Contents - Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
Overall variability in data from 101 dietary calculations
Coefficients of variation
In an attempt to estimate the overall variability of the available data, 101 calculations for regions, countries, and subregions of countries were evaluated as a group. Some have already been tabulated. Others have not yet been considered but are either tabulated or described below. Consideration was given to food energy; total protein; animal, cereal, and legume protein; and the contents of the essential amino acids in milligrams per gram of protein. Also examined were values for lysine both in milligrams per day and in milligrams per 1,000 kcal.
It can be seen from table 9 that the range of values for the various nutrients in the diets evaluated was considerable. Food energy ranged from 1,314 to 3,732 kcal/day, total protein from 34 to 118 g/day, animal protein from 2 to 78 g/day, and legume protein (including soya) from 0 to 35 g/day. Amino acid data (milligrams per gram of protein) also showed considerable variability, with lysine in particular showing a range of 33 to 71 mg/g protein. The coefficient of variation (CV%) for lysine was 18.1%, which was much greater than the average CV% of 5.4% for the other essential amino acids. These differences are also illustrated in figure 2 (see FIG. 2. Coefficient of variation for essential amino acids (milligrams per gram of protein). Determined from 101 dietary calculations. Aaa=aromatic amino acids; Saa= sulphur amino acids), where the much higher CV% value for lysine can be seen. This indicates the relative constancy of these other amino acids in the various diets and the rather special nature of lysine as a dietary variable.
TABLE 9. Means and coefficients of variation (CV%) for food energy, protein, and amino acid data from 101 food balance sheet calculations for regions and countries
|Food energy (kcal/day)||2,537||25 0||3.732||1,314|
|Total protein (g/day)||71.2||30.0||118||34|
|Animal protein (g/day)||27.3||84.0||78||2|
|Cereal protein (g/day)||30.9||34.3||74||9|
|Legume proteins (g/day)||5.5||91.5||35||0|
|Isoleucine (mg/g protein)||44||5.8||48||37|
|Leucine (mg/g protein)||80||8.5||115||70|
|Lysine (mg/g protein)||53||18.1||71||33|
|Sulphur amino acids (mg/g protein)||37||4.2||39||30|
|Aromatic amino acids (mg/g protein)||81||2.4||89||75|
|Threonine (mg/g protein)||37||6.8||41||31|
|Tryptophan (mg/g protein)||12||4.4||13||9|
|Valine (mg/g protein)||51||5.4||5.7||43|
|Lysine (mg/1,000 kcal)||1,498||275||2,414||905|
a. Legume protein includes soya protein.
The much higher CV% for lysine follows because lysine is the limiting amino acid in cereals, with a level per unit of protein that is generally only about one-third that of animal foods. A major difference between the diets of poor and rich countries is in their proportions of animal and cereal proteins. Thus, lysine is significantly affected. For these same data, the values for lysine ranged from 1,741 to 8,173 mg/day, with a CV% of 46.7%. There was also a considerable range when data were expressed as milligrams of lysine per 1,000 kcal; here the range was between 905 and 2,414 mg/1,000 kcal, with a CV% of 27.5%.
Multiple regression analysis
A multiple regression analysis relating dietary protein and lysine value for the same 101 dietary evaluations is shown in table 10. In the first part of the table the dependent variable is lysine (milligrams per gram of protein). For the order-of-magnitude prediction of the lysine value of diets, it is generally quite unnecessary to use data from all the individual protein foods. Since there are major differences between the lysine content of animal foods, legumes, and cereal foods, the proportions of these protein components alone can give a high degree of predictability. For predicting lysine (milligrams per gram of protein), the mere use of animal protein percentage alone (R2=.84) can explain some 84% of the variation. This can increase to 94% when either animal protein and cereal protein percentages, or the combination of animal protein, cereal protein, and pulse protein percent ages are used. Thus, because of the very large differences in the lysine content of animal foods, cereal foods, and pulses when treated as groups, the content of protein from these sources alone can be a very efficient predictor of the lysine value of the diet.
In the second part of table 10, the dependent variable is lysine (milligrams per day). Here, some 95% to 98% of the overall variability is accounted for by the contents (grams per day) of animal protein, cereal protein, and pulse protein. Animal protein (grams per day) alone accounts for 95% of the variability in lysine The best prediction (R2=.98) involved all three major protein components:
Lysine (milligrams per day) = 86.3AP + 19.8CP + 63.6PSP + 599
where AP, CP, and PSP are animal, cereal, and pulse (including soya) protein in grams per day. Cereal protein (grams per day) by itself is a non-significant predictor of daily lysine availability. Although cereal protein may account for a high proportion of the total protein in developing country diets in absolute terms, the amount of cereal protein (in grams per day) is often similar in developing and developed country diets. Although it is not shown in table 10, cereal protein percentage can predict lysine (milligrams per day) at a low to moderate level (R2=.47). It is, however, the best single predictor for lysine (milligrams per gram of protein), with R2 = .88.
TABLE 10. Multiple regression analysis of data from 101 countries and regions relating dietary protein and lysine value
|Dependent variable: lysine (mg/g protein)|
|Animal (% total protein)||Cereal (% total protein)||Pulse -soya (% total protein)||Constant||Rē|
|Dependent variable: lysine (mg/day)|
|Animal protein (g/day)||Cereal protein (g/day)||Pulse-soya protein(g/day)||Constant||Rē|
Relations among diet, wealth, and health in 122 countries
Pearson correlation matrix
The next stage in the analysis was to correlate lysine availability with wider indicators such as wealth (gross national product), life expectancy at birth, and under-five child mortality data. These other indicators were not provided by FAO/Agrostat  but could be found in UNICEF . By examining both sources, complete data for gross national product, life expectancy, and the proportions of the dietary protein from animal, cereal, and legume sources were obtained for 122 countries. The years for data availability (gross national product, life expectancy, and diet composition) were not always identical but were generally between 1991 and 1993. It was therefore considered that broad-based conclusions could be drawn from such data, which do not change rapidly from year to year.
Because of the high degree of predictability (R2 = .98) for dietary lysine from animal, cereal, and legume protein contents, the prediction equation above was used to calculate lysine availability (milligrams per day). Lysine (milligrams per gram of protein) was calculated from the daily lysine by dividing by the total protein. Lysine values are thus not independent variables. Correlation coefficients were, however, calculated between gross national product, life expectancy at birth, total protein per day, animal protein percentage, cereal protein percentage, pulse and soya bean protein percentage, lysine in milligrams per gram of protein, and lysine in milligrams per day.
A Pearson correlation matrix from data for 122 countries is shown in table 11. As gross national product increased, so also did total protein per day, animal protein percentage, lysine in milligrams per gram of protein, and lysine in milligrams per day. In contrast, both cereal protein percentage and pulse and soya bean protein percentage declined as gross national product increased. Life expectancy at birth was also significantly correlated in a positive manner with wealth. Thus, richer countries consume more total protein, more animal protein, and less cereal protein, and in consequence, the lysine value of their diets is also greater.
TABLE 11. Pearson correlation matrix for 122 countries relating GNP, life expectancy, and dietary data
Abbreviations: GNP=gross national product (US$) per capita; LE=life expectancy at birth (years); TP=total protein (g/day); AP% = animal protein (% total protein); CP% =cereal protein (% total protein); PSP% =pulse and soya protein (% total protein); LYSG = lysine (mg/g protein); LYSD = lysine (mg/day).
For df = 120 p < .05 = 0178 and p < .001 = 0.294.
Food availability data for 1992 from FAO/Agrostat 
GNP and life expectancy data from UNICEF; values for 1992 93 .
Lysine values are calculated from food availability data using the equation
Total lysine (mg/day)=86.3AP+ 19.8 CP+63.6 PSP+599 (table 10), where TP, AP, CP, and PSP are, respectively, total, animal, cereal, and pulse and soya protein availability in grams per day Squared multiple R for the equation was 98 from 101 food balance sheet calculations. In this table, therefore, the lysine values are not independent variables.
Since both diet and health are improved with increasing wealth, no causal relationship can be inferred between diet (including lysine value) and life expectancy, although such a relationship is not unreasonable. Further dietary relationships are such that as total protein increases, animal protein percentage increases, cereal protein percentage declines, and lysine value increases. The very high correlations between animal protein percentage and lysine value (animal protein percentage, of course, being a major component in the prediction) are supported by observations made in other sections of this review. Also of note is the strongly negative (-0.93) correlation coefficient between cereal protein percentage and lysine value (milligrams per gram of protein).
Classification of countries by wealth
In table 12 health and dietary data are classified by gross national product. The four classes selected are those with annual per capita gross national products of less than US$500, those with US$500 to US$2,000, those with US$2,000 to US$10,000, and those with more than US$10,000 per year. Additional information in this table includes total population per income group, food energy, lysine and the under-five mortality rate (deaths among children under five years of age per 1,000 live births). There are 37 countries in the lowest gross national product class (<US$500) with a total population of some 2,990 million. These are also the countries with the lowest mean daily availability of food energy (2,070 kcal), total protein (51 g), animal protein percentage (20%), and lysine whether expressed as milligrams per day, milligrams per gram of protein, or milligrams per 1,000 kcal. They also have diets with the highest proportion of their protein from plant sources. These are also the countries with the lowest life expectancy and the highest under-five mortality rate.
In sharp contrast, there are 23 countries, with a total population of 806 million, with a mean per capita gross national product greater than US$10,000. These countries have an average daily availability of 3,335 kcal and 101 g protein, with 61% coming from animal protein sources. In consequence they average 6,555 mg lysine/day compared with 2,405 mg/day for those in the lowest gross national product class. In terms of milligrams per gram of protein, the highest gross national product class has an average of 65 ma/g protein and the lowest an average of 47 mg/g protein. These values can be compared with the 58 mg/g protein estimated by FAD/WHO  as the adult requirement. The average amount of lysine in relation to food energy is 1,966 mg/1,000 kcal in the highest class compared with 1,174 mg/1,000 kcal in the lowest. Different economic classes also show dramatic contrasts in under-five mortality rate and life expectancy.
It is thus clear from tables 11 and 12 that wealth (gross national product) is significantly correlated with both health and the type of diet consumed. The different dietary patterns lead to very different availabilities of lysine in terms of both milligrams per gram of protein and milligrams per day. High correlations do not necessarily indicate causality, but basic questions do arise as to whether lysine fortification of cereal diets, which would improve the overall protein value, would lead directly to improved health. Again, such a supposition is not unreasonable but cannot be proved or disproved from data of the type presented.
TABLE 12. Total population, nutrient availability, mortality rate, and life expectancy for world economic classes (data from 122 countries)
|GNP class US $/person/yr||No. countries||Total population (millions)||Food energy (kcal/day)||Total protein (g/day)||Animal protein (%)||Cereal protein (%)||Pulse-soya protein (%)||Lysine (mg day)||Lysine (ma g protein)||Lysine (mg/1,000 kcal)||Under 5 Mra||Life expectancy|
Lysine values calculated from protein availability data. See footnote to table 11.
a. Under-five mortality rate: deaths at less than five years of age per 1,000 live births.
Contents - Previous - Next