
observed
differences
in
yield
response to
fungicides
among
studies.
This
is
not
surpris-
ing,
given
that
hybrid
corn
yield
is
a
function
of
several
crop-,
environment-,
pest-,
and
management
-related
factors
(13,16,50,
51,57,65,66)
and
complex
interactions
involving
these
factors.
Although
YLD_BASE
can
be
used
as
a
surrogate
risk
factor
for
some
of
these
effects,
as
was
done
here,
the
fact
that
this
moderator
variable
only
explained
2
to
11%
of
the
between
-study
variability
(for
the
fungicide
for
which
its
effect
was
statistically
significant)
suggests
that
a
more
direct
investigation
of
the
influ-
ence
of
factors
such
as
soil
type
and
fertility,
temperature
and
moisture,
weeds,
pests,
diseases,
hybrid
yield
potential,
pest
and
disease
resistance,
and
cropping
practices
on
yield
response
to
fungicides
will
be
needed
to
better
determine
the
conditions
under
which
the
use
of
foliar
fungicides
may
be
more
consistent
and
profitable.
In
addition,
it
is
unclear
what
the
5%
disease
severity
cutoff
means
in
terms
of
yield
loss
and
fungicide
decision
thresholds.
However,
this
information
could
be
used
as
the
basis
for
future
studies
to
develop
such
thresholds
and
to
refine
corn
foliar
disease
risk
assessment
models
(45)
to
predict
late
-season
disease
intensity,
based
on
information
collected
prior
to
making
fungicide
use
decisions
(before
VT/R1).
At
present,
based
on
the
results
from
this
investigation,
one
cannot
recommend
these
fungicides
for
general
use
when
foliar
disease
risk
is
low.
Foliar
fungicides
on
field
corn
may
be
war-
ranted
and
cost
effective
when
disease
severity
would
be
>5%
in
the
absence
of
fungicide,
grain
prices
are
very
high, application
costs
are
low,
and
there
is
reasonable
knowledge
that
yield
would
be
low
without
treatment.
However,
there
is
great
uncertainty,
even
when
disease
severity
is
>5%,
that
a
grower
would
realize
a
profit
in
any
given
year
and
location
when
a
fungicide
is
applied
between
VT
and
Rl.
ACKNOWLEDGMENTS
Salaries
and
additional
research
support
for
P.
A.
Paul
and
L.
V.
Madden
were
provided
by
state
and
federal
funds
to
the
Ohio
Agricultural
Research
and
Development
Center,
The
Ohio
State
University.
Research
in
Mississippi
was
supported
by
the
Mississippi
Corn
Promotion
Board
and
special
thanks
are
extended
to
N.
Buehring,
A.
Henn,
D.
Ingram,
and
E.
Larson
for
assisting
with
research
in
Mississippi.
Salaries
and
additional
research
support
for
A.
Grybauskas
were
provided
by
the
Maryland
Agricultural
Experiment
Station
and
the
Maryland
Grain
Producers
Utilization
Board.
Technical
assistance
in
Maryland
was
provided
by
E.
Reed,
K.
Conover,
M.
Sultenfuss,
J.
Street,
and
T.
Ellis.
Combine
transport
and
maintenance
in
Maryland
was
provided
by
R.
Kratochvil,
P.
Forrestal,
and
P.
Watkins.
Partial
support
for
the
work
done
in
Minnesota
was
provided
by
the
University
of
Minnesota
Agricultural
Experiment
Station.
We
thank
E.
Adee,
K.
Ames,
and
S.
Ebelhar
for
assisting
with
the
University
of
Illinois
fi
eld
trials;
M.
Wallhead,
W.
Bardall,
J.
Davlin,
and
M.
Davis
for
assisting
with
Ohio
trials;
J.
Shriver
for
assisting
with
trials
in
Iowa;
and
J.
Beaty,
G.
Buechley,
and
T.
McCarthy
for
assisting
with
trials
in
Indiana.
This
research
was
partially
funded
by
BASF
Corporation
Agricultural
Products,
Research
Triangle
Park,
NC;
Bayer
CropScience,
Research
Triangle
Park,
NC;
and
Syngenta
Crop
Protection
Inc.,
Greensboro,
NC.
LITERATURE
CITED
1.
Adee,
E.
A.,
Paul,
L.
E.,
Nafziger,
E.
D.,
and
Bollero,
G.
2005.
Yield
loss
of
corn
hybrids
to
incremental
defoliation.
Online.
Crop
Management.
doi:
10.1094/CM-2005-0427-01-RS.
2.
Arends,
L.
R.,
Hoes,
A.
W.,
Lubsen,
J.,
Grobbee,
D.
E.,
and
Stijnen,
T.
2000.
Baseline
risk
as
predictor
of
treatment
benefit:
three
clinical
meta
-
re
-analyses.
Stat.
Med.
19:3497-3518.
3.
Beckman,
P.
M.,
and
Payne,
G.
A.
1982.
External
growth,
penetration,
and
development
of
Cercospora
zeae-maydis
in
corn
leaves.
Phyto-
pathology
72:810-815.
4.
Borenstein,
M.,
Hedges,
L.
V.,
Higgins,
J.
P.
T.,
and
Rothstein,
H.
R.
2009.
Introduction
to
Meta
-Analysis.
John
Wiley
&
Sons,
Chichester,
UK.
5.
Box,
G.
E.
P.,
and
Tiao,
G.
C.
1973.
Bayesian
Inference
in
Statistical
Analysis.
Addison-Wesley,
Reading,
MA.
6.
Bradley,
C.
A.,
and
Ames,
K.
A.
2009.
Evaluation
of
foliar
fungicides
on
corn
in
Auburn,
Illinois,
2008.
Plant
Disease
Management
Reports
3:FC001.
Online
publication.
doi:
10.1094/PDMR03.
7.
Bradley,
C.
A.,
and
Ames,
K.
A.
2009.
Evaluation
of
foliar
fungicides
on
corn
in
Perry,
Illinois,
2008.
Plant
Disease
Management
Reports
3:FC002.
Online
publication.
doi:
10.1094/PDMR03.
8.
Bradley,
C.
A.,
and
Ames,
K.
A.
2009.
Evaluation
of
foliar
fungicides
on
corn
in
Urbana,
Illinois,
2008.
Plant
Disease
Management
Reports
3:FC003.
Online
publication.
doi:
10.1094/PDMR03.
9.
Brand,
R.,
and
Kragt,
H.
1992.
Importance
of
trends
in
the
interpretation
of
an
overall
odds
ratio
in
the
meta
-analysis
of
clinical
trials.
Stat.
Med.
11:2077-2082.
10.
de
Nazareno,
N.
R.
X.,
Lipps,
P.
E.,
and
Madden,
L.
V.
1993.
Effect
of
levels
of
corn
residue
on
the
epidemiology
of
gray
leaf
spot
of
corn
in
Ohio.
Plant
Dis.
77:67-70.
11.
Field,
A.
P.
2001.
Meta
-analysis
of
correlation
coefficients:
A
Monte
Carlo
comparison
of
fixed-
and
random
-effects
methods.
Psychol.
Methods
6:161-180.
12.
Gorman,
D.,
Anderson,
B.
M.,
and
Abad,
L.
1997.
Yield
loss
of
corn
hybrids
with
differing
levels
of
gray
leaf
spot
resistance.
Pages
171-182
in:
Proc.
Annu.
Corn
Sorghum
Res.
Conf.,
52nd.
American
Seed
Trade
Association,
Washington,
DC.
13.
Gower,
S.
A.,
M.
M.
Loux,
J.
Cardina,
and
Harrison,
S.
K.
2002.
Effect
of
planting
date,
residual
herbicide,
and
postemergence
application
timing
on
weed
control
and
grain
yield
in
glyphosate-tolerant
corn
(Zea
mays).
Weed
Technol.
16:488-494.
14.
Grossman,
K.,
Kwiatkowski,
J.,
and
Caspar,
G.,
1999.
Regulation
of
phytohormone
levels,
leaf
senescence
and
transpiration
of
the
strobilurin
kresoxim-methyl
in
wheat
(Triticum
aestivum).
J.
Plant
Physiol.
154:805-
808.
15.
Grossman,
K.,
and
Retzlaff,
G.
1997.
Bioregulatory
effects
of
the
fungicide
strobilurin
kresoxim-methyl
in
wheat
(Triticum
aestivum).
Pestic.
Sci.
50:11-20.
16.
Grover,
K.
K.,
Roth,
G.
W.,
and
Karsten,
H. D.
2009.
Corn
grain
yields
and
yield
stability
in
four
long-term
cropping
systems.
Agron.
J.
101:940-
946.
17.
Hedges,
L.
V.,
Gurevitch,
J.,
and
Curtis,
P.
S.
1999.
The
meta
-analysis
of
response
ratios
in
experimental
ecology.
Ecology
80:1150-1156.
18.
Hedges,
L.
V.,
and
Olkin,
I.
1980.
Vote
-counting
methods
in
research
synthesis.
Psychol.
Bull.
88:359-369.
19.
Hedges,
L.
V.,
and
Vevea,
J.
L.
1998.
Fixed-
and
random
-effects
models
in
meta
-analysis.
Psychol.
Methods
3:486-504.
20.
Higgins,
J.
P.
T.,
and
Thompson,
S.
G.
2002.
Quantifying
heterogeneity
in
a
meta
-analysis.
Stat.
Med.
21:1539-1588.
21.
Hunter,
J.
E.,
and
Schmidt,
F.
L.,
2004.
Methods
of
Meta
-Analysis:
Correcting
Error
and
Bias
in
Research
Findings,
2nd
ed.
Sage
Publications
Inc.,
Thousand
Oaks,
CA.
22.
Jackson,
T.
A.
2006.
Evaluation
of
foliar
fungicides
on
gray
leaf
spot
of
corn
in
Nebraska,
2005.
Fungic.
Nematicide
Tests
61:FC041.
Online
publication.
doi:
10.1094/FN61.
23.
Jackson,
T.
A.,
Behn,
J.
L.,
and
Miller,
D.
W.
2008.
Evaluation
of
foliar
fungicides
on
leaf
diseases
of
corn
in
Nebraska,
2007.
Plant
Disease
Management
Reports
2:FC058.
Online
publication.
doi:
10.1094/
PDMR02.
24.
Jackson,
T.
A.,
Schleicher,
C.
M.,
and
Behn,
J.
L.
2010.
Evaluation
of
foliar
fungicides
on
gray
leaf
spot
of
field
corn
in
Nebraska,
2009.
Plant
Disease
Management
Reports
4:FC089.
Online
publication.
doi:
10.1094/PDMR04.
25.
Jackson,
T.
A.,
Schleicher,
C.
M.,
and
Behn,
J.
L.
2010.
Foliar
fungicides
efficacy
on
gray
leaf
spot
of
field
corn
in
Nebraska,
2009.
Plant
Disease
Management
Reports
4:FC090.
Online
publication.
doi:
10.1094/
PDMR04.
26.
Jenco,
J.
H.
1995.
Epidemiology
of
Cercospora
zeae-maydis
on
Zea
mays
in
Iowa.
M.S.
thesis,
Iowa
State
University,
Ames.
27.
Jeschke,
M.,
and
Doerge,
T.
2007. Management
of
foliar
diseases
in
corn
with
fungicides.
In:
Crop
Insights,
17,
no.
14.
Pioneer
Hi
-Bred,
Johnston,
IA.
28.
Lee,
C.,
P.,
and
Vincelli,
P.
2008.
Evaluation
of
fungicides
for
control
of
gray
leaf
spot
in
corn,
2007.
Plant
Disease
Management
Reports
2:FC096.
Vol.
101,
No.
9,
2011
1131
Online
publication.
doi:
10.1094/PDMR02.
29.
Lee,
C.,
P.,
and
Vincelli,
P.
2010.
Evaluation
of
fungicide
application
on
yield
in
field
corn,
2008.
Plant
Disease
Management
Reports
4:FC019.
Online
publication.
doi:
10.1094/PDMR04.
30.
Lee,
C.,
P.,
and
Vincelli,
P.
2010.
Evaluation
of
fungicide
application
on
disease
intensity
and
yield
in
field
corn,
2009.
Plant
Disease
Management
Reports
4:FCO20.
Online
publication.
doi:
10.1094/PDMR04.
31.
Lipps,
P.
E.
1987.
Gray
leaf
spot
epiphytotic
in
Ohio
corn.
Plant
Dis.
71:281.
32.
Lipps,
P.
E.
1996.
Reaction
of
corn
hybrid
to
gray
leaf
spot.
Pages
163-
180
in:
Proc.
51st
Annu.
Corn
Sorghum
Res.
Conf.
Chicago.
33.
Lipps,
P.
E.
1998.
Gray
leaf
spot:
a
global
threat
to
corn
production.
APSnet
Feature
(May).
Published
online
by
The
American
Phytopathological
Society,
St.
Paul,
MN.
34.
Lipsey,
M.
W.,
and
Wilson,
D.
B.
2001.
Practical
Meta
-Analysis.
Sage
Publications
Inc.,
Thousand
Oaks,
CA.
35.
Littell,
R.
C.,
Milliken,
G.
A.,
Stroup,
W.
W.,
Wolfinger,
R.
D.,
and
Schabenberger,
0.
2006.
SAS
for
Mixed
Models,
2nd
ed.
SAS
Institute,
Cary,
NC.
36.
Madden,
L.
V.,
and
Paul,
P.
A.
2011.
Meta
-analysis
for
evidence
synthesis
in
plant
pathology:
an
overview.
Phytopathology
101:16-30.
37.
Munkvold,
G.
P.,
Doerge,
T.,
and
Bradley,
C.
2008.
IPM
is
still
alive
for
corn
leaf
diseases:
look
before
you
spray.
In:
Proc.
62nd
Annu.
Corn
Sorghum
Res.
Conf.
Chicago.
CD-ROM,
American
Seed
Trade
Associa-
tion,
Alexandria,
VA..
38.
Munkvold,
G.
P.,
Martinson,
C.
A.,
Shriver,
J.
M.,
and
Dixon,
P.
M.
2001.
Probabilities
for
profitable
fungicide
use
against
gray
leaf
spot
in
hybrid
maize.
Phytopathology
91:477-484.
39.
Nason,
M.
A.,
Farrar,
J.,
and
Bartlett,
D.
2007.
Strobilurin
fungicides
induce
changes
to
photosynthetic
gas
exchange
that
do
not
improve
water
use
efficiency
of
plants
grown
under
conditions
of
water
stress.
Pest
Manage.
Sci.
63:1191-1200.
40.
Normand,
S.
L.
T.
1999.
Tutorial
in
biostatistics.
Meta
-analysis:
formulating,
evaluating,
combining,
and
reporting.
Stat.
Med.
18:321-359.
41.
Paul,
P.
A,
Johnson,
A.
L.,
and
Mills,
D.
R.
P.
2008.
Evaluation
of
fungicides
for
gray
leaf
spot
control
in
field
corn
in
Ohio,
2007.
Plant
Disease
Management
Reports
2:FC045.
Online
publication.
doi:
10.1094/PDMR02.
42.
Paul,
P.
A.,
Lipps,
P.
E.,
Hershman,
D.
E.,
McMullen,
M.
P.,
Draper,
M.
A.,
and
Madden,
L.
V.
2007.
A
quantitative
review
of
tebuconazole
effect
on
Fusarium
head
blight
and
deoxynivalenol
content
in
wheat.
Phytopathology
97:211-220.
43.
Paul,
P.
A.,
Lipps,
P.
E.,
Hershman,
D.
E.,
McMullen,
M.
P.,
Draper,
M.
A.,
and
Madden,
L.
V.
2008.
Efficacy
of
triazole-based
fungicides
for
Fusarium
head
blight
and
deoxynivalenol
control
in
wheat:
a
multivariate
meta
-analysis.
Phytopathology
98:999-1011.
44.
Paul,
P.
A.,
Lipps,
P.
E.,
and
Madden,
L.
V.
2005.
Relationship
between
visual
estimates
of
Fusarium
head
blight
intensity
and
deoxynivalenol
accumulation
in
harvested
wheat
grain:
A
meta
-analysis.
Phytopathology
95:1225-1236.
45.
Paul,
P.
A.,
and
Munkvold,
G.
P.
2004.
A
model
-based
approach
to
preplanting
risk
assessment
for
gray
leaf
spot
of
maize.
Phytopathology
94:1350-1357.
46.
Paul,
P.
A.,
and
Munkvold,
G.
P.
2005.
Influence
of
temperature
and
relative
humidity
on
sporulation
of
Cercospora
zeae-maydis
and
expan-
sion
of
gray
leaf
spot
lesions
on
maize
leaves.
Plant
Dis.
89:624-630.
47.
Paul,
P.
A,
Wallhead,
M.
W.,
and
Mills,
D.
R.
P.
2009.
Effects
of
fungicides
on
gray
leaf
spot
and
yield
of
field
corn
in
Ohio,
2008.
Plant
Disease
Management
Reports
3:FC093.
Online
publication.
doi:
10.1094/PDMR03.
48.
Paul,
P.
A,
Wallhead,
M.
W.,
and
Mills,
D.
R.
P.
2009.
Gray
leaf
spot
and
yield
responses
to
foliar
fungicides
in
field
corn
in
Ohio,
2008.
Plant
Disease
Management
Reports
3:FC092.
Online
publication.
doi:
10.1094/PDMR03.
49.
Payne,
G.
A.,
Duncan,
H.
E.,
and
Adkins,
C.
R.
1987.
Influence
of
tillage
on
development
of
gray
leaf
spot
and
number
of
airborne
conidia
of
Cercospora
zeae
maydis.
Plant
Dis.
71:329-332.
50.
Raymond,
E
D.,
Alley,
M.
M.,
Parrish,
D.
J.,
and
Thomason,
W.
E.
2009.
Plant
density
and
hybrid
impacts
on
corn
grain
and
forage
yield
and
nutrient
uptake.
J.
Plant
Nutr.
32:395-409.
51.
Riedell,
W.
E.,
Pikul,
Jr.,
J.
L.,
Jaradat,
A,
A.,
and
Schumacher,
T.
E.
2009.
Crop
rotation
and
nitrogen
input
effects
on
soil
fertility,
maize
mineral
nutrition,
yield,
and
seed
composition.
Agron.
J.
101:870-879.
52.
Ritchie,
S.
W.,
Hanway,
J.
J.,
and
Benson,
G.
0.
1992.
How
a
corn
plant
develops.
Spec.
Rep.
No.
48.
Iowa
State
Univ.
Coop.
Ext.
Serv.
Ames.
53.
Rupe,
J.
C.,
Siegel,
M.
R.,
and
Hartman,
J.
R.
1982.
Influence
of
environment
and
plant
maturity
on
gray
leaf
spot
of
corn
caused
by
Cercospora
zeae-maydis.
Phytopathology
72:1587-1591.
54.
Shah,
D.
A.,
and
Dillard,
H.
R.
2010.
Managing
foliar
diseases
of
processing
sweet
corn
in
New
York
with
strobilurin
fungicides.
Plant
Dis.
94:
213-220.
55.
Sheu,
C.
F.,
and
Suzuki,
S.
2001.
Meta
-analysis
using
linear
mixed
models.
Behay.
Res.
Methods
Ins.
C.
33:102-107.
56.
Shriver,
J.
M.,
and
Robertson
A.
E.
C.
2009.
Comparison
of
fungicides
for
disease
control
and
yield
response
of
corn
in
Nashua,
Iowa
2008.
Plant
Disease
Management
Reports
3:FC014.
Online
publication.
doi:
10.1094/PDMR03.
57.
Stanger,
T.
F.,
and
Lauer,
J.
G.
2008.
Corn
grain
yield
response
to
crop
rotation
and
nitrogen
over
35
years.
Agron.
J.
100:643-650.
58.
Stromberg,
E.
L.,
and
Flinchum,
L.
E.
2003.
Evaluation
of
foliar
fungicides
for
the
control
of
gray
leaf
spot
of
corn
in
Virginia,
2002.
Fungic.
Nematicide
Tests
58:FC001.
Online
publication.
doi:
10.
10.1094/FN58.
59.
Stromberg,
E.
L.,
and
Kenley,
C.
C.
2004.
Evaluation
of
foliar
fungicides
for
the
control
of
gray
leaf
spot
of
corn
in
Virginia,
2003.
Fungic.
Nematicide
Tests
59:CF021.
Online
publication.
doi:
10.1094/FN58.
60.
Stromberg,
E.
L.,
and
Kenley,
C.
C.
2005.
Evaluation
of
foliar
fungicides
for
the
control
of
gray
leaf
spot
of
corn
in
Virginia,
2004.
Fungic.
Nematicide
Tests
60:FC074.
Online
publication.
doi:
10.1094/FN60.
61.
Stromberg,
E.
L.,
and
Kenley,
C.
C.
2006.
Evaluation
of
foliar
fungicides
for
the
control
of
gray
leaf
spot
of
corn
in
Virginia,
2005.
Fungic.
Nematicide
Tests
61:FC061.
Online
publication.
doi:
10.1094/FN61.
62.
Stromberg,
E.
L.,
and
Kenley,
C.
C.
2009.
Evaluation
of
foliar
fungicides
for
the
control
of
gray
leaf
spot
of
corn
in
Virginia,
2008.
Plant
Disease
Management
Reports
3:FC017.
Online
publication.
doi:
10.1094/
PDMR03.
63.
Thorson,
P.
R.,
and
Martinson,
C.
A.
1993.
Development
and
survival
of
Cercospora
zeae-
maydis
germlings
in
different
relative
humidity
environ-
ments.
Phytopathology
83:153-157.
64.
van
Houwelingen,
H.
C.,
Arends,
L.
R.,
and
Stijnen,
T.
2002.
Advanced
methods
in
meta
-analysis:
multivariate
approach
and
meta
-regression.
Stat.
Med.
21:589-624.
65.
Viswakumar,
A.
Mullen,
R.
W.,
Sundermeier,
A.,
and
Dygert,
C.
E.
2008.
Tillage
and
nitrogen
application
methodology
impacts
on
corn
grain
yield.
J.
Plant
Nutr.
31:1963-1974.
66.
Ward,
J.
M.
J.,
Stromberg,
E.
L.,
Nowell,
D.
C.,
and
Nutter,
E
W.,
Jr.
1999.
Gray
leaf
spot:
a
disease
of
global
importance
in
maize
production.
Plant
Dis.
83:884-895.
67.
Wegulo,
S.
N.,
Martinson,
C.
A.,
Rivera
-C.,
J.
M.,
and
Nutter,
F.
W.,
Jr.
1997.
Model
for
economic
analysis
of
fungicide
usage
in
hybrid
corn
seed
production.
Plant
Dis.
81:415-422.
68.
Wegulo,
S.
N.,
Rivera
-C,
J.
M.,
Martinson,
C.
A.,
and
Nutter,
E
W.,
Jr.
1998.
Efficacy
of
fungicide
treatments
for
control
of
common
rust
and
northern
leaf
spot
in
hybrid
corn
seed
production.
Plant
Dis.
82:547-554.
69.
Wise,
K.,
and
Buechley,
G.
2010.
Efficacy
of
early
applications
of
foliar
fungicides
on
corn
diseases
and
yield
in
Indiana,
2009.
Plant
Disease
Management
Reports
4:FC092.
Online
publication.
doi:
10.1094/PDMR04.
70.
Wise,
K.,
and
Buechley,
G.
2010.
Efficacy
of
foliar
fungicides
on
corn
diseases
in
central
and
southeastern
Indiana,
2009.
Plant
Disease
Manage-
ment
Reports
4:FC091.
Online
publication.
doi:
10.1094/PDMR04.
71.
Wise,
K.,
and
Buechley,
G.
2010.
Evaluation
of
foliar
fungicides
for
management
of
corn
diseases
in
Indiana,
2009.
Plant
Disease
Manage-
ment
Reports
4:FC093.
Online
publication.
doi:
10.1094/PDMR04.
72.
Wu,
Y.
-X.,
and
von
Tiedemann,
A.
2001.
Physiological
effects
of
azoxystrobin
and
epoxiconazole
on
senescence
and
the
oxidative
status
of
wheat.
Pest.
Biotechnol.
Phys.
71:1-10.
1132
PHYTOPATHOLOGY