The use of linear programming to determine whether a formulated

complementary food product can ensure adequate nutrients

for 6- to 11-month-old Cambodian infants1–3

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

Jutta KH Skau, Touch Bunthang, Chhoun Chamnan, Frank T Wieringa, Marjoleine A Dijkhuizen, Nanna Roos,
and Elaine L Ferguson

ABSTRACT children, in resource constrained environments are often low in

Background: A new software tool, Optifood, developed by the vitamin A, iron, zinc, and calcium (1, 3, 4).
WHO and based on linear programming (LP) analysis, has been At this time in life, it is critical to provide a nutritionally
developed to formulate food-based recommendations. adequate diet to avoid long-term negative consequences in adult
Objective: This study discusses the use of Optifood for predicting life on health and human resource potential (5). To help develop
whether formulated complementary food (CF) products can ensure population-specific food-based recommendations for the purpose
dietary adequacy for target populations in Cambodia. of improving complementary feeding diets, a mathematical ap-
Design: Dietary data were collected by 24-h recall in a cross- proach based on linear programming (LP)4 analysis was recently
sectional survey of 6- to 11-mo-old infants (n = 78). LP model developed (6–8), which will soon be available to download from
parameters were derived from these data, including a list of foods, the WHO website as the user-friendly software–Optifood. LP
median serving sizes, and dietary patterns. Five series of LP anal- has also proven useful at determining the cost-effectiveness of
yses were carried out to model the target population’s baseline diet food fortification to meet nutrient requirements in populations
and 4 formulated CF products [WinFood (WF), WinFood-Lite (9), at formulating new food products based on local commod-
(WF-L), Corn-Soy-Blend Plus (CSB+), and Corn-Soy-Blend Plus ities (10), and at determining the need for supplements to ensure
Plus (CSB++)], which were added to the diet in portions of 33 g/d adequate nutrient intake in specific vulnerable populations (11).
dry weight (DW) for infants aged 6–8 mo and 40 g/d DW for An option not yet explored is how LP can be used, during initial
infants aged 9–11 mo. In each series of analyses, the nutritionally nutrition intervention program planning phases, to predict
optimal diet and theoretical range, in diet nutrient contents, were whether specially formulated complementary food (CF) prod-
determined.
ucts will cover nutrient gaps in local CF diets.
Results: The LP analysis showed that baseline diets could not
In Cambodia, the prevalence of malnutrition among preschool
achieve the Recommended Nutrient Intake (RNI) for thiamin, ribo-
children has remained unchanged for the past 10 y. The 2010
flavin, niacin, folate, vitamin B-12, calcium, iron, and zinc (range:
Cambodian Demographic and Health Survey concluded that 40%
14–91% of RNI in the optimal diets) and that none of the formu-
of children younger than 5 y were stunted; 28% were underweight
lated CF products could cover the nutrient gaps for thiamin, niacin,
iron, and folate (range: 22–86% of the RNI). Iron was the key
1
limiting nutrient, for all modeled diets, achieving a maximum of From the Department of Nutrition, Exercise and Sports, University of
only 48% of the RNI when CSB++ was included in the diet. Only Copenhagen, Copenhagen, Denmark (JKHS, MAD, and NR); the Depart-
ment of Fisheries Post-harvest Technologies and Quality Control, Fishery
WF and WF-L filled the nutrient gap for calcium. WF-L, CSB+, and
Administration, Ministry of Agriculture, Forestry and Fisheries, Phnom
CSB++ filled the nutrient gap for zinc (9- to 11-mo-olds).
Penh, Cambodia (TB and CC); Institut de Recherche pour le Développe-
Conclusions: The formulated CF products improved the nutrient ade- ment, Montpellier, France (FTW); and the Faculty of Epidemiology and
quacy of complementary feeding diets but could not entirely cover the Population Health, London School of Hygiene and Tropical Medicine, Lon-
nutrient gaps. These results emphasize the value of using LP to evaluate don, United Kingdom (ELF).
2
special CF products during the intervention planning phase. The WF Supported by the Consultative Research Committee for Development
study was registered at controlled-trials.com as ISRCTN19918531. Research, Ministry of Foreign Affairs, Denmark, and the Sustainable Micro-
Am J Clin Nutr 2014;99:130–8. nutrient Interventions to Control Deficiencies and Improve Nutritional Status
and General Health in Asia (SMILING) Project.
3
Address correspondence and reprints requests to JKH Skau, Department
INTRODUCTION of Nutrition, Exercise and Sports, University of Copenhagen, Rolighedsvej
Adequate nutrition in early life is fundamental for physical and 30, 1958 Frederiksberg C, Denmark. E-mail: [email protected].
4
Abbreviations used: ASF, animal-source foods; CF, complementary food;
cognitive development (1). Young children are particularly
CSB+, Corn-Soy-Blend Plus; CSB++, Corn-Soy-Blend Plus Plus; DW, dry
vulnerable to inadequate nutrition during the complementary weight; LP, linear programming; RNI, Recommended Nutrient Intake; WF,
feeding period from 6 to 24 mo of age when nutrient needs are WinFood; WF-L, WinFood-Lite.
high to support rapid rates of growth, and breast milk alone will Received August 13, 2013. Accepted for publication October 4, 2013.
not meet their nutritional requirements (2). The diets of young First published online October 23, 2013; doi: 10.3945/ajcn.113.073700.

130 Am J Clin Nutr 2014;99:130–8. Printed in USA. Ó 2014 American Society for Nutrition
 LINEAR PROGRAMMING AND FOOD-BASED INTERVENTIONS 131
and 11% were wasted. The stunting rate rises as the age at which a scale (Dougguan Machinery Import and Export Co Ltd; pre-
the child is introduced to CF increases (12). cision to 0.1g) and recorded. To convert composite dishes into
The overall aim of the WinFood project was to develop nutri- their individual ingredients, individual recipes were obtained
tionally improved food products for infants and young children from the Cambodian National food-consumption survey, which
based on the use of locally available foods. The WinFood study was carried out 6 mo later in Cambodia (unpublished data
developed 2 different formulated CF products: WinFood (WF) and available from the Fishery Administration, Ministry of Agri-
WinFood-Lite (WF-L). WF contains rice, dried small fish, and culture, Forestry and Fisheries, accessed September 2012). For
edible spiders; WF-L contains rice and dried small fish and is each dish, the author (JKHS) selected 5 similar recipes, which
fortified with a mineral and vitamin premix. These products were had the same name and contained the same raw ingredients as the
compared with Corn-Soy-Blend Plus (CSB+) and Corn-Soy-Blend recalled recipe, although the species of fish in the fish dishes
Plus Plus (CSB++) in a single-blinded randomized trial with growth could differ. The average recipes were created by calculating the

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

and iron status as the primary outcomes (13). CSB+ and CSB++ are mean weight for each raw ingredient divided by the mean
fortified blended foods, distributed by World Food Program and weight of all the cooked dishes. Recipes, for 3 dishes (pork
other organizations, for the prevention and treatment of moderate porridge, chicken porridge, and fish porridge), which were
acute malnutrition, respectively. The objective of this study was to consumed by 37% of the infants, were not available from the
evaluate, with the use of the LP approach, whether these specially national consumption survey. These dishes are often procured as
formulated CF products would ensure a nutritionally adequate CF ready to eat foods from local markets. Therefore, 3 local cooks
diet for 6- to 11-mo-old Cambodian infants. were asked to prepare each of the 3 porridges. All raw ingredients
and the final cooked porridge were weighed and recorded. The
recipe conversion factor, for each ingredient, was calculated (raw
SUBJECTS AND METHODS ingredient weight/cooked porridge weight) and averaged across
Study design cooks to produce the ingredient conversion factors for each
average porridge recipe.
This study was based on secondary cross-sectional dietary and Anthropometric measurements (weight and length) were made
anthropometric data collected from 6- to 11-mo-old infants living in by 2 trained enumerators. Weight was measured to the nearest 100 g
7 communes in PeaRieng district (Prey Veng Province, Cambodia) with an electronic scale (SECA scale), and length was measured to
in November and December 2010. In this study, these data were the nearest 0.1 cm (wooden length board was borrowed from World
analyzed by using LP to theoretically determine whether 4 for- Food Program). General sociodemographic data were also col-
mulated CF products would ensure dietary adequacy for 6- to 8-mo- lected by using a structured questionnaire.
old and 9- to 11-mo-old infants. Five series of LP models were run:
1 baseline model (without formulated food products) and then 1
The CF products
series each for the 4 formulated food product models.
The WinFood products, WF and WF-L, were developed and
produced in Cambodia based on locally available foods. Both
Subjects and sampling
products were based on white rice, which also is used in the rice
In the original cross-sectional survey, all infants aged 6–15 mo porridge bor-bor, traditionally used for complementary feeding.
(n = w800) from the 7 communes in PeaRieng district were in- Both products contained a proportion of animal-source food
vited to participate. Of those who participated (n = 567), 110 in- (ASF). The WF product was 14% dry weight (DW) ASF (small
fants aged 6–15 mo were randomly selected to participate in a indigenous fish species with bones (Esomus longmanus and
dietary-assessment substudy. All dietary data collected from the Paralaupuca typus) and the edible spider (Haplopema sp.). This
6- to 11-mo-old infants in this substudy (n = 78) were included in spider is traditionally consumed in Cambodia and is available in
the current study. Ethical approval was obtained from the National local food markets. The WF-L contained 10% DW ASF (a va-
Ethics Committee for Health Research, Ministry of Health, the riety of small fish species) and was fortified with a mineral and
Royal Government of Cambodia, reference no. 151 NEHR (28 vitamin premix (Table 1). The fish in both products was initially
October 2010), and consultative approval was obtained from the sun dried, and the spiders were heat dried in an oven. Both
Danish National Committee on Biomedical Research Ethics. products were processed by grinding ingredients to powders,
mixing, and precooking by extrusion. The products were tested
Survey data for microbial loads at the Pasteur Institute, Phnom Penh. The
CSB products were provided by the World Food Program. The
Quantitative dietary data were collected by using a single 24-h
CSB+ (14) was distributed to all children in a Mother-Child-
dietary recall by 8 trained assistants who were fluent in the local
Health program, and CSB++ (15) was distributed for the treat-
language. At a population level, the 24-h dietary recalls were
ment of moderately acutely malnourished children in Cambodia
collected on all days of the week except Saturday and Sunday.
(16). The food composition and nutrient composition of all
During the interview, primary caregivers were asked to recall all
4 formulated CF products are summarized in Table 1 and
foods and beverages consumed by their infants in the past 24 h. To
Table 2.
help identify the foods consumed, pictures of food items and food
models were shown to the caregiver so that she or he could point
out the food items. The names of local dishes consumed were also Analysis and data preparation
recorded. The amount of each food item consumed was estimated All analyses were done on infants stratified into 2 age groups
from real food models. Mothers were asked to show the amount (ie, 6–8 and 9–11 mo) because of age group differences in en-
of food consumed by her child, which was then weighed on ergy requirements.
132 SKAU ET AL
TABLE 1 sumed by $5% of the children and nutrient-dense foods con-
Food composition of the intervention foods per 100 g dry weight1 sumed by ,5% of the children (Table 3). The serving size, for
WF2 WF-L2 CSB++ CSB+3 each food, was defined as the median serving size for children
who consumed the food in each age group. To define model
% % % %
parameters for weekly food-consumption patterns, for food
Rice, white, milled 77 79 — —
Fish, Esomus longimanus4 6.1 — — —
groups, and subfood groups with the use of 24-h dietary-recall
Fish, Paralaubuca typus4 6.1 — — — data, 2 assumptions were made: 1) food patterns remain rela-
Spider, Haplopelma sp.4 1.8 — — — tively statistic over 7 d for each individual and 2) overestimation
Fish mix4 — 9.5 — — errors are balanced by underestimation errors at the population
Mineral and vitamin mix — 1.7 1.7 1.4 level. In contrast, constraints on the maximum number of
Vegetable oil 4.8 4.8 3.0 8.5 servings per week, for individual foods, were defined by the
Sugar 4.8 4.8 9.0 8.5

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

percentage of children who consumed them by using the dis-
Maize (white or yellow) — — 58 65
tribution of a sum of binomial random variables. Specifically,
Dehulled soya — — 20 —
Whole soya — — — 20
the maximum number of servings per week was defined as 1, 2,
Skim milk powder — — 8 — 3, 4, 5, 6, or 7 when 0–5%, 6–12%, 13–22%, 23–34%, 35–47%,
1
48–65%, and 66–100%, respectively, of the children consumed
CSB+, Corn-Soy-Blend Plus; CSB++, Corn-Soy-Blend Plus Plus; WF,
the food. The nutrient contents per 100 g, for each modeled food
WinFood; WF-L, WinFood-Lite.
2
WF and WF-L products were precooked by extrusion, similar to the
item, were mainly obtained from the USDA (22), Vietnamese
processing of CSB products. food-composition table (23), and Thai food-composition table
3
Included sugar added by WFP-Cambodia and oil distributed in sepa- (24). Energy constraints were estimated by using the mean body
rate sachets to be added to the daily rations following WFP product speci- weights of each age group, and the FAO/WO/UNU algorithms
fications version 1.1 2010 (14). were used to calculate the average energy requirements of
4
Edible parts of the fish were used, obtained by traditional cleaning breastfed children (25). This would ensure that realistic diets
practices in Cambodia. Edible parts include bones and head (17). For spi- were selected.
ders, all parts were included as edible, following traditional consumption
In the CF product models, the portion size of CF product was
practices.
33 g/d DW for the 6- to 8-mo-olds and was 40 g/d DW for the
9- to 11-mo-olds. These serving sizes were based on the median
Development of modeled diets
All analyses were done by using Optifood (18). Five series of LP
models were developed to model baseline diets and then baseline TABLE 2
diets plus one of each of the intervention foods (WF, WF-L, CSB+, Nutrient composition of the intervention foods per 100 g dry weight1
and CSB++). In each series of analyses, the LP models from
WF2 WF-L CSB++ CSB+3
Optifood’s modules I to III (18) were used to 1) check model
parameters (Module I analyses), 2) formulate the nutritionally Energy (kcal) 474 428 458 482
“best diet” for each target group (Module II analyses), and 3) Protein (g) 15.4 12.6 16.8 14.6
estimate the tails of each nutrient intake distribution at baseline Fat (g) 10.3 9.2 10.7 16
Vitamin C (mg) 0.2 100 100 100
and then after successful adoption of each of the intervention
Thiamin (mg) 0.2 0.1 0.1 0.1
foods [Module III analyses; which minimizes (worst-case sce- Riboflavin (mg) 0.1 0.5 0.5 0.5
nario) and maximizes (best-case scenario) the nutrient content of Niacin (mg) 5.2 4.8 4.8 4.8
modeled diets for each nutrient] (18). The nutrient goals (Module Vitamin B-6 (mg) 0.4 1.7 1.7 1.7
II analyses) and dietary adequacy (Module III analyses) were Folate (mg) 12 60 60 60
defined by using the FAO/WHO Recommended Nutrient Intakes Vitamin B-12 (mg) 2 2 2 2
(RNIs), for all nutrients except zinc (19, 20). For zinc, the In- Vitamin A (mg) 35 166 166 166
ternational Zinc Nutrition Consultative Group RNIs were used Calcium (mg)4 5701 6311 2771 1731
Total iron (mg)4 4.21 6.31 10.51 9.91
(21). Moderate bioavailability was assumed for zinc, and 10%
Zinc (mg)4 4.51 5.21 7.01 6.61
absorption was assumed for iron.
1
“Problem nutrients” were defined as nutrients that did not Values for macronutrients and minerals were analyzed in samples of
achieve 100% of their RNI in the modeled “best-case scenario” all 4 foods. The energy contents were calculated from macronutrient con-
tents. WF-L, CSB++, and CSB+ were fortified with the same mineral and
diets (ie, Module III; upper tail of the modeled nutrient intake
vitamin mix following the specifications for CSB+ and CSB++ from WFP
distribution). Dietary adequacy, for each nutrient (ie, accepted specifications version 1.1 2010 (14, 15). CSB+, Corn-Soy-Blend Plus;
percentage of the population at risk of an inadequate nutrient CSB++, Corn-Soy-Blend Plus Plus; WF, WinFood; WF-L, WinFood-Lite.
intake), was defined as .65% of its RNI in the “worst-case 2
The vitamin values for WF were calculated from food-composition
scenario” analyses (Module III; lower tail of the modeled nu- values of 77% rice and 14% fish. The nutrient composition of rice and fish
trient intake distribution). was obtained from the USDA (rice, USDA:50446; fish, USDA15089) (22).
3
The estimated energy content of CSB+ includes oil distributed to be
Preparation of model parameters added at the time of preparation.
4
Values for calcium, iron, and zinc are analyzed values. WF-L, CSB+,
Data from the dietary assessment survey defined the model and CSB++ were added mineral premix, which should contribute 130 mg Ca
parameters, which included the list of foods modeled, the food (added as mono- or di-calcium phosphate), 6.5 mg Fe (added as 4.0 mg
serving sizes, and food patterns (Tables 3 and 4). The list of ferrous fumerate and 2.5 mg iron-sodium EDTA), and 5 mg Zn (added as
foods included in each model was noncondiment foods con- zinc oxide).
 LINEAR PROGRAMMING AND FOOD-BASED INTERVENTIONS 133
TABLE 3
Foods consumed by 6- to 11-mo-old infants, median serving sizes, and maximum servings per week
6–8 mo 9–11 mo

Serving size1 Maximum consumption2 Serving size1 Maximum consumption2

g/d servings/wk g/d servings/wk

Fruit (snack)
Plantains, raw 47 5 74 3
Pineapples, raw 5 3 — —
Limes, raw — — 4 1
Papayas, unripe, raw — — 6 2
Tamarinds, raw 2 1 2 2

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

Added fats
Oil, sunflower 1 7 1 7
Added sugars
Sugar, white 1 7 1 7
Sugar, brown 1 7 1 7
Staples
Rice, refined, raw 33 14 55 14
Bakery
Bread, wheat, refined 22 2 8 1
Animal-source foods
Eggs, duck, raw 16 1 8 1
Pork, lean, raw 7 6 10 3
Pork, spareribs, raw 18 1 3 1
Chicken, breast meat, raw 3 2 11 4
Beef, lean, raw — — 15 1
Large fish I3 3 3 6 1
Large fish II3 13 4 3 1
Large fish III3 — — 23 3
Shrimp, dried 1 2 1 3
Fermented fish 1 7 1 3
Vegetables
Cabbage, Chinese, raw 23 4 21 2
Morning glory, raw 11 5 8 4
Tamarind leaves, raw 1 5 1 1
Gourd, sponge, raw 11 2 3 3
Gourd, wax, mixed variety, raw 24 2 6 3
Tomatoes, red, ripe, raw 6 2 — —
Water lily, stern, raw — — 12 3
Onion, fresh, raw — — 20 3
Yard long bean, green, raw — — 1 2
Snacks
Candy, jelly pieces 33 1 29 1
Potato chips 7 7 8 3
Human milk4 576 7.1 541 7.1
Formulated CF5 products6 33 14 40 14
1
Values are median serving sizes estimated from the observed intakes of consumers.
2
The lower limit for number of servings/wk for all food items was 0, except for human milk (6.9 servings/wk) and the
intervention products (7 servings/wk).
3
Large fish species eaten without bones include a variety of fish species in Cambodia.
4
The amount of breast milk was estimated from values published by WHO (1998) (3).
5
CF, complementary food.
6
Corn-Soy-Blend Plus, Corn-Soy-Blend Plus Plus, WinFood, and WinFood-Lite.

rice serving size for 6- to 8-mo-olds and 72% of the median rice and 9- to 11-mo-olds, respectively, were entered in all models
serving size for 9- to 11-mo-olds, assuming that the CF product (3).
would replace all rice meals for the younger age group and 75% The dietary data were prepared in Excel 2010 (Microsoft
of the meals for the older age group. The value of 40 g/d DW was Corporation). The median portion sizes for consumers, and the
chosen based on recommendations on the daily portion size of percentage of children who consumed each food were calculated
DW-processed fortified CF for infants aged 6–11 mo (26). in MS Access 2010. The LP models were run by using the WHO
The mean breast-milk intakes for the study population were Optifood tool v.3.1.1. Anthropometric z scores were calculated
unknown. Therefore, published average breast milk intakes of based on WHO’s 2006 Child Growth Standards (27) with the use
403 kcal/d (576 g/d) and 379 kcal/d (541 g/d) for 6- to 8-mo-olds of Anthro v.3.1. A child was categorized as wasted, underweight,
134 SKAU ET AL
TABLE 4
Food-group goals (average servings/wk), food-group constraints (minimum and maximum servings/wk), and food-subgroup constraints (minimum and
maximum servings/wk) for the 6- to 8-mo and 9- to 11-mo age groups1
6–8 mo 9–11 mo

Minimum Average Maximum Minimum Average Maximum

Food group consumption consumption2 consumption consumption consumption2 consumption

servings/wk servings/wk
Added fats 0 0.1 7 0 0.1 7
Added sugars 0 0.1 9 0 7 9
Bakery 0 0.1 2 0 0.1 1
Fruit 0 0.1 7 0 0.1 7

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

Vitamin C–rich fruit 0 — 5 0 — 6
Other fruit 0 — 4 0 — 2
Vegetables 0 7 9 0 7 14
Vitamin A–source dark-green — — 9 — — 5
leafy
Other vegetables — — 11 — — 7
Condiment vegetables — — — — — 3
Savory snack 0 0.1 7 0 7 3
Sweetened snacks and desserts 0 0.1 1 0 0.1 1
ASF 0 7 21 0 14 28
Eggs 0 — 1 0 — 1
Fish without bones 0 — 7 0 — 5
Pork 0 — 7 0 — 4
Red meat 0 — 0 — 1
Poultry 0 — 2 0 — 4
Seafood 0 — 2 0 — 3
Small, whole fish, with bones 0 — 7 0 — 3
Human milk 6.9 7 7.1 6.9 7 7.1
Staple (rice) 4 7 14 4 7 14
Snack 0 7 10 0 7 10
Formulated CF products3 7 11 14 7 11 14
1
ASF, animal-source foods; CF, complementary food.
2
Average serving sizes were based on the observed median number of servings of foods in each food group. Values of 0.1 represent a median of 0.
Nonzero values were entered in the models to avoid a division by zero.
3
Corn-Soy-Blend Plus, Corn-Soy-Blend Plus Plus, WinFood, and WinFood-Lite.

or stunted if his or her z score for weight-for-length, weight-for- infants and 29 food items were included in the models for the 9- to
age, or length-for-age were less than –2 SD. The population- 11-mo-old infants (Table 3). The excluded food items were
descriptive statistics were analyzed in Stata 12 for Windows condiments and rarely consumed food items, which were eaten
(StataCorp). only once and were not of high nutrient value. The most common
foods consumed across the 2 groups were rice (99%), sunflower
RESULTS
oil (55%), potato chips (49%), and pork (40%). For the 9- to
11-mo-old infants, different species of large fish without bones
Characteristics of respondents and various fermented fish products were also commonly con-
sumed (.35%).
The mean body weight and calculated average energy re-
Most infants consumed rice, vegetables, and ASF every day
quirements for the 6- to 8-mo-old (n = 35) and 9- to 11-mo-old
(Table 4). The 9- to 11-mo-old infants also consumed added fats,
(n = 43) infants were 7.5 kg and 578 kcal/d and 8.1 kg and 624
sugar, and savory snacks on a daily basis and twice as many
kcal/d, respectively. Overall, 10% of the infants were wasted,
different types of ASF as the 6- to 8-mo-old infants. Fruit and
17% were underweight, and 14% were stunted. Only 6% of the
sweetened snacks or desserts were rarely consumed by any of
mothers had never attended school, 60% had a primary school-
these infants (Table 4).
level education, and 34% had a junior or senior high school–level
education. The main primary source of income in the households
was farming, mainly rice crops (60%). All infants, except one Serving sizes
were partially breastfed.
In the 6- to 8-mo-old group, the serving sizes varied from 1 g/d
for white sugar to 47 g/d for plantain, but most foods (n = 17;
Food patterns 71%) had serving sizes ,15 g/d. In the 9- to 11-mo-old group,
Overall, 60 different food items were reported in the dietary the serving sizes varied from 1 g/d for white sugar to 74 g/d for
recalls collected from both age groups. From the list of 60 foods, plantain, and again, most of the foods (n = 24; 80%) had serving
24 food items were included in the LP models for 6- to 8-mo-old sizes ,15 g/d (Table 3).
 LINEAR PROGRAMMING AND FOOD-BASED INTERVENTIONS 135
TABLE 5
Number of servings per week by food group in the best diet for the baseline diet and each diet with a CF product1
6–8 mo 9–11 mo

Baseline WF WF-L CSB++ CSB+ Baseline WF WF-L CSB++ CSB+

servings/wk servings/wk
Fruit 7 3 3 0 3 7 5 5 3 5
Added sugars 0 0 0 0 0 0 0 0 0 0
Vegetable 9 9 9 9 9 14 14 14 14 14
Savory snacks 7 0 0 0 0 3 0 0 0 0
Bakery and breakfast cereals 2 0 0 0 0 1 1 1 0 0
Meat, fish, and eggs 21 21 21 20 21 21 16 11 12 15

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

Grains and grain products 4 0 0 0 0 5 0 0 0 0
Human milk 7 7 7 7 7 7 7 7 7 7
Formulated CF products2 — 7 8 8 7 — 8 9 8 8
1
The “best diet” was the diet that came as close as possible to meeting the nutrient Recommended Nutrient Intake
goals for protein, vitamin C, thiamin, riboflavin, niacin, vitamin B-6, folate, vitamin A, calcium, iron, and zinc while
respecting modeled energy and food-pattern constraints. CF, complementary food; CSB+, Corn-Soy-Blend Plus; CSB++,
Corn-Soy-Blend Plus Plus; WF, WinFood; WF-L, WinFood-Lite.
2
WF, WF-L, CSB++, and CSB+.

The food pattern of the optimal modeled diet (except WF), niacin, calcium (except CSB+), and zinc by using
A total of 16 food items were selected in the nutritionally optimal the criteria of achievement of .65% RNI in the worst-case sce-
baseline diet for the 6- to 8-mo-old group, and 22 food items were nario (Table 6 and Table 7).
selected in this diet for the 9- to 11-mo-old infants. The number of
servings selected of ASF, vegetables, fruit, and bakery products DISCUSSION
exceeded the median number observed for these 2 populations We showed how LP can be used to investigate whether 4
(Table 4 and Table 5), whereas fewer than the median numbers of different CF products could contribute to filling nutrient gaps in
servings of rice were selected. The number of servings of savory the local diets of 6- to 11-mo-old Cambodian infants and thereby
snacks selected, in the nutritionally optimal diet, was above the help ensure dietary adequacy. The LP modeling indicated that the
median number observed for the 6- to 8-mo-old but below this products provided in realistic servings would improve the nu-
number observed for the 9- to 11-mo-old (Tables 4 and 5). When tritional quality of the diet but not ensure adequate intakes of all
the special complementary feeding products were included in the micronutrients. Iron, thiamin, and folate remained problem nu-
models, in the optimal diets the number of servings/wk of fruit for trients in all modeled diets. They also suggest that WF-L and CSB
both age groups and of ASF for the 9- to 11-mo-old group de- ++ are slightly superior to WF and CSB+ at ensuring dietary
clined; for both age groups, the special complementary feeding adequacy, because they ensured .65% of the daily requirement
products replaced rice in the diets. Between 7 and 9 servings/wk of for 8 of 11 micronutrient, whereas WF and CSB+ ensured di-
these formulated CF products were selected in each optimal etary adequacy for 7 of 11 micronutrients. These findings high-
modeled diet across both groups (Table 5). light the value of modeling specific CF products, when planning
a nutrition intervention, to determine whether they are likely to
meet nutritional requirements for multiple micronutrients in a
Nutrients target population’s diet.
Vitamins A, B-6, and C were the only nutrients that, in the The modeled optimal baseline diets clearly suggest that, for
nutritionally optimal baseline diet, met their daily nutrient re- any improved CF product, it would be a challenge to cover the
quirements (19) in both age groups. Without special comple- nutrient gaps. The nutrient contents of even the nutritionally best
mentary feeding products, all other micronutrients were problem diets, for these specific populations, were low for a range of
nutrients. The highest levels achieved (ie, best-case scenario nutrients. These results agree with those of a previous dietary
levels) for these micronutrients ranged from 14% (iron) to 93% survey done in villages on the outskirts of Phnom Penh, which
(vitamin B-12) of their RNIs (Table 6). The formulated com- showed that calcium, iron, and zinc were consistently below the
plementary feeding products increased the number of micro- RNI (28).
nutrients that achieved recommended levels in the optimal diets Remarkably, vitamin A was not identified as a problem nutrient
from 4 micronutrients at baseline to 6 (WF and CSB+) or for this particular population. The baseline diet in both age groups
7 (WF-L and CSB++) of the 11 micronutrients modeled, for the would ensure .81% of the required vitamin A intake in the
6- to 8-mo-old and to 7 (WF), 8 (CSB+ and CSB++), or 9 (WF- worst-case scenario, which increased to .98% when the diets
L) of the 11 micronutrients modeled for the 9- to 11-mo-olds. included WF-L, CSB++, or CSB+. One of the main food sources
Iron, thiamin, and folate remained below the RNI in all modeled of vitamin A in these modeled diets was breast milk, which was
diets, ie, remained “problem nutrients.” However, for other nu- included in every modeled diet in a specified amount. The re-
trients, the results suggest that daily consumption of the special sults for vitamin A will therefore depend on maternal vitamin A
CF products would ensure that the daily requirements were met status, because it will influence breast-milk vitamin A concen-
for vitamin A, vitamin C, vitamin B-12, vitamin B-6, riboflavin trations (29). Furthermore, actual breast-milk concentrations
136 SKAU ET AL
TABLE 6
The optimal, best-case scenario, and worst-case scenario diets for each nutrient expressed as a percentage of their recommendations (Recommended Nutrient
Intakes) for the baseline diet and each diet with the formulated food products in 6- to 8-mo-old infants1
Vitamin Vitamin Vitamin
Protein2 Vitamin C3 Thiamin3 Riboflavin3 B-33 B-63 Folate3 B-123 Vitamin A3 Calcium3 Iron4 Zinc5

Baseline, optimal diet6 141 135 69 80 64 103 69 91 140 52 14 48

Best-case scenario7 147 149 71 80 66 111 71 93 144 57 14 50
Worst-case scenario8 95 97 36 57 42 42 41 41 87 47 6 39
WF, optimal diet6 168 133 74 74 86 94 55 183 121 100 22 73
Best-case scenario7 170 148 77 75 88 102 60 193 140 107 24 73
Worst-case scenario8 127 97 48 59 69 67 41 140 90 93 17 66
WF-L, optimal diet6 161 223 67 106 86 257 80 191 148 102 33 84

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

Best-case scenario7 165 257 73 109 88 276 83 199 155 109 34 86
Worst-case scenario8 118 164 42 89 66 210 61 135 100 90 25 72
CSB++, optimal diet6 175 227 65 103 83 245 78 185 147 77 47 97
Best-case scenario7 178 259 71 106 86 260 81 191 152 80 50 100
Worst-case scenario8 133 164 41 89 66 210 61 135 100 69 40 86
CSB+, optimal diet6 163 211 65 101 81 234 77 179 146 67 43 90
Best-case scenario7 165 243 69 104 83 248 79 186 151 70 45 93
Worst-case scenario8 124 164 41 87 65 209 61 135 100 60 37 83
1
CSB+, Corn-Soy-Blend Plus; CSB++, Corn-Soy-Blend Plus Plus; WF, WinFood; WF-L, WinFood-Lite.
2
WHO/FAO/UNU protein and amino acid requirements in human nutrition (20).
3
FAO/WHO-recommended nutrient intake (19).
4
FAO/WHO-recommended nutrient intake, assuming 10% absorption (19).
5
International Zinc Nutrition Consultative Group, assuming moderate absorption (21).
6
The optimal diet formulated by using goal programming (Optifood Module II).
7
Each diet sequentially maximizes each micronutrient (Optifood Module III; “best-case scenario”).
8
Each diet sequentially minimizes each micronutrient (Optifood Module III; “worst-case scenario”).

were not measured in this study. An overestimation of average recent national data for vitamin A status is lacking. The other
breast milk intakes would affect these estimates of dietary vi- important food source of vitamin A was morning glory. Morning
tamin A adequacy. For these reasons, the results for vitamin A glory is a commonly consumed green leafy vegetable in Cam-
should be interpreted with caution, particularly given some ev- bodia. Traditional CF made from food items rich in vitamin A is
idence of vitamin A deficiency in the country (30), although the reported to be moderately high in Cambodia (12), which would

TABLE 7
The optimal, best-case scenario, and worst-case scenario diets for each nutrient expressed as a percentage of their recommendations (Recommended Nutrient
Intakes) for the baseline diet and each diet with the formulated food products in 9- to 11-mo-old infants1
Vitamin Vitamin Vitamin
Protein2 Vitamin-C3 Thiamin3 Riboflavin3 B-33 B-63 Folate3 B-123 Vitamin A3 Calcium3 Iron4 Zinc5

Baseline, optimal diet6 154 138 64 75 83 118 71 89 135 52 14 51

Best-case scenario7 159 143 64 76 84 119 72 89 144 53 14 54
Worst-case scenario8 98 97 39 56 47 48 40 39 81 44 7 42
WF, optimal diet6 178 137 76 73 101 108 62 220 136 114 28 86
Best-case scenario7 190 143 77 75 117 136 69 228 147 121 29 88
Worst-case scenario8 130 97 52 58 78 77 40 158 85 100 21 72
WF-L, optimal diet6 170 247 68 118 102 335 93 225 153 117 42 101
Best-case scenario7 183 270 70 120 117 356 96 235 162 123 44 104
Worst-case scenario8 120 164 45 95 76 252 65 153 98 96 30 80
CSB++, optimal diet6 188 252 66 114 100 321 90 217 144 83 61 119
Best-case scenario7 200 271 67 116 113 334 94 225 161 84 65 123
Worst-case scenario8 136 163 44 94 75 250 64 153 98 71 48 97
CSB+, optimal diet6 176 239 65 111 100 305 88 209 143 69 55 109
Best-case scenario7 185 254 68 114 111 319 92 218 160 70 59 113
Worst-case scenario8 126 164 44 94 74 249 64 153 98 60 45 93
1
CSB+, Corn-Soy-Blend Plus; CSB++, Corn-Soy-Blend Plus Plus; WF, WinFood; WF-L, WinFood-Lite.
2
WHO/FAO/UNU protein and amino acid requirements in human nutrition (20).
3
FAO/WHO-recommended nutrient intake (19).
4
FAO/WHO-recommended nutrient intake, assuming 10% absorption (19).
5
International Zinc Nutrition Consultative Group, assuming moderate absorption (21).
6
The optimal diet formulated by using goal programming (Optifood Module II).
7
Each diet sequentially maximizes each micronutrient (Optifood Module III; “best-case scenario”).
8
Each diet sequentially minimizes each micronutrient (Optifood Module III; “worst-case scenario”).
 LINEAR PROGRAMMING AND FOOD-BASED INTERVENTIONS 137
agree with the model results suggesting that vitamin A intakes these products should be integrated with approaches that en-
were adequate. courage changes to the traditional dietary practices for infants
Iron was identified as a key problem nutrient in all modeled and young children. Before promoting improved CF products and
diets. For the 4 CF products evaluated, total iron contributed from realistic changes to dietary practices, they should be evaluated by
the modeled diets reached ,50% of the desired levels, even in using approaches such as the Trial of Improved Practices (42) or
the best-case scenario analyses. CSB++ achieved the highest Integrated Management of Childhood Illness (43). LP modeling
dietary iron contents and WF the lowest dietary iron contents. in combination with other approaches can support the devel-
These findings support the results from the WinFood inter- opment and promotion of realistic interventions to improve di-
vention trial. After a 9-mo intervention period in which children etary adequacy when nutrient gaps are too wide to be covered by
were randomly assigned to be fed 1 of the 4 CF products, both food supplementation.
their hemoglobin and ferritin concentrations declined from base- The Optifood LP approach is a potentially powerful instrument

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

line to endpoint for all groups (13). The prevalence of anemia was for predicting the impact of a nutritional intervention on dietary
highest in the group that had received WF compared with the adequacy. However, this study had several limitations. First, the
other CF product groups (data not published). The Cambodian sample size for the dietary data was small, so rarely consumed
DHS 2010 reported that 55% of children younger than 5 y were nutrient-dense foods might have been missed. In addition, the
anemic (12). Even though the high prevalence of anemia might be portion sizes of some foods were estimated from ,5 individuals.
partially attributable to parasitic infections or endemic occurrence Second, the dietary data were collected by using average recipes
of genetic hemoglobin disorders (31), our modeling results sug- of mixed-food dishes. The mixed dishes in Cambodia contain
gest that low dietary intakes are a contributing factor, despite the many different ingredients, and it is difficult to assess exactly
use of iron-fortified CF products. Other intervention strategies, which parts of the mixed dishes the infants consumed. Third, the
such as the use of micronutrient powder (32–34), might be needed nutrient composition of the modeled food items was obtained
to eliminate iron deficiency and anemia. from different food-composition databases (22–24). There is
WF and WF-L ensured adequate dietary nutrient intakes of considerable variability in the nutrient content of foods related
calcium in both age groups. WF and WF-L contain whole small to local conditions, such as soil, cultivation, and food processing
fish. Whole small fish, for which the bones are edible, has proven (11), which might have introduced errors. Fourth, most of the
to be a good calcium source with a calcium bioavailability similar RNIs are based on Adequate Intake estimates, except for iron
to that of milk (35). CSB++ contains dried skim milk powder, but and zinc, which are based on Recommend Dietary Allowances.
its calcium content is half the amount of the 2 WinFood products. Adequate Intakes are based on intakes of a healthy population
In a highly productive aquatic environment, such as Cambodia and therefore might overestimate the actual nutrient needs (11).
(36), infants should frequently consume fish. However, in our To address this potential source of error, we selected a more
study, infants were served large fish rather than small fish, so they conservative criterion, 65% of the RNI, for comparing the base-
did not consume calcium-rich bones. Perhaps caregivers are line diets and special CF products. These limitations are impor-
afraid that a fish bone will get stuck in the throat of infants. The tant to take into account when interpreting results from diet
advantage of the processed WinFood products was that small fish modeling.
species were ground, which makes them an acceptable dietary These findings emphasize the value of nutrition program
source of calcium for infants. Note that the RNI for calcium, planners using LP to evaluate special CF products during the
primarily based on studies in whites, has been argued to be too intervention planning phase. With the LP approach, we were able
high for ethic Asian populations (37); therefore, the results from to make a distinction across the 4 formulated CF products in
Optifood could overestimate the calcium as a problem nutrient in terms of their potential for ensuring dietary adequacy of selected
the current population. micronutrients for the target groups. Such information is valuable
The results showed that the CSB products and the WF-L for planning a nutritional-intervention program or for under-
ensured adequate dietary intakes of zinc in the 9- to 11-mo-old standing the reasons for an unexpected lack of response to a
group. Zinc is characterized as a growth nutrient (type II nutrient) program.
(38) and is vital for nearly all biochemical pathways in the human
We thank Reproductive And Child Health Alliance (RACHA) for support
body. Therefore, deficient intakes of any of these type II nutrients with implementing the study in PeaRieng, Prey Veng, and assistance with
will impair child growth (38). Meat and other ASF are the best facilitating the contact with the mothers in the villages in PeaRieng; field as-
source of bioavailable zinc. Studies have shown that increased sistants and staff at the Department of Fisheries Post-harvest Technologies
meat consumption is associated with a lower risk of stunting and and Quality Control, Fishery Administration, Ministry of Agriculture, For-
improves cognitive development and general health (39–41)—an estry and Fisheries, Cambodia; the children and their families for participat-
association that is partly a result of an improved intake of bio- ing; and the SMILING project for funding.
available zinc. The population in the current study consumed ASF The authors’ responsibilities were as follows—JKHS, TB, CC, FTW,
MAD, NR, and ELF: designed the study; JKHS and TB: conducted the
(mainly large fish and pork) frequently, but in very small amounts.
study; JKHS and ELF: developed the concepts for the LP analyses and
WF-L and the CSB products are all fortified products. The current analyzed the data; and JKHS: wrote the first draft of manuscript. All authors
results indicate that access to fortified CF products may be needed edited the manuscript and approved the final version. None of the authors
in this population to achieve sufficient zinc intake. had any conflicts of interest.
The extent of the deficit in multiple micronutrients in the
modeled baseline diets is large and difficult to meet with a single
intervention product. Making formulated nutritionally improved REFERENCES
CF products available will contribute to improving overall dietary 1. Black RE, Victora CG, Walker SP, Bhutta ZA, Christian P, de Onis M,
quality; however, to eliminate undernutrition and poor health, Ezzati M, Gratham-McGregor S, Katz J, Martorell R, et al. Maternal
138 SKAU ET AL
and child undernutrition and overweight in low-income and middle- deficiency in populations and options for its control. Food Nutr Bull
income countries. Lancet 2013;382:427–51. 2004;25:S99–203.
2. PAHO. Guiding principles for complementary feeding of the breastfed 22. US Department of Agriculture, Agricultural Research Service. USDA
child. Washington, DC: Pan American Health Organization, 2001. national nutrient database for standard reference, release 25. Wash-
3. WHO. Complementary feeding of young children in developing ington, DC: Nutrient Data Laboratory, US Department of Agriculture,
countries: a review of current scientific knowledge. Geneva, Switzer- 2013.
land: World Health Organization, 1998. 23. National Institute of Nutrition. Vietnamese food composition table.
4. Dewey KG, Brown KH. Update on technical issues concerning com- Hanoi, Vietnam: Medical Publishing House, 2007.
plementary feeding of young children in developing countries and 24. Puwastien P, Raroengwichit M, Sungpuag P, Judprasong K. Thai food
implications for intervention programs. Food Nutr Bull 2003;24:5–28. composition tables. Bangkok, Thailand: Institute of Nutrition Mahidol
5. Bhutta ZA, Das JK, Rizvi A, Gaffey MF, Walker N, Horton S, Webb P, University, 1999.
Lartey A, Black RE, The Lancet Nutrition Interventions Review 25. FAO/WHO/UNU Expert Consultation. Human energy requirements.
Group, et al. Evidence-based interventions for improvement of ma- Rome, Italy: Food and Agriculture Organization, 2004.
ternal and child nutrition: what can be done and at what cost? Lancet 26. Lutter CK, Dewey KG. Proposed nutrient composition for fortified

Downloaded from https://academic.oup.com/ajcn/article-abstract/99/1/130/4577245 by 82211954 user on 01 March 2019

2013;382:452–77. complementary foods. J Nutr 2003;133:3011S–20S.
6. Briend A, Darmon N, Ferguson E, Erhardt JG. Linear programming: 27. World Health Organization. WHO child growth standards: length/height-
a mathematical tool for analyzing and optimizing children’s diets for-age, weight-for-age, weight-for-length, weight-for-height and body
during the complementary feeding period. J Pediatr Gastroenterol Nutr mass index-for-age: methods and development. Geneva, Switzerland:
2003;36:12–22. World Health Organization, 2006.
7. Santika O, Fahmida U, Ferguson EL. Development of food-based 28. Anderson VP, Cornwall J, Jack S, Gibson RS. Intakes from non-
complementary feeding recommendations for 9- to 11-month-old peri- breastmilk foods for stunted toddlers living in poor urban villages of
urban Indonesian infants using linear programming. J Nutr 2009;139: Phnom Penh, Cambodia, are inadequate. Matern Child Nutr 2008;4:
135–41. 146–59.
8. Ferguson EL, Darmon N, Fahmida U, Fitriyanti S, Harper TB, 29. Stoltzfus RJ, Underwood BA. Breast-milk vitamin A as an indicator of
Premachandra IM. Design of optimal food-based complementary the vitamin A status of women and infants. Bull World Health Organ
feeding recommendations and identification of key “problem nutrients” 1995;73:703–11.
using goal programming. J Nutr 2006;136:2399–404. 30. Anderson VP, Jack S, Monchy D, Hem N, Hok P, Bailey KB, Gibson
9. Frega R, Lanfranco JG, De Greve S, Bernardini S, Geniez P, Grede N, RS. Co-existing micronutrient deficiencies among stunted Cambodian
Bloem M, de Pee S. What linear programming contributes: world food infants and toddlers. Asia Pac J Clin Nutr 2008;17:72–9.
programme experience with the “cost of the diet” tool. Food Nutr Bull 31. George J, Yiannakis M, Main B, Devenish R, Anderson C, An US,
2012;33:S228–34. Williams SM, Gibson RS. Genetic hemoglobin disorders, infection,
10. Dibari F, Diop EHI, Collins S, Seal A. Low-cost, ready-to-use thera- and deficiencies of iron and vitamin A determine anemia in young
peutic foods can be designed using locally available commodities with Cambodian children. J Nutr 2012;142:781–7.
the aid of linear programming. J Nutr 2012;142:955–61. 32. Adu-Afarwuah S, Lartey A, Brown KH, Zlotkin S, Briend A, Dewey
11. Dewey KG, Cohen RJ, Rollins NC. WHO technical background paper: KG. Randomized comparison of 3 types of micronutrient supplements
feeding of nonbreastfed children from 6 to 24 months of age in de- for home fortification of complementary foods in Ghana: effects on
veloping countries. Food Nutr Bull 2004;25:377–402. growth and motor development. Am J Clin Nutr 2007;86:412–20.
12. National Institute of Statistics, Ministry of Planning Cambodia. De- 33. Lundeen E, Schueth T, Toktobaev N, Zlotkin S, Hyder SMZ, Houser R.
mographic and Health Survey 2010—Cambodia. Phnom Penh, Cam- Daily use of Sprinkles micronutrient powder for 2 months reduces
bodia: National Institute of Statistics and Ministry of Planning, 2011. anemia among children 6 to 36 months of age in the Kyrgyz Republic:
13. Skau JKH, Chhoun C, Touch B, Chea M, Unni U, Filteau S, Wieringa a cluster-randomized trial. Food Nutr Bull 2010;31:446–60.
FT, Dijkhuizen MA, Wells JCK, Berger J, et al. Effect of animal-source 34. Kounnavong S, Sunahara T, Mascie-Taylor CGN, Hashizume M,
foods and micronutrient-fortification complementary foods on body Okumura J, Moji K, Boupha B, Yamamoto T. Effect of daily versus
composition, linear growth, iron status—the WinFood project in weekly home fortification with multiple micronutrient powder on
Cambodia. Ann Nutr Metab 2013;63:163–4 (abstr). haemoglobin concentration of young children in a rural area, Lao
14. WFP. Technical specifications for the manufacture of Corn Soya Blend People’s Democratic Republic: a randomised trial. Nutr J 2011;10:129.
for young children and adults—CSB plus—version 2.1. Rome, Italy: 35. Larsen T, Thilsted SH, Kongsbak K, Hansen M. Whole small fish as
United Nations World Food Programme, 2010. a rich calcium source. Br J Nutr 2000;83:191–6.
15. WFP. Technical specifications for the manufacture of Corn Soya Blend 36. Halwart M. Biodiversity and nutrition in rice-based aquatic ecosys-
for young children—CSB plus plus—version 2.1. Rome, Italy: United tems. J Food Compost Anal 2006;19:747–51.
Nations World Food Programme, 2010. 37. Lee WTK, Jiang J. Calcium requirements for Asian children and ad-
16. WFP-Cambodia. Available from: http://www.wfp.org/countries/cambodia/ olescents. Asia Pac J Clin Nutr 2008;17:33–6.
operations (cited May 2013). 38. Golden MH. Proposed recommended nutrient densities for moderately
17. Roos N, Thorseng H, Chamnan C, Larsen T, Gondolf UH, Bukhave K, malnourished children. Food Nutr Bull 2009;30:S267–342.
Thilsted SH. Iron content in common Cambodian fish species: per- 39. Krebs NF, Mazariegos M, Tshefu A, Bose C, Sami N, Chomba E,
spectives for dietary iron intake in poor, rural households. Food Chem Carlo W, Goco N, Kindem M, Wright LL, et al. Meat consumption is
2007;104:1226–35. associated with less stunting among toddlers in four diverse low-
18. Daelmans B, Ferguson E, Lutter CK, Singh N, Pachón H, Creed- income settings. Food Nutr Bull 2011;32:185–91.
Kanashiro H, Woldt M, Mangasaryan N, Cheung E, Mir R, et al. 40. Neumann C, Harris DM, Rogers LM. Contribution of animal source
Designing appropriate complementary feeding recommendations: tools foods in improving diet quality and function in children in the de-
for programmatic action. Matern Child Nutr 2013;9:116–30. veloping world. Nutr Res 2002;22:193–220.
19. FAO/WHO. Human vitamin and minerals requirements. Reports of 41. Dror DK, Allen LH. The importance of milk and other animal-source
a joint FAO/WHO Expert Consultation. Rome, Italy: Food and Nutri- foods for children in low-income countries. Food Nutr Bull 2011;32:
tion Division, Food and Agriculture Organization, 2002. 227–43.
20. WHO/FAO/UNU. Protein and amino acid requirements in human 42. FAO. Trials of Improved Practices (TIPs)—guiding notes for TIPs
nutrition—report of a joint WHO/FAO/UNU Expert Consultation. trainers and implementers. Phnom Penh, Cambodia: Food and Agri-
Geneva, Switzerland: World Health Organization, 2007. culture Organization, 2011.
21. International Zinc Nutrition Consultative Group, Brown KH, Rivera 43. WHO. Integrated management of childhood illness IMCI adaptation
JA, Bhutta Z, Gibson RS, King JC, Lönnerdal B, Ruel MT, Sandström guide. Part 3: the study protocols. Geneva, Switzerland: Department of
B, Wasantwisut E, Hotz C. International Zinc Nutrition Consultative Child and Adolescent Health and Development, World Health Orga-
Group (IZiNCG) technical document #1. Assessment of the risk of zinc nization, 2002.