Intakes and breast-milk concentrations of essential fatty acids are low among
Bangladeshi women with 24–48-month-old children
Elizabeth A. Yakes1,2*, Joanne E. Arsenault1, M. Munirul Islam3, Mohammad B. Hossain3,
Tahmeed Ahmed3, J. Bruce German4, Laura A. Gillies4, Ahmed Shafiqur Rahman3, Christiana Drake2,5,
Kazi M. Jamil3, Bess L. Lewis6and Kenneth H. Brown1,2
1Department of Nutrition, University of California, One Shields Avenue, Davis, CA 95616, USA
2Graduate Group in Epidemiology, University of California, Davis, CA 95616, USA
3International Centre for Diarrhoeal Disease Research (ICDDR,B), Dhaka, Bangladesh
4Department of Food Science and Technology, University of California, Davis, CA 95616, USA
5Department of Statistics, University of California, Davis, CA 95616, USA
6International Agricultural Development Graduate Group, University of California, Davis, CA 95616, USA
(Received 17 May 2010 – Revised 26 October 2010 – Accepted 1 November 2010)
Maternal fat intake and adipose reserves are major sources of PUFA during lactation. The present study examined the cross-sectional
relationship between prolonged breast-feeding and maternal BMI, assessed adequacy of fat intake among lactating and non-lactating
mothers of children 24–48 months of age and determined breast-milk fatty acid composition. Multi-stage sampling was used to select a
representative sample of mothers from two rural districts in Bangladesh (n 474). Dietary data were collected during two non-consecutive
24h periods via 12h in-home daytime observations and recall. The National Cancer Institute method for episodically consumed foods was
used to estimate usual intake distributions. Breast milk samples were collected from ninety-eight women, and breast-milk fatty acid methyl
esters were quantified using GC. Approximately 42% of lactating v. 26% of non-lactating mothers were underweight (BMI ,18·5kg/m2;
P¼0·0003). The maternal diet was low in total fat (approximately 8% of mean total energy) and food sources of PUFA, including oil and
animal source foods, resulting in a low estimated mean total consumption of PUFA (5·1g/d). Almost all women were estimated to consume
less than the recommended intake levels for total fat, total PUFA, a-linolenic acid (ALA) and DHA. Median breast-milk linoleic acid (8·5%
weight) and ALA (0·2%) concentrations were among the lowest reported in the literature, in contrast with arachidonic acid (0·5%) and
DHA (0·3%) concentrations, which were mid-range. Bangladeshi women in general, and especially those who practise prolonged
breast-feeding, may benefit from increased consumption of food sources of PUFA.
Key words: Breast-feeding: PUFA: DHA: Arachidonic acid
To promote normal child growth, development and
immune function, mothers must supply their children
with adequate amounts of PUFA during pregnancy and lac-
tation(1,2). In particular, maternal fat stores and dietary
intake must be adequate to allow for sufficient transfer of
linoleic acid (LA), arachidonic acid (ARA), a-linolenic
acid (ALA), EPA and DHA to the child. In resource-poor
countries, breast milk composition is especially important,
as children often depend on breast milk as an important
source of fat for as long as they continue to breast-feed(3,4).
The 2008 Joint WHO/FAO Expert Consultation on Fats
and Fatty Acids in Human Nutrition recommended that
all women of reproductive age, including pregnant and
lactating women, consume 20–35% of total energy as
fat, with 6–11% of total energy from PUFA and at least 2
and 0·5% of total energy from LA and ALA, respectively(5,6).
Estimated minimum intakes of LA and ALA that will
prevent essential fatty acid (EFA) deficiency are 1 and
0·2% of total energy, respectively(7,8). Both the WHO/
FAO Expert Consultation and the Perinatal Lipid Intake
Working Group specifically recommend that pregnant
and lactating women consume a usual daily intake of
*Corresponding author: E. A. Yakes, fax þ1 530 752 3406, email firstname.lastname@example.org
Abbreviations: ALA, a-linolenic acid; ARA, arachidonic acid; EFA, essential fatty acid; FAME, fatty acid methyl ester; ICDDR,B, International Centre for
Diarrhoeal Disease Research, Bangladesh; LA, linoleic acid; SES, socio-economic status.
British Journal of Nutrition (2011), page 1 of 11
q The Authors 2011
British Journal of Nutrition
Maternal fat stores are another source of PUFA for trans-
fer to the child, and there is consensus that both adequate
pre-pregnancy weight (BMI $18·5kg/m2) and appropriate
gestational weight gain are important for maternal and
child health(10,11). Energy intake during lactation must ade-
quately support milk production and prevent lean tissue
loss and excessive depletion of adipose stores. During
the first 6 months of lactation, the estimated additional
energy needs for poorly nourished women are about
711·3kJ/d (170kcal/d) higher than those for well-nour-
ished women(12). The energy requirements for lactation
beyond 6 months differ depending on the amount of
milk being produced(12).
The amount of PUFA delivered in human milk varies in
different populations. Milk PUFA content is influenced by a
number of factors, including maternal dietary intake,
maternal fat stores, endogenous synthesis of fats and regu-
latory mechanisms affecting fatty acid transport and
synthesis in the mammary
women of child-bearing age may have inadequate fat
intake and fat stores. Dietary surveys of households and
lactating women in Bangladesh have broadly characterised
the rural Bangladeshi diet as being very low in fat and high
in carbohydrates, but specific intakes of PUFA have not
been examined(17,18). Rural Bangladeshi women store
very little fat during pregnancy and consistently lose
weight during the first 2 years of lactation(17,19,20). The
median duration of breast-feeding in Bangladesh is 32·8
months, with 78% of children still receiving some breast
milk at 24 months of age(21).
The present study examined data from Bangladeshi
mothers of children 24–48 months of age. Our aims
were to examine the relationship between prolonged
breast-feeding and maternal BMI, to compare the fat and
energy intakes of breast-feeding and non-breast-feeding
mothers, to assess the adequacy of fat intake in rural
Bangladeshi mothers and to determine breast-milk fatty
acid composition in a subsample of mothers. We also
examined the relationships between maternal fat intake
season and maternal age, BMI, education level, socio-econ-
omic status (SES) and site of residence. These data are
unique in that they characterise maternal dietary intake
and breast-milk PUFA content during prolonged lactation;
the vast majority of the work done previously on this
topic has focused on the first 6 months of lactation.
Based on the anthropometric status, dietary fat intake
and prolonged lactation of these mothers, we hypothesised
that the PUFA content of their breast milk would be lower
than that observed for breast milk that was previously
reported in the literature.
Subjects and methods
The study was designed as a cross-sectional, represen-
tative, multi-stage survey of mothers of children 24–48
months of age who resided in one of two rural districts
of the northern poverty belt of Bangladesh. The present
laid down in the Declaration of Helsinki, and all proce-
dures involving human subjects were approved by the
California, Davis and the Ethical Review Committee
(ERC) of the International Centre for Diarrhoeal Disease
consent was obtained from all study participants.
Study sites and selection of study participants
We selected two rural ‘upazilas’, Trishal and Pirgacha,
in northern Bangladesh, based on their high prevalence
of poverty and food insecurity and the presence of
some existing research infrastructure established by the
Within each ‘upazila’, all villages or ‘mauzas’ (administra-
tive subdivisions) identified by the 2001 census were
included in the sampling universe. There were a total of
159 clusters (‘mauzas’ or villages) in Trishal and 169
clusters in Pirgacha. During the first sampling stage,
twenty-five clusters were selected from each study site
using systematic sampling with the probability of selection
proportional to estimated population size. At the second
stage of sampling, ten households within each cluster
were selected using a global positioning system sampling
method(22). The study sites and our sampling methods
have been previously described in detail(23).
Approximately 92% of the originally selected eligible
households agreed to participate in the study. Data were
collected from 240 households in Trishal from October
2007 to May 2008 and from 240 households in Pirgacha
from January to June 2008. Forms for one household in
Trishal were lost during transit from the field site to the
ICDDR,B. The present analysis excluded data from house-
holds in which the primary carer was the father (n 1) or a
woman older than reproductive age (age $55 years; n 4).
Thus, the total sample size was 474 women of reproductive
age, 4% of whom were grandmothers of the study
Data collection and processing
Dietary data were collected by fieldworkers posted in
recruited households for 12h periods on two non-consecu-
tive days during the course of a 1-week period. A fieldwor-
ker observed food preparation and consumption, and used
a frequently standardised food scale (^1g, MyWeigh
KD7000; MyWeigh, Phoenix, AZ, USA) to measure the
E. A. Yakes et al.2
British Journal of Nutrition
weights of all foods included in recipes, the weights of
recipes before and after cooking and the weights of
foods and mixed preparations served and leftover after
consumption. Animal source foods included in mixed
preparations (e.g. fish in a vegetable curry) were weighed
separately at the point of consumption to estimate actual
intake. On the morning after the 12h observation period,
a fieldworker returned to the household to elicit infor-
mation on any food consumed during the 12h period
after the previous day’s observation ended. A standardised
plate, cup and spoon were used to assist with portion size
estimation, and fieldworkers reviewed all information with
the carer after the initial list of foods was completed to
probe for food items that may have been forgotten.
Child’s breast milk intake was estimated by the test-
weighing procedure, which has been previously described
Foods consumed were converted to nutrients using the
USDA Nutrient Database for Standard Reference (Release
20)(24)and the International Minilist(25). Food sources of
nutrients were calculated by summing the total nutrient
intakes for all observation days, summing the nutrient
intakes from each food source and calculating the percen-
tage of total nutrient intake from each food.
A trained field-worker measured maternal height using a
portable stadiometer (^0·1cm, ShorrBoard; Shorr Pro-
ductions, Olney, MD, USA) and maternal weight using a
frequently standardised electronic scale (^100g, Seca
840 Digital Floor Scale; Seca, Hamburg, Germany). If two
height measurements differed by more than 0·5cm, a
third height measurement was taken. Average weight and
the average of the two closest height values were used to
calculate BMI (kg/m2).
Information on SES and agricultural practices was gath-
ered during interviews with the carer and the head of the
Biological samples were collected from six enrolled
households in a cluster. Breast milk was collected from lac-
tating mothers starting in December 2007 (2 months after
the start of the study) due to a delay in obtaining Ethical
Review Committee approval. Of the 474 carers of repro-
ductive age enrolled in the study, 242 were breast-feeding
a child between 24 and 48 months of age and 128 of these
women were from households who participated in biologi-
cal sample collection in the period after breast milk collec-
tion started. These women were asked to provide a breast
milk sample, and ninety-eight (77%) gave a sample.
Reasons for not providing a sample included refusal and
difficulties with producing a milk sample via hand
Efforts were made to standardise both the time of day
and the time elapsed since the last feeding for breast
milk collection(26). Carers were brought to a central collec-
tion site (health complex or school) early in the morning
and were asked to breast-feed their children upon arrival
and then again 60min later. During this latter feed,
a female fieldworker assisted the mother with hand expres-
sing approximately 3·5ml of human milk into a sterile coni-
cal vial from the last breast suckled during the previous
feed. Breast milk samples were transported to the field
laboratory on crushed ice and then gently inverted several
times and aliquoted to storage vials. The breast milk
aliquots were stored in a freezer (2208C) with a back-up
generator for 1 week to 6 months (depending on the
date of collection) until being shipped to ICDDR,B and
then to University of California, Davis on dry ice. At the
ICDDR,B and University of California, Davis, the samples
were stored at 2808C until extraction and derivatisation.
A sample of 30ml (approximately 0·03g) human milk
was used for fatty acid methyl ester (FAME) analysis.
Both the human milk and an authentic internal standard
(TAG (17:0); Avanti Polar Lipids, Alabaster, AL, USA)
were measured by weight, and then the Folch method
was used for lipid extraction(27). The organic phase was
then evaporated under N2. The dried organic extracts
were methylated by adding 1ml of 3 M-methanolic HCl
and incubating in a sealed vial for 12h at 608C. FAME
were extracted in hexane with 0·01% butylated hydroxyto-
luene after neutralisation of the methanolic hydrochloric
acid with 5% potassium bicarbonate. FAME extracts were
dried under vacuum and then dissolved in hexane for
analysis by GC.
FAME were separated via capillary GC using an Agilent
Technologies (Santa Clara, CA, USA) gas chromatograph
model 7890 equipped with a 30m DB225MS capillary
column (J&W Scientific, Folsom, CA, USA) and a flame
ionisation detector. An authentic GLC reference standard
(461; Nu-Chek Prep, Elysian, MN, USA) and Agilent Tech-
nologies GC ChemStation software were used to identify
and quantify the FAME peaks. Every chromatogram was
reviewed to check for proper peak integration and identi-
fication. Percentage of fatty acids by weight was calculated
by dividing the peak area for a particular fatty acid by the
total sum of the peak areas for all identified fatty acids.
We used the National Cancer Institute method for episodi-
cally consumed foods to estimate distributions of usual
food and nutrient intake(28,29). This method was used to
better approximate the variance of the intake distributions
by removing within-person variation and to improve esti-
mates of intake for foods and nutrients that were episodi-
cally consumed during the 2d of dietary data collection.
Briefly, the National Cancer Institute method uses data
from two or more 24h estimates of intake in a two-part
model that first estimates the probability of consumption
using logistic regression and then estimates the amount
consumed using linear mixed regression on a trans-
The covariates of interest for both the probability of
consumption and the amount of consumption models
PUFA intake among Bangladeshi women3
British Journal of Nutrition
included maternal age, BMI, education level, household
access to electricity and housing quality (as proxies of
household wealth), season and site of residence. The hous-
ing quality variable was a composite continuous measure
assessing the type of floors, walls and sanitary facilities in
the family home, and the type of cooking fuel used.
Maternal education was a continuous variable that was
categorised to minimise the effects of measurement error.
All other continuous variables (maternal age, BMI and
house quality score) were used as continuous variables
in the models. The first model (probability of consump-
tion) was of the form(28):
The second model (amount of consumption) was of
Both models were used to estimate distributions of
intake for foods and nutrients that were episodically con-
sumed, including oil, eggs, fish, meat/poultry, ARA, EPA
and DHA. Only the amount of consumption model was
used for all other foods and nutrients because they were
consumed in some amount by all participants on all obser-
vation days. The percentage of energy derived from fat,
carbohydrate and protein was calculated for each woman
on each day; these distributions estimate the long-term
mean of the daily ratio of intakes (the usual ratio of
intakes)(30). Distributions of intakes were estimated for
the entire maternal population who had complete covari-
ate information (n 455) and for the breast-feeding and
non-breast-feeding subgroups. Covariates were assessed
as being significantly associated with intake based on test
statistics from the fitted models.
The SURVEYLOGISTIC procedure in SAS (version 9.2;
SAS, Cary, NC, USA) was used to analyse the relationships
between maternal characteristics and breast-feeding status,
and the SURVEYREG procedure was used to analyse the
relationships between maternal characteristics and total
milk fat content. These procedures allow for correct esti-
Characteristics of women and study households
Table 1 presents the characteristics of Bangladeshi mothers
included in the study sample and their households. The
women ranged in age from 16 to 50 (median 25) years, and
their BMI ranged from 14·5 to 36·7 (median 19·1)kg/m2.
Maternal weight ranged from 30·0 to 82·0 (median 43·1)kg,
and height ranged from 131·8 to 168·6 (median 150·2)cm.
Approximately, 40% of the mothers had not completed
any formal education, and 70% did not have access to
electricity. A small number of households (8%) were sur-
veyed before the major rice harvest (pre-Aman harvest;
October–November); the remaining households were sur-
veyed after the main harvest (50%; post-Aman harvest;
December–March) and during the secondary rice harvest
(42%; Aus/Boro harvest; April–June). A total of 274
women (58% of total female carers of reproductive age)
were currently breast-feeding, with 241 women breast-
feeding a 24–48-month-old child and the rest (n 33) breast-
feeding a younger sibling. About 76% of 24–35-month-old
children and 38% of 36–48-month-old children nursed
an average of $5 times/d on the observation days.
Relationships between breast-feeding status and
covariates of interest
Bivariate relationships between breast-feeding status and
the covariates of interest were examined (Table 1), and
breast-feeding status was not significantly associated with
access to electricity, housing quality, maternal age, maternal
education level, season of data collection or district of resi-
dence. Maternal BMI, however, was significantly associated
with lactation status. Approximately 42% of breast-feeding
mothers were underweight (BMI ,18·5kg/m2) compared
with 26% of non-breast-feeding mothers (P¼0·0003). The
significant relationship between breast-feeding status and
maternal BMI persisted in a multivariate model that
adjusted for access to electricity, housing quality, maternal
education, maternal age and season. On average, mothers
who were breast-feeding had a BMI 0·8kg/m2less than
non-breast-feeding mothers (P¼0·001).
Energy intake and macronutrient profile of diet
Table 2 shows estimated distributions of usual energy,
macronutrient and EFA intake, expressed as a percentage
of total energy. The mean (5th–95th percentile) daily
energy intake for all women was 7870 (6782–8979)kJ/d.
Almost all of the women in the study population consumed
more than 75% of their total energy from carbohydrates.
Intakes from protein had a relatively narrow range, with
E. A. Yakes et al.4
British Journal of Nutrition
most women consuming about 10% of their energy from
protein. The maternal diet averaged 7·8% of total energy
from fat and 2 and 0·33% of total energy from LA and
After adjusting for maternal age, BMI, education level,
SES and site of residence in separate linear mixed
regression models, intakes of total energy, carbohydrates
and protein as a percentage of total energy did not differ
by lactation status. Intakes of total fat (b ¼ 20·07;
P¼0·04), LA (b ¼ 20·10; P¼0·03) and ALA (b ¼ 20·12;
P¼0·01), expressed as a percentage of total energy, were
slightly lower in breast-feeding than in non-breast-feeding
women in the same model. SES was significantly associated
with higher intakes of protein (electricity access: b ¼ 0·03;
P¼0·01; better housing quality: b ¼ 0·03; P¼0·002) and
fat (better housing quality: b ¼ 0·20; P,0·0001; maternal
access to formal education: b ¼ 0·11; P¼0·003) and
lower intakes of carbohydrates (better housing quality:
b ¼ 21·7; P¼0·02) as a percentage of total energy in
separate linear mixed regression models.
Based on the intake distributions, it is estimated that
99% of the women in the study sample had usual fat
intakes less than 20% of total energy, and 84% had usual
fat intakes of less than 10% of total energy. Almost all
women (99%) were estimated to consume less than 6%
of total energy from PUFA. About 60% of women con-
sumed less than 2% of total energy from LA, and 90%
consumed less than 0·5% of total energy from ALA.
Table 1. Characteristics of rural Bangladeshi mothers included in the study and their households, by the women’s current breast-
(Numbers of subjects and percentages)
Maternal age (years)
Maternal BMI† (kg/m2)
Grade 3 CED (BMI , 16)
Grade 2 CED (BMI 16–16·99)
Grade 1 CED (BMI 17–18·49)
Normal (BMI 18·5–22·99)
At-risk for overweight
Overweight (BMI $25)
Maternal height (cm)
Maternal weight (kg)
$ 40– , 50
Maternal completion of any formal education
Low score (,2)
Middle score (2–3)
High score (.3)
Site of residence
CED, chronic energy deficiency.
*The P value for the effect in a logistic regression model, with characteristic as the independent variable and maternal breast-feeding status (yes/no)
as the outcome.
†BMI categories defined according to Shafique et al.(31)and WHO Expert Consultation(54).
‡Three women in this category were pregnant. One was 3 months pregnant (BMI ¼ 22kg/m2), one was 7 months pregnant (BMI ¼ 21kg/m2) and one
was 9 months pregnant (BMI ¼ 19kg/m2).
PUFA intake among Bangladeshi women5
British Journal of Nutrition
An estimated 7 and 13% of mothers had a usual intake of
LA and ALA less than 1 and 0·2% of total energy, respect-
ively, placing them at a potential risk of EFA deficiency.
Maternal consumption of fatty acids
Table 3 presents the women’s estimated distributions
of usual fat intake expressed as g/d. The mean intake of
total fat for all women was 16·3g/d. The average ratio
of LA:ALA intake in the maternal diet was 6·7, with a
5th–95th percentile range of 4·4–9·7. After adjusting
for maternal age, BMI, education level, SES and site of
residence in a linear mixed regression model, breast-feed-
ing status was not a significant predictor of total fat intake
or intake of any of the fat subclasses. Total fat intake was
significantly associated with several other factors in the
model, including season (lowest in the pre-Aman harvest:
b ¼ 20·08; P¼0·04), maternal age (b ¼ 20·01; P¼0·02),
maternalBMI(b ¼ 0·02;P¼0·02),housingquality
(b ¼ 0·22; P¼,0·0001), maternal level of formal education
(b ¼ 0·11; P¼0·02) and site of residence (lower in Trishal:
b ¼ 20·16; P¼0·001). Similar associations were observed
for intake of the fat subclasses.
Food sources of fatty acids in the maternal diet
Vegetable oil was the source of about 50% of total fat
intake for the entire carer population. Rice was also a
prominent source of fat, providing 27% of the total fat
in the maternal diet. Animal source foods, including
freshwater fish, eggs, meat/poultry and dairy products,
provided about 15% of total maternal fat intake. These
patterns did not differ for breast-feeding and non-breast-
The dietary patterns for PUFA were similar to those for
total fat, with vegetable oil accounting for 57% of total
PUFA intake and rice providing another 26%. Vegetables
and beans supplied about 6% of the total PUFA. Oil, rice,
Table 2. Estimated distributions of usual energy intake and percentage of energy derived from fat, carbohydrate and
protein among rural Bangladeshi women, by current breast-feeding status*
(Mean values and 5th–95th percentiles)
All women (n 455)
women (n 196)
women (n 259)
kJ/d per kg
Total energy (%)
LA, linoleic acid; ALA, a-linolenic acid.
*Percentage of energy derived from fat, carbohydrate and protein was calculated for each woman on each day; these distributions estimate
the long-term mean of the daily ratio of intakes (the usual ratio of intakes)(30).
†Reported P -values are for the effects of lactation status on intake after adjusting for maternal age, BMI, education, socio-economic status
and site of residence in a linear mixed regression model.
Table 3. Estimated distributions of usual fat intake (g/d) by rural Bangladeshi women, by current breast-feeding status
(Mean values and 5th–95th percentiles)
All women (n 455)
women (n 196)
women (n 259)
Type of fatMean
LA, linoleic acid; ARA, arachidonic acid; ALA, a-linolenic acid.
*Reported P-values are for the effects of lactation status on intake after adjusting for maternal age, BMI, education, socio-economic
status and site of residence in a linear mixed regression model.
E. A. Yakes et al.6
British Journal of Nutrition
beans and vegetables (particularly leafy vegetables) were
the most important sources of both LA and ALA. The pri-
mary food sources of ARA included freshwater fish
(49%), eggs (42%) and meat/poultry (7%). Almost all of
the EPA and DHA were provided by freshwater fish.
Again, the pattern of food intake was similar in breast-
feeding and non-breast-feeding women.
Estimated usual intake distributions for food sources of
PUFA (g/d) are shown in Table 4. There were significant
differences in the probability of consumption of food
sources of PUFA between the two study sites in a linear
mixed model that adjusted for lactation status, maternal
age, BMI, education level and SES. The probabilities of
consuming soyabean oil (b ¼ 4·0; P,0·0001), meat/poul-
try (b ¼ 0·87; P¼0·002) and eggs (b ¼ 1·2; P,0·0001)
were higher in Pirgacha, whereas the probabilities of con-
suming mustard oil (b ¼ 2·1; P,0·0001) and fish (b ¼ 1·2;
P,0·0001) were higher in Trishal.
Breast-milk fatty acid composition
The characteristics of mothers who provided a breast milk
sample and their households were similar to those of the
overall maternal population. The mean BMI for breast-
feeding women who provided a sample was 18·8kg/m2,
while the mean BMI for breast-feeding women who did
not provide a sample because of lack of opportunity
(n 114) or refusal/difficulty with producing a sample
from hand expression (n 30) was 19·6kg/m2(P¼0·01).
Table 5 lists the fatty acid content (% wt) of the breast
milk samples. The median total fat content for the samples
was 3·5g/100g of breast milk (5th–95th percentile range,
1·4–7·3). Total milk fat content was not significantly associ-
ated with maternal BMI (P¼0·22) or with breast-feeding
frequency (P¼0·42) or duration (P¼0·18). The majority
of the fatty acids present in the human milk were SFA,
followed by MUFA and then PUFA. The most abundant
fatty acids were 16:0, 18:1, 14:0 and 18:2n-6 (LA). The
median ratio of LA:ALA in the milk was 35, while the
median ratio of long-chain n-6:long chain n-3 fatty acids
was 1·6. Breast-milk ARA and DHA concentrations were
highly correlated (r 0·7; P,0·001).
Our data support the concerns raised by earlier studies
regarding the adequacy of fat stores in lactating Bangladeshi
women. Overall, the BMI patterns for the women in the
representative 2000–4 Nutritional Surveillance Project(31).
However, we found that women who were breast-feeding
Table 4. Estimated usual consumption (g/d) of food sources of PUFA by rural Bangladeshi women, by current breast-feeding status
(Mean values and 5th–95th percentiles)
All women (n 455)
women (n 196)
women (n 259)
*Reported P -values are for the effects of lactation status on intake after adjusting for maternal age, BMI, education, socio-economic status and site of
residence in a linear mixed regression model.
Table 5. Fatty acid composition of breast milk from rural Bangladeshi
(Median values and 5th–95th percentiles, n 98)
Total fatty acids by wt (%)
Fatty acidsMedian5th–95th Percentiles
LC n-6:LC n-3 ratio
LA, linoleic acid; ALA, a-linolenic acid; LC, long-chain.
PUFA intake among Bangladeshi women7
British Journal of Nutrition
a 24–48-month-old child had a considerably higher
prevalence of underweight (BMI ,18·5kg/m2) than non-
breast-feeding women with a 24–48-month-old child.
As our data are cross-sectional, we cannot determine the
postpartum period during which the weight discrepancy
between these groups was established, or definitively
establish that the energy demands of lactation were the
cause of this weight discrepancy. Longitudinal studies in
both well-nourished American women and poorly nouri-
women experienced the most weight loss from 3 to 6
months postpartum(17,32). For the American mothers, the
weight differences that were observed between breast-
feeding and formula-feeding mothers were no longer sig-
nificant after 12 months, indicating that the breast-feeding
mothers probably experienced some weight recovery as
breast-feeding intensity decreased(32). The Bangladeshi
mothers in the longitudinal study were not followed past
13 months postpartum. A cross-sectional study of rural
Bangladeshi women in the Jhenaidah district found that
the weights of lactating women were lower than those of
non-lactating women throughout 48 months of lactation
after controlling for height, education and food consump-
tion; however, the differences were only significant up
to 24 months postpartum(20). Milk production was not
quantified in that study, but the authors hypothesised
that decreased milk production after 24 months may have
allowed maternal weight to recover(20).
There was no significant difference in energy intake
between lactating and non-lactating women after adjusting
for maternal age, BMI, education level, SES and site of
residence. Based on estimated mean 24h breast milk con-
sumption for the children in the present study (145g/d for
24–35-month-old children; 90g/d for 36–48-month-old
children) and assuming that maternal fat stores were not
covering additional energy needs, the breast-feeding
women in the present study needed about 301–502kJ/d
in excess to support milk production (range of maternal
energy needs based on 1st and 99th percentiles of the
child’s breast milk intake: 29–1799kJ/d in excess)(12,33).
We may not have had the power to detect a difference in
intake of 418kJ between the two groups; a post hoc
power analysis showed that the power to detect this differ-
ence in the present study is 48%. It is also possible that we
may have underestimated energy intake, although this
would not affect our ability to find a difference in energy
intake between the two groups unless it was differential
between breast-feeding and non-breast-feeding women.
The women in the present study spent most of the day
engaged in household activities (cooking, cleaning, child
care and grain processing). Bangladeshi women with
similar activity profiles were found to have physical activity
levels of 1·59 (lightly active) in a study that used the
doubly labelled water method(34)and 1·76 (moderately
active) in a study that used the FAO/WHO/UNU factorial
method(25). Based on these physical activity levels, the
total estimated energy expenditures for a woman weighing
43kg with an estimated BMR of 4700kJ/d would be 7473
and 8272kJ, respectively(12). These total estimated energy
expenditures correspond to approximately the 25th and
75th percentiles of estimated energy intake for both
lactating and non-lactating women in the present study.
Our data also support the concerns raised by earlier
studies regarding the adequacy of fat intake in Bangladeshi
women(17,18). We found that women in rural Bangladesh
consumed a very low-fat, high-carbohydrate diet, with an
estimated average of just 7·8% total energy from fat.
Intakes of fat were similar among breast-feeding and
non-breast-feeding women. The main sources of fat in
the diet were fairly limited, with vegetable oil and rice
contributing the most to overall fat intake. There were
significant geographical differences in the consumption
of the animal source foods that are good sources of long-
chain PUFA. These differences are probably due to the
differences in availability and local preference and have
important implications for programme planning. Based
ondistributions of usual
women of child-bearing age consume very small quantities
of foods that are important sources of fat, and conse-
quently, small quantities of PUFA.
The total fat and ALA intakes of Bangladeshi mothers, as
a percentage of total energy, are estimated to be almost
universally less than the respective 20 and 0·5% minimums
recommended by the 2008 Joint WHO/FAO Expert Consul-
tation on Fats and Fatty Acids(5,6). The estimated intake of
LA was also below the minimum recommended intake
levels in approximately 60% of women. Approximately,
10% of the women in our sample may have LA and ALA
consumption below the levels considered to be necessary
to prevent EFA deficiency. Although it is likely that most
women with this low intake would be protected from
EFA deficiency by their fat stores, women with low body
fat may be at increased risk(35). Also, there should be par-
ticular concern for breast-feeding women with this low
level of intake, as they are continuously transferring LA
and ALA to their children in breast milk, and our BMI
data suggest that depleted fat stores may be more
common in breast-feeding women. A similar concern
might be raised for DHA, as the women in the present
study consumed far less DHA than the recommended
usual daily intake of 200mg/d for lactating women(5,9,36).
Overall, the pattern of fatty acids that we observed in the
milk from rural Bangladeshi mothers reflected the dietary
patterns in the population. Consistent with the findings in
a number of other studies in populations with high-carbo-
hydrate, low-fat diets, we observed a high level of myristic
acid (14:0), a medium-chain SFA that is synthesised from
acetyl-CoA in the mammary gland(15,37–40). The breast
milk from the Bangladeshi mothers also had low levels
of oleic (18:1n-9) and stearic (18:0) acids; the concen-
trations of these fatty acids are generally lower in human
milk with increased medium-chain SFA levels(13,16).
E. A. Yakes et al.8
British Journal of Nutrition
Levels of LA and ALA in the breast milk from the present
study sample were among the lowest reported in the
extensive literature characterising breast-milk fatty acid
composition. Prentice & Paul(41)summarised data on the
breast milk composition of mothers from separate studies
in eight African and South American countries with diets
that could be generally characterised as high carbohydrate
and low fat. When compared with women from those
countries, the women from the present study delivered
the lowest median LA levels (0·103g/g fat; range for eight
countries: 0·110–0·238) and the second lowest ALA level
(0·003g/g fat; range: 0·001–0·014)(41). Breast-milk LA and
ALA levels from the Bangladeshi mothers were also
below those observed in mothers from more developed
countries(42,43). Human milk from mothers in Nepal, the
Philippines, the Congo and Pakistan exhibited comparably
low LA levels, but the very low ALA level observed in the
breast milk from Bangladeshi women was observed only in
milk from the Pakistani women(43–46). As a result of the
very low ALA levels, breast milk from the Bangladeshi
mothers had a median LA:ALA ratio of 35, which is much
higher than the recommended range of 5–15:1(46). The
authors of the studies that included the Filipina and Pakis-
tani mothers both concluded that inadequate LA and ALA
intake might be the cause of the low EFA levels in the
breast milk(43,44). The present study provides support
for this conclusion, as we found that the Bangladeshi
women were consuming low levels of EFA compared
with international recommendations.
In contrast to the present findings regarding breast-
milk LA and ALA, the breast-milk ARA and DHA con-
centrations for the Bangladeshi mothers fell in the middle
of reported ranges for women from many different
countries(41–43,45,47,48). The median ARA and DHA percen-
tage weights for breast milk from Bangladeshi mothers
were very close to the mean concentrations of DHA
(0·32%) and ARA (0·47%) that were determined in a
descriptive meta-analysis of milk from 2474 women in
It is possible that regulatory mechanisms play an import-
ant role in determining DHA and ARA levels in human
milk, as the ARA and DHA levels in the breast milk of
the Bangladeshi mothers were maintained at levels that
fall near the middle of ranges observed in other popu-
lations, despite low intake of fat and prolonged lactation.
There are several potential mechanisms that could be up-
or down-regulated to control breast-milk PUFA content,
including the release of fatty acids from lipoproteins by
lipase, the transport of fatty acids from the blood into the
mammary gland and the synthesis of milk TAG by acyl
transferases(16). ARA, in particular, seems to be tightly con-
trolled in breast milk, with a limited range of ARA concen-
trations observed across populations with very different
diets(39,50–52). Future longitudinal studies that examine
changes in white adipose tissue or erythrocyte fatty acid
composition in mothers over the course of prolonged
lactation could help to elucidate these issues.
The low LA intake of the women in the present study
may have allowed for increased conversion of ALA to
DHA. However, subjects in an intervention trial who con-
sumed a diet with an EFA profile similar to the women in
the present study (3% of energy from LA and 0·4% of
energy from ALA) converted more dietary ALA into EPA
but did not increase the absolute amount of DHA
synthesised compared with subjects on a control diet
(7% of energy from LA and 0·4% of energy from ALA)(53).
Fish is a regular part of the Bangladeshi diet, but the
breast-milk DHA concentrations in the present study
were not as high as those observed in other countries
where fish is regularly consumed(51). This may be
explained by the fact that the vast majority of fish intake
was from freshwater fish, which generally contain less
DHA than ocean fish. In addition, the estimated amount
of usual fish intake was low.
Several limitations of the present study deserve com-
ment. Our observations about the relationship between
PUFA intake and breast-milk concentrations would have
been strengthened by presenting data on maternal bio-
chemical PUFA status. However, we were unable to collect
blood samples from the mothers, and there is no previous
data on biochemical PUFA status available for a compar-
able Bangladeshi population. It is possible that we may
have under- or overestimated the percentage of women
who would fall below the stated fat intake cut-off values
because we collected only 2d of dietary data for each indi-
vidual. However, we improved our estimates for the popu-
lation intake distribution by using statistical methods to
adjust for episodic consumption and remove within-
individual variation(28). Although the in-home observations
allowed us to collect reliable quantitative data on actual
food intake, it is conceivable that our presence may have
influenced the women’s food consumption on the obser-
vation days. However, if the women did include more
desirable foods (e.g. oil and animal source foods) in their
cooking during our observation days, we may have overes-
timated their true usual fat intakes. Another potential
source of error is that at least one major meal (approxi-
mately 22% of total energy) was generally obtained via
recall, as it was consumed outside of the 12h observation
window. However, we did find that estimated intakes of
energy were similar between recalled and observed
meals. Additional error may have been introduced when
using food composition tables to convert food to nutrients,
as the nutrient composition of foods can vary significantly
across seasons and geographical regions, and budgetary
constraints prevented us from directly analysing foods
from the study area to determine fatty acid composition.
To partially address this limitation, we have presented
quantitative data on food intake to support the nutrient
PUFA intake among Bangladeshi women9
British Journal of Nutrition
Overall, breast milk from Bangladeshi mothers provides
24–48-month-old children with easily absorbed medium-
chain SFA, some LA and ALA, and adequate amounts of
ARA and DHA. However, prolonged lactation combined
with very low fat intake may lead to significant demands
on the maternal body pool of fatty acids. Increasing the
dietary intake of PUFA in these mothers thus has the poten-
tial to positively affect breast milk composition, and conse-
quently, children’s health and development. Research into
this area needs to consider not only the fatty acid content
of breast milk, but also to examine the effects of breast-
feeding on maternal fat stores, maternal health outcomes
and the amount of fatty acids supplied to subsequent off-
spring during pregnancy and lactation. Women who are
breast-feeding at high frequency for an extended period
of time may require greater fat intake to provide the fatty
acids for milk and the additional energy needed to prevent
lean tissue loss and to promote replenishment of maternal
fat stores. Bangladeshi women in general, and especially
those who breast-feed for more than 2 years, may benefit
from an increased percentage of total energy consumed
from lipids and greater intake of food sources of PUFA.
The authors thank the study participants who welcomed us
into their homes and the staff from the ICDDR,B who were
instrumental to the data collection process. We thank Dr
Daniel Tancredi from the University of California (UC)
Davis School of Medicine, Department of Pediatrics, for
statistical consultation. We are also grateful to Diego
Vargas and Andrea Eaton for their invaluable assistance
during the laboratory analyses and to Dr Susan Ebeler
from UC Davis Department of Viticulture and Enology,
for generously allowing us to use the GC in her laboratory.
The present study was supported by the Bristol-Myers
Squibb Foundation, Inc., the Harvest Plus Challenge Pro-
gram (coordinated by the Centro Internacional de Agricul-
tura Tropical and the International Food Policy Research
Institute), the National Institute of Environmental Health
Sciences (NIEHS) grant R37 ES02710, the NIEHS Superfund
Basic Research Program P42 ES04699, the UC Davis Center
for Children’s Environmental Health, NIEHS grant P01
ES11269 and the University of California Discovery Pro-
gram. None of the authors had any financial or personal
interest in the organisations sponsoring the present study.
The authors’ responsibilities were as follows: E. A. Y. par-
ticipated in the study design, study implementation, data
analysis, interpretation of results and prepared the manu-
script; J. E. A. contributed to the study design, study
implementation and critical review of the manuscript;
M. M. I., M. B. H., A. S. R., T. A., K. M. J. and B. L. L.
were involved in the study implementation and critical
review of the manuscript; J. B. G. assisted in for funding,
L. A. G. was responsible for method development for the
laboratory analysis; C. D. performed the critical review of
the manuscript; K. H. B. had the primary responsibility for
funding, study design, interpretation of results and critical
review of the manuscript.
1. Das UN (2006) Essential fatty acids: biochemistry, physiology
and pathology. Biotechnol J 1, 420–439.
Innis SM (2007) Dietary (n-3) fatty acids and brain develop-
ment. J Nutr 137, 855–859.
Brown KH, Black RE, Becker S, et al. (1982) Consumption
of foods and nutrients by weanlings in rural Bangladesh.
Am J Clin Nutr 36, 878–889.
Prentice AM & Paul AA (2000) Fat and energy needs of chil-
dren in developing countries. Am J Clin Nutr 72, Suppl.,
Brenna JT & Lapillonne A (2009) Background paper on fat
and fatty acid requirements during pregnancy and lactation.
Ann Nutr Metab 55, 97–122.
Elmadfa I & Kornsteiner M (2009) Fats and fatty acid require-
ments for adults. Ann Nutr Metab 55, 56–75.
Holman RT (1960) The ratio of trienoic:tetraenoic acids in
tissue lipids as a measure of the essential fatty acid require-
ment. J Nutr 70, 405–410.
Bjerve KS, Fischer S & Alme K (1987) Alpha-linolenic acid
deficiency in man: effect of ethyl linolenate on plasma and
erythrocyte fatty acid composition and biosythesis of prosta-
noids. Am J Clin Nutr 46, 570–576.
Koletzko B, Lien E, Agostoni C, et al. (2008) The role of long-
chain polyunsaturated fatty acids in pregnancy, lactation and
infancy: review of current knowledge and consensus
recommendations. J Perinat Med 36, 5–14.
Salihu HM, Mbah AK, Alio AP, et al. (2009) Low pre-preg-
nancy body mass index and risk of medically indicated
versus spontaneous preterm singleton birth. Eur J Obstet
Gynecol Reprod Biol 144, 119–123.
Siega-Riz AM, Viswanathan M, Moos MK, et al. (2009)
A systematic review of outcome of maternal weight gain
according to the Institute of Medicine recommendations:
birthweight, fetal growth, and postpartum weight retention.
Am J Obstet Gynecol 201, 339.e1–339.e14.
Joint FAO/WHO/UNU Expert Consultation (2004) Human
energy requirements. http://www.fao.org/docrep/007/y5686e/
y5686e00.HTM (accessed 20 September 2010).
Vuori E, Kiuru K, Makinen SM, et al. (1982) Maternal diet and
fatty acid pattern of breast milk. Acta Paediatr Scand 71,
Hachey DL, Silber GH, Wong WW, et al. (1989) Human
lactation II: endogenous fatty acid synthesis by the mammary
gland. Pediatr Res 25, 63–67.
hydrate-rich diet not only leads to incorporation of medium-
chain fatty acids (6:0-14:0) in milk triglycerides but also in
Schmeits BL, VanderJagt DJ, Okolo SN, et al. (1999) Selective
retention of n-3 and n-6 fatty acids in human milk lipids in
the face of increasing proportions of medium chain-length
(C10:14) fatty acids. Prostaglandins Leukot Essent Fatty
Acids 61, 219–224.
Vinoy S, Rosetta L & Mascie-Taylor CGN (2000) Repeated
measurements of energy intake, energy expenditure and
E. A. Yakes et al. 10
British Journal of Nutrition
energy balance in lactating Bangladeshi mothers. Eur J Clin
Nutr 54, 579–585.
Hels O, Hassan N, Tetens I, et al. (2003) Food consumption,
energy and nutrient intake and nutritional status in rural
Bangladesh: changes from 1981–1982 to 1995–96. Eur J
Clin Nutr 57, 586–594.
during pregnancy and lactation: consequences for maternal
nutrition in rural Bangladesh. Eur J Clin Nutr 57, 151–156.
Sarkar NR & Taylor R (2005) Weight loss during prolonged
lactation in rural Bangladeshi mothers. J Health Popul Nutr
National Institute of Population Research and Training
(NIPORT) & Mitra and Associates, Macro International
(2009) Bangladesh Demographic and Health Survey 2007.
Dhaka, Bangladesh and Calverton, MD: National Institute
of Population Research and Training, Mitra and Associates,
and Macro International.
Grais RF, Rose AMC & Guthman JP (2007) Don’t spin the
pen: two alternative methods for second-stage sampling in
urban cluster surveys. Emerg Themes Epidemiol 4, 8.
Arsenault JA, Yakes EA, Hossain MB, et al. (2010) The current
high prevalence of dietary zinc inadequacy among children
rated by zinc biofortification of rice. J Nutr 140, 1683–1690.
U.S. Department of Agriculture: Agricultural Research Ser-
vice (2007) Release 20: USDA nutrient database for standard
reference. http://www.ars.usda.gov/ba/bhnrc/ndl (accessed
23 August 2009).
Kramer EM, Peterson KE, Rogers BL, et al. (1997) Intrahou-
sehold allocation of energy intake among children under
five years and their parents in rural Bangladesh. Eur J Clin
Nutr 51, 750–756.
Ruel MT, Dewey KG, Martinez C, et al. (1997) Variation of
single daytime samples of human milk to estimate the 24-h
concentration of lipids in urban Guatemalan mothers. Am J
Clin Nutr 65, 439–444.
Folch J, Lees M & Sloane Stanley GH (1957) A simple method
for the isolation and purification of total lipides from animal
tissues. J Biol Chem 226, 497–509.
Tooze JA, Midthune D, Dodd KW, et al. (2006) A new statisti-
cal method for estimating the usual intake of episodically
consumed foods with application to their distribution. J Am
Diet Assoc 106, 1575–1587.
National Cancer Institute (2009) Usual Dietary Intakes: SAS
Macros for the NCI Method. Bethesda, MD. http://riskfactor.
Freedman LS, Guenther PM, Dodd KW, et al. (2010) The
population distribution of ratios of usual intake of dietary
components that are consumed every day can be estimated
from repeated 24-hour recalls. J Nutr 140, 111–116.
Shafique S, Akhter N, Stallkamp G, et al. (2007) Trends of
under- and overweight among rural and urban poor
women indicate the double burden of malnutrition in
Bangladesh. Int J Epidemiol 36, 449–457.
Dewey KG, Heinig MJ & Nommsen LA (1993) Maternal
weight-loss patterns during prolonged lactation. Am J Clin
Nutr 58, 162–166.
Butte NF & King JC (2005) Energy requirements during
pregnancy and lactation. Public Health Nutr 8, 1010–1027.
Rosetta L, Kurpad A, Mascie-Taylor CGN, et al. (2005) Total
energy expenditure (H218O), physical activity level and milk
output of lactating rural Bangladeshi tea workers and nontea
workers. Eur J Clin Nutr 59, 632–638.
Hibbeln JR, Nieminen LRG, Blasbalg TL, et al. (2006) Healthy
intakes of n-3 and n-6 fatty acids: estimations considering
Koletzko B, Cetin I, Brenna JT, et al. (2007) Dietary fat intakes
for pregnant and lactating women. Br J Nutr 98, 873–877.
Insull W, Hirsch J, James T, et al. (1959) The fatty acids of
human milk. II. Alterations produced by manipulation of
caloric balance and exchange of dietary fats. J Clin Invest
ary carbohydrates and fat on the fatty acids of mature milk.
A study in four ethnic groups. Am J Clin Nutr 17, 180–183.
Smit EN, Martini IA, Mulder H, et al. (2002) Estimated biolo-
gical variation of the mature human milk fatty acid compo-
sition. Prostaglandins Leukot Essent Fatty Acids 66, 549–555.
Thiombiano-Coulibaly N, Rocquelin G, Eymard-Duvernay S,
et al. (2003) Seasonal and environmental effects on breast
milk fatty acids in Burkina Faso and the need to improve
the omega-3 PUFA content. Acta Paediatr 92, 1388–1393.
in developing countries. Am J Clin Nutr 72, 1253S–1265S.
Xiang M, Harbige LS & Zetterstrom R (2005) Long-chain poly-
unsaturated fatty acids in Chinese and Swedish mothers: diet,
breast milk and infant growth. Acta Paediatr 94, 1543–1549.
Yuhas R, Pramuk K & Lien EL (2006) Human milk fatty acid
composition from nine countries varies most in DHA. Lipids
Smit EN, Oelen EA, Seerat E, et al. (2000) Breast milk
docosahexaenoic acid (DHA) correlates with DHA status of
malnourished infants. Arch Dis Child 82, 493–494.
Glew RH, Huang YS, VanderJagt TA, et al. (2001) Fatty acid
composition of the milk lipid of Nepalese women: corre-
lation between fatty acid composition of serum phospholi-
pids and melting point. Prostaglandins Leukot Essent Fatty
Acids 65, 147–156.
Rocquelin G,Tapsoba S, Kiffer J, etal. (2003) Human milkfatty
acids and growth of infants in Brazzaville (The Congo) and
Ouagadougou (Burkina Faso). Public Health Nutr 6, 241–247.
Marangoni F, Agostoni C, Lammardo AM, et al. (2000) Poly-
unsaturated fatty acid concentrations in human hindmilk are
stable throughout 12-months of lactation and provide a sus-
tained intake to the infant during exclusive breastfeeding:
an Italian study. Br J Nutr 84, 103–109.
Peng YM, Zhang TY, Wang Q, et al. (2007) Fatty acid compo-
sition in breast milk and serum phospholipids of health term
Chinese infants during first 6 weeks of life. Acta Paediatr 96,
Brenna JT, Varamini B & Jensen RG (2007) Docosahexaenoic
and arachidonic acid concentrations in human breast milk
worldwide. Am J Clin Nutr 85, 1457–1464.
Sanders TAB (1999) Essential fatty acid requirements of
vegetarians in pregnancy, lactation and infancy. Am J Clin
Nutr 70, Suppl., 555S–559S.
Lauritzen L, Jorgensen MH, Hansen HS, et al. (2002)
Fluctuations in human milk long-chain PUFA levels in
relation to dietary fish intake. Lipids 37, 237–244.
Kuipers RS, Fokkema MR, Smit EN, et al. (2005) High
contents of both docosahexaenoic and arachidonic acids
in milk of women consuming fish from lake Kitangiri
(Tanzania) targets for infant formulae close to our ancient
diet? Prostaglandins Leukot Essent Fatty Acids 72, 279–288.
Goyens PLL, Spilker ME, Zock PL, et al. (2006) Conversion of
a-linolenic acid in humans is influenced by the absolute
amounts of a-linolenic acid and linoleic acid in the diet
and not by their ratio. Am J Clin Nutr 84, 44–53.
WHO Expert Consultation (2004) Appropriate body-mass
index for Asian populations and its implications for policy
and intervention strategies. Lancet 363, 157–163.
diversity. AmJClinNutr 83, Suppl.,
PUFA intake among Bangladeshi women11
British Journal of Nutrition