Chronic Diseases of Lifestyle Research Unit

Report on South African food consumption studies undertaken amongst different population groups (1983-2000): Average intakes of foods most commonly consumed

Executive summary
The primary objective of this study was to generate a reference table of "most commonly" consumed food items and average intakes of these items in the diet of South Africans. The table is required to be representative of foods eaten by children and adults from all age and ethnic groups in South Africa. The table will serve as a reference table for the Department of Health who would undertake analyses of (a) toxic chemicals, such as pesticides, heavy metals and environmental contaminants; (b) naturally occurring toxins; and (c) food additives in the commonly consumed food items. The goal is to estimate the actual dietary intake of toxic chemicals, naturally occurring toxins and food additives for comparison with their corresponding toxicological reference intakes, such as the Acceptable Daily Intake (ADI) or provisional tolerable weekly intake (PTWI). A secondary objective of the study was to derive average (mean) weights of South Africans in different age groups in order for the calculation of dietary exposure of selected contaminants according to:

Dietary Exposure=(Food Chemical Concentration*Consumption)/(Body Weight).

Secondary data-analysis was conducted on existing dietary databases (raw data) obtained from surveys undertaken in South Africa between 1983 and 2000. The National Food Consumption Survey (NFCS) served as a framework for compiling data on children since this was a national representative survey of 1-9 year-old children in South Africa. However there has never been a national dietary survey on adults in South Africa. Consequently the data had to be extrapolated from existing isolated surveys on adults. In this process the following databases were utilised: Lebowa Study; Dikgale Study; Black Risk Factor Study (BRISK); Transition, Health and Urbanisation Study (THUSA); THUSA Bana Study; First Year Female Student (FYFS) Project; Weight and Risk Factor Study (WRFS); and the Coronary Risk Factor Study (CORIS). The dietary intake for the groups 1-5 years and 6-9 years were calculated only from the NFCS, and were not supplemented by other databases. The substantiation for treating age 10+ as a unit (and calling it an adult group), was the finding that average consumption of adolescents (10 - 15 years) did not differ significantly from that of adults when comparing mean energy intakes of age groups in the studies analysed.

It is important to evaluate the results of this report in the context of the databases used. The estimates generated represent crude portions of food items consumed and should not be compared with the methods generally used in dietary surveys to evaluate macro- and micro-nutrient intakes of specific age groups. Although an attempt was made to include as many databases as possible to represent the average South African population, it was not realistic or feasible to include every study which has been undertaken in the specified period

The dietary data from different studies were firstly coded into GEMS/Food Commodities (main food groups); then into GEMS/Food subgroups; then into food items having a description and a method of processing (i.e. dried/canned/fresh). The latter step involved utilising the MRC food groups and the EURO codes. The final tables generated comprised the following data with regard to food items consumed: main food group (i.e. cereals); the subgroup where appropriate (i.e. maize); a description of the item where appropriate (i.e. maize porridge); the percentage of the sample consuming that item; the portion consumed per day by those individuals who actually consumed the item and the average portion (per capita) consumed per day by all individuals in the relevant sample. The latter portion is smaller because it represents the total quantity consumed divided by the size of the relevant sample.

The procedures used to generate adult data were based on factor analyses of all databases. Data were analysed in terms of percentage of the group consuming specific food subgroups/main groups/food items and on average per capita portion size. Factor analyses were done to establish the relationship between NFCS 6-9 year-olds in 9 provinces, urban and rural separately, with those databases having adult participants, namely: BRISK, Lebowa Study, CORIS urban and rural (adults), Dikgale Study (adults) and THUSA Bana (urban and rural).

The results implied that Factor 1 reflected portion size and Factor 2 variety of items consumed. Dikgale and Lebowa data clustered together with most NFCS rural groups to form a group (group 1). CORIS, BRISK and the Western Cape NFCS data clustered together (group 2). Data of the main urban areas clustered together in a third corresponding group (group 3), which lay between Lebowa/Dikgale on the one hand and BRISK on the other hand. Group 1 was regarded as the cluster of studies that consumed large portions of food (specifically maize) and included: Northern Province (urban and rural), Free State (urban and rural), North West (urban and rural) and rural areas of Mpumalanga, Eastern Cape, Gauteng, and KwaZulu-Natal. Group 2 on the other hand included studies where participants consumed smaller portion sizes yet consumed a large variety of food items. This group included the Western Cape urban and rural areas. Group 3 formed a cluster, which lay between group 1 and 2. Group 3 included all the remaining urban areas: Gauteng, Eastern Cape, KwaZulu-Natal, Mpumalanga and Northern Cape.

Equations were developed to determine combined estimates for different population groups by 2 different methods.

Method 1:
Estimation of group 1: adult consumption was estimated by taking the average values of Dikgale and Lebowa adult data. This data formed a pivotal point of Group 1. Dikgale and Lebowa data complemented each other, since the latter included adolescents and the former adults.

Estimation of Group 2: CORIS data represented the white population of the Western Cape, and BRISK data represented the black population of the Western Cape. Because of the similarities between CORIS urban and rural data, the combined databases were used in further analyses. It was accepted that white and "coloured" populations in Western Cape have similar dietary patterns (Steyn 1988). Adult dietary intake for the Western Cape were calculated as the weighted average of CORIS and BRISK data, using the ratio of black versus non-black residents in the Western Cape as described in Census 1996 data (Central Statistical Services 1999).

Estimation of Group 3: The average of BRISK and the combined average of Lebowa and Dikgale data were used to estimate adult consumption for this group.

Urban and rural intakes were combined to produce a single adult estimate per province, using the ratios between urban and rural per province, as calculated from the 1996 Census data. Adult intakes (average per capita portion size and percentage of adults consuming the item) in South Africa (RSA) were estimated by applying weights according to the proportion of populations in each province (Central Statistical Services 1999), as follows:

RSA =0.155*EC + 0.065*FS + 0.181*GP + 0.207*KZ + 0.069*MP + 0.021*NC
+ 0.122*NP + 0.083*NW + 0.097*WC;

RSA(rural) = 0.21*EC + 0.05*FS + 0.01*GP + 0.26*KZ + 0.09*MP + 0.01*NC
+ 0.23*NP + 0.12*NW + 0.02*WC;

RSA(urban) = 0.11*EC + 0.08*FS + 0.33*GP + 0.17*KZ + 0.05*MP +0.03*NC
+ 0.02*NP + 0.05*NW + 0.16*WC.

This equations were further simplified and adjusted for sample size, in terms of the original databases to:

RSA = 4.365*Lebowa + 5.901*Dikgale + 0.575*BRISK + 0.152*CORIS;

RSA(rural) = 5.932*Lebowa + 8.019*Dikgale + 0.011*BRISK + 0.032*CORIS

RSA(urban) = 3.002*Lebowa + 4.059*Dikgale + 1.078*BRISK + 0.250*CORIS.

Method 2: Adult data were calculated by using proportions of urban and rural data for black and non-black ethnic groups according to the Census '1996 results (Central Statistical Services 1999). BRISK represented urban blacks, and the average of Lebowa and Dikgale represented rural blacks, CORIS-urban represented non-black urban and CORIS-rural represented non-black rural. In terms of the original databases, after adjusting for sample size, the following equations were obtained:

RSA = 0.642*CORIS urban + 0.152*CORIS rural + 0.948*BRISK
+ 2.628*Lebowa + 3.553*Dikgale;

RSA(rural) = 0.110*CORIS(rural) + 1.874*Lebowa + 2.534*Dikgale;

RSA(urban) = 0.800*CORIS(urban) + 1.181*BRISK.

Method 1 results corresponded with results from the NFCS, which was over-sampled for lower socio-economic areas, whereas the results from method 2 ignored relationships with NFCS data and was based on the ethnic proportions of the population in South Africa.

Results: Food items consumed by at least 3% of one age group (for the 3 age groups) (adults = method 1) were: maize porridge, maize-based snacks, maize-based breakfast cereals such as corn flakes, samp, oats, white rice, maltabella porridge (sorghum), cookie/cake, fat cakes, wheat breakfast cereal, white bread, brown bread, canned fish, fried fish, orange juice, apples, bananas, breast milk, beef steak, beef stew, minced meat dish, beef sausage (wors), beef gravy, chicken meat, chicken stew, chicken giblets, chicken heads and feet, mutton, full cream milk, full cream milk (reconstituted), full cream processed milk, high fat cheese, buttermilk, white cooking fat, chicken eggs, peanut butter, dried beans, carrots, potatoes, soup, coffee, tea, white sugar, jam, boiled sweets, carbonated cold drinks, orange squash, cabbage, pumpkin, fresh tomato, tomato and onion stew, wild green leaves/spinach, pickled vegetables (atchar), non-dairy creamer, brick margarine, medium fat spread, sunflower oil and tap water.

Additional items, which were consumed by at least 3% of the older group when using method 2, are included. These are items appearing over and above the existing food items: beer, spirit drinks, custard/maizena, rusks, grapes, peaches, beef gravy (flour type), beef offal, skim milk, sweet potatoes, fried onions, green beans and peas. The tables generated in the text present data on percentage children and adults consuming the food items, average potion size of consumers, and per capita portion size of the whole group. It is recommended that average portion size of consumers, and not per capita portion size, be used by the Department of Health to compile their shopping list. The reason for this being the fact that per capita portion size would greatly under-estimate contaminants determined.

Mean weights and standard deviations calculated for the different age groups, were for 1-5 years: 14.2 (3.5) kg; 6-9 years: 22.5 (4.8) ; adults method 1: 55.7 (19.5) and method 2: 60.9 (19.7).

A final table is included which provides the 97.5th percentile of the consumption figures (consumers only), as well as the corresponding gram per kilogram body weight consumed. These figures represent the most popular food items consumed as described above, for the following age groups: 1-5 years, 6-9 years and age 10+ (adults).