Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

An analytical approach for exploring the importance of dietary quality versus quantity in the growth of Mexican children

Lindsay H. Allen, Anne K. Black, Jeffrey R. Backstrand, Gretel H. Pelto, Richard D. Ely, Elsa Molina, and Adolfo Chávez

Abstract

The average annual intake of specific nutrients, foods, food groups, and proxies for nutrient bioavailability of 87 Mexican preschoolers and 110 schoolchildren were compared with their anthropometry. Median intakes of energy, protein, thiamine, and iron were adequate; calcium and zinc were low; and other nutrients were very inadequate. Anaemia and low serum retinol were common. Intake of individual nutrients failed to predict size. Correlation matrices, median traces, and principal-components analysis illustrated a dietary continuum ranging from a high dependence on tortillas to more animal products and fruit. Children consuming a lower proportion of tortillas and legumes and more animal products were taller and heavier. Even though the high-tortilla dietary pattern provided more of most nutrients, these were less available. In conclusion, children's size was predicted by dietary quality - not quantity - measured either as a high intake of animal products or as a lower intake of factors inhibiting nutrient bioavailability.

Introduction

Growth faltering of children is common in developing countries. The reasons vary from one context to another, but this growth failure is most often explained by childhood infections [1] and/or inadequate quantity or quality of weaning foods. Where it is attributed to features of the diet, the most limiting nutrients for growth have not yet been determined. In the past, interest focused on the quantity and quality of dietary protein. In more recent years, attention has shifted to the adequacy of dietary energy.

The data analysed in the present article were collected during the Nutrition Collaborative Research Support Program (Nutrition CRSP), which was funded in 1981 by the US Agency for International Development to investigate the relationship between food intake and functional outcomes that are important for the development of individuals, their households, and communities [2]. In designing the programme, it was reasoned that insufficient food intake would manifest itself as inadequate energy intake. Therefore, energy intake was considered a major independent variable in the anticipated data analyses. Also, the intake of other nutrients could be calculated and examined for their relationship to functional outcomes.

The analyses presented in this paper compare the intake of nutrients (including energy), specific foods and food groups, and nutrient bioavailability to the attained size of preschoolers and schoolchildren in a Mexican rural community. The main purpose was to explore different methods of expressing food intake data for their ability to predict children's size.

Data collection and analysis

The Nutrition CRSP in Mexico was conducted from January 1984 to May 1986 in the Solis Valley, a rural highland area in the north-west of the state of Mexico. The 290 participating households were selected on the basis of their having male and female heads of household and meeting one of the following demographic criteria: (1) the mother was pregnant, (2) the parents had a preschool child 18 months old, or (3) the parents had a schoolchild 7-8 years old. The study design called for each target individual to be followed over 12 months. A large amount of data were collected on each target individual and household, of which only the following are relevant here: individual food intakes by a combined recall/weighing/record method on two days each month, anthropometry measured monthly (preschoolers) or every three months (schoolchildren), and household-level socio-economic variables measured twice (at study entrance and exit). Details of the study design and preliminary results are published elsewhere [3].

Several analyses were performed on data from 87 preschoolers and 110 schoolchildren, for each of whom at least eight days of food intake data were available. Daily nutrient and food intake variables were calculated for each child. Foods were converted to nutrients using the Mexican food composition tables [4] augmented by other sources of information as described below.

Summary measures of habitual intake were then obtained by taking the mean of the daily intake variables. In the case of the preschoolers, the means were calculated for the 12 months preceding the 30-month measurement of weight and height. Similarly, the mean intake of the schoolchildren was calculated for the 12 months preceding the final measurement of weight and height (at between 8 and 9 years of age). To correct for differences in age at the time of the anthropometry measurements, weight, height, and weight-for-height data were transformed to Z scores using FAO/WHO/UNU reference values and an algorithm supplied by the US Centers for Disease Control. Statistical analyses were performed using SAS and SPSS-X.

Findings and discussion

Children's allots

The Solis diet is predominantly vegetarian. Major plant foods include maize, pulque (a native alcoholic drink made from the maguey plant), legumes, pasta, potatoes, rice, bread, tomatoes, onions, garlic, various chillies (dried and fresh), and gathered plants [5]. Important animal foods are chicken eggs and meat, fresh cow's milk, and beef. Home production of these foods is fairly low, reflecting the meagre economic and agricultural resources of these households. Also, consumption of animal products is low because of high purchase costs that make them unaffordable for many families. Consumption of animal foods is further complicated by limited household access to refrigeration and the poor availability of these perishable foods in local, community-based shops.

Table 1 describes the food sources of dietary energy for the two groups of children. The preschoolers obtained 88% and the schoolchildren 93% of their energy from plant sources. Of these, 51% and 65% were from tortillas, with smaller amounts of energy coming from legumes, bread, pasta, and rice. The small percentages of energy from animal products (7% and 12%) were fairly equally distributed among meat, dairy products, and eggs, with the preschoolers deriving a larger proportion of energy from these animal foods than the schoolchildren.

In general the mean intakes of energy per kilogram of body weight were 96% of the US Recommended Dietary Intake (RDA) [6] for preschoolers and 115% for schoolchildren (table 2). Protein intakes were twice the RDA for both groups of children, and iron and thiamine intakes also exceeded the RDA. Calcium and zinc intakes were around 60% of the RDA for preschoolers; most schoolchildren met or nearly met the RDA for these nutrients. The median intakes of riboflavin, niacin, retinol, and ascorbic acid were only one-third to two-thirds of the RDA for both groups. It should be noted that these children had a high prevalence of anaemia (45% for preschoolers and 40% for schoolchildren), probably caused by both iron and vitamin-B12 deficiency [3]. In one of the communities, analyses of serum retinol revealed values below 20 mg/ml in 29% of the children. Thus, the nutrients lacking in quantity in Solís diets are minerals and vitamins rather than energy and protein.

TABLE 1. Sources of dietary energy (percentages of total energy)

Q1 = 25th percentile of intake of energy from plant foods; Q2 = 50th percentile; Q3 = 75th percentile.

TABLE 2. Nutrient intakes (percentages of recommended dietary allowances)

Relationships between diets and size

Nutrient intake

Spearman rank-order correlations were used to explore the relationships between the intake of specific nutrients and the children's attained size, since these correlations are most resistent to outliers, distribution problems, and nonlinear relationships. Food intake data are rarely distributed normally.

Previous analyses had demonstrated a marked growth stunting (poor growth in weight and height but normal weight for height) that started around 3 months of age. As a result, children by or before 18 months of age were small compared to the international reference percentiles [7]. The results in table 3 indicate that neither the average daily energy intake nor the average daily protein intake during the previous 12 months was related to the attained weight, length, or weight for length of the children. Dietary retinol was positively related to the Z score for height at both ages, and for weight in the schoolchildren. As discussed below, these significant correlations can probably be explained by a dietary pattern that contains more retinol, rather than by the actual intake of this nutrient. Weight-for-height Z scores were not associated with the intake of any nutrient.

Intake of specific foods

It would appear from the above analyses that the intake of single nutrients is a poor predictor of children's size. An exception is retinol, which may be weakly related to the anthropometric measures. In earlier analyses [3], however' we observed that a number of functional indicators of nutritional status (specifically growth and cognitive performance) were significantly correlated with the children's intake of animal products (expressed as either the percentage of total energy or of total protein from these foods). Energy and protein proved to be interchangeable as variables, since the intake of one is highly correlated with intake of the other.

These associations between intake of animal foods and children's growth raise many questions. For example, are there specific nutrients in animal products that correlate with children's growth? Are the nutrients contained in animal products more available for absorption? Alternatively, are animal products simply a marker for a better diet in general? With these questions in mind, the following analyses were designed to explore in greater detail why the intake of animal products appears to be an important marker of nutritional status.

TABLE 3. Spearman's correlations between nutrient intake and size (Z scores)

*p<.05.
**p<.01.

The food-group variables explored here were the total percentage of energy obtained from all animal products as well as from meat, dairy products, and eggs separately; and the total percentage of energy from all plant sources as well as from tortillas, legumes, other vegetables, fruits, and other plants. The latter category included wheat products (bread, pasta), rice, root crops, oil, candies, and non-dairy beverages, none of which on average contributed more than 7% to dietary energy.

From the correlations between the preschoolers' Z scores at 30 months and their average intakes of specific food groups during the previous 12 months shown in table 4, it is evident that the usual diet of taller children provided more energy from animal products (especially meat and milk) and less from tortillas than that of shorter children. Neither weight nor weight for height was associated with the intake of these foods. In the schoolchildren, animal products and fruits predicted greater weight and weight for height, legume intake was negatively correlated with all three anthropometric outcomes, and the overall proportion of energy from plant sources predicted lower weight.

TABLE 4. Spearman's correlations between energy from foods (percentage of total kilocalories from each source) and size (Z scores)

*p<.05.
**p<.01.
***p<.001.

TABLE 5. Spearman's correlations among food-group variables (mean percentage of energy intake from each source)

*p<.05.
**p<.01.
***p<.001.

To summarize these analyses, we see a much stronger association between the intake of specific foods and children's size than between the nutrients and the anthropometric measures. The general pattern is for animal products. and sometimes fruits and other plant products, to predict better growth, whereas children eating a greater proportion of tortillas and legumes grew less well.

Analysis of dietary patterns

The above results suggest a link between dietary pattern and functional nutritional outcomes, so further analyses were done to provide a better description of the dietary patterns of Solís children. This task was approached in three separate ways.

First, we calculated the Spearman's rank-order correlations among the food-group variables (table 5). A higher intake of animal products signifies greater consumption of meat, dairy products, and, to a lesser extent, eggs. Dairy products are most strongly associated with the intake of calories from animal sources. It is also apparent that children eating more animal products, and especially meat and dairy products, tend to consume fewer tortillas and legumes but more fruits and other plant products.

As a second step, the mean percentage of energy from the different food groups was plotted against the percentage of energy from tortillas (figs. I and 2). The plots are so-called median traces, a graphic technique advocated by Tukey [8] and others. To obtain the plotted values, the X values were ordered according to value and partitioned into groups of approximately equal size. For each group, the medians of X and Y were then calculated. Next, the median Y values were smoothed to improve interpretability (first using running groups of three and then running groups of two).

To improve presentation, the Y values have been resealed in the figures. Given the widely varying contribution of the different food groups to dietary energy (see table 1), resealing is necessary if all the food groups are to be plotted on the same graph [9]. The Y values were resealed by dividing the child's food-group value by the sample food-group inter-quartile range (third quartile minus first quartile). The resulting values may be plotted on a single graph, although they are presented here on two graphs for clarity.

The X axis represents the mean percentage of energy consumed as tortillas by each schoolchild. The Y axis is the average percentage of each of the other food groups consumed. Figure 1 shows that the children who ate proportionately more tortillas consumed remarkably less dairy products and fruit, and substantially less meat and other plant products. It is notable that the intake of dairy products falls to almost nothing at the point where the energy from tortillas exceeds approximately 55%, which is the case in about 75% of the schoolchildren. By comparison, figure 2 shows that the proportion of dietary energy from legumes, eggs, and vegetables stays fairly stable with respect to the percentage of energy from tortillas, although eggs do show a pattern similar to that seen for meat and dairy products (the effect being less dramatic). In the case of vegetables (tomatoes, onions, chillies, gathered plants) and legumes, increased reliance on tortillas for dietary energy is associated with a somewhat increased percentage of dietary energy from these two food groups. The patterns for preschoolers were similar to those shown in the figures for schoolchildren.

FIG. 1. Schoolchildren's intake of energy from meat, dairy products, fruit, ?and "other plant" foods in relation to tortillas (Values on the vertical axis are smoothed medians, and have been resealed)

FIG. 2. Schoolchildren's intake of energy from eggs, legumes, and vegetables in relation to tortillas (Values on the vertical axis are smoothed medians, and have been resealed)

As a third step in investigating the interrelationships between the food-group variables, principal-components analysis was employed. The first analyses for both the preschoolers and the schoolchildren resulted in two factors with eigenvalues greater than 1.0. After plotting the initial eigenvalues and examining the three-factor solutions, twofactor models were selected as the best.

In table 6, the first factor can be seen reflecting the continuum of dietary patterns that were observed in the former analyses. In the case of the preschoolers, those who have a lower intake (as indicated by the negative loadings) of tortillas (-0.88), and to a lesser extent of legumes (-0.44), have a higher intake of dairy products (0.82), other plant foods (0.75), fruit (0.60), and meat (0.44). The second factor has high loadings for meat and vegetables.

TABLE 6. Rotated factor patterns for principal-components analyses with VARIMAX rotations: food-group variables (mean percentage of daily energy intake from each source)

TABLE 7. Spearman's correlations of nutrient intake with the factor-l score and with the percentage of energy from tortillas

*p<.05.
**p<.01.
***p<.001.
No correlations with protein, riboflavin, or ascorbic acid were statistically significant (p<.05).

The pattern for the schoolchildren is similar to that seen for the preschoolers. A low proportion of energy from tortillas is associated with greater consumption of other plant foods (0.90), dairy products (0.85), and fruit (0.52). The second factor has high loadings for vegetables (0.75), fruit (0.54), meat (0.51), and eggs (0.44).

In general, all three approaches (the correlation matrices, the plots, and the principal-components analysis) show that a continuum of diets exists in the Solís Valley, ranging from that of individuals who are very highly dependent on tortillas to a much more complex diet in which dairy products, meat, fruit, and other plant products are consumed in larger quantities.

Given its value in the factor loading, it appears that the percentage of a child's energy intake that is consumed as tortillas might act as an indicator of poor diet quality in general. To test this, we compared each child's factor-1 score and intake of tortillas against nutrient intake (table 7). As expected, the nutrients showed similar correlations with the factor-l scores and tortilla intake (although the signs are reversed because of the negative correlation between consumption of tortillas and factor 1). Surprisingly, tortilla intake is strongly positively correlated with the intake of iron, thiamine, zinc, energy, calcium, and niacin for preschoolers, and of iron, thiamine, zinc, and calcium for schoolchildren. Retinol, on the other hand, is consumed in lower amounts when tortillas supply a larger proportion of dietary energy. The positive relationship between retinol intake and children's size is probably explained by the lower intake of retinol that results from consuming a larger percentage of energy as tortillas. The intake of protein, riboflavin, and ascorbic acid shows little association with the amount of tortillas consumed or with the factor-l score.

It is surprising that tortilla intake is associated with a higher intake of several nutrients, since the analyses above show that tortilla intake predicts poorer growth. The reason for this apparent contradiction may lie in the bioavailability of the nutrients: the nutrient composition values for tortillas may not reflect the bioavailability of the nutrients. Also, the bioavailability of nutrients from other foods may be poorly modelled in a diet that is heavily dependent on tortillas.

Bioavailability variables and size

To investigate these bioavailability issues, estimated values for the fibre, phytate, and zinc content of foods were added to the Mexican food composition nutrient database [4]. Available iron was calculated by Monsen's formula [10], which considers the intake of haem iron, meat, and ascorbic acid but does not include any correction for dietary phytate. As the phytate content of the Mexican diets is high (table 8) and phytate interferes with iron absorption, the calculated values for available iron must be substantially overestimated. Nevertheless, assuming a requirement for absorbable iron of 1 mg per day [6], the intake of this by preschoolers is particularly low. Since the available-iron variable did not account for the negative effects of phytate, the ratio of fibre to iron (g:mg) was also calculated. Variables that should reflect the bioavailability of zinc include the phytate-to-zinc molar ratio [11] and the ratio of the product of phytate times calcium to zinc [12]. In the diets of Solís children, these ratios exceed the limits published in the literature above which zinc absorption is impaired [11-13]. Although calcium bioavailability from Solís diets is poor [14], no proxy for calcium availability was employed here because there is no theoretical basis on which to believe that it would affect children's growth.

TABLE 8. Distributions of bioavailability variables

Q1 = 25th percentile; Q2 = 50th percentile; Q3 = 75th percentile
a. Sources: available iron, ref.10; fibre/iron, ref. 13; phytate/zinc, ref. 11;(phytate x calcium)/zinc, ref. 12.
b. N=87.
c. N=110.

TABLE 9. Spearman's correlations between nutrient availability and size (Z scores)

*p<.05.
**p<.01.

Associations between the bioavailability variables and the nutritional outcomes were then explored. Higher intakes of fibre and phytate relative to iron and zinc were associated with more growth stunting among the preschoolers as reflected in lower height and weight Z scores at 30 months (table 9). However, preschooler weight for height showed little correlation with the bioavailability measures. For the schoolchildren, negative associations were seen between weight and the ratios of fibre to iron and of phytate to zinc.

Not surprisingly, both the factor-1 score and the tortilla-intake measure were strongly associated with these bioavailability variables. Children eating more tortillas, or who had a lower factor-l score, had diets with more phytate and fibre, less absorbable iron as indicated by the higher fibre-to-iron ratio, less available zinc, and less haem iron (table 10). These negative relationships between indicators of poor nutrient availability and tortilla intake may be the explanation for the negative associations seen between tortilla intake and children's size.

Potential confounding effect of socio-economic status

The relationships described above suggest that the growth stunting of Solís children reflects the poor quality, rather than quantity, of the diet. However, the possible confounding effect of socio-economic status cannot be ignored. Better dietary quality in the Solís Valley is associated with higher socio-economic status. As a result, it is possible that the apparent effects of dietary quality on children's growth are in fact a reflection of some other correlate of socioeconomic status.

TABLE 10. Spearman's correlations of the bioavailability variables with the factor-l score and with the percentage of energy from tortillas

*p<.05.
**p<.01.

TABLE 11. Partial correlations (Pearson's) between nutrient availability and size (Z scores), controlling for socioeconomic status

*p<.05.
**p<.01.

To investigate this possibility, the partial correlations between selected dietary and nutrient intake variables and children's size were examined controlling for socio-economic status (SES). The measure of SES employed here is derived from data on household ownership of material goods and the construction of the house [3]. This use of partial correlations is a conservative approach, since SES is allowed to explain variation in children's size that also could be explained by diet. In other words, this approach over-controls for the effects of SES, since the effects of SES-related, higher-quality diet are attributed to SES.

Despite overcontrolling for SES, table 11 shows that the relationships between the dietary and bioavailability variables and children's size persist. For the preschoolers, sizeable partial positive correlations are seen between the length Z scores and the percentage of animal calories, and negative correlations between the length Z scores and the percentage of tortillas and the phytate variables. In the case of the schoolchildren, the negative correlations of the tortilla and the phytate and zinc measures with the weight Z scores remain significant. These results suggest that the apparent effects of dietary quality on children's growth are not the result of some unmeasured correlate of socio-economic status.

Conclusion

We have identified a continuum of dietary patterns in children living in the Solís Valley that ranges from a high intake of tortillas and beans (the lowest-quality diets) to a proportionately higher intake of animal products (especially milk), fruits, and plant products such as wheat and rice, oil, root crops, and non-dairy beverages. Children at the more diverse end of this continuum have significantly better growth. However, although the diet that favours these better outcomes would be accepted in traditional nutritional terminology as being of higher quality, it was not identifiably higher in any specific nutrient except retinol, as calculated from food composition tables. Rather, dietary variables that were models of nutrient bioavailability were the best predictors of children's size. Although proxies for bioavailable iron and zinc were used here, these may or may not be the growth-limiting nutrients for the Mexican children, a the bioavailability of other nutrients is likely to be concomitantly low.

These results, if they are at all representative of the situation in developing countries, have several implications. First, the potential for the intake of specific nutrients to function as proxies for more general features of the diet should be taken into consideration when designing or evaluating programmes or projects that include measurement of nutrient intake as a goal. For example, the quantity of nutrients consumed was not related to children's size. The only exception to this pattern was retinol, but this association is most likely explained by the higher intake of retinol in dietary patterns that have a greater nutrient bioavailability in general. Therefore, higher retinol intake may be a marker for a better quality diet rather than a causal agent.

Second, dietary quality, measured either as the percentage of dietary energy from animal products or as a lower intake of factors inhibiting nutrient bioavailability, was a positive predictor of attained size. These variables should be included in future analyses of the relationship between diet and nutritional outcomes. It is also evident that it is risky to interpret relationships between specific foods (or nutrients) and nutritional status outcomes without analysing complex patterns of food substitutions and additions. Dietary patterns will vary from one cultural setting to another. However, once the dietary pattern is understood, it is likely that the intake of key foods, such as tortillas in Mexico, can serve as an adequate predictor of children's growth and development.

Acknowledgements

This study was supported by USAID grant no. DAN-1309-G-SS-1070-00 and the National Live Stock and Meat Board.

References

1. Rivera J, Martorell R. Nutrition, infection and growth. Part 1: Effects of infection on growth. Clin Nutr 1988;7:15662.

2. Calloway DH, Murphy SP, Beaton GH. Food intake and human function: a cross-project perspective. Berkeley, Calif, USA: University of California Press, 1988.

3. Allen LH, Chavez A, Pelto GH. The collaborative research and support program on food intake and human function. Mexico project final report Storrs, Conn, USA: University of Connecticut Department of Nutritional Sciences, 1987.

4. Hernandez M, Chavez A, Bourges H. Valor nutritivo de los alimentos Mexicanos. Mexico City: Instituto Nacional de la Nutrición Salvador Zubiran, 1980.

5. Backstrand JR. Patterns of household food consumption in rural, central Mexico. Doctoral dissertation, University of Connecticut, Storrs, Conn, USA, 1990.

6. National Research Council. Recommended dietary allowances. 10th ed. Washington, DC: National Academy Press, 1989.

7. Allen LH, Pelto GH. Chavez A, Martinez H, Ely RD, Capacchione C. Maternal correlates of infant growth in rural Mexico. In: Atkinson SA, Hanson LA, Chandra RK, eds. Breastfeeding, nutrition, infection and infant growth in developed and emerging countries. St. John's, Newfoundland, Canada: ARTS Biomedical Publishers and Distributors, 1990:299-306.

8. Tukey JW. Exploratory data analysis. Reading, Mass, USA: Addison-Wesley, 1977.

9. Cleveland WS. The elements of graphing data. Monterey, Calif, USA: Wadsworth Advanced Books and Software, 1985.

10. Monsen ER, Hallberg L, Layrisse M et al. Estimation of available dietary iron. Am J Clin Nutr 1978;31:134-41.

11. Oberleas D, Harland B Phytate content of foods: effect on dietary zinc bioavailability. J Am Diet Assoc 1981; 79:433-36.

12. Ellis R, Kelsay JL, Reynolds RD. Morris ER, Moser PB, Frazier CW. Phytate:zinc and phytate x calcium:zinc millimolar ratios in self-selected diets of Americans, Asian Indians and Nepalese. J Am Diet Assoc 1987;87:104347.

13. Bindra GS, Gibson RS, Thompson LU. Phytate, calcium, zinc ratios in Asian immigrant lacto-ovo vegetarian diets and their relationship to zinc nutriture. Nutr Res 1986;6:475-83.

14. Rosado, JL . Bioavailability of nutrients from rural vs. urban Mexican diets. Doctoral dissertation, University of Connecticut, Storrs, Conn, USA, 1988.

Contents - Previous - Next