Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Evaluation of the impact of messages to improve the diets of lactating rural Bangladeshi women on their dietary practices and the of their breast-fed infants

Laurine V. Brown, Marian F. Zeitlin, Leisa H. Weld, Beatrice L. Rogers, Karen E. Peterson. Nasreen Huq. and Stanley N. Gershoff



A community-based nutrition education intervention taught 48 Bangladeshi families with breast-feeding infants how to improve the mothers' diet. The energy adequacy of the women's diets and of 30 comparable controls averaged 65% + 14% of the FAD/WHO/ UNU requirement at baseline and declined to 55% + 7% immediately after the education (Post1) and to 52% + 6% after eight months of study (Post2). This decline was probably a seasonal effect resulting from lower food availability at Post1 and Post2. The adjusted declines in adequacy of treatments and controls did not differ at Post1 ( - 9.9% v. - 9.5%; p = .806) when behavioural changes were expected.

Adjusted declines from baseline to Post2 were significantly less for treatments than controls ( - 10.1% v. - 15.5%; p =.001), but results may have been influenced by flooding that affected food distribution and production. Arm circumferences (MUAC) of both groups remained along the fifth percentile of the international reference. No significant differences were found between the average weight for age (WAZ) or MUAC of the breast-fed children in the two groups, although a greater percentage of control children became severely malnourished (p =.011). The evaluation raises concerns about the effectiveness of nutrition education for improving the diets of poor women if given in isolation of programmes that make improvements affordable.

Nutrition condition of women and children in Bangladesh

Limited maternal breast-milk outputs, combined with limited feeding of complementary foods to infants of weaning age, contribute to the poor growth of Bangladeshi children in the first and second years of life. The children grow well within the first few months after birth, but before age 4 to 6 months weight gains falter relative to international standards [1-3]. One study documents 90% of rural girls and boys to be below the fifth percentile of the National Center for Health Statistics (NCHS) reference standards by ages 8 and 15 months respectively [4].

The little information available on maternal nutrition status and trends in Bangladesh highlights the poor condition of these women. Bangladesh is one of three countries in the world where women have a shorter life expectancy than men [5]. The overall poor maternal nutrition status is worsened by seasonal food scarcity and frequent pregnancies [6]. Widespread maternal undernutrition in Bangladesh extends to unborn children, and the vicious cycle of chronic nutritional deprivation entraps each successive generation.

Over one-fourth of all infants are born with low birth weight [7] to mothers who produce, on average, less breast-milk than well-nourished women [8-10]. Supporting these observations, a national survey found pregnant and lactating women consumed only 70% of their energy requirement and 65% of their estimated protein requirement [11]. Recent trends suggest that nutrient intakes and nutrition status have declined in recent years [12,13]. Expansion of rice and wheat production has been at the expense of other sources of food, such as pulses, fruit, vegetables, and fish in the post-Green Revolution period, resulting in a less varied and balanced diet [13].

Numerous cultural food practices compound the problem. Despite the tremendous nutritional needs of the often pregnant and lactating women, men and children eat first and the women eat the remaining food, if any [14, 15]. Whether they receive proportionately less food than others is not clear. There is convincing evidence from Bangladesh and India for food discrimination against girls under age four years [3,16-19]; however, the evidence is not strong for sex discrimination for adults, when considering total food intake (energy) throughout the day compared with requirements [16, 17, 20-24]. Studies do not always take the additional needs for pregnancy and lactation into consideration [24], however, or the quality of the diet aside from energy intake [25 26].

A rural woman's day is consumed with cooking, collecting firewood and water, cleaning and caring for children and elders. Women have no control over the food budget - the men buy and sell food in the market. Nearly 80% of women have never been to school and cannot read, write, or understand numbers at a functional level [7]. Adding to this powerlessness, the newlywed goes to live with her husband's family, often a less supportive environment than her own parents' home.

Nutrition education

Exposure to education and information can empower a woman to maximize the few resources around her for the health of her family [27, 28]. An education intervention to improve the diets of weaning-age children, conducted in the same setting as this study, resulted in greater energy intakes and weight gains of treatment children compared with controls [29]. This study, however, questions the sources of nutrition education alone in improving maternal diets under conditions of extreme poverty and possible discrimination against women in intrahousehold food distribution. Nutrition messages were designed for the Bangladesh Rural Advancement Committee's (BRAC) Child Survival Programme to improve lactating women's diets. Field observations suggested that, in general, the mothers wished to eat more during lactation. Yet, despite the project's promotion of low-cost traditional foods and inclusion of other family members in the teaching, the mothers reported that financial barriers limited their ability to comply with messages. This study evaluates the impact of the messages on the dietary adequacy and nutrition status of the women, and the growth of their breast-fed infants. A discussion of the food costs imposed by these nutrition messages is available elsewhere [30, 31].


Intervention design

From February to November 1987, BRAC and Tufts University conducted a lactation intervention in a rural site in the Harirampur subdistrict in the Manikganj district of Bangladesh. This low-lying area on the banks of the Padma (Ganges) River was selected because of its high rates of malnutrition and isolation from other development projects [32]. The intent of the intervention was to develop feasible messages through community-based behavioural trials, that would ameliorate the lactating mother's diet to improve her milk supply, her health, and the growth of the breast-fed infant. The procedures followed were in accordance with the guidelines of the Tufts University Human Subjects Committee.

The field trial model is described in detail elsewhere [31, 33]. Briefly, eight village workers (VWs) were hired from the local community. Additional staff were a supervisor, anthropologist, and nutritionist. The preliminary messages, which were developed by the VWs and technical staff in response to a community needs assessment, encouraged the lactating mother to eat an extra serving of all foods she normally ate at each meal, consume adequate fluids, wash hands and utensils before eating, and offer both breasts to the baby when breast-feeding.

The messages were then disseminated by volunteer mothers to other lactating mothers tested, and revised on the basis of observations and outcomes during the intervention. The home-based approach to teaching was participatory and family oriented, involving the mother, her husband, and her mother-in-law. Following cultural norms, male VWs met with and taught the male household members.



Using a BRAC census conducted in January 1987, we identified treatment mothers in five villages with breast-fed infants living near other families with young infants to whom they could disseminate the educational messages. Control mothers with breast-fed infants were recruited from five villages located approximately one hour by foot away from the nearest intervention village, using a BRAC baseline nutrition status survey conducted in January as a sampling frame. A total of 78 lactating mothers and their breast-fed infants (48 treatment, 30 control) were followed for approximately eight months. The infants" ages ranged from 0 to 11 months at baseline (over 80% were under 6 months) to 8 to 18 months at the final measure.

Missing data were primarily due to the child or mother being absent on follow-up visits (e.g., visiting maternal relatives) or the family's out-migration from the study area. In addition, one treatment child died from an acute illness early in the study.


Data-collection and survey instruments

Baseline anthropometric, dietary, and socio-economic data were collected on all sample households, and anthropometric, dietary, and child morbidity data were collected monthly except during serious flooding in August and September. Workers were trained in anthropometry using methods developed by Zerfas [34]. Child weight was measured with a portable 25 kg dial scale, and mother and child arm-circumference measurements were taken with the TALC tape [35]. The mother's and child's monthly 24-hour food recall was recorded on a pre-tested instrument containing common Bangladeshi foods [36, 37]. Child morbidity data consisted of mother's monthly recall of the type of illness and approximate number of days the child was sick.


Data analysis variable construction

The impact of the intervention was evaluated on the basis of changes in pre- and post-education indicators for dietary adequacy and nutrition status, after adjusting for baseline (pre-education) differences between treatment and control groups. The baseline measure was obtained during the rolling recruitment in February-March. Two post-education indicators were used, Post1 in May-June, and Post2 in October-November. Evaluating at Post1, when the average age of the children was only seven months, allowed assessment of any changes in the mothers' diet on the growth outcomes of the younger children who were predominantly breast- fed, measured against the World Health Organization (WHO) definitions [38]. This measurement was also before the severe August floods, which may have altered the potential impact of the nutrition education. Post2 allowed assessment of impact after a longer follow-up.



The changes in women's and children's mid-upper arm circumference (MUAC) measurements were evaluated from the baseline to the Post2 end point, but not at Post1 owing to a large number of missing measurements. Age and sex of children were controlled for in the child MUAC analyses.

Children's weights (available only on infants under six months at recruitment) converted to Z scores (WAZ) according to the NCHS and WHO standards for age and sex [39], were used for statistical comparisons. Actual weights and percentage of median weight-for-age (WAPM), using the same standards, were displayed for descriptive purposes.

The baseline anthropometric variables were the first measurements available during the February-March recruitment. For controls, the first anthropometric measurements were obtained during the last few days in January; however, for simplicity, the baseline is referred to as beginning in February. Post1 and Post2 variables were the last measures available during May-June and October-November respectively.


Dietary adequacy

The women's intakes of energy, protein, fat, vitamin A, and iron were computed from the monthly 24hour recalls using Bangladesh food composition tables [40] and US Department of Agriculture tables [41], and expressed as a percentage of the international recommendations [42, 43]. Because all mothers were lactating throughout the study, and test weighings for breast-milk output were not feasible, milk production and (child's consumption) was estimated for each age in months from research on a comparable Bangladeshi sample [8, 9].

In estimating the daily energy requirement, a conservative estimate of the women's basal metabolic needs (BMR) was first made on the basis of FAO/ WHO/UNU guidelines [42], using a reference 40.4 kg woman [6]. The BMR was multiplied by a factor of 1.7 for energy expenditure, derived by estimating energy requirements for the daily activities of the rural Bangladeshi woman [42]. An additional 500 kcal (2,094 kJ) was added for adequate lactation [42, 44, 45] for a total daily requirement of 2,405 kcal (10,070 kJ).

The safe daily protein requirement for the non-lactating woman was estimated to be 30.3 g, with an additional factor for lactation varying from 16 to 11 g based on the age of her child [42]. The protein content of non-animal foods was adjusted assuming 85% digestibility and 75% amino acid score [42]. The daily iron requirement was estimated at 15 mg [43]. The vitamin A requirement was 3,900 IU (1,300 g RE) if the mother's child was under six months and 3,600 IU (1,200 g RE) if the child was over six months [43, 46].

Energy intakes from complementary foods were also computed for the infants from 24-hour recalls, and expressed as a percentage of ideal requirements (with and without breast-milk estimates), using the NCHS median weight for the child's age and sex [39]. Energy requirements per kilogram of body weight for infants 0 to 12 months were derived from the FAO/WHO/UNU equation: kcal/kg/day=[(123 - 8.9 (age)) + U.59 (age2) x 1.05. For children 12 to 24 months, the estimated daily requirement was 105 kcal/kg (440/kJ/kg) [42].

The baseline dietary variables were taken as the first 24-hour recall available. The Post1 and Post2 variables were created by averaging all monthly 24hour recalls available during May-July and October-November respectively.


Child morbidity

For analysis, the mother's monthly recall of the type and duration of her breast-fed child's illness was categorized into gastrointestinal illness (diarrhoea, dysentery, vomiting), respiratory and oral infections (cold, cough, throat or mouth infections), fever, measles, and other (skin or eye infections, other minor illness). Combining all the available observations for each child, the mean percentage of days per month the child was sick with the illness in a respective category was computed and used to control for the effects of morbidity on child growth. Despite acknowledged limitations of the monthly recall for morbidity reporting, the data can be expected to distinguish very sick children from the generally healthy when it is used as a control variable.

TABLE 1. Characteristics of sample mothers and their families

Characteristic Treatment (n = 48) Control
(n = 30)
Mother's age (yrs) (mean + SD) 26 6 28 6 NSa
No. of household members (mean + SD) 7 3 6 2 NSa
No. of mother's children (mean + SD) 3 2 3 2 NSa
Mother's education (%)     NSa
none 71 77  
some primary 4 0  
completed primary 17 13  
some secondary 8 10  
Religion (%)     ***b
Moslem 85 47  
Hindu 15 53  
Owns house and lot (%) 98 60 ***b
Landless ( % ) 42 93 ***b
Tube well water source (%) 98 97 NSb
Bamboo, non-pit latrine (%) 90 90 NSb
Major source of income (%)     NSb
farmer 33 0  
agricultural/skilled labour 19 20    
government service 17 30  
business 15 20  
other 16 30  
  1. Calculated by t test.
  2. Calculated by chi-square tests.

*p < .05; ***p <.001.

Socio-economic indicators

Socio-economic indicators collected at baseline were education and literacy of household members, occupation, housing type, amount of land, solvency, source of drinking water, latrine type, and possessions. Scales for monetized household possessions or traditional wealth, agricultural wealth, and modernization/education were created for each household. Acceptable internal consistencies of the scales were confirmed using SPSSX's reliability program with Cronback's a [47]. Characteristics of sample mothers and families are presented in table 1.


Statistical methods

The impact of the education intervention was determined by examining the differences in the baseline, Post1, and Post2 measurements using analysis of covariance. Adjusted mean values were calculated for treatment and controls for anthropometric and dietary variables, controlling for wealth, mother's education, morbidity, child age and sex, mother's age and baseline nutrition status or nutrient intake. The McNemar non-parametric test for changes from pre- to post-intervention was used to determine the significance of changes in the number of severely undernourished children in each treatment group. SPSSX was used for statistical analysis. Statistical significance was defined as p < .05.


Dietary adequacy

Table 2 summarizes the adjusted energy adequacies of the mother's diets relative to estimated requirements. At baseline, mothers in the control group were consuming significantly more energy than those in the treatment group (p <.001). The majority of controls were recruited during the beginning of February, which was shortly after the major rice harvest. Owing to the continuing enrolment in the education programme, the majority of treatment mothers entered the study in March, which was generally a less plentiful month, with greater food scarcity, higher food prices, and lack of available work [15, 31]. For both groups, however, the energy adequacy of the women's diets was very poor, at only 65% + 14% (mean + SD) of their estimated needs.

The adequacy of the mothers' energy intakes declined for both groups over the first three months of the education programme (Post1 55% + 7%). Although the decline was greater for the controls, the difference was not statistically significant (p= .806), even after controlling for other significant factors. The mothers' dietary energy adequacy declined further from Post1 to Post2 measure (52% + 6%). The (adjusted) decline over the total eight months of the study was significantly greater (p= .001) for the controls than for the treatment mothers (15.5% v. 1 0.1 % ). The estimated decline in the mother's breast-milk production (based on age of child) was not significant in predicting the decline in her energy intake over time.

TABLE 2. Adjusted mean energy adequacy of lactating mothers' diets before and after nutrition education

  Base line Post 1 Post2 Change B-Post 1 Change B-Post2 Average
Energy adequacy (adjusted means)  
treatment (n = 39) (%) 60.1 54.6 55.5 -9.9 -10.1 55.8
control (n = 28) (%) 70.6 55.1 47.3 -9.4 -15.5 56.2
F for main effects 12.34** 0.06 20.99*** 0.06 12.11** 0.13
R2 0.38 0.22 0.40 0.83 0.89 0.36
F for model 12.87*** 4.35** 8.23*** 73.46*** 165.01*** 11.78***
F for covariates 13.13*** 5.78** 5.04** 97.93*** 241.47*** 17.61***
child's age 4.82* 4.25* 12.72** 4.25* 12.51** 9.02**
mother's age - - 4.67* - - -
mother's education 2456* ** 15.70***   15.70***   31.16***
mother's baseline energy - 1.96 2.10 269.33*** 482.45 *** -
agricultural wealth - - 0.34 - - -

Means adjusted by analysis of covariance for significant covariants and potential confounding variables (as listed above). Baseline measurements taken in February-March, Post1 in May-July. and Post2 in October-November. B-Post1 and B-Post2 compute the difference from baseline to Post1 and Post2 respectively The average column averages all months available.

*p < .05; **p < .01; ***p < .001;- not entered.


General characteristics of diets

The percentage of diet energy contributed by each food and nutrient group was similar for both treatment and control mothers. Figure 1 (see FIG. 1. Percentage of diet energy from various food groups in diets of lactating mothers by season, treatments, and controls combined) combines both groups to illustrate the very high percentage contributed by grain, which included rice and wheat (>80%). The proportion from grain was lowest during February-March, when the diet included more vegetables. Fruits, fish, other animal foods, and lentils were consumed in very small quantities. Fats from all food sources (oils, animal food, etc.) contributed only 5% of the total energy in the womens' high-bulk diets.

Protein and iron intakes paralleled trends in energy intake; thus additional statistical comparisons between treatments and controls for these nutrients were of no further value. In general, protein adequacy was slightly better than energy adequacy, even after adjusting for the low protein quality (60% + 8% v. 56% + 6% overall). Dietary iron intake was adequate, averaging 116% + 20% over all the seasons, assuming a requirement level of 15 ma/ day [43]. In addition to food sources analysed in this study, the drinking water, often obtained from iron tube wells, may have provided additional iron, although the absorption may be low.

Vitamin A intakes were very low, averaging only 30% + 16% of requirements throughout the study. Vitamin A consumption was highest during Post1, when carotene-rich foods such as mangoes were in season, and many dark green leafy vegetables were plentiful and inexpensive. The increase in vitamin A intake from the baseline to Post1 was significantly greater for the treatment mothers than for the controls ( p = .023); however, by Post2 the change in consumption was not significantly different between the two groups (p= .333).


Changes in nutritional status

Mothers' arm circumference

Mothers' adjusted MUAC data are shown in table 3. The average MUAC of the mothers in the treatment and control groups was similar before education, with the average measurement approximating the fifth percentile of the reference standard [48]. Over the eight months of study, MUAC remained relatively stable, with no significant group differences, controlling for other significant factors.

TABLE 3. Adjusted mean mid-upper arm circumference (MUAC) of lactating mothers before and after nutrition education

  Baseline Post2 Change B-Post2 Average
MUAC (cm) (adjusted means)
treatment (n = 39) 22.8 22.9 0.2 23.0
control (n = 25) 22.4 22.9 0.2 22.6
F for main effects 0.66 0.00 0.00 0.53
R2 0.27 0.63 0.29 0.70
F for model 4.347** 24.91** 6.05** 1.50
F for covariates 5.27** 33.21*** 8.06*** 1.99
child's age. 10.74** - - 3.51
mother's age 7.90** - - -
mother's education 4.30* - - 1.14
mother's baseline energy 9.78** - - -
mother's B-Post2 energy change 14.09*** 14.09*** - -
mother's baseline MUAC - 98.62*** 3.86 -
agricultural wealth - 0.01 0.01 -

Means adjusted by analysis of covariance for significant covariates and potential confounding variables (as listed above).
*p < .05; **p < .01; ***p < .001; - not entered.

Child weight and arm circumference

There were no significant differences in nutrition status at baseline, measured by either MUAC or WAZ, between the children whose mothers received nutrition education and those who did not. The nutrition status of both groups deteriorated over the study period, as indicated by a decline in WAZ, but the changes were not significantly different between the treatment and control children for either WAZ or MUAC (tables 4 and 5). Table 6 provides descriptive data on the adequacy of complementary food intake with and without breast-milk estimates for the sample of children with weights at all three measures. No significant differences in complementary food energy intake were found between control and treatment children. Details of the traditional complementary food composition and adequacy for a similar sample are described elsewhere [29, 31].

TABLE 4. Adjusted mean mid-upper arm circumference (MUAC) of breast-fed children before and after nutrition education

  Baseline Post2 Change B-Post
MUAC (cm) (adjusted means)  
treatment (n = 38) 12.5 14.0 1.6
control (n = 25) 12.3 13.4 1.0
F for main effects 0.13 2.21 2.21
R2 0.24 0.41 0.67
F for model 4.65** 6.61 *** 19.22***
F for covariates 6.15** 7.49*** 22.62***
child's age 16.52*** 1.28 1.28
child's sex 1.22 9.19** 9.19**
baseline MUAC _ 1.63 85.95***
mother's av. energy _ 18.29*** 18.29***
av. no. days B-Post2 _ 1.04 1.04
agricultural wealth 3.33 _ _

Means adjusted by analysis of covariance for significant covariates and potential confounding variables (as listed above). Children's mean SD ages at baseline treatment 5.2 2.7 mo, controls 2.9 1.4 mo; Post2 treatment 13.1 2.5 mo, controls 12.2 1.3 mot
**p < .01; ***p< .001;- not entered.


TABLE 5. Adjusted mean weight-for-age Z scores (WAZ) for breast-fed children before and after nutrition education

  Baseline Post1 Post2 Change B-Post1 Change B-Post2
WAZ (adjusted means)          
treatment (n = 21) -0.69 -1.96 -2.33 0.76 - 1.39
control (n = 28) -1.12 - 1.70 -2.38 - 1.02 -1.44
F for main effects 2.10 3.18 0.13 3.18 0.13
R2 0.26 0.73 0.59 0.67 0.60
F for model 3.77* 18.71*** 10.14*** 14.06*** 10.64***
F for covariates 4.32** 21.82*** 12.14*** 16.23*** 12.74***
child's age 3.68 9.20** _ 9.20** _
child's sex 5.30* 0.88 2.92 0.88 2.92
baseline WAZ _ 27.77*** 31.93*** 35.42*** 32.52***
diarrhoea _ _ 7.58* * _ 7.58**
child's Post2 food energy _ _ 4.965* _ 4.965*
mother's Post1 energy _ 2.90 _ 2.90 _
mother's av. energy _ _ 19.69*** _ 19.69***
av. no. days B-Post1 _ 23.88*** _ 23.88*** _
agricultural wealth 4.92* _ _ _ _

Means adjusted by analysis of covariance for significant variates and potential confounding variables (as listed above). See table 6 for sample children's ages and dietary energy adequacy.

*p < .05; **p < .01 ***p < .001;- not entered.

We define moderate and severe malnutrition using the UNICEF cutoff of -2 SD and -3 SD respectively below the international reference for the child's weight for age and sex [49]. Figure 2 (see FIG. 2. Weight gain in breast-fed children of families receiving nutrition education compared with controls, and the UNICEF cut-off for moderate malnutrition, which averages the median weights of the male and female reference child [39] (average child age Feb. Mar. 3.21.4 months, May-July 7.31.9 months, Oct.-Nov. 11.81.6 months)) illustrates the actual weight gain for treatment and control children relative to the cut-off for moderate malnutrition for the reference population. Figure 3 (see FIG. 3. WAPM in breast-fed children of families receiving maternal nutrition education compared with controls (average child age Feb.-Mar. 3.21.4 months, May -July 7.31.9 months, Oct.-Nov. 11.81.6 months)) illustrates the change in WAPM. Of the 49 children present for all three weighings, only 6% were moderately or severely malnourished at baseline. This increased to 51% by Post1 and 71% by Post2. The increase in the number of children who became severely malnourished from baseline to Post2 (0-25%; p = .016) was significant in the control group, but not in the treatment group (0-10%; p = .500).

No significant treatment group differences were apparent in child anthropometry, but the energy adequacy of the mother's diet was significantly related to her child's WAZ and MUAC. Sex of the child was a more significant prediction of child WAZ at baseline (i.e., in the 3-month child) than at Post2 (average age 12 months), with boys faring slightly better than girls. Girls, however, showed less decline in WAZ over the study period than males. At Post2, no significant sex differences in WAZ were apparent.

For MUAC, which includes a larger sample of children, there were no significant sex differences at baseline after controlling for other significant factors. By Post2, however, sex differences were apparent, with boys having larger MUACs. This is expected, since the measures are not standardized against a norm for age and sex. At each point, the average MUAC for the treatment group was slightly (although not significantly) higher than the controls simply as an artefact of the slightly older age in this sample (also not significant).

Contents - Previous - Next