Contents - Previous - Next


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


Some potential problems with incomplete and non-standardized nutrient data bases

Food composition data is required on many different levels of complexity and in forms that are readily computerized. In addition, users of such nutrient information need computerized composition data in a variety of formats that are not readily compatible with currently available data tapes. Thus, many epidemiologists are developing specialized data bases, usually by modifying or adding to the USDA tapes.

Table 1. Selected epidemiologic studies of dietary factors and colon cancer reported after 1977

Ref no. Year
pubIished
Place of
study
Type of
study
Study
method
Source of
data base
Risk factor
association

Dietary fibre

12 1977 Finland,
Denmark
Cross-sectional Dietary record D10 +
23 1977 India Cross-sectional Food frequency - +
5 1978 USA Case-control Food frequency - +
21 1978 Utah, USA Ecological study Food consumption - 0
22 1978 Scandinavia Cross-sectional Dietary record D7 +
37 1978 South Africa Ecological study - - +
20 1979 Many countries Ecological study Food consumption D11 0
27 1979 4-country Ecological study Food consumption D11 +
1 1979 UK Ecological study Food consumption D4 0
15 1979 Scandinavia Prospective study Food frequency D1 0
29 1979 Israel Case-control Food frequency - +
13 1979 Canada Case-control Diet History D7 0
3 1980 Kenya Prospective study Food frequency - +
26 1980 Australia Migrant study Food consumption - +
35 1981 Israel Case-control Food frequency D1, D2, D3 +
16 1982 Denmark, Finland Cross-sectional Dietary record D4, D5, D6 +
6 1982 4-country Cross-sectional Dietary record D4, D5, D5 +
28 1983 Canada Case-control Food frequency D7 +
33 1983 3-country Cross-sectional Diet history D5, D9 +
31 1984 7-country Ecological study Food consumption D11 +
19 1984 SDA in USA Ecological study Dietary record D7 +

Fat and/or meat

12 1977 Finland, Denmark Cross-sectional Dietary record D10  
5 1978 USA Case-control Food frequency    
40 1978 Buffalo, USA Case-control Food frequency - 0
21 1978 Utah, USA Ecological study Food consumption - 0
22 1978 Scandinavia Cross-sectional Dietary record    
27 1979 4-country Ecological study Food consumption D11 0
1 1979 UK Ecological study Food consumption D4  
20 1979 Many countries Ecological study Food consumption D11 0
40 1979 Many countries Ecological study Food consumption D11  
9 1980 Japan Case-control Diet history - 0
13 1980 Canada Case-control Diet history D7  
26 1980 Australia Migrant study Food consumption    
32 1980 USA Ecological study Food consumption    
35 1981 Israel Case-control Food frequency D1, D2, D3  
18 1981 Hawaii, USA Cross-sectional Food frequency D7 0
16 1982 Denmark, Finland Cross-sectional Dietary record D4, D5, D6 0
17 1982 UK Ecological study Food consumption - 0
24 1983 Greece Case-control Food frequency    
28 1983 Canada Case-control Food frequency D7  
4 1984 USA Cross-sectional Dietary record D7  
36 1984 Hawaii, USA Prospective study 24-hour recall - -

Beer and/or alcohol

7 1977 USA Ecological study Food consumption    
22 1978 Scandinavia Ecological study Dietary record    
27 1979 4-country Ecological study Food consumption    
1 1979 UK Ecological study Food consumption D4 0
15 1979 Scandinavia Prospective study Food frequency D10 0
11 1980 Hawaii, USA Cross-sectional Food frequency - 0
24 1983 Greece Case-control Food frequency - 0
28 1983 Canada Case-control Food frequency D7 0
14 1983 SDA in Denmark Prospective study Food consumption    
30 1984 Nebraska, USA Case-control Diet history D7, D8  

Cruciferous vegetables

8 1978 Buffalo, USA Case-control Food frequency - +
9 1980 Japan Case-control Diet history - +
24 1983 Greece Case-control Food frequency - +
28 1983 Canada Case-control Food frequency D7 0

a. Risk factor: "-" = harmful effect; "0" = no effect; i`+" = protective effect.

Sources of data bases:

D1. Y. Guggenheim, N. Kaufman, and A. Reshaf, Food Composition Tables (Ministries of Health and Culture, Government School of Home Economics and Nutrition, Romema, Jerusalem, 1980).

D2. R. M. Freely, P. E. Criner, and B. K. Watt, "Cholesterol Content of Foods," J. Am. Diet. Assoc., 61: 134148 (1972).

D3. R. M. Narayana and M. Polacchi, Food Composition Table for Use in East Asia, part 2 (NIAMDD; NIH, DHWS, Bethesda, Md., 1972), pp. 298 301.

D4. D. A. T. Southgate, "Dietary Fibre: Analysis and Food Sources," A.J.C.N., Suppl. 31: s107-s110 (1978).

D5. R. A. McCance and E. M. Widdowson, eds., The Composition of Foods, 4th ed. (HMSO, London, 1978).

D6. W. P. T. James and O. Theander, eds., Analysis of Dietary Fiber in Foods (Marcel Dekker, New York, 1981).

D7. US Department of Agriculture, "Composition of Foods: Raw, Processed, Prepared," Agriculture Handbook No. 8 (Science and Education Administration, USDA, Washington, D.C., 1968; expansion, 1972).

D8. US Department of Agriculture, Nutritive Value of American Foods in Common Units, Agriculture Handbook No. 456 (US Government Printing Office, Washington, D.C., 1975).

D9. Consumer and Food Economics Institute, Nutrition Value of Foods (USDA' Washington, D.C., 1971).

D10. Laboratory analysis.

D11. FAO, Food Balance Sheets (FAO, Rome, 1977, 1980).

For example, almost every researcher who begins nutrition-related clinical or population-based studies begins by finding and purchasing a data base that must then he modified (usually by a review of the literature) for the specific foods or nutrients under study. However, the uncoordinated creation of such data bases makes it virtually impossible to compare nutritional studies that utilize different data bases even when those data bases are relatively well known and documented. In addition, the repeated modification of existing USDA tapes duplicates effort and increases costs that could be minimized by having an available standardized data system.

Table 2. Selected nutrient composition of several varieties of cabbage reported by two different food composition tables (amount per 100 g)

  Kcal Fibre
(g)
Fat
(g)
Vit. A
(IU)
Vit. C
(mg)

Bowes and Church

Headed 24 0.8 0.2 130 57
Red 31 1.0 0.2 40 61
Savoy 24 4.6 0.2 200 55
Chinese 14 0.6 0.1 150 25
Spoon 16 0.6 0.2 3,100 25

USDA

Headed 24 0.8 0.18 126 47
Red 27 1.0 0.26 40 57
Savoy 27 0.8 1.0 1,000 31
Chinese 13 0.6 0.20 3,000 45
Spoon 16 0.6 0.20 1,200 27

Sources: A. dc P. Bowes and C. F. Church, eds., Food Value of Portion Commonly Used, 12th ea., rev. C. F. and H. N. Church (J. B. Lippincott, Philadelphia, Pa., 1975); US Department of Agriculture, "Composition of Foods: Raw, Processed, Prepared`" Agricultural Handbook No. 8-11 (Science and Education Administration USDA, Washington, D.C., 1984).

With reference to the colon cancer literature, ten different nutrient data base sources were cited in the studies listed in table 1. The variability of data can be demonstrated by the differences in data on the nutrients contained in foods in even well-known food composition tables, as shown in table 2. Here, reported vitamin A levels show differences between source A and B for savoy, Chinese, and spoon (pi-tsai) cabbage.

Epidemiological investigations could also be improved if the foods chosen for nutrient analysis were representative of those foods consumed by the study population. For example, the colon cancer studies identified in table 1 were conducted in numerous regions all over the world. However, because there has been no systematic sampling frame, it is difficult to determine how well the values in food composition tables represent various regional and national food supplies. Note that the nutrient content of the varieties of cabbage indigenous to various world regions may differ. Consider the differences in fibre and vitamin A content for the four different kinds of cabbage shown in table 2. There is more than a fourfold difference in the fibre content between the savoy cabbage and the Chinese and spoon varieties, while the vitamin A values ranged from 40IU per l00g for red cabbage to 3100IU per l00g for spoon cabbage. Thus, food sampling is a key issue in developing food composition tables, especially since the world supply is constantly expanding and the product on offer changing. Sampling should include new strains of edible plants and animals [10]. At present, none of the major food composition tables are based on sampling that is representative of the foods offered to consumers in defined geographic regions. Instead, data is compiled from food industry, government and independent laboratories, and from the scientific and technical literature, with each covering a different geographic area [2, 10, 25, 34, 38, 39]. The current practice is to weight the averaged analytical values of foods that are similar but not identical. Weighting schemes reflect geographic production of samples, and seasonal availability or production figures [381. However, without a representative sample of the food at the retail level, weighted models for many foods must remain empirical. Systematic sampling is also required to determine the variance of nutrients in foods consumed by specific regional populations.

As shown in table 3, food sampling variance differs from one nutrient to another and from food to food. For example, note that the standard error for the mean is large for the iron content of apples selected from the retail food supply of Utah, but is small for the fibre content. A statistically significant nutrient difference in consumption may be observed between two groups, but these differences are not meaningful if the difference is less than the food sampling variance.

In addition to regional, seasonal, and maturational variations and differences between various parts of a foodstuff, variability in reported food composition data may also be caused by differences in analytical method. The new methodological advances in the field of nutrient analysis, including widespread use of radio-immunoassay (RIA), radiobioassay, fluorometry, atomic absorption, neutron activation, high performance liquid chromatography (HPLC), stable isotope electrophoresis, and auto analysis techniques, among others, are creating masses of new data which need to be rapidly incorporated into existing data bases if these are to be kept current and relevant.

However, there are differences in reported food composition data due to intra- and interlaboratory variance even when samples are assayed by the same analytical techniques. Such analytical errors could bias study outcomes. The difference in the fibre content of foods analysed by three different analytical methods (shown in table 4) illustrates the point. Note that crude fibre values generally underestimate dietary fibre as measured by the newer assays. Neutral detergent fibre values from two sources, Van Soest and Mahoney, show interlaboratory variation. It should be noted that assays were performed on different food samples. However, inter-laboratory differences between different food samples are less than those observed between different analytic methods. Such differences point out the need for suitable standard reference materials that can be distributed to laboratories as part of the quality-control process.

The effects of processing may significantly alter the nutrient content of foods. Processing includes harvesting, mechanical and heat treating, packaging, and storage procedures. Processing food products together also alters the nutrient content of the products. For example, deep frying potatoes in vegetable oil increases the fat content of the product as eaten. Table 5 shows the effects of boiling on selected cruciferous vegetables. Although calories remain constant, vitamins A and C decrease with cooking.

Cruciferous vegetables have characteristics as a group that appear to be protective for some types of cancer, including bowel cancer. Note, however, that the nutrient content of these vegetables is quite different. Thus, using cruciferous vegetables as a class reduces the quantitative power of a study unless the proportions of the individual vegetable consumed are known; one should therefore document manipulations of collapsing of data in a data base used for specific studies.

Going one step further, the potential protection of crucifers may be conferred by the nonnutrient compounds, aromatic isothiocyanates. There is no quantitative information about the concentration of these compounds in foods, but there is no reason to believe that their levels in foods are any more constant than those of essential nutrients. The study of the relationship of diet to health and disease may therefore

Table 3. Selected examples: Nutrient composition of foods in Utah (amount per 100 edible material). Retail food sample variance

Food and description Index number Watera
(g)
Fat
(g)
Protein
(g)
Neutral detergent fibre
(g)
Iron
(
mg)
Copper
(
mg)
Zinc
(
mg)
Mn
(
mg)
Ash
(g)
Apples 13 87.9 0.72 0.21 1.1 433 58 13 117 0.20
Raw,commercialvarieties:   0.75 0.35 0.04 0.08 188 13 16 50 0.03
not pared   (87.1-88.7) (0.33-1.1) (0.16-0.25) (0.94-1.2) (197-717) (43-78) (0.00-40) (40-175) (0.16-0.25)
Apple sauce,canned 29 77.0(4) 0.49 0.14 0.71 522 55 55 55 0.17
Sweetened   1.53 0.92 0.01 0.08 257 11 18 4.2 0.3
    (74.9-78.6) (0.00 1.87) (0.13-0.17) (0.62-0.81) (160- 761) (40-60) (31-74) (50-60) (0.12-0.21)
Apricots 30 85.2 0.39 1.86 1.32 1,220 164 245 175 1.03
Raw   5.46 0.38 1.50 0.33 510 115 262 43 0.35
    (77.4 89.9) (0.09-1.01) (1.08-4.54) (0.77-1.59) (137-1,990) (68-358) (73-608) (135-217) (0.66 138)
Apriots   72.9 0.56 0.94 0.81 678 114 190 90 0.29
Canned,heavy syrup:   2.46 0.43 0.19 0.20 351 36 46 49 0.17
drained solids   (69.9-75.9) (0.00-1.2) (0.62-1.1) (0.61-1.1) (476-1,300) (69-158) (147-258) (27-156) (0.00-0.46)
Asparagus 52 94.0(4) 0.64 2.14 1.06 1,830 96 404 175 0.76
Canned spears:   1.3 0.12 0.29 0.08 1,570 23 45 132 0.54
green,regular peck:   (92.7-95.2) (0.49-0.80) (1.85-2.63) (0.99-1.43) (601-3,960) (68-117) (340-477) (97-408) (0.28-1.25)
drained solids                    
Asparagus 63 92.4(4) 0.41 2.95 1.1 638 170 556 172 0.60
Frozen spears:   0.31 0.15 0.21 0.23 131 10 114 35 0.08
cooked, boiled, drained   (91.9-92.6) (0.29-0.64) (2.68-3.17) (0.82-1.4) (546-830) (161-180) (431-707) (131-214) (0.53-0.73)
Bacon, cured 126 14.9 44.0 31.6   2,180 360 3,140 103 7.50
Cooked,broiledor fried,   5.33 5.91 6.83   640 78 574 18 1.64
drained   (11.1-24.3) (34.9-49.4) (25.6-42.3)   (1,500- (245-458) (2,430- (77-119) (5.11-9.72)
            2,960)   4,010)    
Bananas, raw 141 77.7(6) 0.70 0.96 1.1 382 130 155 152 0.74
Common   3.00 0.66 0.17 0.48 38 20 18 44 0.07
    (73.9-82.1) (0.22-1.9) (0.73-1.1) (0.68-1.8) (346-452) (111-167) (131-184) (98-208) (0.61-0.84)

a. Data reported as mean, standard error of the mean (SEM) and the range.

Table 4. Fibre content of various foods by different analytic methods (amount per 100g)

  Crude fibre (g)a Dietary fibre (g)c Neutral detergent fibre (g)c
      1 2
All bran cereal 7.80 26.7 32.98  
Whole wheat bread 1.60 8.50 1.55 2.60
Apple 0.40 3.71 0.89 1.10
Broccoli 1.50 4.10 1.34 1.42
Cabbage 0.80 2.83 1.11 1.12
Potato 0.50 3.51 2.33 0.67

a. US Department of Agriculture, "Composition of Foods: Crude Fiber,"Agriculture Handbook No. 8 (Science and Education Administration, USDA, Washington, D.C., 1983).
b. D. A. T. Southgate, "Dietary Fiber," J. Hum. Nutr., 30: 303 (1976).
c. Neutral detergent fibre: (1) P. J. Van Soest, "Fiber Analysis Table: By the Amylase Modification,"A.J.C.N., 31: s281-s284 (1978): (2) A. W. Mahoney, S. K. Collinge, B. H. Byland, and A. W. Sorenson, Nutrient Composition of Foods Contained from Retail Outlets in Utah, Utah Agricultural Experiment Station, Research Report, 53 (Utah State University, 1980).

Table 5. Selected nutrient composition of cruciferous vegetablesa by different cooking methods (amount per 100 g)

Type

Raw

Cooked (boiled, drained)

  Kcal Fibre
(g)
Fat
(g)
Vit. A
(IU)
Vit. C
(m)
Kcal Fibre
(g)
Fat
(g)
Vit. A
(IU)
Vit. C
(m)
Cabbage (common) 24 0.80 0.18 126 47 21 0.60 0.25 86 24
Cauliflower 24 0.85 0.18 16 72 24 0.82 0.17 14 55
Brussels sprouts 43 1.51 0.30 883 85 39 1.37 0.51 719 62
Broccoli 28 1.11 0.35 1,542 93 29 1.20 0.28 1,409 63

a. Colon cancer protection may be conferred by a non-nutrient(s) component of food, aromatic isothiocyanates. There is no quantitative information on the proposed protective agent.

Source: US Department of Agriculture, "Composition of Foods: Raw, Processed, Prepared," Agriculture Handbook No. 8-11 (Science and Education Administration, USDA, Washington, D.C., (1984). require accurate information on the non-nutrient as well as the nutrient components of food. However, there is very little non-nutrient food composition data available.

In addition to non-nutrient data, future epidemiologic research will demand more information on subunits of nutrients, including data on biologically active forms of compounds found in food with chemically or physically different components. Pyridoxal/pyridoxine dietary fibre and carotenoids and retinoids are examples of difficult biological forms of nutrients.

A major hindrance to epidemiologic studies are data sets with missing values. Even though a computerized system is designed to update and expand food composition data, there will always be incomplete nutrient or food information which will necessitate users' judgement for dealing with missing data. Epidemiologists, like other users, are forced to fill in "zeros" in data bases with imputed values. Estimates of missing values may come from data on similar items, recipe calculations, or even values based on educated guesses. Raw values are often substituted for food usully consumed cooked, for example in relation to meat. And sometimes a food or food group is used as a surrogate for the nutrient content of diets: milk has been used to estimate retinoid values of diets while selected fruit and vegetables have been used as estimators of beta-carotene. Decisions regarding inputting missing values would be better made on standardized criteria developed by panels of experts in the fields of nutrition and data base management.


Contents - Previous - Next