Content uploaded by Maria Grillet
Author content
All content in this area was uploaded by Maria Grillet on Oct 13, 2014
Content may be subject to copyright.
Acta
Tropica
129 (2014) 52–
60
Contents
lists
available
at
ScienceDirect
Acta
Tropica
jo
ur
n
al
hom
epa
ge:
www.elsevier.com/locate/actatropica
The
periodicity
of
Plasmodium
vivax
and
Plasmodium
falciparum
in
Venezuela
María-Eugenia
Grilleta,∗,
Mayida
El
Soukia,
Francisco
Lagunaa,
José
Rafael
Leónb
aLaboratorio
de
Biología
de
Vectores
y
Parásitos,
Instituto
de
Zoología
y
Ecología
Tropical,
Facultad
de
Ciencias,
Universidad
Central
de
Venezuela,
Apartado
Postal
47072,
Caracas
1041-A,
Venezuela
bEscuela
de
Matemáticas,
Facultad
de
Ciencias,
Universidad
Central
de
Venezuela,
Apartado
Postal
47072,
Caracas
1041-A,
Venezuela
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
1
November
2012
Received
in
revised
form
27
September
2013
Accepted
4
October
2013
Available online 19 October 2013
Keywords:
Plasmodium
dynamics
Malaria
epidemiology
Wavelet
analyses
Rainfall
ENSO
Venezuela
a
b
s
t
r
a
c
t
We
investigated
the
periodicity
of
Plasmodium
vivax
and
P.
falciparum
incidence
in
time-series
of
malaria
data
(1990–2010)
from
three
endemic
regions
in
Venezuela.
In
particular,
we
determined
whether
disease
epidemics
were
related
to
local
climate
variability
and
regional
climate
anomalies
such
as
the
El
Ni˜
no
Southern
Oscillation
(ENSO).
Malaria
periodicity
was
found
to
exhibit
unique
features
in
each
studied
region.
Significant
multi-annual
cycles
of
2-
to
about
6-year
periods
were
identified.
The
inter-annual
variability
of
malaria
cases
was
coherent
with
that
of
SSTs
(ENSO),
mainly
at
temporal
scales
within
the
3–6
year
periods.
Additionally,
malaria
cases
were
intensified
approximately
1
year
after
an
El
Ni˜
no
event,
a
pattern
that
highlights
the
role
of
climate
inter-annual
variability
in
the
epidemic
patterns.
Rainfall
mediated
the
effect
of
ENSO
on
malaria
locally.
Particularly,
rains
from
the
last
phase
of
the
season
had
a
critical
role
in
the
temporal
dynamics
of
Plasmodium.
The
malaria–climate
relationship
was
complex
and
transient,
varying
in
strength
with
the
region
and
species.
By
identifying
temporal
cycles
of
malaria
we
have
made
a
first
step
in
predicting
high-risk
years
in
Venezuela.
Our
findings
emphasize
the
importance
of
analyzing
high-resolution
spatial–temporal
data
to
better
understand
malaria
transmission
dynamics.
© 2013 Elsevier B.V. All rights reserved.
1.
Introduction
Malaria,
one
of
the
most
serious
parasitic
diseases
of
tropical
ecosystems,
is
caused
by
parasites
of
the
genus
Plasmodium
(Apicomplexa:
Plasmodidae)
and
transmitted
among
human
hosts
by
the
bites
of
infected
female
Anopheles
mosquitoes
(Diptera:
Culicidae).
In
2010,
malaria
was
responsible
for
219
million
cases,
causing
nearly
700,000
deaths
(WHO,
2012).
Epidemiologic
patterns
of
malaria
can
be
highly
heterogeneous
and
caused
by
a
complex
set
of
interactions
among
parasites,
vectors,
and
hosts
occurring
at
specific
locations,
and
at
specific
times.
In
low
endemic
and
epidemic
areas,
Plasmodium
incidence
exhibits
regular
seasonal
cycles
and
multiyear
oscillations
over
time
(Hay
et
al.,
2000).
Annual
changes
in
rainfall
and
temperature
may
directly
or
indirectly
affect
Anopheles
reproduction
and
mortality
rates,
the
blood
feeding
frequency
of
the
mosquito
female
and
the
extrinsic
incubation
period
of
Plasmodium
and
thereby
cause
seasonal
variations
in
both
vectors
and
parasites
(Stresman,
2010).
Longer-term
or
inter-annual
cycles
of
the
parasite
might
be
driven
by
extrinsic
climatic
factors
(Bouma
and
Dye,
1997;
Bouma
et
al.,
∗Corresponding
author.
Tel.:
+58
2126051404;
fax:
+58
2126051204.
E-mail
address:
maria.grillet@ciens.ucv.ve
(M.-E.
Grillet).
1997;
Poveda
et
al.,
2001),
intrinsic
mechanisms
associated
with
epidemiological
dynamics
such
as
host
immunity
(Hay
et
al.,
2000),
or
both
factors
(e.g.,
Pascual
et
al.,
2008).
In
the
Americas,
malaria
is
still
a
serious
health
concern,
with
almost
20%
of
the
total
population
at
some
degree
of
risk,
espe-
cially
in
countries
such
as
Venezuela,
where
the
reported
morbidity
has
increased
significantly
in
the
last
decade
(WHO,
2012).
In
Venezuela,
Plasmodium
vivax
malaria
accounts
for
82%
of
all
cases,
followed
by
P.
falciparum
(16%),
P.
malariae
(<1%)
and
P.
vivax/P.
falciparum
mixed
(1.4%)
infections
(Cáceres,
2011).
The
pattern
of
malaria
transmission
varies
regionally,
depending
on
climate,
biogeography,
ecology,
and
anthropogenic
activities.
Whereas
P.
falciparum
malaria
occurs
mostly
in
the
lowland
rain
forests
of
the
Venezuelan
Guayana
region,
P.
vivax
malaria
is
endemic
in
the
coastal
plains
and
savannas
as
well
as
the
lowland
Guayana
forests
(Rubio-Palis
and
Zimmerman,
1997).
Before
the
successful
malaria
eradication
campaign
in
the
early
20th
century
in
Venezuela,
recurrent
epidemics
occurred
every
five
years,
particularly
in
the
savannas
landscapes
and
coastal
plains
where
Anopheles
darlingi
was
the
main
vector
of
P.
falciparum
(Gabaldon,
1949).
This
author
observed
that
malaria
cycles
apparently
coincided
with
periodic
fluctuations
of
the
vector
population.
Later,
Bouma
and
Dye
(1997)
associated
these
epidemics
of
malaria
with
the
El
Ni˜
no
Southern
Oscillation
(ENSO).
This
previous
work
analyzed
malaria
at
the
0001-706X/$
–
see
front
matter ©
2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.actatropica.2013.10.007
M.-E.
Grillet
et
al.
/
Acta
Tropica
129 (2014) 52–
60 53
country
level,
yearly
timescales,
and
overall
malaria
incidence
(P.
vivax
+
P.
falciparum).
However,
no
study
has
addressed
temporal
patterns
in
malaria
infections,
specially
their
inter-annual
cycles,
by
resolving
the
species
and
the
malaria
eco-regions
of
Venezuela.
Such
downscaling
in
space
and
parasite
taxonomy
could
reveal
significant
heterogeneity
in
malaria
periodicity.
Since
malaria
has
become
again
a
serious
health
problem
in
this
country
(Cáceres,
2011),
year-to-year
variation
in
the
size
of
epidemics,
are
of
par-
ticular
concern.
Understanding
this
inter-annual
variability
in
the
population
dynamics
of
malaria
can
provide
useful
insights
for
malaria
elimination
programs.
Furthermore,
a
better
knowledge
of
the
malaria
temporal
patterns
would
allow
the
development
of
more
effective
surveillance
and
early
warning
systems
to
predict
disease
risk
in
response
to
changes
in
climate.
In
this
paper,
we
re-examine
the
question
of
malaria’s
multi-
year
cycles
in
Venezuela
by
using
primarily
a
statistical
method
of
time-series
analysis
well
suited
for
transient
patterns
in
diseases
dynamics
and
environmental
conditions
over
time
(non-stationary
patterns).
We
specifically
address
the
following
questions:
(i)
Is
there
evidence
for
particular
frequencies
in
the
temporal
dynamics
of
malaria?
(ii)
Is
malaria
periodicity
species-specific
and
geo-
graphically
variable?
(iii)
Is
the
inter-annual
pattern
of
malaria
in
Venezuela
associated
with
climate
variability?
(iv)
If
so,
does
rainfall
mediate
the
effect
of
ENSO
on
malaria
locally?
To
do
this,
we
analyze
the
monthly
incidence
of
P.
vivax
and
P.
falciparum
(1990–2010)
from
three
endemic
regions
of
the
country.
We
show
that
ENSO
has
played
a
role
in
the
long-term
malaria
dynamics
during
the
last
20
years
in
Venezuela,
but
that
the
disease–climate
relationship
is
complex,
varying
in
characteristic
periodicities
and
strength
according
to
region
and
parasite
species.
2.
Materials
and
methods
2.1.
Study
area
Venezuela
is
located
in
the
northern
coast
of
South
America
with
a
surface
area
of
contrasting
landscapes
including
a
north-
ern
Caribbean
coastal
plain
and
the
Venezuelan
Guayana
in
the
south
(Fig.
1).
Malaria
is
a
major
public
health
problem
in
different
endemic–epidemic
eco-regions
of
the
country
such
as
the
lowland
rain
forest
and
savannas
of
Guayana
(<200
m),
and
the
north-
eastern
coastal
plains.
Currently,
the
lowland
Venezuelan
Guayana
is
considered
a
region
of
high-risk
of
stable
malaria
mainly
caused
by
P.
vivax
(∼76–84%
of
cases)
and
P.
falciparum
(∼21–15%
of
cases),
and
largely
transmitted
by
An.
darlingi
and
An.
marajoara
(Magris
et
al.,
2007;
Moreno
et
al.,
2007).
Anopheles
darlingi
is
mainly
a
river-
ine
and
forest-dwelling
species,
while
An.
marajoara
is
a
mosquito
species
associated
with
wetlands,
secondary
forests,
and
human
intervention
(Moreno
et
al.,
2007).
The
whole
Guayana
region
cov-
ers
an
extensive
area
of
the
country
(530,145
km2),
however,
the
population
density
is
very
low
and
heterogeneously
distributed
in
two
administrative
areas
(Fig.
1):
the
Amazonas
State
(0.86
inhabi-
tants
per
km2)
and
the
Bolívar
State
(6.74
inhabitants
per
km2).
Most
of
the
inhabitants
of
Amazonas
live
in
the
north-western
corner
of
the
state
(in
the
Atures
and
Autana
municipalities)
and
belong
to
predominantly
indigeneous
ethnic
groups
(Metzger
et
al.,
2009).
Here,
the
savanna
ecosystem
is
the
dominant
landscape
of
malaria
transmission
(Rubio-Palis
and
Zimmerman,
1997)
and
An.
darlingi
is
the
main
species
vector.
In
Bolívar
State,
the
population
at
risk
is
mostly
localized
in
the
south-east
(e.g.,
in
the
Sifontes
Municipality),
where
economic
activities
are
agriculture,
gold
and
diamond
mining,
and
forest
exploitation.
In
this
endemic
area,
the
lowland
forest
ecosystem
is
the
dominant
malaria
landscape,
with
An.
darlingi
and
An.
marajoara
as
the
main
species
vectors
(Moreno
et
al.,
2007).
In
the
malaria
coastal
eco-region
(Sucre
State),
along
the
Caribbean
Sea
(Fig.
1),
the
infection
is
caused
by
P.vivax
and
transmitted
by
Anopheles
aquasalis.
This
area
is
largely
composed
of
mangroves,
herbaceous
and
woody
swamps.
Anopheles
aquasalis
is
mainly
associated
with
brackish
and
freshwater
wetlands
(Grillet,
2000).
Economic
activities
of
the
population
are
mainly
fishing,
subsistence
agriculture,
and
tourism.
Semi-annual,
annual
and
inter-annual
cycles
strongly
character-
ize
weather
and
climate
variability
in
Venezuela
(Pulwarty
et
al.,
1992).
The
large-scale
spatial
features
of
rainfall
are
primarily
influenced
by
the
annual
location
of
the
Atlantic
Inter-tropical
Convergence
Zone
(ITCZ),
whereas
its
local
spatial
variability
is
con-
trolled
by
the
mountain
ranges,
the
influence
of
the
atmospheric
circulation
over
the
Amazon
basin,
and
the
local
vegetation
and
land
surface.
At
the
inter-annual
scale,
the
Southern
Oscillation
is
the
main
forcing
mechanism
of
rainfall
variability
(Pulwarty
et
al.,
1992).
Overall,
the
April–November
season
carries
88%
of
the
mean
annual
rainfall
in
Venezuela.
In
the
Amazonas
stud-
ied
region,
the
annual
rainfall
is
around
2300
mm
and
the
rainier
months
are
those
from
May
through
August,
followed
by
lower
rains
during
September,
October
and
November
(Fig.
A.1.a,
Sup-
plementary
material-SM).
Drier
months
include
December–March
and
the
annual
mean
temperature
is
around
25–28 ◦C.
In
Bolívar
State,
the
malaria
region
is
characterized
by
an
average
tem-
perature
of
24–26 ◦C
and
1000–1300
mm
annual
rainfall
(Huber,
1995).
Here,
rains
display,
in
particular,
a
primary
peak
from
May
to
August
followed
by
a
secondary
but
lower
peak
from
November
to
January
(Fig.
A.1.d,
SM).
In
the
malaria
coastal
eco-
region,
the
annual
mean
temperature
is
around
27 ◦C
and
the
total
annual
rainfall
is
close
to
1000
mm,
with
a
rainy
season
from
May
to
November
and
a
dry
season
from
December
to
April
(Fig.
A.1.g,
SM).
Northeastern
Venezuelan
rainfalls
(e.g.,
the
Boli-
var
and
Sucre
States)
are
more
directly
influenced
by
the
ITCZ
than
the
southwestern
Venezuelan
rainfalls
(e.g.,
the
Amazonas
State).
Supplementary
material
related
to
this
article
can
be
found,
in
the
online
version,
at
http://dx.doi.org/10.1016/j.actatropica.2013.
10.007.
2.2.
Epidemiological
and
socio-demographic
data
State-
and
municipality-level
cases
of
P.
vivax
and
P.
falci-
parum
were
obtained
from
the
Malaria
Control
Program
database,
Venezuelan
Ministry
of
Health.
Malaria
treatment
in
the
country
is
exclusively
carried
out
by
the
public
health
system
which
uses
artemisinin-based
combination
therapy
(ACT)
as
the
first-line
treatment
for
P.
falciparum,
whereas
chloroquine
and
primaquine
are
applied
for
the
treatment
of
P.
vivax.
Local
health
services
are
obliged
to
compile
all
notifications
of
malaria
on
a
weekly
basis.
Symptomatic
cases
are
detected
by
passive
and
active
surveillance
as
well
as
reported
according
to
geographic
origin.
We
analyzed
regional
monthly
records
for
P.
vivax
and
P.
falciparum
incidence
from
1990
to
2010
with
data
aggregated
across
municipalities
by
state.
For
Amazonas
State,
we
pooled
the
malaria
data
from
the
north-western
part
of
the
state,
and
we
carried
out
the
same
procedure
with
the
incidence
data
from
the
south-eastern
side
of
Bolívar
State
(Fig.
1).
Complete
time
series
were
not
available
for
the
coastal
eco-region.
Therefore,
malaria
data
were
studied
from
1998
to
2008
from
the
Cajigal
and
Benitez
Municipalities
(Fig.
1).
We
calculated
malaria
incidence
rates
per
1000
inhabitants
(No.
of
new
cases
×
1000/population
at
risk
per
time)
by
spatial
region
and
by
species
(Plasmodium
species).
We
assumed
that
the
entire
population
of
the
studied
area
was
exposed
to
the
risk
of
contract-
ing
malaria;
that
is,
each
person
contributed
exactly
1
person-time
of
exposure.
The
malaria
incidence
rate
was
calculated
by
taking
into
account
the
human
population
growth
rate
predicted
for
54 M.-E.
Grillet
et
al.
/
Acta
Tropica
129 (2014) 52–
60
Fig.
1.
Map
of
Venezuela
showing
the
main
malaria
regions:
Amazonas,
Bolívar
and
Sucre
States.
Malaria
cases
were
aggregated
across
the
municipalities
within
each
state.
In
the
low-land
forest
eco-region,
malaria
is
caused
by
Plasmodium
vivax
(84%)
and
P.
falciparum
(21%),
and
mainly
transmitted
by
Anopheles
darling.
In
the
coastal
eco-region,
the
infection
is
caused
by
P.
vivax
and
transmitted
by
An.
aquasalis.
each
studied
period
according
to
the
demographic
data
(at
risk
population)
from
the
National
Statistics
Office
of
Venezuela.
2.3.
Climate
data
Contemporaneous
meteorological
data
were
obtained
from
the
nearest
meteorological
stations.
The
stations
by
region
were:
Ama-
zonas
(Puerto
Ayacucho:
05◦39N,
67◦38W);
Bolívar
(Anacoco:
06◦43N,
61◦08W;
El
Dorado:
06◦42N,
61◦38W;
Tumeremo:
07◦17N,
61◦30W);
and
Sucre
(Guiria:
10◦34N,
62◦17W;
Irapa:
10◦34N
62◦34W;
Carupano:
10◦39N,
63◦15W).
We
aver-
aged
the
data
of
all
stations
by
each
region.
Data
included
mean
temperature
and
rainfall
by
month;
however,
we
mainly
focused
on
rainfall
due
to
the
low
annual
oscillation
of
temperature
in
Venezuela
(Huber,
1995).
We
used
the
monthly
sea
surface
tem-
peratures
(SSTs)
of
the
eastern
and
central
tropical
Pacific
as
an
index
of
the
“El
Ni˜
no”
phenomenon
(for
the
Ni˜
no
Regions
known
as
3
+
4),
the
main
inter-annual
climatic
event
in
the
northern
coast
of
South
America
(Poveda
and
Mesa,
1997)
and
Venezuela
(Pulwarty
et
al.,
1992).
The
SST
time-series
were
obtained
from
the
Climate
Prediction
Center
of
the
National
Oceanic
and
Atmospheric
Admin-
istration
(NOAA,
2011).
The
atmospheric
component
linked
to
El
Ni˜
no
is
termed
the
Southern
Oscillation
(ENSO)
with
both
atmo-
sphere
and
ocean
phenomena
acting
together.
El
Ni˜
no
corresponds
to
the
warm
phase
of
ENSO,
whereas
the
opposite
“La
Ni˜
na”
phase
consists
of
a
basin-wide
cooling
of
the
tropical
Pacific
and
thus
the
cold
phase
of
ENSO
(Trenberth,
1999).
In
general,
there
is
a
coher-
ent
pattern
of
climatic
and
hydrological
anomalies
over
the
region
during
extreme
phases
of
ENSO
(Poveda
and
Mesa,
1997).
Nega-
tive
anomalies
in
rainfall
(below-normal),
soil
moisture
and
river
flows,
as
well
as
warmer
air
temperatures,
occur
during
El
Ni˜
no
for
Venezuela
(Pulwarty
et
al.,
1992;
Poveda
and
Mesa,
1997).
The
reverse
is
true
for
the
cold
phase
La
Ni˜
na.
El
Ni˜
no
is
a
climatic
oscilla-
tion
with
an
average
recurrence
varying
from
2
to
10
years,
with
an
average
of
about
every
4
years
(Trenberth,
1999).
The
events
usu-
ally
include
two
calendar
years,
and
are
generally
characterized
by
SSTs
positive
anomalies
that
increase
during
the
Northern
hemi-
sphere
spring
and
fall
of
the
first
year
(Ni˜
no0),
with
the
maximum
SSTs
anomalies
occurring
during
the
winter
of
the
following
year
(Ni˜
no+1),
and
SSTs
anomalies
receding
during
the
spring
and
sum-
mer
of
the
year +1 (Poveda
and
Mesa,
1997).
Six
El
Ni˜
no
(1991–1992,
1994–1995,
1997–1998,
2002–2003,
2004,
2009–2010)
and
three
La
Ni˜
na
(1995–1996,
1998–2000,
2007–2008)
events
occurred
dur-
ing
our
study
period
according
to
the
criteria
of
Trenberth
(1999).
The
1991–1995
was
considered
a
very
long
and
extended
(5-year)
El
Ni˜
no
“event”,
while
the
1997–1998
has
been
one
of
the
strongest
El
Ni˜
no
episodes
in
the
last
30
years
(Fedorov
and
Philander,
2000).
2.4.
Time
series
analysis
of
P.
vivax
and
P.
falciparum
malaria
Fourier
analysis
(FA)
and
the
associated
Fourier
power
spectrum
has
been
one
of
the
most
common
statistical
analyses
to
decom-
pose
the
variance
of
a
time-series
into
dominant
frequencies,
and
separate
seasonal
(annual)
from
longer-term
(multi-annual)
cycles.
FA
is
not
able
however
to
characterize
signals
whose
frequency
content
changes
with
time
in
a
transient
manner.
Thus
FA
cannot
provide
information
on
when
particular
frequencies
are
present
(Cazelles
et
al.,
2007).
Because,
epidemiological
and
environmental
time-series,
as
well
as
their
associations,
can
be
strongly
non-
stationary
(varying
in
time),
a
different
spectral
method
known
as
wavelet
analyses
(WA)
has
been
applied
to
analyze
their
peri-
odic
and
dominant
components
and
how
they
change
over
time
(Cazelles
et
al.,
2007;
Torrence
and
Compo,
1998).
Here,
WA
M.-E.
Grillet
et
al.
/
Acta
Tropica
129 (2014) 52–
60 55
was
performed
using
the
Morlet
wavelet
transform
which
can
be
regarded
as
a
generalization
of
the
Fourier
transform
that
allows
the
localization
in
time
of
the
analyses;
by
analogy
with
spectral
approaches,
one
can
compute
the
global
power
spectrum
by
aver-
aging
the
local
wavelet
spectrum
over
time
(Cazelles
et
al.,
2007).
Thus,
the
wavelet
power
spectrum
(WPS)
estimates
the
distribu-
tion
of
variance
between
different
frequencies
at
different
time
locations.
The
time-series
were
first
filtered
with
a
low-pass
filter
to
remove
the
seasonality
and
focus
on
their
inter-annual
variabil-
ity.
Also,
data
were
square-root
transformed
and
normalized.
In
addition,
wavelet
coherency
(WC)
was
used
to
compare
the
fre-
quency
components
of
the
Plasmodium
and
climate
time-series
in
the
three
regions
and
quantify
the
statistical
(linear)
associ-
ation
between
variables
locally
in
time
(Cazelles
et
al.,
2007).
WC
provides
local
information
on
when
two
non-stationary
signals
are
linearly
correlated
and
at
what
particular
fre-
quency.
Finally,
the
Kulldorff’s
scan
statistics
allowed
us
to
identify
sig-
nificant
excesses
of
cases
(e.g.,
the
most
likely
cluster)
of
P.
vivax
and
P.
falciparum
incidence
in
time
(Grillet
et
al.,
2010).
The
lin-
ear
relationship
of
seasonal
anomalies
in
malaria
with
seasonal
anomalies
of
El
Ni˜
no
and
rainfall
were
explored
through
cross-
correlation
functions.
We
obtained
the
standardized
anomaly
by
subtracting
the
long-term
mean
value
(e.g.,
1990–2010
period)
of
a
particular
season
(3-month
running
mean)
and
by
scaling
this
anomaly
using
the
seasonal
standard
deviation.
This
stan-
dardization
method
filters
out
the
annual
cycle
in
each
variable,
and
highlights
changes
between
years
(inter-annual
cycles).
We
performed
all
the
time-series
analyses
using
original
algorithms
developed
in
Matlab
(V.7.5,
The
MathWorks,
Natick,
Massachusetts,
United
States)
by
Cazelles
et
al.
(2007).
The
software
SaTScan
V.9.0.1
(Kulldorff,
1997)
was
used
for
the
Kulldorff’s
Scan
analy-
sis.
3.
Results
3.1.
Temporal
patterns
of
P.
vivax
and
P.
falciparum
incidence
The
temporal
pattern
of
P.
vivax
and
P.
falciparum
cases
in
the
three
malaria
regions
of
Venezuela
is
shown
in
Fig.
2.
In
general,
both
parasite
species
have
exhibited
a
significant
rise
in
incidence
in
the
south-eastern
region
of
Bolívar
State.
By
contrast,
disease
incidence
has
gradually
declined
in
the
coastal
eco-region.
Par-
ticularly,
for
the
Amazonas
State,
the
annual
incidence
of
P.
vivax
increased
from
12
(1998)
to
64
(2004)
cases
per
1000
inhabitants,
depicting
a
significant
time
clustering
of
malaria
cases
(Fig.
2a)
dur-
ing
2001–2008
(Relative
Risk:
RR
=
2.0,
log-likelihood
ratio
=
4846,
P
<
0.001).
The
annual
incidence
of
P.
falciparum
varied
from
5
cases
per
1000
in
2010
to
25
cases
per
1000
during
2004,
when
a
signif-
icant
epidemic
peak
was
detected
(Fig.
2a;
RR
=
2.2,
log-likelihood
ratio
=
1073,
P
<
0.001).
In
the
Bolívar
State,
cases
of
P.
vivax
(Fig.
2b)
ranged
from
1
case
(1993)
to
29
cases
per
1000
in
2010,
show-
ing
a
significant
and
positive
trend
from
2003
onwards
(linear
model,
R2=
0.77,
P
<
0.05),
and
an
epidemic
period
from
2004
to
2010
(RR
=
4.2,
log-likelihood
ratio
=
51,243,
P
<
0.001).
Similarly,
P.
falciparum
cases
had
a
significant
positive
trend
during
the
study
period
(linear
model,
R2=
0.77,
P
<
0.05)
and
an
epidemic
peak
dur-
ing
2010
(Fig.
2b),
when
the
annual
incidence
rate
reached
10
cases
per
1000
(RR
=
5.1,
log-likelihood
ratio
=
11,860,
P
<
0.001).
In
the
northern
Sucre
State
(Fig.
2c),
P.
vivax
showed
an
epidemic
peak
during
2002
(18
cases
per
1000
inhabitants,
RR
=
4.5,
log-likelihood
ratio
=
8579,
P
<
0.001),
although
malaria
infection
has
displayed
a
significant
negative
trend
(linear
model,
R2=
0.67,
P
<
0.05)
during
the
last
years.
Fig.
2.
Monthly
time
series
of
malaria
cases
in
(a)
Amazonas,
(b)
Bolívar,
and
(c)
Sucre
regions.
3.2.
Periodicity
of
malaria,
ENSO
and
rainfall
Overall,
multi-annuals
oscillations
of
Plasmodium
at
the
2-year
and
3–6-year
bands
were
detected
and
dominated
the
malaria
dynamics
in
the
three
regions.
However,
these
parasite
cycles
were
transient
and
varied
in
time
and
space.
The
global
spectrum
of
P.
vivax
time-series
in
the
Amazonas
State
showed
(Fig.
3a,
right
panel)
a
significant
and
dominant
period
of
4-year.
Locally
in
time,
the
wavelet
power
spectrum-WPS
(Fig.
3a,
left
panel)
revealed
that
this
oscillatory
mode
was
most
pronounced
after
2002
and
that
a
weaker
2-year
period
was
detected
in
the
late
1990s.
A
similar
pat-
tern
was
found
for
P.
falciparum
cases
after
1999,
except
that
the
longer
cycle
had
a
greater
amplitude
(Fig.
3b,
right
panel),
and
the
shorter
one
had
a
stronger
signal
(Fig.
3b,
left
panel).
In
Fig.
3c,
the
periodicity
of
ENSO
is
depicted.
The
SST
time-series
exhibited
inter-
annual
variability
for
the
3–6
and
2-years
bands,
but
the
first
cycle
had
higher
power,
variance
and
significance
from
the
end
of
1990s
and
the
beginning
of
the
2000s
(Fig.
3c,
left
panel),
coinciding
with
56 M.-E.
Grillet
et
al.
/
Acta
Tropica
129 (2014) 52–
60
Fig.
3.
Patterns
of
inter-annual
variability
of
the
monthly
malaria
cases
of
Amazonas
region
(a
(top):
Plasmodium
vivax,
b
(center):
P.
falciparum)
and
monthly
sea
surface
temperatures
(c
(bottom):
SSTS).
Left:
Wavelet
power
spectrum
(WPS).
Right:
Global
spectrum
(GS).
The
y-axis
describes
period
(in
years),
the
x-axis
(GS)
describes
the
power
at
a
given
frequency
(continuous
line)
with
its
significant
threshold
value
of
5%
(dashed
line).
In
the
WPS,
the
color
code
for
power
values
ranges
from
dark
blue,
for
low
values,
to
dark
red,
for
high
ones.
The
dotted-dashed
lines
show
the
˛
=
5%
significant
levels
(see
Cazelles
et
al.,
2007).
The
cone
of
influence
(continuous
line)
in
the
WPS
indicates
the
region
not
influenced
by
edge
effects.
To
remove
seasonality,
variables
were
filtered
with
a
low-pass
filter.
Different
scales
(years)
are
used
due
to
the
different
extent
of
the
time
series.
(For
interpretation
of
the
references
to
color
in
this
figure
legend,
the
reader
is
referred
to
the
web
version
of
this
article.)
the
3
Ni˜
no
events
of
that
period.
In
the
Bolívar
region,
the
P.
vivax
time-series
showed
a
continuous
oscillation
period
of
high
ampli-
tude
around
the
3–6-year
mode
(Fig.
4a).
By
contrast,
a
transient
multi-annual
cycle
around
the
3-year
periodic
band
was
detected
for
P.
falciparum
incidence
(Fig.
4b).
For
the
coastal
region,
a
bien-
nial
cycle
dominated
the
P.
vivax
periodicity
(Fig.
4c,
right
panel)
but
the
power
of
this
oscillation
mode
was
most
intense
before
2005
(Fig.
4c,
left
panel).
Finally,
we
investigated
the
patterns
of
inter-annual
variability
of
rainfall
in
the
three
regions
(Fig.
5).
In
general,
rainfall
spectra
exhibited
power
at
periods
of
3–6
and
2-
years
bands.
The
long-term
periodicity
of
rainfall
in
the
Amazonas
region
(Fig.
5a)
was
characterized
by
two
areas
of
high
significance
for
the
2-year
and
5–6-year
periodic
bands,
especially
after
1997.
By
contrast,
rainfall
in
the
Bolívar
region
showed
inter-annual
vari-
ability
for
the
2–4-year
band
(Fig.
5b)
before
2006,
whereas
rains
in
Sucre
exhibited
one
main
oscillation
around
the
2–3-year
mode
(Fig.
5c).
3.3.
Malaria
periodicity
and
ENSO
We
evaluated
the
correspondence
of
the
wavelet
spectra
for
malaria
and
ENSO
through
the
cross-coherence
spectrum
(WC).
As
Fig.
6
shows,
there
was
a
significant
but
transient
cross-coherence
between
both
variables
across
the
three
regions,
especially
for
the
3–6
year
scale.
This
coupling
was
most
marked
with
the
Plasmodium
time-series
of
Bolívar
(Fig.
6b
and
Fig.
A.2.b,
SM).
Additionally,
a
Fig.
4.
Patterns
of
inter-annual
variability
of
the
monthly
malaria
cases
of
Bolívar
(a
(top):
Plasmodium
vivax,
b
(center):
P.
falciparum)
and
Sucre
(c
(bottom):
P.
vivax)
regions.
Left:
Wavelet
power
spectrum
(WPS).
Right:
Global
spectrum
(GS).
The
y-axis
describes
period
(in
years),
the
x-axis
(GS)
describes
the
power
at
a
given
frequency
(continuous
line)
with
its
significant
threshold
value
of
5%
(dashed
line).
In
the
WPS,
the
color
code
for
power
values
ranges
from
dark
blue,
for
low
values,
to
dark
red,
for
high
ones.
The
dotted-dashed
lines
show
the
˛
=
5%
significant
levels
(see
Cazelles
et
al.,
2007).
The
cone
of
influence
(continuous
line)
in
the
WPS
indicates
the
region
not
influenced
by
edge
effects.
To
remove
seasonality,
variables
were
filtered
with
a
low-pass
filter.
Different
scales
(years)
are
used
due
to
the
different
extent
of
the
time
series.
(For
interpretation
of
the
references
to
color
in
this
figure
legend,
the
reader
is
referred
to
the
web
version
of
this
article.)
M.-E.
Grillet
et
al.
/
Acta
Tropica
129 (2014) 52–
60 57
Fig.
5.
Patterns
of
inter-annual
variability
of
the
monthly
rainfall
of
Amazonas
(a:
top),
Bolívar
(b:
center),
and
Sucre
(c:
bottom)
regions.
Left:
Wavelet
power
spectrum
(WPS).
Right:
Global
spectrum
(GS).
The
y-axis
describes
period
(in
years),
the
x-axis
(GS)
describes
the
power
spectrum
(continuous
line)
with
its
significant
threshold
value
of
5%
(dashed
line).
In
the
WPS,
the
color
code
for
power
values
ranges
from
dark
blue,
for
low
values,
to
dark
red,
for
high
ones.
The
dotted-dashed
lines
show
the
˛
=
5%
significant
levels
(see
Cazelles
et
al.,
2007).
The
cone
of
influence
(continuous
line)
in
the
WPS
indicates
the
region
not
influenced
by
edge
effects.
To
remove
seasonality,
variables
were
filtered
with
a
low-pass
filter.
Different
scales
(years)
are
used
due
to
the
different
extent
of
the
time
series.
(For
interpretation
of
the
references
to
color
in
this
figure
legend,
the
reader
is
referred
to
the
web
version
of
this
article.)
weaker
cross-coherence
between
malaria
and
SST
time-series
was
observed
for
the
2
year
scale,
especially
for
Amazonas
(Fig.
6a
and
Fig.
A.2.b,
SM)
and
Sucre
(Fig.
6c).
The
role
of
ENSO
on
the
malaria
temporal
dynamics
in
Bolívar
was
further
supported
by
the
posi-
tive
correlations
we
found
between
the
annual
P.
vivax
cases
and
the
seasonal
SST
anomalies
from
July
through
September
(r
=
0.52;
P
<
0.05)
and
October
through
December
(r
=
0.52;
P
<
0.05)
of
the
previous
year.
We
also
observed
significant
correlations
between
annual
malaria
by
P.
falciparum
and
the
seasonal
SST
anoma-
lies
of
the
previous
year
(July–September:
r
=
0.75;
P
<
0.001
and
October–December:
r
=
0.67;
P
<
0.001).
This
suggests
that
number
of
malaria
cases
intensifies
in
the
year
following
an
El
Ni˜
no
event
in
this
endemic
area.
Similarly,
a
high
correspondence
between
sea-
sonal
malaria
and
seasonal
SST
anomalies
of
the
9
previous
months
(r
=
0.57;
P
<
0.05)
was
observed
for
the
coastal
eco-region.
Supplementary
material
related
to
this
article
can
be
found,
in
the
online
version,
at
http://dx.doi.org/10.1016/j.actatropica.2013.
10.007.
3.4.
Malaria
seasonality
and
rainfall
At
the
seasonal
scale,
we
explored
the
role
of
rainfall
as
a
driver
of
the
annual
dynamics
of
Plasmodium.
The
average
seasonal
cycle
of
malaria
showed
a
bimodal
pattern
in
the
Amazonas
and
Sucre
regions
(Fig.
A.1.b,
c
and
h,
SM):
a
large
peak
in
January–February
(dry
season)
and
a
smaller
peak
in
October–November.
Rainfall
in
both
regions
displays
a
main
season,
with
their
maximum
levels
varying
in
time
and
amplitude
(Fig.
A.1.a
and
g,
SM).
We
correlated
accumulated
rains
for
different
windows
during
the
year
with
accumulated
Plasmodium
cases
separately
for
the
two
malaria
peaks
to
find
the
“rainfall
window”
with
the
strongest
association
to
cases
in
a
given
malaria
season.
Rainfalls
from
October
to
December
had
the
best
correlation
with
the
main
peak
of
malaria
in
the
Amazonas
region
(P.
vivax,
r
=
0.55;
P.
falciparum,
r
=
0.57;
P
<
0.05).
Additionally,
the
observed
cases
of
malaria
from
October
to
November
were
significantly
associated
with
the
first
malaria
peak
of
January–February
(P.
vivax:
r
=
0.89,
P.
falciparum:
r
=
0.47;
P
<
0.05),
and
the
rains
from
the
wetter
months
(e.g.,
the
JJA
season,
Fig.
A.1.a,
SM)
showed
a
negative
association
with
the
second
annual
peak
of
malaria
(P.
vivax:
r
=
−0.54;
P.
falciparum:
r
=
−0.50;
P
<
0.05).
Together,
these
results
suggest
that
the
rains
from
the
last
phase
of
the
season
play
a
critical
role
in
the
annual
variation
of
Plasmodium
cases,
whereas
the
excess
of
rains
results
in
less
malaria
than
expected.
Comparable
results
were
observed
in
the
Sucre
region,
where
the
accumulated
rains
from
October
to
December
(Fig.
A.1.g,
SM)
accounted
for
the
first
peak
of
malaria
(r
=
0.68;
P
<
0.05),
whereas
malaria
cases
from
August
to
September
were
explained
by
the
previous
malaria
of
February
(r
=
0.92;
P
<
0.001).
Here,
rains
of
the
wetter
months
(JAS
season)
did
not
show
a
sig-
nificant
effect
on
Plasmodium
cases
at
the
end
of
the
year
albeit
the
association
was
negative
(r
=
−0.36;
P
>
0.05).
Unlike
Amazonas
and
Sucre,
malaria
in
the
Bolívar
region
had
a
less
marked
seasonality
Fig.
6.
Wavelet
coherence
analysis
of
malaria
incidence
with
the
sea
surface
temperatures
(SSTs)
time
series
across
three
regions:
Amazonas
(a
(top):
Plasmodium
vivax),
Bolívar
(b
(center):
P.
falciparum)
and
Sucre
(c
(bottom):
P.
vivax)
States.
The
colors
are
coded
as
dark
blue,
for
low
coherence
and
dark
red,
for
high
coherence.
The
dotted-
dashed
lines
show
the
˛
=
5%
significant
levels
(see
Cazelles
et
al.,
2007).
The
cone
of
influence
indicates
the
region
not
influenced
by
edge
effects.
Different
scales
(years)
are
used
due
to
the
different
extent
of
the
time
series.
(For
interpretation
of
the
references
to
color
in
this
figure
legend,
the
reader
is
referred
to
the
web
version
of
this
article.)
58 M.-E.
Grillet
et
al.
/
Acta
Tropica
129 (2014) 52–
60
(Fig.
A.1.e
and
f,
SM),
especially
malaria
by
P.
vivax,
which
showed
an
almost
constant
incidence
the
whole
year
round.
The
rainy
season
in
this
region
is
more
long
and
sustained
in
time
compared
to
the
other
two
regions
(Fig.
A.1.d,
SM)
and
P.
falciparum
cases
seem
to
exhibit
a
delayed
response
to
rainfall
(Fig.
A.1.f,
SM).
Nevertheless,
we
found
that
rains
from
October
to
December
could
account
for
the
cases
of
P.
vivax
during
the
January–February
period
(r
=
0.58;
P
<
0.05).
These
lower
but
sustained
rains
also
explained
the
malaria
by
P.
falciparum
from
the
same
period
(r
=
0.78,
P
<
0.05)
and
from
the
January–February
period
(r
=
0.56,
P
<
0.05).
Overall,
these
results
highlight
the
critical
role
of
the
last
rains
of
the
season
for
the
annual
malaria
dynamics
in
the
three
studied
regions.
3.5.
Rainfall
and
ENSO
At
the
inter-annual
scale,
we
evaluated
the
correspondence
of
the
wavelet
spectra
for
rainfall
and
SSTs.
In
the
Amazonas
region,
a
significant
cross-coherence
between
both
variables
was
found
for
the
3–6
and
2-years
bands
(data
not
shown).
Here,
the
El
Ni˜
no0event
was
associated
with
below-normal
rainfall
(r
=
−0.14,
P
<
0.05),
but
interestingly,
when
we
explored
different
time
win-
dows,
only
the
last
rains
(SON
season)
had
a
significant
correlation
with
the
seasonal
SST
anomalies
from
April
through
June
(r
=
−0.45,
P
<
0.05).
Surprisingly,
we
also
observed
positive
rainfall
anoma-
lies
during
the
El
Ni˜
no+1 year,
when
the
event
was
receding
(July–September)
or
almost
gone
(October
and
November).
Indeed,
the
SON
accumulated
rains
during
the
El
Ni˜
no+1 years
were
higher
compared
to
those
of
“neutral”
years
in
this
region
(Kruskal–Wallis
test,
H
=
4.08,
P
<
0.05).
Particularly,
we
observed
this
pattern
during
the
1992,
1993,
1998,
and
2003
years.
In
the
Bolívar
region,
a
signif-
icant
cross-coherence
between
rainfall
and
ENSO
was
found
for
the
3–6
year
band
(data
not
shown).
In
this
area,
the
El
Ni˜
no0years
were
associated
with
below-normal
rainfall
(r
=
−0.49,
P
<
0.05),
and
the
rains
from
the
OND
months
were
also
particularly
affected
by
the
seasonal
SST
anomalies
from
July
through
September
(r
=
−0.45,
P
<
0.05)
and
October
through
December
(r
=
−0.46,
P
<
0.05).
As
for
the
Amazonas
region,
positive
anomalies
in
the
late
rainy
sea-
son
of
the
El
Ni˜
no+1 years
(1993,
1994,
1998)
generated
higher
rains
compared
to
those
of
“neutral”
years
(Kruskal–Wallis
test,
H
=
5.33,
P
<
0.05).
Finally,
rainfall
in
the
Sucre
region
showed
a
sig-
nificant
cross-coherence
with
the
SST
time-series
for
the
2
and
5–6
years
bands
(data
not
shown).
The
El
Ni˜
no0years
were
associated
with
below-normal
rainfall
(r
=
−0.24,
P
<
0.05),
and
positive
rain-
fall
anomalies
were
also
observed
at
the
end
of
the
El
Ni˜
no+1 events
but
without
a
significant
association.
4.
Discussion
In
this
study,
we
were
able
to
identify
the
periodicity
of
malaria
cases
by
Plasmodium
species
in
three
malaria
endemic
regions
of
Venezuela.
The
identified
cycles
in
malaria
dynamics
were
tran-
sient
and
differed
across
regions
and
species.
Cycles
within
the
3–6-year
band
dominated
the
dynamics
of
the
disease
in
the
Guayana
region,
whereas
biennial
cycles
were
dominant
in
the
coastal
eco-region,
and
occasionally,
in
the
Amazonas.
Our
results
partially
agree
with
the
previous
works
of
Gabaldon
(1949)
and
Bouma
and
Dye
(1997)
on
the
existence
of
multi-annual
fluctu-
ations
in
malaria
incidence
in
Venezuela.
In
particular,
Bouma
and
Dye
(1997)
described
a
clear
5-year
cycle
previous
to
the
1990s
for
the
coastal
zone,
where
periodicity
showed
a
more
reg-
ular
pattern
than
in
the
interior
of
the
country.
In
contrast,
we
observed
significant
heterogeneity
for
the
period
1990–2010
in
the
malaria
periodicity
patterns
across
regions
and
species,
highlight-
ing
the
importance
of
analyzing
regional
spatial–temporal
disease
data.
Inter-annual
epidemic
cycles
of
malaria
have
been
previously
detected
and
accounted
for
by
extrinsic
and/or
intrinsic
factors
(e.g.,
Bouma
and
Dye,
1997;
Chowell
et
al.,
2009;
Gagnon
et
al.,
2002;
Hay
et
al.,
2000;
Pascual
et
al.,
2008;
Poveda
et
al.,
2001).
Here,
we
focused
mainly
on
the
effects
of
climatic
variability.
At
the
regional
scale,
ENSO
has
been
the
most
commonly
studied
driver
of
cyclic
climate
phenomena
in
human
diseases
(Kovats
et
al.,
2003).
Indeed,
we
found
that
the
inter-annual
variability
of
malaria
cases
was
coherent
with
that
of
SSTs
(ENSO),
mainly
at
temporal
scales
within
the
3–6
year
bands.
These
results
suggest
that
climate
inter-
annual
variability
has
played
a
role
in
the
multi-annual
cycles
of
malaria
in
Venezuela
for
the
last
20
years.
Although
the
malaria-
ENSO
coupling
appears
more
complicated
than
that
reported
by
Bouma
and
Dye
(1997),
we
should
be
cautious
when
we
compare
recent
associations
with
previously
reported
ones
(before
1990)
given
that,
first,
we
are
working
with
different
epidemiological
settings,
and
second,
ENSO
is
a
quasi-periodic
phenomenon
whose
influence
on
local
weather
is
not
continuous
in
time,
often
chang-
ing
on
longer
time
scales.
Indeed,
we
observed
that
the
effect
was
most
marked
during
the
interval
of
local
maxima
of
particular
El
Ni˜
no
events
(1991–1994,
1997–1998).
The
influence
of
ENSO
on
malaria
dynamics
was
further
supported
by
the
findings
that
an
increased
malaria
burden
followed
elevated
SSTs
(associated
with
El
Nino
conditions)
with
a
delay
of
9–12
months.
Similar
associa-
tions
were
previously
reported
for
Colombia
(Bouma
et
al.,
1997;
Poveda
et
al.,
2001),
India
(Bouma
and
van
der
Kaay,
1996),
and
Venezuela
(Bouma
and
Dye,
1997;
Gagnon
et
al.,
2002),
among
others.
Regarding
malaria’s
seasonality
in
each
region,
we
found
two
contrasting
patterns:
bimodal
(Amazonas
and
Sucre)
and
weak
(Bolívar)
seasonality.
In
the
first
pattern,
disease
does
not
seem
to
follow
the
observed
annual
cycle
of
rains.
Main
peaks
of
malaria
in
Amazonas
and
Sucre
were
observed
during
the
dry
or
transitional
periods,
whereas
rains
in
both
areas
were
mostly
concentrated
from
May
to
September–November.
Nevertheless,
we
observed
that
the
minor
rains
from
the
last
phase
of
the
season
play
a
critical
role
in
the
annual
variation
of
Plasmodium
cases
by
hav-
ing
a
positive
influence
on
the
main
malaria
peak.
In
contrast,
malaria
is
suppressed
in
the
period
with
most
rainfall,
especially
in
the
Amazonas
region.
Rains
promote
mosquito
breeding
sites;
but,
excessive
rainfall
can
destroy
aquatic
habitats
and
flush
out
the
mosquito
larvae.
In
particular,
wetlands
or
large
rivers
may
cause
such
wash
out
or
become
too
deep
during
months
with
high
precipitation.
Anopheles
aquasalis,
the
malaria
vector
in
the
coastal
eco-region,
is
mostly
associated
to
wetlands,
and
its
pre-
adult
populations
are
most
abundant
during
the
end
of
the
rainy
season
(October)
and
at
the
beginning
of
the
dry
season
(Grillet,
2000).
Anopheles
darlingi,
the
main
vector
in
Amazonas
(Magris
et
al.,
2007)
and
Bolívar
(Moreno
et
al.,
2007)
is
most
abundant
during
the
least
rainy
months
(October–November;
Berti
et
al.,
2008;
Moreno
et
al.,
2007),
periods
when
river
levels
(and
there-
fore,
the
main
aquatic
breeding
sites)
are
more
stable.
Therefore,
the
ecology
of
the
local
vectors
can
explain
why
the
last
rains
of
the
season
would
be
critical
for
the
annual
malaria
dynamics
in
given
areas.
In
the
Bolívar
region,
our
results
revealed,
first,
that
malaria
has
a
delayed
response
to
rainfall,
and
second,
that
the
minor
rains
of
the
season
are
also
critical
for
Plasmodium
species
dynamics
at
the
beginning
of
the
following
year.
The
almost
constant
annual
disease
incidence
(especially
P.
vivax
malaria)
is
mainly
explained
by
the
differences
in
seasonal
prevalence
and
relative
dominance
of
species
vectors
throughout
the
year.
In
this
region,
the
mining
activity
and
its
associated
process
of
forests
clearing
creates
sustained
favorable
conditions
for
the
breeding
of
species
such
as
An.
darlingi
and
An.
marajoara.
Moreno
et
al.
(2007)
have
shown
that
the
small
water-bodies
generated
by
the
min-
ing
activities
are
very
productive
and
provide
permanent
aquatic
M.-E.
Grillet
et
al.
/
Acta
Tropica
129 (2014) 52–
60 59
habitats
for
these
species
during
the
year,
especially
An.
mara-
joara.
How
ENSO
drives
the
long-term
periodicity
of
malaria
remains
unexplained.
The
most
obvious
and
plausible
pathway
for
the
influ-
ence
of
this
regional
climatic
driver
would
be
through
local
changes
in
rainfall
and
temperature.
It
would
follow
that
the
spatial
dif-
ferences
in
the
malaria
cycles
among
the
three
studied
regions
could
be
understood
based
on
the
particular
relations
of
ENSO
with
the
local
climate.
By
analyzing
this
pathway,
we
mainly
focused
on
rainfall
since
it
is
the
main
climate
variable
displaying
signifi-
cant
inter-annual
variability
in
Venezuela
(Pulwarty
et
al.,
1992)
and
as
we
previously
showed,
the
main
seasonal
driver
of
the
annual
dynamics
of
Plasmodium.
Rainfall
periodicity
corresponded
to
some
degree
to
the
observed
periodicity
in
ENSO,
but
interest-
ingly,
each
region
responded
to
particular
forcing
periods.
As
an
example,
rains
in
the
Amazonas
region
exhibited
variability
around
the
periods
of
2
and
4–6
years,
while
in
the
Bolívar
region,
rain-
fall
mainly
showed
cycles
at
the
3-year
period.
Malaria
periodicity
corresponded
to
some
degree
to
that
of
rainfall.
In
the
Amazonas
region,
malaria
exhibited
oscillations
of
periods
2
and
4,
while
in
the
Bolívar
region
it
did
at
the
3
and
6
periods.
Thus,
all
these
results
taken
together
suggest
that
rainfall
mediates
the
effect
of
ENSO
on
malaria
locally.
In
addition,
the
impact
of
the
El
Ni˜
no
on
local
rains
differed
regionally.
It
was
most
apparent
for
the
Bolívar
region
which
displayed
the
higher
negative
correlations
between
seasonal
rains
and
SSTs
over
several
time
lags.
In
contrast,
Amazonas
was
the
least
affected
region,
by
showing
lower
negative
rainfall
anomalies
during
just
a
few
months.
What
determines
whether
one
particular
region
responds
to
a
particular
forcing
period
that
may
not
be
the
strongest
is
an
open
question.
Some
explanations
could
be
found
at
the
regional
scale,
by
exploring
the
relation
between
ENSO
and
the
regional
rainfall
in
South-America
and
Venezuela.
Of
particular
interest
is
that
rainfall
is
more
directly
influenced
by
the
Atlantic
Inter-tropical
Convergence
Zone
(ITCZ)
in
Bolívar
and
Sucre
than
in
the
Amazonas
(Pulwarty
et
al.,
1992).
Although
there
is
an
expected
coherent
pattern
of
climatic
and
hydrological
anomalies
in
tropical
South-America
during
phases
of
ENSO,
regional
differences
in
tim-
ing,
amplitude,
sign
and
intensity
of
the
event
have
been
previously
reported
within
Venezuela
(Pulwarty
et
al.,
1992)
and
Colombia
(Poveda
et
al.,
2001).
Lastly,
the
biennial
oscillation
detected
in
the
malaria
series
of
Sucre
and
Amazonas
opens
other
questions
on
the
probable
influence
of
the
quasi-biennial
oscillation
(QBO),
which
is
another
large-scale
ocean-atmospheric
phenomenon
affecting
the
hydro-climatology
of
South-America
at
inter-annual
timescales
(Poveda
and
Mesa,
1997).
ENSO
itself
also
exhibits
an
important
quasi-biennial
(QB)
component,
which
was
detected
in
our
analy-
ses.
How
the
El
Ni˜
no
and
related
rainfall
patterns
affect
disease
transmission
is
also
a
matter
of
conjecture.
We
found
that
during
El
Ni˜
no
years,
above-normal
(positive)
SST
anomalies
were
associ-
ated
with
below-normal
(negative)
rainfall
anomalies
in
the
three
regions
as
it
was
expected
based
on
Poveda
and
Mesa
(1997).
Sur-
prisingly,
positive
seasonal
anomalies
in
rains
were
also
described
from
July
to
December
in
the
last
phase
of
El
Ni˜
no+1 year
result-
ing
in
wetter
months
compared
to
similar
periods
for
the
non-Ni˜
no
years.
These
findings
suggest
that
the
impact
of
the
El
Ni˜
no
event
on
local
rains
can
be
complex
(with
both
negative
and
positive
effects).
Also,
there
was
a
differential
impact
of
ENSO
on
rains,
with
an
effect
mostly
at
the
end
of
the
rainy
season.
Bouma
and
Dye
(1997)
did
not
observe
these
post-El
Ni˜
no
anomalous
rains
in
Venezuela,
even
though
Poveda
and
Mesa
(1997)
describe
the
climate
conditions
under
which
anomalous
rains
can
develop
during
the
El
Ni˜
no+1
in
Venezuela.
Some
hypotheses
can
be
proposed.
As
we
observed,
the
last
rains
of
the
season
were
associated
with
the
main
peak
of
malaria
at
the
beginning
of
the
following
year.
Then,
the
num-
ber
of
cases
in
this
first
outbreak
influenced
in
turn
disease
burden
later
in
the
year.
Thus,
higher
than
expected
anomalous
rains
in
this
phase
of
the
season
could
increase
the
size
of
the
main
malaria
peak,
and
the
number
of
cases
in
the
following
year.
Indeed,
malaria
epi-
demics
were
detected
several
months
after
these
anomalous
wetter
months.
Otherwise,
these
anomalous
rains
could
promote
bet-
ter
breeding
conditions
for
vector
populations,
and
consequently,
more
malaria
cases
several
months
afterwards.
As
suggested
by
Gabaldon
(1949),
larger
seasonal
abundances
of
the
vectors
would
explain
the
interannual
cycles
of
malaria.
Future
studies
should
address
these
hypotheses.
In
particular,
the
role
of
the
positive
anomalous
rains
during
the
El
Ni˜
no+1 event
can
be
elucidated
by
using
a
more
extensive
time-frame
(longer
malaria
and
climate
series).
One
alternative
conjecture
would
be
that
when
aquatic
habitats
are
re-established
after
dry
years,
mosquito
populations
can
increase
to
higher
than
usual
numbers
because
predators
of
larvae
have
been
reduced
(Bouma
and
Dye,
1997;
Grillet
et
al.,
2002).
Another
hypothesis
would
invoke
additional
ENSO-related
climate
variables
such
as
changes
in
wind
direction
(Poveda
and
Mesa,
1997)
which
might
affect,
with
a
delay
of
several
months,
the
dispersal
and
distribution
of
malaria
vectors
(Bouma
and
Dye,
1997).
Also,
a
reduction
in
transmission
during
a
dry
year
is
likely
to
reduce
population
immunity
and
hence
increase
the
size
of
the
vul-
nerable
population
in
the
following
transmission
season
(Bouma
and
Dye,
1997).
Further
research
into
the
ecology
of
Anopheles
and
Plasmodium
species
are
needed
to
fully
understand
the
causal
rela-
tionships
between
ENSO
and
malaria
transmission.
Our
results
further
showed
that
disease
periodicity
can
be
modified
by
local
malaria
control
measures
(Anderson
and
May
(1992).
Irregular
dynamics
in
the
coastal
eco-region
were
proba-
bly
affected
by
the
falling
of
P.
vivax
cases
after
2003
as
a
result
of
massive
control
efforts
(Cáceres,
2008).
Although
the
similar-
ity
in
malaria
periodicity
across
regions
supports
the
hypothesis
of
a
regional
environmental
driver,
our
findings
also
indicated
differences
between
Plasmodium
species.
We
observed
stronger
associations
between
the
seasonal
malaria
peaks
for
P.
vivax
than
for
P.
falciparum,
a
pattern
suggesting
a
role
of
relapses,
a
fea-
ture
of
the
life-cycle
of
the
former
parasite
but
not
the
latter.
We
also
detected
irregular
dynamics
for
P.
falciparum
in
Bolívar
and
for
P.
vivax
in
the
Amazonas
that
we
were
not
able
to
explain.
Finally,
the
significant
upward
trend
of
malaria
during
the
last
10
years
in
the
Bolívar
region
raises
the
question
of
whether
this
epidemic
has
contributed
to
the
changes
in
P.
vivax
periodicity
during
the
study
period.
Social
(mining
activities),
epidemiolog-
ical
and
environmental
changes
(deforestation)
occurring
lately
in
southern
Venezuela
could
explain
this
malaria
growth.
Thus,
it
would
be
interesting
to
know
if
epidemiological
and
demo-
graphic
processes
account
for
the
different
periodicities
of
malaria
cases.
Future
research
should
investigate
how
climate
forcing
inter-
acts
with
relapses,
host
immunity,
and
demographic
and
land-use
change
to
determine
the
population
dynamics
of
the
disease
in
Venezuela.
Conflict
of
interest
The
authors
declare
they
have
not
competing
interests
Acknowledgements
We
acknowledge
the
logistic
support
provided
by
the
Minis-
ter
of
Health
of
Venezuela
(A.
Girón,
A.
Martinez
and
M.
Herrera),
Y.
Rangel,
J.
Moreno,
M.
Magris,
A.
Mejía,
E.
Navarro,
O.
Noya,
A.
Zorrilla,
F.
del
Ventura,
F.
Marichal,
N.
Moncada
and
V.
Behm.
J.
Williams,
R.
Barrera,
M.
Pascual
and
two
anonymous
review-
ers
made
valuable
comments
on
the
manuscript.
The
study
was
supported
by
the
Venezuelan
“Fondo
Nacional
de
Investigaciones
60 M.-E.
Grillet
et
al.
/
Acta
Tropica
129 (2014) 52–
60
Científicas”
(FONACIT,
UC-2008000911-3)
and
the
Council
for
Sci-
ences
and
Humanities
Development
(CDCH-PG-0382182011).
References
Anderson,
R.M.,
May,
R.M.,
1992.
Infectious
Diseases
of
Humans:
Dynamics
and
Control.
Oxford
University
Press,
Oxford,
United
Kingdom.
Berti-Moser,
J.,
Gonzales-Rivas,
J.,
Navarro,
E.,
2008.
Fluctuaciones
estacionales
y
temporales
de
la
densidad
larvaria
de
Anopheles
darlingi
root
(Diptera:
Culicidae)
y
familias
de
insectos
asociados
al
hábitat
en
El
Granzón,
Parroquia
San
Isidro,
Municipio
Sifones
del
estado
Bolívar,
Venezuela.
Bol.
Dir.
Malariol.
Salud
Amb.
48,
177–189.
Bouma,
M.J.,
Dye,
C.,
1997.
Cycles
of
malaria
associated
with
El
Ni˜
no
in
Venezuela.
J.A.M.A.
278,
1772–1774.
Bouma,
M.J.,
van
der
Kaay,
H.J.,
1996.
El
Ni˜
no
Southern
Oscillation
and
the
historic
malaria
epidemics
on
the
Indian
subcontinent
and
Sri-Lanka.
Trop.
Med.
Int.
Health.
1,
86–96.
Bouma,
M.J.,
Poveda,
G.,
Rojas,
W.,
Chavasse,
D.,
Qui˜
nones,
M.,
Cox,
J.,
Patz,
J.,
1997.
Predicting
high-risk
years
for
malaria
in
Colombia
using
parameters
of
El
Ni˜
no
Southern
Oscillation.
Trop.
Med.
Int.
Health
2,
1122–1127.
Cáceres,
J.L.,
2008.
Malaria
antes
y
después
de
la
cura
radical
masiva
en
el
estado
Sucre,
Venezuela.
Bol.
Dir.
Malariol.
San.
Amb.
48,
83–90.
Cáceres,
J.L.,
2011.
La
malaria
en
el
estado
Bolívar,
Venezuela:
10
a˜
nos
sin
control.
Bol.
Dir.
Malariol.
San.
Amb.
51,
207–214.
Cazelles,
B.,
Chavez,
M.,
Constantin
de
Magny,
G.,
Geguan,
J.F.,
Hales,
S.,
2007.
Time-
dependent
spectral
analysis
of
epidemiological
time
series
with
wavelets.
J.
R.
Soc.
Interface
4,
625–636.
Chowell,
G.,
Munayco,
C.V.,
Escalante,
A.A.,
McKenzie,
E.,
2009.
The
spatial
and
tem-
poral
patterns
of
falciparum
and
vivax
malaria
in
Perú:
1994–2006.
Mal.
J.
8,
1–19.
Fedorov,
A.V.,
Philander,
G.,
2000.
Is
El
Ni˜
no
changing?
Science
288,
1997–2001.
Gabaldon,
A.,
1949.
The
nation-wide
campaign
against
malaria
in
Venezuela.
Trans.
R.
Soc.
Trop.
Med.
Hyg.
43,
113–164.
Gagnon,
A.S.,
Smoyer-Tomic,
K.E.,
Bush,
A.,
2002.
The
El
Ni˜
no
southern
oscil-
lation
and
malaria
epidemics
in
South
America.
Int.
J.
Biometeorol.
46,
81–89.
Grillet,
M.E.,
2000.
Environmental
factors
associated
with
the
spatial
and
tem-
poral
distribution
of
Anopheles
aquasalis
and
Anopheles
oswaldoi
in
wetlands
of
an
endemic
malaric
area
in
north-eastern
Venezuela.
J.
Med.
Entomol.
37,
231–238.
Grillet,
M.E.,
Barrera,
R.,
Martínez,
J.,
Berti,
J.,
Fortin,
M.J.,
2010.
Disentangling
the
effect
of
local
and
global
spatial
variation
on
a
mosquito-borne
infec-
tion
in
a
neotropical
heterogeneous
environment.
Am.
J.
Trop.
Med.
Hyg.
82,
194–201.
Grillet,
M.E.,
Legendre,
P.,
Borcard,
D.,
2002.
Community
structure
of
Neotropical
wetland
insects
in
Northern
Venezuela.
II.
Habitat
type
and
environmental
fac-
tors.
Arch.
Hydrobiol.
155,
437–453.
Hay,
S.I.,
Myers,
M.F.,
Burke,
D.S.,
Vaughn,
D.W.,
Endy,
T.,
Ananda,
N.,
Shanks,
G.D.,
Snow,
R.W.,
Rogers,
D.J.,
2000.
Etiology
of
inter-epidemic
periods
of
mosquito-
borne
disease.
Proc.
Natl.
Acad.
Sci.
U.
S.
A.
97,
9335–9339.
Huber,
O.,
1995.
Geographical
and
physical
features.
In:
Steyermark,
J.A.,
Berry,
P.E.,
Holst,
B.K.
(Eds.),
Flora
of
the
Venezuelan
Guayana,
Vol.
1.
Missouri
Botanical
Garden.
St.
Louis
&
Timber
Press,
Oregon,
pp.
1–61.
Kovats,
R.S.,
Bouma,
M.J.,
Hajat,
S.,
Worrall,
E.,
Haines,
A.,
2003.
El
Ni˜
no
and
health.
Lancet
362,
1481–1489.
Kulldorff,
M.,
1997.
A
spatial
scan
statistic.
Comm.
Stat.
Theor.
Meth.
26,
1481–1496.
Magris,
M.,
Rubio-Palis,
Y.,
Menares,
C.,
Villegas,
L.,
2007.
Vector
bionomics
and
malaria
transmission
in
the
Upper
Orinoco
River,
Southern
Venezuela.
Mem.
Inst.
Oswaldo
Cruz
102,
303–311.
Metzger,
W.G.,
Giron,
A.M.,
Vivas-Martínez,
S.,
González,
J.,
Charrasco,
A.J.,
Mord-
müller,
B.G.,
Magris,
M.,
2009.
A
rapid
malaria
appraisal
in
the
Venezuelan
Amazon.
Malar.
J.
8,
291.
Moreno,
J.,
Rubio-Palis,
Y.,
Paez,
E.,
Perez,
E.,
Sanchez,
V.,
2007.
Abundance,
biting
behavior
and
parous
rate
of
anopheline
mosquito
species
in
relation
to
malaria
incidence
in
gold-mining
areas
in
southern
Venezuela.
Med.
Vet.
Entomol.
21,
339–349.
NOAA,
2011.
National
Centers
for
Environmental
Prediction
Center
(Sea
Sur-
face
Temperature
Historical
Archives),
http://www.cpc.ncep.noaa.gov/
products/analysis
monitoring/ensostuff/ensoyears.shtml
(Accessed
date:
08/02/2011).
Pascual,
M.,
Cazelles,
B.,
Bouma,
M.J.,
Chaves,
L.F.,
Koelle,
K.,
2008.
Shifting
patterns:
malaria
dynamics
and
rainfall
variability
in
an
African
highland.
Proc.
Biol.
Sci.
75,
123–132.
Poveda,
G.,
Mesa,
O.J.,
1997.
Feedbacks
between
hydrological
processes
in
tropical
South
America
and
large-scale
ocean-atmospheric
phenomena.
J.
Climate
10,
2690–2702.
Poveda,
G.,
Rojas,
W.,
Quinones,
M.L.,
Velez,
I.D.,
Mantilla,
R.I.,
Ruiz,
D.,
Zuluaga,
J.S.,
Rua,
G.L.,
2001.
Coupling
between
annual
and
ENSO
timescales
in
the
malaria
climate
association
in
Colombia.
Environ.
Health
Perspect.
109,
489–493.
Pulwarty,
R.S.,
Barry,
R.G.,
Riehl,
H.,
1992.
Annual
and
seasonal
patterns
of
rainfall
variability
over
Venezuela.
Erdkunde
46,
273–289.
Rubio-Palis,
Y.,
Zimmerman,
R.H.,
1997.
Ecoregional
classification
of
malaria
vectors
in
the
neotropics.
J.
Med.
Entomol.
34,
499–510.
Stresman,
G.H.,
2010.
Beyond
temperature
and
precipitation:
ecological
risk
factors
that
modify
malaria
transmission.
Acta
Trop.
116,
167–172.
Torrence,
C.,
Compo,
G.P.,
1998.
A
practical
guide
to
wavelet
analysis.
Bull.
Am.
Meteorol.
Soc.
79,
61–78.
Trenberth,
K.E.,
1999.
The
definition
of
El
Ni˜
no.
Bull.
Amer.
Met.
Soc.
78,
2771–2777.
World
Health
Organization,
2012.
World
Malaria
Report
2012.
World
Health
Orga-
nization,
Geneva.