Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
ARTICLE IN PRESS JOURNAL OF FOOD COMPOSITION AND ANALYSIS Journal of Food Composition and Analysis 17 (2004) 291–300 www.elsevier.com/locate/jfca Original Article Protein, carbohydrate and fibre in cereals from Mali—how to fit the results in a food composition table and database I. Barikmoa,*, F. Ouattarab, A. Oshauga b a Akershus University College, Kjeller, Norway Institut National de Recherche en Sant!e Publique, Bamako, Mali Received 4 July 2003; received in revised form 14 February 2004; accepted 26 February 2004 Abstract During the past 5 years, the main staple foods (cereals) used in Mali have been collected to develop a food composition table and database. We present recent results of protein, carbohydrate and fibre content for some cereals. Samples were collected from five different regions. To reduce laboratory costs, composite samples (cs) were made. The cereals analysed were sorghum (Sorghum bicolor) (cs ¼ 142), millet (Pennisetum glaucum) (cs ¼ 163), maize (Zea mays) (cs ¼ 107), wheat (Triticum aestivum) (cs ¼ 123), rice (Oryza sativa) (cs ¼ 151) and fonio (Digitaria exilis) (cs ¼ 104). Fonio is an old cereal cultivated across the dry savannahs in West Africa, and is very popular in Mali. All samples were cleaned and processed (ready to cook) before analysis. Detailed sampling plans were used. For total nitrogen, Kjeldahl and Dumas combustion methods were used. Methods used for carbohydrate (sugar and starch) were polarimetric, spectrophotometric and HPLC, and a gravimetric method was used for fibre. The mean7s.d. content of protein for 100 g cereal was: in millet 7.971.4 g, sorghum 10.370.7 g, maize 7.671.1 g, rice 6.370.3 g, wheat 10.671.1 g and fonio 7.270.4 g. The mean7s.d. content of carbohydrate and fibre per 100 g cereal was: in millet 65.8710.1 and 6.272.3 g, sorghum 73.574.3 and 4.770.1 g, maize 73.0710.2 and 4.671.3 g, rice 83.777.8 and 1.170.0 g, wheat 75.171.8 and 3.070.0 g and fonio 74.370.1 and 2.270.3 g, respectively. As indicated by the standard deviations there were considerable geographical differences in nutrient content for the same cereal. There is no apparent explanation for these differences. Until this is explored further, it is necessary to develop separate tables for different regions. r 2004 Elsevier Ltd. All rights reserved. Keywords: Food composition table; Database; Africa; Mali; Cereal; Protein; Carbohydrate; Fibre *Corresponding author. Akershus University College, Box 423, Lillestr^m, Norway. Tel.: +47-64849182; fax: +4764849002. E-mail address: [email protected] (I. Barikmo). 0889-1575/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.jfca.2004.02.008 ARTICLE IN PRESS 292 I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 1. Introduction The composition of nutrients in food can vary considerably between regions within a country as well as between countries. Such differences can be caused by variation in temperature, rainfall and access to water, use of fertilizer, nutrient content of the soil, etc. (Greenfield and Southgate, 1992). A 5-year initiative has been taken to develop a food composition table and database for Mali by analysing staple foods and other foods that are frequently eaten, such as dry green leaves, some vegetables and wild fruits (Barikmo et al., 2002). The staple foods in Mali are cereals and the most commonly eaten are millet, sorghum, rice, maize, wheat and fonio. The cereals accounted for 74% of the energy and plant foods provided 74% of the protein in the Malian diet (FAO/WAICENT, 2003). In the past, protein quality of wild gathered foods from Mali (Nordeide et al., 1994) and nutrient composition of green leaves (Nordeide et al., 1996) have been examined, and the results are included in the Malian food composition table and database. It became clear early in the work that nutrient content varied considerably. A food composition table and databases with average figures on nutrient content in foods would therefore have been of limited use. A key question is how to ensure access to representative results of nutrient content of foods in a national database, taking into account ecological and seasonal variations. One response is to analyse foods using a sampling plan that captures these variations. Mali is a land-locked country in West Africa which covers a surface area of 1,241,248 km2. Its total population in 2000 was 11.9 million inhabitants with a 3% growth per year (UN, 2003). The country can be divided into three ecological zones: the Sahara zone which covers 56% of the total surface and is characterized by less than 200 mm of annual rainfall (regions of Timbuktu, Gao and Kidal); the Sahel zone which covers 19% of the land mass with an annual rainfall between 200 and 700 mm (region of Mopti and Kayes); and the Sudan zone that receives 700–1400 mm of rain per year and is covered by significant vegetation such as savannah and forests (regions of Se! gou, Koulikoro, Bamako and Sikasso) (FAO/WAICENT, 2003). In addition to this ecological variation Mali has three climatic seasons: the cold season from October to January; the hot season from February to May/June; and finally the rainy season from June/July to September. The considerable ecological and climatic variations cause varied soil and growing conditions for plants and animals that may affect the nutrient content of the local food. Anecdotal data indicate that use of fertilizer in Mali is very low, and is basically limited to the area where agriculture is less risky and integrated in the market economy. The purpose of this article is to present results of protein, carbohydrate and fibre content for some cereals in five different regions in Mali with considerable ecological variation, and to suggest how to ensure the correct use of such results. 2. Materials and methods The collected cereal samples are all produced in Mali, except for the wheat flour. The contribution from each region to the national cereal production varies a lot. In 1996/1997 the Se! gou region (the Sudan zone) accounted for 35% of the national production, followed by the ARTICLE IN PRESS I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 293 region of Sikasso also in the Sudan zone (about 22%), Koulikoro in the Sudan zone (13%) and Mopti and Kayes in the Sahel zone (17% and 8%). The proportion from the regions of the Sahara zone to the total national grain production was 5% from Timbuktu and 1% from Gao (SISEI Mali, 2003). With this background, sampling plans were worked out (Barikmo and Ouattara, 1998; Ouattara and Barikmo, 2002). For the cereals the main collection started in the region of Se! gou and Mopti in 1999, followed by a new collection of the same food items in the capital, Bamako, in 2000. In Bamako, most of the cereals on the market came from the regions of Se! gou, Sikasso and Koulikoro, the last one being located close to the capital. In 2002 the collection continued, this time in the Timbuktu region. Timbuktu was chosen because of its special conditions in the desert and the fact that no samples from this region had been analysed before. 2.1. Sampling plan and methods All the collections followed a major sampling plan. Because we wanted foods from different selected regions, stratified sampling1 combined with randomly selected counties in each region was considered adequate (Greenfield and Southgate, 1992). The main feature was that in each region, counties were randomly chosen and the samples were collected from different retailers at different markets in each county (Table 1). How many counties, markets and retailers to select depended on the population figures, but as a general rule, 10–12 primary samples of each food item were collected from each county. In order to minimize the costs and transportation, composite samples2 prepared from, e.g., 12 different primary samples with multiple composition3 from one county, were mixed with composite samples from the other counties in the region. Thus all the samples in each region were mixed. 2.2. Food collection The food collected was the most commonly produced and used cereals in Mali, namely millet (Pennisetum glaucum), sorghum (Sorghum bicolor), maize (yellow) (Zea mays), rice (Oryza sativa), wheat (Triticum aestivum) and fonio (Digitaria exilis). Fonio is probably the oldest African cereal, and is cultivated across the savannahs in West Africa. It is very popular in Mali (NRC, 1996). Because the food collection was done from retailers at different markets, no cultivar or variety names were available. Table 1 shows that between 163 (millet) and 104 (fonio) primary samples of each cereal were taken to make reduced composite samples, and samples sent for analyses represented 3–5 cereal grains. The number of samples varied from region to region. In Kayes, only maize was collected, but for the other regions the variation in the number of samples was due to the availability of the different cereals in the area. 1 A sample consisting of portions obtained from identified subparts (strata) of the parent population. Within each stratum, the samples were taken randomly. 2 A representative part of the primary sample obtained by a division and reduction process. 3 Different cultivars but the same food, e.g., millet from 12 different retailers. ARTICLE IN PRESS 294 I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 Table 1 Number of samples of cereals and origin in Mali, collected in 1998–2002 Region Countiesa Millet (Date of collection) No csb Kayes (Nov 1998) S!egou (Nov 1999) Mopti (Nov 1999) Bamako (Nov 2000) Timbuktu (Jun 2002) 2 4 4 6 4 — 40 39 71 13 Primary samples Analysed samples Sorghum Maize nc csb nc csb — 1 1 1 1 — 37 28 67 10 — 1 1 1 1 19 13 4 61 10 163 142 4 Rice nc 1 1 1 1 1 107 4 Wheat csb nc csb nc csb nc — 44 24 72 10 — 1 1 1 1 — 27 18 68 10 — 1 1 1 1 — 23 11 70 0 — 1 1 1 0 151 5 Fonio 123 5 104 4 3 a Counties=number of counties in each region where the primary food samples were collected. cs=numbers of primary samples used for preparing composite samples. c n=numbers of composite samples sent for analysis, made from cs. b 2.3. Sample handling After pooling the samples in the field, they were prepared for laboratory analysis in Bamako. Millet, sorghum, maize and fonio were bought on the local market as whole grain but rice was bought in its polished form. Wheat was bought as flour in all the regions, except in Timbuktu where it was bought as whole grain and grounded to flour by a local woman. The wheat was locally grown in the region of Timbuktu, but in all the other regions it was imported from abroad, mostly from Europe. The cereals were analysed as ready to cook and different procedures had to be followed according to type. For millet, maize and sorghum the bran was separated from the grain. After the separation, the grain was washed, and then grounded to flour by hand or by small mills. The flour was kept frozen at 20 C until analysis. Fonio and rice grains were cleaned and kept frozen. The wheat flour was kept dry in a dark place but not frozen. Two hundred grams of each cleaned cereal were taken for laboratory analyses. The frozen samples were transported to the laboratories on ice. 2.4. Laboratory analyses None of the laboratories in Mali was accredited. The analyses were therefore done in South Africa and Norway. For the analyses of protein, carbohydrate and fibre, three different laboratories were used. The methods used are shown in Table 2. For the analysis of protein in maize collected in 1998 (Kayes region) the National Institute of Nutrition and Seafood Research in Norway was used. For carbohydrate and fibre on samples collected in 1998, and for all the analyses done on the 1999 (Se! gou and Mopti regions) and 2000 (Bamako) samples, another Norwegian laboratory, LabNett St^rdal A/S, was used. The last samples collected in Timbuktu in 2002 were analysed in co-operation with the Food and Agricultural Organization of the United Nations (FAO). The analyses of these samples were performed by the ARC-Irene Analytical Service in South Africa. For total nitrogen, Kjeldahl and Dumas combustion methods were used. ARTICLE IN PRESS I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 295 Table 2 Analytical methods used for the different parameters and laboratories Parameters Description Protein Determination of crude protein Carbohydrate Determination of sugar Determination of starch Analytical methods used, Assay name (Eurofood Codea) Institute of Nutrition and Seafood research, Norway LabNettac St^rdal A/S, Norway ARC-Irene Analytical Services, South-Africa PROT0916 (ME61— Kjeldahl method) 93/28EØF (ME61— Kjeldahl method) ASM 041 Dumas Combustion (ME73—protein from total nitrogen) HPLC (ME52— HPLC) Enzymatic hydrolysis 99/79/EU (ME70— polarimetry) (ME85— spectrophotometry) Enzymatic hydrolysis (ME85— spectrophotometry) Fibre Total dietary fibre AOAC (ME50— gravimetric method) a Guidelines notes for preparing and exporting food composition data according to the common formats of export files (2000). Vignat J., Ireland J., M^ller A. and Charrondi"ere U.R. (poster). The methods used for carbohydrate (sugar and starch) were polarimetric, spectrophotometric and HPLC, and a gravimetric method was used for fibre. All the laboratories were accredited for the analyses, but as shown in Table 2, they used different methods. 2.5. Conversion factors The conversion factors for nitrogen to protein were 6.31 for millet, 6.25 for sorghum, maize and fonio, 5.95 for rice and 5.70 for wheat flour (Greenfield and Southgate, 1992). 2.6. Water content For making comparisons and averaging values, the water content in the cereals was standardized by calculation. The water content was assumed to be 12% for millet and fonio, and 11% for sorghum, maize and rice. Nutrient contents were then calculated from dry weight values assuming these moisture contents. 2.7. Statistics Statistical Package for Social Sciences (SPSS for Windows, 2001) was used to estimate the mean and standard deviations of the samples. ARTICLE IN PRESS 296 I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 3. Results The results revealed large differences in the content of protein, carbohydrate and fibre, and they occurred both in the different regions and in the different cereals. Tables 3a and b show differences between the regions in the protein content of millet, maize and wheat. The mean values were 8.0 g/100 g in millet, 10.4 g/100 g in sorghum, 7.6 g/100 g in maize, 6.1 g/100 g in rice, 10.3 g/100 g in wheat and 7.2 g/100 g in fonio. For millet Mopti was the region with the lowest values, being 26% lower than the mean and Se! gou had the highest content, with 10% above the mean. The maize in Mopti had the lowest protein value, namely 13% lower than the mean (7.6 g/100 g cereal) while Timbuktu had 24% higher value than the mean. The protein content in wheat flour from Timbuktu had the highest value, namely 15% higher than the mean. The lowest content of protein in wheat flour was in Bamako, with 10% lower than the mean (10.3 g/100 g cereal). The mean values for carbohydrate were 65.7 g/100 g in millet, 73.5 g/100 g in sorghum, 73.0 g/ 100 g in maize, 83.7 g/100 g in rice, 75.1 g/100 g in wheat and 74.3 g/100 g in fonio. For carbohydrate the difference was most obvious for millet and maize (Table 3a). For millet, both minimum and maximum values differed around 15% from the mean, lowest content in Mopti and highest in Timbuktu. For maize the Mopti region again had the lowest value, namely 19% lower than mean while Timbuktu had the highest, with 13% above the mean. The mean values for fibre were 6.2 g/100 g in millet, 4.7 g/100 g in sorghum, 4.6 g/100 g in maize, 1.1 g/100 g in rice, 3.0 g/100 g in wheat and 2.2 g/100 g in fonio. Percentage differences from mean were from 18% to 25% in millet, maize and fonio and all the values were from cereals from Mopti. On the other hand, the differences were close to zero in sorghum, rice and wheat (see Tables 3a and b). 4. Discussion 4.1. Possible reasons for variation in the data The collection of food items was done in accordance with the sampling plan. As a collection method, primary food samples from different cultivars, obtained by division and reduction processes, were used. The disadvantage with this method is that the information on the variation between the same food in the region is lost, although it permits to reduce the cost of nutrient analysis. The ample variations as reported here, even though this method was used, were not expected. It appears that the variation is real; the differences between Se! gou and Mopti regions are examples of this. The sample collection was done in the same year and season (cold season and just after the harvest), by the same fieldworkers; the samples were handled in the same way, and analysed at the same laboratories using the same methods. The collection in Bamako was done one year later, but in the same season. The collection method was the same as for Se! gou and Mopti, the same fieldworkers were used together with some new, the handling of the samples was the same and so were the laboratory and analytical methods. The last collection, in the Timbuktu region, was done 2–3 years later. It was the hot season and the stocks were almost depleted. This may have had an influence on the results. Also, another laboratory was chosen and thereby also (a) Millet, sorghum and maize Region Millet, flour Sorghum, flour Maize, flour Protein % Diffa CHO % Diffa Fibre % Diffa Protein % Diffa CHO % Diffa Fibre % Diffa Protein % Diffa CHO % Diffa Fibre % Diffa — 8.8 5.9 8.5 8.7 8.0 1.4 10.0 26.3 6.3 8.7 n.r. — 66.5 55.3 n.a. 75.4 65.7 10.1 1.2 15.8 14.8 n.r. — 7.8 4.6 n.a. n.a. 6.2 2.3 25.8 25.8 n.r. — 9.8 10.4 10.0 11.3 10.4 0.7 5.8 0.0 3.8 8.7 n.r. — 71.5 70.5 n.a. 78.6 73.5 4.4 2.7 4.1 6.9 n.r. — 4.7 4.6 n.a. n.a. 4.7 0.1 1.1 1.1 n.r. 7.2 7.1 6.6 7.8 9.4 7.6 1.2 5.3 6.6 13.2 2.6 23.7 n.r. 77.8 72.6 59.0 n.a. 82.5 73.0 11.8 6.6 0.5 19.2 13.0 n.r. 4.7 5.5 3.7 n.a. n.a. 4.6 1.3 1.4 18.7 20.1 n.r. (b) Rice, wheat and fonio Region Rice Wheat, flour Fonio Protein % Diffa CHO % Diffa Fibre % Diffa Protein % Diffa CHO % Diffa Fibre % Diffa Protein % Diffa CHO % Diffa Fibre % Diffa Kayes S!egou Mopti Bamako Timbuktu Mean s.d. — 6.3 6.2 6.6 5.8 6.1 0.3 3.3 1.6 8.2 4.9 n.r. — 77.1 77.1 n.a. 92.3 83.7 7.8 7.9 7.9 10.3 n.r. — 1.1 1.1 n.a. n.a. 1.1 0.0 0.0 0.0 n.r. — 9.8 10.2 9.3 11.8 10.3 1.1 4.9 1.0 9.7 14.6 n.r. — 73.7 74.4 n.a. 77.2 75.1 1.8 1.9 0.9 2.8 n.r. — 3.0 3.0 n.a. n.a. 3.0 0.0 0.0 0.0 n.r. —=not collected, n.a.=not analysed, n.r.=not relevant, CHO=carbohydrate. Numbers in bold shows biggest percentage difference from mean for each cereal and nutrient. a % Diff=percent differences from mean. — 7.4 7.5 6.7 — 7.2 0.4 2.8 4.2 6.9 n.r — 74.3 74.4 n.a. — 74.3 0.1 0.0 0.1 n.r. — 1.9 2.6 n.a. — 2.2 0.3 13.6 18.2 n.r. ARTICLE IN PRESS Kayes S!egou Mopti Bamako Timbuktu Mean s.d. I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 Table 3 Content of protein, carbohydrate and fibre (g/100 g) (a) in millet, sorghum and maize; (b) rice, wheat and fonio, from five different regions in Mali 297 ARTICLE IN PRESS 298 I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 some of the methods were different. For protein it seems that the methods had no influence on the data, as both methods measure crude protein. For millet the large difference in protein content was between Se! gou and Mopti regions (analysed with the same method) and there were no large differences for sorghum and rice (Tables 3a and b). For maize there are results from five regions. Timbuktu had the highest value and it should be investigated if maize from this region actually contains more protein than the others and why. The same also applies to wheat flour. Wheat is cultivated in this region and the protein content in wheat flour was higher than in the imported flour collected in the other regions. All the retailers were asked if fertilizer was used, but most of them, in all the regions, did not know. For carbohydrate content the differences in millet and maize were considerable between Mopti and Timbuktu. Different methods were used but all samples were analysed by hydrolysis. So these differences may well be real but further research is needed. The fibre content in the cereals varied for some cereals but not for others. Tables 3a and b show no values from Timbuktu because crude fibre was analysed by mistake and therefore is not included. 4.2. How to fit the results into a database How to handle these differences in a food composition table and database is a crucial question. There is no agreed upon range of acceptable values above and below the mean. This is a common problem in food database compilation and for clarity maybe one should come to an agreement on a standardized cut-off point. Are 20% differences from the mean a sensible cut-off point when considering including or rejecting values? Removing outliers and using mean figures is not an adequate approach. Removing the ‘‘trouble protein’’ from millet, maize and wheat to make acceptable mean values (less than 20% difference from the mean) will eliminate the existing diversity in the country. For carbohydrate all the differences were less than 20% and the variation would be considered as acceptable according to this cut-off point but not if the cut-off point was 10%. For fibre the type of food seems to be the most important factor but the number of analyses are too few to conclude. These analyses point out the importance of adequate cut-off points. Both mean values and separate tables with figures from each region showing the variation should be included in food composition databases and printed tables. The data should be compiled into databases open for general use, preferentially within the INFOODS system. 4.3. The end users and need for different quality criteria for the data Food composition tables give detailed information on the nutrient composition of foods. It is used as a tool in assessing the nutritional quality of diet, food and nutrition security work, food and nutrition training, education and research, food production, regulations, control and trade, consumer protection and information, as well as epidemiological research. The end users want as good quality on the data as possible, to ensure correct use of the data. It should be obvious that epidemiological research and dietary assessment need representative data of the best quality, meaning that it is important that the figures in the food database reflect the food composition that investigated persons have eaten. Using average values from central regions can create problems since they may not be representative for foods grown locally in Mali, a country where people may ARTICLE IN PRESS I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 299 eat considerable quantities of staple foods; e.g., a survey in 1997 (Diallo et al., 1998) showed that the average intake of cereals among adults was 650 g per day. Using maize as the staple and the figures for protein from Table 3a, the average intake of protein would have varied from 61 g per day by the highest value to 43 by the lowest. The mean value would give 50 g. For the Mopti population, e.g., where the protein content was 6.6 g/100 g maize, this could lead to incorrect assessment of a diet; a diet too low in protein would be assessed as adequate just by using the mean value from the food composition table. This shows the disadvantage of using average values and can lead to erroneous conclusions and wrong advice, with possible harmful effects, especially for children. In epidemiological surveys when researchers try to find a relationship between nutrient intake and health, relationships can be hidden because the survey under- or overestimates intakes of different nutrients (Willet and Buzzard, 1998). This can be one of the reasons why it is difficult to find evidence for nutrient and disease associations. In other situations, e.g., in teaching about food and nutrition in general the regional differences are not so important and representative average figures for the nation can be used. Professionals working on food composition table and database have tried to inform the epidemiological researchers about this issue for years (Southgate, 2002), but still it is easily overlooked and it is hard to convince research funders of the necessity for food collection and analyses in order to improve food composition databases. Food analyses should be included in plans for epidemiological surveys, even though it will increase the budget. 5. Conclusion In the Food Composition Table for Mali, data for different geographical and ecological areas are included. This will make use of the data for different purposes more relevant and meet the needs of various users. It would in particular be important when data are used for dietary advice in different regions, and in local epidemiological research. But still not enough is known about the nutrient content of the foods in Mali and their large differences, although local ecological variations may be important. More research and enhanced co-operation between different experts in nutrition, analytical laboratories, industry, and agencies in charge of food composition tables and database development are necessary and should be strengthened. Acknowledgements The authors would like to thank the Str^mme Foundation in Norway and Bamako for their financial support and co-operation. We also acknowledge the Action d’Appui aux Initiatives de De! veloppement de Bafoulabe! (AIDEB) in Bafoulabe! and the Care Mali and fieldworkers who have taken part in the collection of the samples. We also acknowledge the support from FAO for collection of some of the food samples. ARTICLE IN PRESS 300 I. Barikmo et al. / Journal of Food Composition and Analysis 17 (2004) 291–300 References Barikmo, I., Ouattara, F., 1998–2002. Protocol for Sample Collection. Akershus University College, Norway (Personal communication). Barikmo, I., Ouattara, F., Oshaug, A., 2002. Competence building in Mali, linked up to food—and nutrition security with food composition table, database management system and data program as tools. Report, Akershus University College, Norway. Diallo, F., Diarra, M., Ouattara, F., Hatl^y, A., Oshaug, A., Torheim, L.E., Barikmo, I., 1998. Rapport de l’enqu#ete de base. Oussoubidiania et Ouassala, Bafoulab!e Cercle, 1997. Programme de collaboration PIDEB/INRSP/Universit!e d’Oslo Mali/Norv"ege 1997–2000, Oslo, Norway. FAO/WAICENT, 2003. Division des Statistiques FAOSTAT (WAICENT). ESN—Aper@us nutritionnels par pays—Mali 26 Janvier 1999. FAO, Rome, http://www.fao.org/es/ESN/nutrition/mli-e.stm. Greenfield, H., Southgate, D.A.T., 1992. Food Composition Data—Production, Management and Use. Elsevier Science publishers LTD, Barking. Nordeide, M.B., Hatl^y, A., F^lling, M., Lied, E., Oshaug, A., 1996. Nutrient composition and importance of green leaves and wild food resources in an agricultural district, Koutiala, in southern Mali. International Journal of Food Sciences and Nutrition 47, 455–468. Nordeide, M.B., Holm, H., Oshaug, A., 1994. Nutrient composition and protein quality of wild gathered foods from Mali. International Journal of Food Sciences and Nutrition 45, 276–286. NRC, 1996. National Research Council. Board on Science and Technology for International Development. In: Lost Crops of Africa: Vol. I: Grains. National Academy Press, Washington, DC, pp. 59–76. Ouattara, F., Barikmo, I., 2002. Protocol for Sample Collection, 1999. National Institute for Research in Public Health. Bamako, Mali. SISEI Mali, 2003. Syste" me d’information et de suivi de l’environnement sur Internet au Mali, http://www.sisei.net/ nationaux/mali. Southgate, D.A.T., 2002. Data quality in sampling, analysis, and compilation. Journal of Food Composition and Analysis 15, 507–513. SPSS for Windows, 2001. Release 11.0.0. SPSS Inc., Chicago. UN, 2003. Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat. World Population Prospects: The 2002 Revision and World Urbanization Prospects: The 2001 Revision, http:// esa.un.org/unpp. Willet, W., Buzzard, I.M., 1998. In Nutrition Epidemiology, 2nd Edition. Oxford University Press, New York, pp. 23–32.