current as of October 27, 2010.
Online article and related content
http://jama.ama-assn.org/cgi/content/full/304/15/1675
. 2010;304(15):1675-1683 (doi:10.1001/jama.2010.1507) JAMA
Maria Makrides; Robert A. Gibson; Andrew J. McPhee; et al.
Children: A Randomized Controlled Trial
Maternal Depression and Neurodevelopment of Young
Effect of DHA Supplementation During Pregnancy on
Supplementary material
http://jama.ama-assn.org/cgi/content/full/304/15/1675/DC1
eTables
Correction Contact me if this article is corrected.
Citations
Contact me when this article is cited.
This article has been cited 2 times.
Topic collections
Contact me when new articles are published in these topic areas.
Outcomes
Health; Pregnancy and Breast Feeding; Randomized Controlled Trial; Prognosis/
Malnutrition; Pediatrics; Child Development; Psychiatry; Depression; Women's
Nutritional and Metabolic Disorders; Lipids and Lipid Disorders; Nutrition/
http://pubs.ama-assn.org/misc/permissions.dtl
permissions@ama-assn.org
Permissions
http://jama.com/subscribe
Subscribe
reprints@ama-assn.org
Reprints/E-prints
http://jamaarchives.com/alerts
Email Alerts
at University of California - San Francisco on October 27, 2010 www.jama.comDownloaded from
ORIGINAL CONTRIBUTION
Effect of DHA Supplementation
During Pregnancy on Maternal Depression
and Neurodevelopment of Young Children
A Randomized Controlled Trial
Maria Makrides, BSc, BND, PhD
Robert A. Gibson, BSc, PhD
Andrew J. McPhee, MBBS
Lisa Yelland, BSc
Julie Quinlivan, MBBS, PhD
Philip Ryan, MBBS, BSc
and the DOMInO Investigative Team
EPIDEMIOLOGICAL INVESTIGA-tions from theUnited States andEurope demonstrate that higherintakes of n-3 long-chain poly-
unsaturated fatty acids (LCPUFA) from
fish and seafood during pregnancy are
associated with a reduced risk of de-
pressive symptoms in the postnatal pe-
riod,1 as well as improved developmen-
tal outcomes in the offspring.2,3 Of the
n-3 LCPUFA, it is hypothesized that
docosahexaenoic acid (DHA) may be
responsible for the observed associa-
tions based on estimates of dietary re-
quirements during pregnancy and the
results of experimental animal stud-
ies.4 However, n-3 LCPUFA interven-
tion trials in human pregnancy have re-
portedmixed results and have not been
conclusive largely because of method-
ological limitations. Studies focused on
perinatalmood have had open-label de-
signs, small sample sizes, or large at-
trition, andmost did not analyze by in-
tention-to-treat.5 Similarly, trials
focused on the developmental out-
comes of the children have made post-
See also p 1717 and Patient Page.
Author Affiliations: Child Nutrition Research Centre,
Women’s and Children’s Hospital, Flinders Medical Cen-
tre, and Women’s and Children’s Health Research In-
stitute, Adelaide (Drs Makrides and Gibson); Schools
of Pediatrics and Reproductive Health (Dr Makrides),
Agriculture, Food and Wine (Dr Gibson), and Popula-
tion Health and Clinical Practice (Dr Ryan and Ms Yel-
land), University of Adelaide, Adelaide; Department of
Neonatal Medicine, Women’s and Children’s Hospi-
tal, Adelaide (Dr McPhee); and School of Medicine,
University of Notre Dame Australia, and Sunshine Hos-
pital, Melbourne (Dr Quinlivan), Australia.
The DOMInO Investigative Teammembers and their
author affiliations are listed at the end of this article.
Corresponding Author: Maria Makrides, BSc, BND,
PhD, Child Nutrition Research Centre, Women’s and
Children’s Health Research Institute, Women’s and
Children’s Hospital, 72 King William Rd, North Ad-
elaide SA 5006, Australia (maria.makrides@health.sa
.gov.au).
Context Uncertainty about the benefits of dietary docosahexaenoic acid (DHA) for
pregnant women and their children exists, despite international recommendations that
pregnant women increase their DHA intakes.
Objective To determine whether increasing DHA during the last half of pregnancy
will result in fewer women with high levels of depressive symptoms and enhance the
neurodevelopmental outcome of their children.
Design, Setting, and Participants A double-blind, multicenter, randomized con-
trolled trial (DHA to Optimize Mother Infant Outcome [DOMInO] trial) in 5 Austra-
lian maternity hospitals of 2399 women who were less than 21 weeks’ gestation with
singleton pregnancies and who were recruited between October 31, 2005, and Janu-
ary 11, 2008. Follow-up of children (n=726) was completed December 16, 2009.
Intervention Docosahexaenoic acid–rich fish oil capsules (providing 800 mg/d of
DHA) or matched vegetable oil capsules without DHA from study entry to birth.
Main Outcome Measures High levels of depressive symptoms in mothers as indi-
cated by a score of more than 12 on the Edinburgh Postnatal Depression Scale at 6 weeks
or 6 months postpartum. Cognitive and language development in children as assessed
by the Bayley Scales of Infant and Toddler Development, Third Edition, at 18 months.
Results Of 2399 women enrolled, 96.7% completed the trial. The percentage of
women with high levels of depressive symptoms during the first 6 months postpar-
tum did not differ between the DHA and control groups (9.67% vs 11.19%; adjusted
relative risk, 0.85; 95% confidence interval [CI], 0.70-1.02; P=.09). Mean cognitive
composite scores (adjusted mean difference, 0.01; 95% CI, −1.36 to 1.37; P=.99) and
mean language composite scores (adjusted mean difference, −1.42; 95% CI, −3.07
to 0.22; P=.09) of children in the DHA group did not differ from children in the con-
trol group.
Conclusion The use of DHA-rich fish oil capsules compared with vegetable oil cap-
sules during pregnancy did not result in lower levels of postpartum depression in moth-
ers or improved cognitive and language development in their offspring during early
childhood.
Trial Registration anzctr.org.au Identifier: ACTRN12605000569606
JAMA. 2010;304(15):1675-1683 www.jama.com
©2010 American Medical Association. All rights reserved. (Reprinted) JAMA, October 20, 2010—Vol 304, No. 15 1675
at University of California - San Francisco on October 27, 2010 www.jama.comDownloaded from
randomization exclusions,6,7 had high
attrition rates, and lacked power.6-9 De-
spite the paucity of evidence, recom-
mendations exist to increase intake of
DHA in pregnancy,4,10 and the nutri-
tional supplement industry success-
fullymarkets prenatal supplementswith
DHA to optimize brain function of
mother and infant. Before DHA supple-
mentation in pregnancy becomeswide-
spread, it is important to knownot only
if there are benefits, but also of any risks
for either themother or child. TheDHA
to Optimize Mother Infant Outcome
(DOMInO) trial was designed primar-
ily to assess whether DHA supplemen-
tation during the last half of preg-
nancy reduced the risk of depressed
maternal mood during the postpar-
tum period and improved early cogni-
tive development in the offspring.
METHODS
Study Design
We conducted a double-blind, multi-
center, randomized controlled trial in
5 Australian perinatal centers. Ap-
proval was granted by the local insti-
tutional review boards (human re-
search ethics committees) of each center
and written informed consent was ob-
tained from each participant. An inde-
pendent serious adverse event commit-
tee reviewed all deaths, admissions to
level III care, and major congenital
abnormalities.
Women with singleton pregnancies
at less than 21weeks’ gestationwere ap-
proached by study research assistants
while attending routine antenatal ap-
pointments to participate in the trial.
Women were excluded if they were al-
ready taking a prenatal supplementwith
DHA, their fetus had a knownmajor ab-
normality, they had a bleeding disor-
der in which tuna oil was contraindi-
cated, were taking anticoagulant
therapy, had a documented history of
drug or alcohol abuse, were participat-
ing in another fatty acid trial, were un-
able to give written informed consent,
or if English was not the main lan-
guage spoken at home. Recruitment for
the trial began October 31, 2005, and
ended January 11, 2008.
Randomization and Trial Entry
Women were randomly assigned a
unique study number and treatment
group allocation through a computer-
driven telephone randomization ser-
vice according to an independently gen-
erated randomization schedule, with
balanced variable-sized blocks. Strati-
fication was by center and parity (first
birth vs subsequent birth). Baseline
characteristics, includingmaternal age,
medical diagnosis of previous or cur-
rent depression, current treatment for
depression, social support using theMa-
ternal Social Support Index,11 weight,
highest level of education, occupa-
tion, and smoking status, were re-
corded.
Dietary Treatments
Women allocated to the DHA group
were asked to consume three 500-
mg/d capsules of DHA-rich fish oil con-
centrate, providing 800 mg/d of DHA
and 100mg/d of eicosapentaenoic acid
(EPA, 20:5n-3; Incromega 500 TG,
Croda Chemicals, East Yorkshire, En-
gland); andwomen in the control group
were asked to take three 500-mg/d veg-
etable oil capsules without DHA. The
dose of 800 mg/d was chosen because
this was above the estimated thresh-
old associated with lower risk of de-
pressed maternal mood and higher
scores on developmental outcomes of
children,1 as well as being consistent
with the estimated requirement to cover
97% of the population.12 The veg-
etable oil capsules contained a blend of
3 nongenetically modified oils (rape-
seed, sunflower, and palm) in equal
proportions. This blend of oils was de-
signed to match the polyunsaturated,
monounsaturated, and saturated fatty
acid profile of the average Australian
diet.13Womenwere asked to take their
assigned capsules daily, from study en-
try until birth of their child. All cap-
sules were similar in size, shape, and
color and donated by Efamol, Surrey,
England.
Treatment Phase Monitoring
During the intervention period, trial as-
sistants telephonedwomen 2weeks af-
ter enrollment (approximately 22
weeks’ gestation) and at 28 and 36
weeks’ gestation to document adverse
gastrointestinal or bleeding events and
to monitor and encourage adherence.
The concentration of DHA in cord
bloodwasmeasured using capillary gas
chromatography14 to provide an inde-
pendent biomarker of adherence. An-
tenatal hospitalizations, antenatal hem-
orrhage, pregnancy andbirth outcomes,
and postpartum hemorrhage were re-
corded from a review of medical rec-
ords.
Outcome Assessments
Women completed a self-adminis-
tered Edinburgh Postnatal Depression
Scale (EPDS) at 6 weeks and 6months
postpartum, and the primary mater-
nal outcomewas a high level of depres-
sive symptoms documented as a score
ofmore than 12 on the EPDS at 6weeks
or 6 months postpartum. Although a
high EPDS score cannot in itself con-
firm a diagnosis of depression, a score
of more than 12 is widely used to in-
dicate a probable depressive disorder.
Validation studies indicate high sensi-
tivity (68%-95%) and high specificity
(78%-96%) of the EPDS against a clini-
cal psychiatric diagnosis of depres-
sion.15-17 Women with a score of more
than 12 on the EPDS were referred to
their general practitioners formore for-
mal medical assessment. Secondary
analyses compared the percentage of
women medically diagnosed with de-
pression or receiving treatment for de-
pression, as reported by women dur-
ing pregnancy and at 6 weeks and 6
months postpartum, between the 2
groups.
The primary childhood outcome of
neurodevelopment at 18 months was
assessed by 1 of 4 study psychologists
using the Cognitive and Language
Composite Scales of the Bayley Scales
of Infant and Toddler Development,
Third Edition (BSID-III). The cogni-
tive scale evaluates abilities, such as sen-
sorimotor development, exploration
and manipulation, object relatedness,
concept formation,memory, and simple
problem solving, and the language scale
DHA SUPPLEMENTATION DURING PREGNANCY AND MATERNAL DEPRESSION
1676 JAMA, October 20, 2010—Vol 304, No. 15 (Reprinted) ©2010 American Medical Association. All rights reserved.
at University of California - San Francisco on October 27, 2010 www.jama.comDownloaded from
is a composite of receptive communi-
cation (verbal comprehension, vocabu-
lary) and expressive communication
(babbling, gesturing, and utterances).
Themotor scale, which evaluates both
gross and fine motor functioning, as
well as the parental report scales of so-
cial-emotional behavior and adaptive
behavior were assessed as secondary
outcomemeasures. The raw scores for
each of the scales are standardized to a
mean of 100 with an SD of 15 (range,
50-150). The standardized scores were
also classified into the categories of ac-
celerated performance (�115), within
normal limits (85-115), and delayed
performance (�85).
All 96 preterm children and 630 ran-
domly selected term children fromAd-
elaide, Australia, were chosen for BSID-
III assessment (N=726). The families
and trial staff did not knowwhich chil-
dren had been randomly selected for
follow-up until the infants were 12
months old. At this time, trial assis-
tants in each center were supplied with
birthday cards for all children, and those
selected for follow-up were informed
and invited for a BSID-III appoint-
ment. The last BSID-III assessment was
completed on December 16, 2009.
Sample Size and Statistical
Analysis
Epidemiological data suggest a 7% to
8% absolute reduction (from approxi-
mately 16% to 9%) in the prevalence
of high levels of depressive symptoms
when n-3 LCPUFA intake is increased
from the typically low-level intakes
commonly observed in Westernized
diets to more than 1 g/d.1 Although
these data were derived from a well-
controlled cohort study, we expected
that any effect size of a DHA-rich in-
terventionwould be smaller because of
possible residual confounding. We
therefore powered our trial to detect an
absolute reduction of 4.2% (from16.9%
to 12.7%) in depressive symptomswith
80%power (�=.05), requiring a sample
size of 1121 women per group. The
control rate of depressive symptoms
was estimated from Australian popu-
lation data,18 which was also consis-
tent with the epidemiological data from
England.1 We planned to enroll 2280
women in total, allowing for 2% loss to
follow-up.
A minimum clinically meaningful
difference in developmental scores is
considered to be of the order of 4
points.19 Studies showing differences
between nutritional or environment in-
terventions of 4 to 5 points or greater
have been catalysts for changes inhealth
policy.20,21 To detect a difference of 5
points between groups (mean [SD], 100
[15])with 80%power (�=.05) for boys
and girls separately, we required a total
sample size of 572 children.We there-
fore randomly sampled 630 term chil-
dren such that half were male, allow-
ing for 10% loss to follow-up. The
selection process occurred between
birth and 1 year. All children born pre-
termwere included in the follow-up to
enable modeling of the effect of DHA
supplementation in pregnancy on all
children; those born preterm being
more nutritionally and more develop-
mentally vulnerable than children born
at term.22
All analyses were performed accord-
ing to the intention-to-treat principle.
Multiple imputation was used to deal
with missing data (outcomes and co-
variates) and create 50 complete data
sets for analysis. Sensitivity analyses
were performedon the original data and
on imputed data using different seeds
for imputation and different imputa-
tion models. All produced similar re-
sults; therefore, we reported the re-
sults of the imputed analyses.
Continuous outcomes were ana-
lyzed by using linear regression mod-
els, following log transformationswhere
appropriate, with treatment effects ex-
pressed asmeandifferences. Binary out-
comes were analyzed using log bino-
mial regressionmodels, with treatment
effects expressed as relative risks (RRs)
or Fisher exact test for rare outcomes.
Time-to-event outcomes were ana-
lyzed by using stratified log-rank tests.
For outcomes measured at multiple
time points, dependence was ac-
counted for using generalized estimat-
ing equations. Models initially in-
cluded a treatment� time interaction.
Separate estimates of treatment effect
are presented at each time point if the
interaction was significant or if sepa-
rate estimateswere prespecified.Where
the interactionwas not significant, this
was removed from the model and an
overall estimate of treatment effect is
presented. Analysis of the primaryma-
ternal depression outcome was per-
formed on all women and on the sub-
group with a previous or current
diagnosis of depression at trial entry.
Outcomes derived from the BSID-III as-
sessment took into account both the
sampling design and probability
weights, calculated as the inverse of the
probability of selection. A priori sec-
ondary analyses were performed to test
for effect modification by sex and re-
sults are presented both overall and by
sex, because previous studies suggest
that boys and girls may respond differ-
ently to DHA supplementation.
Both unadjusted and adjusted analy-
ses were performed, with adjustment
for the stratification variables, center,
and parity, as well as any prespecified
potential confounders. Statistical sig-
nificance was assessed at the 2-sided
P� .05 level. No adjustment wasmade
formultiple comparisons and results for
secondary outcomes should be inter-
preted with caution unless they are
highly significant. Analyses were per-
formed by using SAS version 9.2 (SAS
Institute Inc, Cary,NorthCarolina) and
Stata Release 11 (Statacorp LP, Col-
lege Station, Texas).
RESULTS
The number of women screened and
assessed for the trial, randomly
assigned to receive either DHA or con-
trol supplementation, are shown in
the FIGURE. A total of 2399 women
were enrolled and adequate data for
the analysis of the primary outcome
were available for 2320 women
(97.3% in the DHA group and 96.1%
in the control group). In addition,
694 ch i ld ren (95 .6% of those
selected for follow-up) were assessed
at 18 months. Other outcomes and
covariates were generally available
DHA SUPPLEMENTATION DURING PREGNANCY AND MATERNAL DEPRESSION
©2010 American Medical Association. All rights reserved. (Reprinted) JAMA, October 20, 2010—Vol 304, No. 15 1677
at University of California - San Francisco on October 27, 2010 www.jama.comDownloaded from
for more than 95% of participants.
The demographic and clinical char-
acteristics of the women at random-
ization were comparable between the
2 groups, both overall (TABLE 1) and
for the subset assessed at 18 months
(eTable 1). The families of children se-
lected for follow-up also had similar
baseline characteristics to those of all
other children in the trial (eTable 2).
Docosahexaenoic acid concentration in
the plasma phospholipids of cord blood
from women in the high-DHA group
was greater than control (median,
7.22%vs 6.09% total phospholipid fatty
acids, P� .001 based on a comparison
ofmean log-transformed values). At 28
weeks’ gestation, 35.6% of mothers re-
ported that they had not missed tak-
ing any capsules and an additional
34.5% reported that they hadmissed 3
or less per week (from a total of 21 cap-
sules per week). Fewer than 2% of
mothers in each group chose not to take
any capsules.
Efficacy
The percentage of women reporting
high levels of depressive symptoms
(EPDS score �12) during the first 6
months postpartum did not differ be-
tween the DHA and control groups
(9.67% vs 11.19%; adjusted RR, 0.85;
95% CI, 0.70-1.02; P=.09) (TABLE 2).
Depressive symptomsweremore com-
mon among womenwith a previous or
current diagnosis of depression at trial
entry but did not differ between groups.
The percentage of women with a new
medical diagnosis for depression dur-
ing the trial or a diagnosis requiring
treatment also did not differ between
groups.
Mean cognitive scores of children
from women allocated to the DHA
group did not differ from mean scores
of children of