• Food Description Effects
After the foods and components have been determined, the individual and specific products which represent a food must be identified. For a single food (e.g. beef, pizza, or eggs) the investigator can define the characteristics of the product which may influence the composition and variability for the component(s) of interest. Relevant characteristics will include the primary food source and scientific name for a product (e.g. wheat v. corn, coconut v. sunflower, beef v. pork), the part of the plant or animal used, preservation state, food processing treatments, added ingredients, etc. For some components, geographical source (e.g. broccoli, above) and ripening practices will be important. As mentioned, carotenoid values for broccoli cultivars grown in Guatemala were different than values for fresh broccoli grown in California (24). For others, packaging type, pH, or storage conditions will be sources of variability. In recent years, several food description systems have been developed which provide classifications of important descriptors (29, 30). The specific products and components of interest will determine the preliminary list of descriptors for the products.
Following the definition of relevant descriptors for a food, it is necessary to identify the specific major sources of that food which are consumed by the population of interest. In addition, the distribution and marketing schemes need to be identified. For branded products, sales volume data and product information are important to the selection of representative units (9). For commodity products, such as meats, eggs, milk, etc., it is possible to identify the major breeds or cultivars, as well as major commercial purveyors of the products and an approximation of their sales ranking (31). In regions where food production is localized the major outlets for products (butcher, bakery) or ingredients (flour mill, refineries) can be identified. Some products may be manufactured in one location and distributed nationwide while others may be formulated in many regions from different sources of raw ingredients. In a recent study of selenium levels in bread, ninety samples of white bread were selected in nine major population areas across the US (32). Bread is baked in regional or local bakeries in or near cities and towns. Yet most of the wheat in the US is grown in the north central plains area, an area of relatively high selenium levels, and then transported to major metropolitan areas to supplement smaller regional supplies grown near those areas. Selenium levels in bread in several of the cities varied as the local source of the flour was supplemented by the supply grown in the north central area. The study demonstrated that selenium levels in breas were more closely related to the source of wheat levels of selenium in soil where the grain was grown than to the selenium levels in soils where the bread samples were purchased. By defining the form of the product and its sources, the investigator can begin to determine which specific products will need to be selected as well as the time and location for sampling.
• Food Consumption Patterns
After marketing and distribution variables have been defined, consumption patterns should be assessed to determine where to select the samples. If the objective is to determine estimates for foods in a national database then it is necessary to sample food products on the basis of the population distribution and product use. Several questions should be answered: Is the food consumed frequently and in significant amounts by the population of interest? In which regions or populations is the food consumed? Is the food consumed more in rural areas than towns? If the food is widely consumed by many sub-groups, what is the distribution of the population in the country or region of interest? Major population centers within a country can be identified and used as locations for sample selection. In the US the majority of the population is concentrated in a number of metropolitan areas called Metropolitan Statistical Areas (MSAs) and defined by the US Office of Management and Budget as cities which have at least 50,000 persons or an urbanized area of at least 50,000 with a total population of at least 100,000 individuals (33). The top ten cities, their percent of the population, and their respective proportion of grocery sales are given in Table I. The percent of the population represented by the top 100 MSAs as well as the number of supermarkets is also given. In most major cities two to four supermarket chains dominate each city. Most are significant regional vendors.
Table I. Top 10 US MSAa markets by populationb
|% of US Totals|
|1||Los Angeles-Long Beach, CA||3.59||2.39|
|9||Riverside-San Bernadino, CA||1.14||0.90|
|Top 100 MSA Market||59.86||50.86|
|All Other U.S.||40.14||49.14|
|U.S. Total Figures||254,926,669||30,552|
a Metropolitan Statistical Area
b Adapted from Progressive Grocer's Market Scope (40)
While many sample mixes are selfweighted–that is, the available products are similar to the number and kind needed to mimic sales volume, it is possible to weight the sample estimates after analysis of equal numbers of individual units/per brand or region by applying pre-determined weighting factors (10). In view of the nationwide distribution and market share of many products and the concentration of the population in major MSAs the USDA and others have selected representative units of foods from retail grocery stores and/or restaurants in three to ten cities across the country (22, 27, 28, 34). For example, for the recent study of selenium in approximately 200 foods, sample units were purchased in two major supermarkets in each of nine cities (Holden, unpublished data). Two to three cities were selected in each of four regions of the country; two major supermarkets were sampled in each city. For each of the major contributors of dietary selenium (beef, white bread, pork, chicken, eggs) approximately 100 analytical samples were randomly selected and prepared. For minor contributors five to 25 analytical samples were chosen. By choosing units of the highest volume brands within the largest supermarkets in major metropolitan areas it was assumed that the most frequently consumed and representative products were selected for a specific food.
To determine selenium in beef, it was necessary to determine the major categories of beef products in the US diet. Marketing and production data obtained from the US Livestock and Meat Board, the private sector trade association for the meat producers, indicated that the per capita consumption of beef was 72.7 lbs. Fresh beef cuts including steaks and roasts, and ground beef, including bulk ground beef purchased in supermarkets as well as hamburger sandwiches sold in fast food restaurants were the major forms of beef consumed (35). Using this information a sampling plan was developed. Ninety four units of five primal beef cuts were obtained from a larger study of beef composition conducted by Texas A & M (27). The samples had been collected from major retail stores in ten cities. In addition, 58 samples of ground beef collected from a USDA nationwide study were analyzed (34). Finally, 27 samples of hamburger sandwiches were collected in nine cities from each of three prevalent chains (Holden, unpublished data). Mean values and standard deviations were calculated and published in the recently released USDA Provisional Table of Selenium in Foods (36).
• How Many Sample Units Are Needed?
The number of sample units analyzed will determine, in part, the statistical power of the estimate. Although statistical models for calculating the required number of units can be complex and multi-tiered, the following equation indicates the most important facets of the computation for homogeneous populations (10):
The appropriate number of units is based on four parameters. The first is “t”, the abscissa of the normal curve that cuts off an area “a” at the tails of the distribution, indicating the desired confidence level. The standard error of the estimate is denoted by “s” while the sample mean is denoted by . This mean and standard error can be obtained from previously published data or pilot studies, if available. Some existing handbooks of food composition data publish standard deviations or standard errors of the mean and can be used as rough estimates of sample size. Previously published estimates and the scientific objectives for the study should serve as the basis for sample number calculations. The reader should note that the coefficient of variation, if known, can substitute for s/. The limit of the desired relative error in the estimate is indicated by “r.” That is, the proximity of the estimate to the “true” mean, e.g. within 10 per cent, is represented by “r.” The calculation of the appropriate number of samples is an iterative process which begins with an approximation of the number of samples determined by the investigator as a “guess.” The “guess” can be based on preliminary cost estimates or capabilities of the analytical laboratory. After the initial calculation the estimate of number of samples is further refined by recalculation until successive trial values of “n” yield similar values. The cost of sampling can be included in the equation as well. Table II demonstrates the effect of increasing the coefficient of variation on the number of samples required to obtain the same level of confidence. The “t” value was set at 2.00 while r=0.1 for the purpose of the illustration. Further information is given by Cochran (10).
In the past, the mean or average value has been used as the estimate of the level of a component in the food. However, the use of the mean presumes that the statistical distribution of all values for that component in a specific food follows the Gaussian or normal distribution (37). Recently, the USDA Food Composition Laboratory, in collaboration with the US National Cancer Institute, complied and published a food composition table of the levels of five carotenoids in important fruit and vegetable contributors (16). The values were collected from published and unpublished analytical sources. Due to the apparent skewed distribution for several foods and the limited amount of available data (one to 14 acceptable sources per food) the median value was used in the table. However, the use of the median precludes the calculation of a variance indicator. More research is needed to evaluate the characteristics of statistical distributions which result from broad-based original sampling as well as those which result from compilations of data from different sources. Furthermore, the robustness of traditional statistical techniques should be evaluated to determine how appropriate these techniques are for food composition data. The impact of using means v. medians in food composition databases on conclusions drawn from dietary studies, must be tested. In particular, caution is required when estimating food composition values from small data sets.
Table II. Effect of increasing the coefficient of variation on sample sizea
|If CV equals||then n equals|
a α = .05, t = 2.00, r=0.1
b If α = .10 then n = 19
After the sample is defined individual items or units within the sample can be identified and procured to be prepared for analysis. Once the units, packages, or containers arrive in the laboratory their handling (e.g. preparation, homogenization) and the selection of aliquots must be carefully planned to maintain the representativeness and integrity of the material. Since the developer of the project design and sampling strategy may not be the laboratory analyst the importance of communication between these individuals or groups cannot be overestimated. At this point it is important to emphasize the use of standardized nomenclature with regard to sampling at the laboratory level. According to the 1990 recommendations for nomenclature for sampling in analytical chemistry submitted to the International Union of Pure and Applied Chemistry (IUPAC) Horwitz defines the “sample” as “a portion of material selected in some manner to represent a larger body of material. The result obtained from the sample is merely an estimate of the quantity … of constituent … of the parent material.” Previously, the term sample has often been used to refer to the portion (e.g. extract, diluted or not) being analyzed at various points in the analytical process. Other terms such as “test” or “analytical” should be used to describe those portions to avoid inconsistencies or ambiguities and subsequent misinterpretation of the results. The reader is referred to reference (8) for further information.
• How Good Do The Data Have To Be?
Food composition data must be “good enough” to permit the careful assessment of food consumption patterns and their impact on the health of population groups and subgroups. Similarly, the data must be “good enough” to accomplish other scientific and economic objectives defined by investigators. The quality of a specific estimate is based, in part, on the accuracy and precision of the measurement process. The generation of accurate food composition data requires that variability inherent to the food be accurately quantified while variability inherent to the measurement process be minimized. In general, the major sources of statistical variability in dietary estimates are the food consumption data captured by the dietary assessment tool, and the food composition data. Variability for food composition data includes all variability attributable to sampling plans, sample handling, analytical method, and analytical quality control. Each of these sources can be partitioned into the sources of variability and can be quantified by an analysis of variance (37). The assessment of the sources and magnitude of variability for food composition data can indicate areas where improvement in the measurement process needs to be made (38). While sampling is only one source of variability, the lack of representative sampling can increase the degree of bias in the estimates of central tendency and cause errors in the estimates of variance. As previously mentioned, for a specific component, a small number of foods may contribute the majority of that component to the diet of the population. Therefore, it is recommended that sampling resources be dedicated to obtaining statistically sound estimates for those major contributors.
The author wishes to express her appreciation to the First International Food Data Base Conference for generous financial support to attend the conference.
(1) Life Sciences Research Office (1989) Nutrition Monitoring in the United States: An Update Report on Nutrition Monitoring, US Dept. of Health and Human Services, Hyattsville, MD
(2) Steinmetz, K.A., & Potter, J.D. (1991) Cancer Causes and Control, 2, 427–442
(3) Hegsted, D.M., & Ausman, L.M. (1988) J. Nutr. 118, 1184–1189
(4) Katan, M.B., Van Gastel, A.C., de Rover, C.M., van Montfort, M.A.J., & Knuiman, J.T. (1988) Eur. J. Clin Invest. 18, 644–647
(5) Judd, J.T., Clevidence, B.A., Muesing, R.A., Wittes, J., Sunkin, M.E., & Podczasy, J.J. (1994) Am. J. Clin. Nutr. 59, 861–868
(6) Vanderveen, J.E., & Pennington, J.A.T. (1983) Food Nutr. Bull. 5, 40–45
(7) Holden, J.M., Schubert, A., Wolf, W.R., & Beecher, G.R. (1987) in Food Composition Data: A User's Perspective, W.M. Rand, C.T. Windham, B.W. Wyse & V.R. Young (Eds.), UNU Press, Tokyo, pp. 177–193
(8) Horwitz, W. (1990) Pure Appl. Chem. 62, 1993–1208
(9) Nielsen, A.C. Co. (1990) Nielsen Scantrack Data, Northbrook, IL
(10) Cochran, W.G. (1977) Sampling Techniques, 3rd Ed., Wiley, New York, NY, pp. 1–78
(11) Greenfield, H., & Southgate, D.A.T. (1992) Food Composition Data: Production, Management and Use, Elsevier Applied Science, London
(12) Stewart, K.K. (1981) in Beltsville Symposia in Agricultural Research IV Human Nutrition Research, Allenheld, Osmun Publication, Totowa, NJ
(13) Beecher, G.R., & Matthews, R.H. (1990) in Present Knowledge in Nutrition, 6th Ed., International Life Sciences Institute, Washington, DC
(14) Le Marchand, L., Yoshizawa, C.N., Kolonel, L.N., Hankin, G.H., & Goodman, M.T. (1989) J. Nat. Cancer Inst. 81, 1158–1164
(15) Moore, T. (1957) Vitamin A, Elsevier, Amsterdam
(16) Mangels, A.R., Holden, J.M., Beecher, G.R., Forman, M.L., & Lanza, E. (1993) J. Am. Diet. Assoc. 93, 284–296
(17) West, C.E., & Poortvliet, E.J. (1993) The Carotenoid Content of Foods with Special Reference to Developing Countries, USAID/VITAL, Washington, DC
(18) Khachik, F., Beecher, G.R., Goli, M.B., & Lusby, W.R. (1992) Methods Enzymol. 213, 347–359
(19) Rodriguez-Amaya, D.B. (1989) J. Micronutr. Anal. 5, 191–225
(20) Chug-Ahuja, J.K., Holden, J.M., Forman, M.R., Mangels, A.R., Beecher, G.R., & Lanza, E. (1993) J. Am. Diet. Assoc. 93, 318–323
(21) Schneeman, B.O., & Gallaher, D.D. (1990) in Present Knowledge in Nutrition, 6th Ed., International Life Sciences Institute, Washington, DC
(22) Li, B.W., Holden, J.M., Brownlee, S.G., & Korth, S.G. (1987) J. Am. Diet. Assoc. 87, 740–743
(23) Hepburn, F.N. (1988) Proceedings of the 12th National Nutrient Data Bank Conference, The CBord Group, Inc., Ithaca, NY, pp. 31–33
(24) Schubert, A., Holden, J.M., & Wolf, W.R. (1987) J. Am. Diet. Assoc. 87, 285–299
(25) Tonucci, L.H., Holden, J.M., Beecher, G.R., Khachik, F., Davis, C.S., & Mulokozi, G. (1995) J. Agric. Food Chem. (in press)
(26) Micozzi, M.S., Brown, E.D., Edwards, B.K., Bieri, J.G., Taylor, P.R., Khachik, F., Beecher, G.R., & Smith, J.C. (1992) Am. J. Clin. Nutr. 55, 1120–1125
(27) Savell, J.W., Harris, J.J., Cross, D.S. Hale, D.S., & Beasley, L.C. (1991) J. Anim. Sci. 69, 2883–2893
(28) Buege, D., Held, J.E., Smith, C.A., Sather, L.K., & Klatt, L.V. (1990) Research Bulletin R-3509, College of Agriculture and Life Sciences, University of Wisconsin, Madison, WI
(29) McCann, A., Pennington, J.A.T., Smith, E.C., Holden, J.M., Soergel, D., & Wiley, R.C. (1988) J. Am. Diet. Assoc. 88, 336–341
(30) Kohlmeier, L., & Poortvliet, E. (1992) Report of the FLAIR Eurofoods-Enfant Project Second Annual Meeting, Wageningen Agricultural University, Wageningen
(31) Honikel, K.O. (1994) Report, FLAIR Eurofoods-Enfant Project Third Annual Meeting, Wageningen Agricultural University, Wageningen
(32) Holden, J.M., Gebhardt, S., Davis, C.S., & Lurie, D.G. (1991) J. Food Comp. Anal. 4, 183–195
(33) Progressive Grocer's Market Scope (1993) Progressive Grocer's Trade Dimension Division, Maclean, Hunter Media, Inc., Stamford, CT, pp. 18, 348–549
(34) Holden, J.M., Lanza, E., & Wolf, W.R. (1986) J. Agric. Food Chem. 34, 302–308
(35) Knutson, J. (1989) Meatfacts 88, American Meat Institute, Washington, DC, p. 17
(36) Gebhardt, S.E., & Holden, J.M. (1992) Provisional Table on the Selenium Content of Foods, USDA, Washington, DC
(37) Sokal, R.R., & Rohlf, F.J. (1981) Biometry, 2nd Ed., W.H. Freeman and Company, San Francisco, CA
(38) Beaton, G.H., Milner, J., Corey, P., McGuire, V., Cousins, M., Stewart, E., de Ramos, M., Hewitt, D., Grambsch, P.V., Hassim, N., & Little, J.A. (1979) Am. J. Clin. Nutr. 32, 2546–2559
Guangya Wang, Xiaolin Li
Institute of Nutrition and Food Hygiene, Chinese Academy of Preventive Medicine, 29 Nan Wei Road, Beijing 100050, China
A nationwide collaborative project on the analysis of food composition for China was organized and conducted by the Institute of Nutrition and Food Hygiene between 1987 and 1990. In order to assure the quality of analytical data from all 20 participating laboratories, a quality assurance system was conducted involving five procedures: a written manual of analytical methods; technical training courses for laboratory technicians; the use of identical methodological protocols for sampling and handling of food samples; analytical duplicate or replicates for unknown samples; standard reference materials and quality control materials. The results were monitored by means of a control chart to check the reliability of technical performance. Data were evaluated by logic and statistical tests and then compiled into new Chinese food composition tables. The total number of food items is 1358, comprising 3280 separate food samples.
Anationwide collaborative project to revise and update the food composition data of China was organized and conducted by the Institute of Nutrition and Food Hygiene (INFH) in 1987–1990. In order to assure the quality of the analytical data provided by each of the 20 participating laboratories, an analytical quality assurance system was designed and carried out. The data obtained in this project were the basis of the new edition of the Chinese food composition tables (FCT) published in 1991.
The first edition of the Chinese FCT published by INFH in 1952 included only 12 nutrients, crude fiber and energy value for about 300 food items. In the following years, INFH updated the FCT with three editions having been published. The last printing was in 1981 and its English version was published in 1990 (1). A new edition of the food composition tables has been needed since the early 1980s, because food composition may have changed, due to the changes in crop cultivation and animal husbandry as well as food storage, transportation and marketing during the recent decades; also newer and better analytical methods are now available; and data on a number of important micronutrients (vitamins, trace elements) were missing from the previous editions. In this project, both the nutrients and food items were increased. The food items were selected based on the knowledge of frequency and amount of food consumption obtained from several national dietary surveys, and newer methods were used in the laboratory analyses. All the nutrients were analyzed by AOAC methods (2) and official Chinese methods (3, 4). The analytical data were categorized as follows: proximate composition (moisture, energy, protein, fat, carbohydrate, dietary fiber and ash); vitamins (ascorbic acid, thiamin, riboflavin, niacin, retinol, carotenes, and tocopherols); minerals (calcium, iron, magnesium, phosphorus, potassium, sodium, zinc, copper, manganese and selenium); lipids (fatty acids and cholesterol); and, amino acids. Foods items were divided into 28 groups including cereals, dried legumes, fresh and sprouted legumes, roots, tubers and stems, fresh leafy vegetables, melons, squashes and gourds, fruit-bearing vegetables, pickled, salted and preserved vegetables, fungi and algae, fruits, nuts and seeds, meats, poultry, milk and milk products, infant foods, eggs, fish, molluscs and crustaceans, fats and oils, confections and snacks, tea and beverages, alcoholic beverages, sugars and sweets, starch and its products, condiments and spices, edible Chinese medicinal herbs, and miscellaneous items. The total number of food items analyzed in this project was 1358. Food composition analysis was performed by 20 laboratories located in 15 provinces. Among them, there were 11 provincial and five municipal Institutes of Food Safety Inspection, three provincial Medical Institutes and one provincial Medical College. These provinces and municipalities covered half of the areas of China and more than two-thirds of the total Chinese population (Figure 1).
• Working Procedures
The following system was introduced to assure the quality of data generated by all the participating laboratories.
Validation of Analytical Methods
A written manual of analytical methods including food sampling and handling was prepared by INFH to ensure laboratories adhered to the same methods. All the methods were validated for accuracy and precision according to published guidelines (5, 6). Each analytical method was evaluated by three to six selected laboratories using standard reference materials (SRMs), i.e. bovine liver (National Bureau of Standard, USA), bread crumbs (a gift from Dr. Harry G. Lento, Campbell Institute for Research Technology, USA), and pig liver (China National Standard Bureau), as well as quality control materials (QCMs) prepared by the central laboratory in INFH, i.e. wheat flour, whole milk powder and carrot paste. Accuracy and precision of analytical methods between laboratories were determined daily to validate the methods. The three QCMs were used to measure the level of precision and recovery. The detectability and correlation coefficient of standard curves were used as additional indices for method validation.
Training Courses for Participating Laboratories
In order to assure that the analytical procedures would be carried out correctly and consistently by all the participating laboratories, several technical training courses were organized by INFH. The first training course was conducted in 1986 and attended by more than 50 technicians from the 20 laboratories. The methods for determination of six vitamins, amino acids, fatty acids, dietary fiber and selenium were demonstrated by instructors and then practiced by the trainees in the training laboratories. The second course was conducted in 1987. Two specialists in food analysis, Dr Gary R. Beecher and Dr Joseph T. Vanderslice from USDA, were invited to give lectures and to introduce new technologies in food nutrient analysis. During 1988 to 1989 secondary training courses were organized at the local level to train more technicians with the instructors from INFH.
Figure 1. Outline map of China showing provinces included in food composition program (shaded)
Sampling and Handling of Food Samples
It was critical to ensure that identical protocols of sampling and handling of foods for analysis were used in each participating laboratory in order to eliminate both intrinsic and extrinsic sources of variation which could affect the measured levels of nutrients in foods. The sampling scheme was designed to reflect representativeness of the food with regard to the brand or cultivar and geographic origin of the food as well as the differences in food consumption in the different areas. Foods were collected according to the priority of quantity consumed. The sample size for each collection was 1.5 kg by weight or by pieces. If the weight of each piece was over 500 g, three pieces were collected. The same variety of food was collected in three places located in an urban district and/or county area. After the food was homogenized, one-third of each homogenate was pooled into one analytical sample. The analyses for vitamins were carried out as soon as the foods were collected. A set protocol for homogenization, temperature control and other aspects of sample preparation was followed.
Interlaboratory Quality Control
Any interlaboratory variation will affect the variability of the compiled data. The values produced by each laboratory were evaluated by using quality control materials (QCMs). The maximum acceptable relative standard deviation (RSD or CV) was between 5 per cent and 10 per cent. The coefficient of correlation of the regression curve for each standard curve of an analytical method should ideally be 0.999. Recovery tests of fortified QCMs were used as an index of accuracy. Recoveries between 90–110 per cent were defined as satisfactory. The analyzed mean value was expected to fall within plus or minus one standard deviation of the certified value. For the QCMs used in this program a mean certified value and standard deviation was determined by six of the selected laboratories. In general, values within two standard deviations from the mean were acceptable. Data of QCM analyses produced by participating laboratories were evaluated using the Youden pairs method (8, 9) to test whether the value fell within the 95 per cent confidence interval. The outlier data were examined in order to identify problems. Sample exchanges, replicate analyses, calculation checks and further training of technicians in INFH were carried out to improve analytical accuracy and precision. For unknown samples the results of duplicate analyses had to be within 10 per cent of their mean. Otherwise, a third or further replication was required to re-determine the mean.
Assessment of Analytical Data Reported from Different Laboratories
The analytical data for foods from each laboratory were evaluated with respect to their reliability. Some statistical tests were used such as the Dixon and Grubbs test to reveal the outlier values. The t test was used to determine whether data were significantly different, and the F test was used to determine whether the variances of the data were different (8, 9). Validated values were compiled into the new FCT.
• Results and Discussion
SRMs are ideal tools for analytical quality control, but they are too expensive to be used throughout an entire project. Therefore, QCMs prepared by INFH were used by each laboratory. Whole milk powder and wheat flour were easy to obtain in large amounts and very homogeneous, so they were suitable for use as QCMs. On the other hand, carrot paste proved to be difficult to stabilize and was readily spoiled during transportation and storage. Results from carrot paste showed large variations, and are not included in this paper. The certified values of the QCMs were the mean values calculated from the individual values from six laboratories. Each individual value produced by a laboratory was the mean value calculated from six duplicate determinations on different days. These six selected laboratories passed the quality control test. The certified nutrient values of the QCMs for wheat flour and whole milk powder are shown in Table I.
Table I. Certified nutrient values of quality control materials per 100 g (mean±SD)
|Whole Milk Powder||Wheat flour|
|Dietary fiber (g)||-||2.4±0.2|
|Vitamin E (mg)||0.44±0.02||1.56±0.39|
- = not applicable
Bovine liver, bread crumbs and pig liver SRMs were used to validate the analytical methods. The accuracy and percent recovery of analyses were used as indices for evaluation. For example, the accuracy of the fluorometric method for selenium (Se) analysis is shown in Table II. The reported mean values were close to the certified values and the coefficients of variation (CV) of the analytical values were between 2.7 per cent and 6.3 per cent. The percent recoveries of the analysis were between 95.2–99. 1 per cent and the CVs of the results were between 2.8 per cent and 6.2 per cent. The data in Table II indicate that the fluorometric method was a valid method for Se.
Other indices for evaluating analytical methods were also used and the results are shown in Table III. Using the data collected from the selected laboratories, all the methods were evaluated. The recoveries of these methods ranged from 87 per cent to 110 per cent, most of them being in the range 90 per cent to 110 per cent. The analytical precision of the methods shows that the CV of repeatability within each laboratory was around 2–7 per cent and the CV of reproducibility between each laboratory was larger (Table III). The CV for proximate analyses (not shown in Table III) was between 1 per cent to 8 per cent, but for vitamin analyses there were larger variations. In general, the methods for vitamin determination had somewhat lower precision than mineral and proximate analyses. The above results showed that all the methods were satisfactory. In order to monitor analytical performance, the data for nutrient analyses of QCMs were collected and evaluated using three statistical tests. A simple method was the control chart test (10). All the QCMs data from each laboratory were plotted on the control chart. The certified value (X) of a given nutrient was assigned as the central line (CL), the mean value plus or minus one standard deviation (S) as the upper and lower auxiliary lines (XS), respectively, the mean 2S as the upper and lower warning limit line, and the mean 3S as the control or confidence limits. Because the QCMs are biological materials and are unstable, their composition could change with time and be affected by factors such as oxidation, temperature and light etc. So we preferred to use the mean value 3S as the largest acceptable limit. The percentages of acceptable values from the participating laboratories are shown in Table IV and Figures 2 and 3.
Table II. Accuracy of the fluorometric method for determination of selenium
|Standard reference material||Certified value||Reported mean value||Reported recovery|
|Bovine Liver 1577a||1.1±0.1||1.04±0.03(11)||2.7|
Numbers in parentheses are the total number of determinations
Table III. Indices and results for methods validation
|Method||Nutrientananlysed||Recovery||Repeatability||Reproducibility||Linearity of std curve||Limit of detection|
|Atomic absorption spectometry||Ca||93.7–108.3||3.8–5.1||2.0–5.7||1.1–5.7||0.9996||0.1μg/ml|
|Paper chromatography||Total carotenes||88.8–102.9||5.2||1.7||5.9||0.9996||0.1μg|
NDF = Neutral detergent fiber
- = Not determined
NA = Not applicable
Table IV. Percentages of acceptable values from participating laboratories
|Nutrient||Wheat flour||Whole milk|
|No. of labs||No. of labs accepted||Acceptability|
|No. of labs||No. of labs accepted||Acceptability|
NDF = Neutral detergent fibre
- = Not applicable
Some laboratories failed to submit the results to INFH in time and their data were not included in Table IV. According to the results in Table IV, 87–100 per cent of the laboratories passed the quality control tests, except that around half of the laboratories failed in the determination of thiamin and calcium. Most of the calcium values of wheat flour were much higher than the certified values. The errors came from technical mistakes such as not adding the 8-hydroxyquinoline to eliminate interference from reagents. Some thiamin values of wheat flour were higher and some were lower than the central line. Problems included low recovery after column filtration or interfering substances from reagents. Most of the niacin and protein values of both wheat flour and milk powder fell within UWL and LWL (X2S) (shown in Figures 1 and 2). The over-range data were questioned and the problems identified by means of replicating the analyses, making new standard curves and checking the calculations. To calculate the representative value for each analyte in each food item, two standard deviations from the mean value after deleting the suspect data were used to eliminate the values outside the range limits and then the mean value was recalculated. This mean value was used for the food composition table. Some results considered as unreasonable were checked for the causes. In some cases, re-analysis of foods was carried out through exchanges with other laboratories or IFNH. Some unreasonable data which could not be validated were eliminated during data compilation. In practice, some values were difficult to judge based on the current knowledge of food and nutrition, and were, therefore, retained in the FCT.
Figures 2 and 3. Examples of a quality control chart for two QCMs with different mean values of protein obtained from collaborative laboratories
Quality control is costly and time consuming, but it is essential. We have conducted an efficient analytical quality assurance system in a nationwide project of food composition analysis of 1358 food items, and involving 20 collaborating laboratories. According to our experience, the critical parts in this analytical quality assurance system were the validation of analytical methods, the availability and use of reference materials and the training of the technicians. Large variations existed in the conditions of the collaborating laboratories as well as in the technical background of the technicians. There were some inadequacies in this approach, for example, analytical methods for minerals were not included in training courses, except for selenium, and a few technicians were not familiar with the LC and GLC techniques. The question of how to ensure the comparability of the Chinese food composition data with those of other nations is still an unresolved problem.
This project was supported by National Science Foundation of China and Ministry of Public Health and Jia Li Bao company.
(1) Ershow, A.G., & Wang, Chen, K. (1990) J. Food Comp. Anal. 3, 191– 442
(2) Official Methods of Analysis (1984) 14th Ed., AOAC, Arlington. VA, secs 14.002–14.004, 31.005–31.008, 7.009, 43.275– 43.277, 24.037–24.040, 7.093– 7.103, 43.024–43.038, 43.069– 43.081
(3) People's Republic of China Standard GB 12388-12399-90 (1990) Methods for Determination of Nutrient Composition in Foods (in Chinese), Chinese Standard Publishing House, Beijing
(4) Institute of Nutrition and Food Hygiene (1990) Methods of Food Analysis, 3rd Ed. (in Chinese), People's Medical Publishing House, Beijing
(5) Uriano, G.A., & Cali, J.P. (1977) in Role of Reference Materials and Reference Methods in the Measurement Process, J.R. DeVoe (Ed), ACS Symposium Series 63, American Chemical Society, Washington, DC, Chap. 4
(6) Holden J.M., Schubert A., Wolf, W.R., & Beecher, G.R. (1987) Food Nutr. Bull., Suppl. 12, 177– 193
(7) People's Republic of China Standard G.B. 6379–86 (1986) Precision of Test Methods for Determination of Repeatability and Reproducibility for a Standard Test Method by Interlaboratory Tests (in Chinese), Chinese Standard Publishing House, Beijing
(8) Pan, X.R. (1989) Assurance and Evaluation of the Accuracy of Chemical Analysis (in Chinese), Chinese Measurement Publishing House, Beijing
(9) Gerrit, K., & Frans, W.P. (1981) Quality Control in Analytical Chemistry, John Wiley & Sons Inc., New York, NY
Kent K. Stewart, Margaret R. Stewart
Virginia Polytechnic Institute & State University, Blacksburg, VA, 24061-0308, USA
Scientific journals are a primary vehicle for the transmission of original food composition data and critical reviews of food composition data to the scientific community. Publication of composition data in a scientific journal implies that the data are accurate, precise, and meaningful. To publish data meeting these attributes it is necessary to establish criteria for data quality control. Quality control is achieved by critical evaluation of all aspects of a scientific manuscript by expert reviewers. The key attribute of a good quality control in a manuscript is adequate documentation. In a good publication, those items that should be documented include the purpose of the study; description of the sampling plan for selection of the food items to be assayed; descriptions of the food items; descriptions of the sample preparation, homogenization, and storage; description of the analyte extraction; descriptions of the identification and measurement of the analyte; and description of the analytical quality control measures used to validate the data sets. Reviewers also evaluate the quantitative data including their statistical components and the discussion of how the new data relate to existing knowledge on the composition of foods.
It is almost an article of faith in the scientific community that “good” data will aid in the development of wise decisions and that “bad” data will lead to the development of unwise decisions. In cases of conflicting data, the perceptions of which are good data and which are bad data may well be as important as the actual fact of the quality of the data.
These are not just issues of academic concern to those working on food composition data. The current public concerns about the impact of diet on health will inevitably lead to the promulgation of new policies and regulations on the composition of foods. The public in many countries is concerned about the possibility of inadequate intakes of essential nutrients, problems related to inadequate or excessive intake of energy, the possibility of intakes of toxic levels of man-made chemicals such as pesticides and herbicides, and the perceived dangers of the use of biotechnology in the production and processing of the food supply. Given the current level of knowledge on the composition of foods, a great deal of new food composition data will be needed if wise policies and regulations on the issues of diet and public health are to be made.
The discussions in this paper about quality control for food composition data in scientific journals are extensions of opinions from editorials originally published in the Journal of Food Composition and Analysis (1–10). Scientific journals are a primary vehicle for the transmission of original food composition data and critical reviews of food composition data to the scientific community. Publication of composition data in a scientific journal specifically implies that the data and their attributes have been evaluated and reviewed prior to publication by independent experts in the field. The responsibilities of publishing credible, good quality composition data are spread among the authors of the manuscripts, the reviewers of the manuscripts, and the editors of the journals. It is the authors' responsibility to carry out the study properly and then to provide adequate documentation on how the study was done.
While the editor selects the reviewers and ensures that conflicts between authors and reviewers are resolved, it is the reviewers who are the key to quality control of journal articles through critical evaluation of all aspects of a scientific manuscript. Given the chemical complexity of foods and their matrices, the enormous size of the food distribution system, and the frequent technical complexity and difficulty of modern analytical assay techniques, reviewers of food composition data need special expertise as well as significant knowledge of a broad range of subjects.
The reviewers' first responsibility is to determine whether or not adequate documentation (a key aspect of food composition data quality control) was provided in the manuscript reporting the study. Without adequate documentation there can be no critical evaluation of the science, and its lack is a fundamental failure of manuscript quality control and, if not rectified, should ultimately result in rejection of the manuscript.
Once the reviewers ascertain that the documentation is present, then they should determine that the appropriate techniques were used for the acquisition of the food composition data. A review of the appropriateness of various techniques for food composition data acquisition is very complex and requires a great deal of technical discussion beyond the scope of this paper. Finally, it is the reviewer's responsibility to determine whether or not the composition data are accurate, precise, novel, and credible. These are primarily issues of quality control and are the main topic of this paper.
The editor's primary responsibility is to ensure to the readers of a journal that the data published therein are accurate, precise, and meaningful. The goal of a journal is to have the data review and publication done in an authoritative manner so that the burden of proof will be on those who challenge the published assay data of the composition of specific foods. Thus in many ways, while authors are the source of scientific knowledge, scientific journals can be viewed as “gates” for transmission of knowledge, and the reviewers and editors can be viewed as “gatekeepers”.
The primary focus of this paper is documentation needed in a food composition paper. The underlying theme is that adequate documentation is required for good quality control of published food composition data.
• Documentation Needed in Manuscripts
The documentation for a food composition data manuscript includes the sampling plan, a description of the foods and the laboratory samples, a description of the assay methodologies, the actual composition data, the quality control information for those data, and a comparison of the data presented with those in the published scientific literature.
A primary goal in the analysis of food samples is the description of the nutrient content of the foods that are encountered in the real world. This seemingly trivial and almost tautological statement is unfortunately not often followed to its logical conclusion: that the design of the analytic protocol (especially the choice of food items and the number of analyses run) should be directed towards gathering as much information as possible about the distribution of the nutrients in food in the real world. Users of food composition data need information about the average or “usual” level of that nutrient and the range of values that would likely be encountered.
A key issue is thus which food item should be assayed? Another way to put this is the question “Were the assayed items representative of the foods for which composition information is presented?” It should be intuitively obvious that representative composition data can only be obtained from the assay of representative lots. Thus the determination of which lots to assay may well be the most important of all the questions facing the analyst in the design of a food composition assay program. The goal of a good sampling plan should be to have a protocol which indicates how many lots should be sampled and when and where they should be obtained, and which provides other details on how individual food items should be selected.
Once a good sampling plan is selected and documented, then the means used for distinguishing the sources of variability in analytical data, i.e., those arising from inherent biologic variability that reflect differences in genotype, phenotype, environment, processing, etc., and those arising from analytic variability, introduced in the process of preparation and assay of the laboratory sample, should be documented. In most cases the total number of analyses to be performed is strongly influenced by economics and each laboratory sample should be assayed a minimum number of times. It might seem that a general implementation of this design strategy would be to assay each laboratory sample only once. However, in general, the need for protection against major blunders in the assay of an individual sample leads to the suggestion that each sample be assayed in duplicate.
There are a very large number of foods in international and domestic market places. Different species are used as food sources; various growing conditions are used; the processing, packaging, and storage technologies vary. Cultural differences in food recipes are common. At the same time, there is a great commonality within some foodstuffs due to the worldwide availability of some brand name items (e.g., soft drinks and fast foods). Given the complexity of the world food supply, the readers of papers on the composition of foods need to be given the information to identify the foods for which the composition data are being presented. For even with the best of analytical techniques, food composition data are no better than the description of the products or foods being analyzed. The analyst should describe the foods so that another professional in the field can identify the foods. The sources and unique descriptions of the foods should be given. Those foods which were enriched and/or fortified should be so identified. Identification of market share can be useful. The analysts should specify numbers of items collected and the number of units in a composite. The dates of food acquisition should always be given. Frequently, most of the needed description of manufactured foods can be provided by identifying the brand name and place and date of purchase. The post-selection transportation and storage of the food items should be described. Any further fractionation of samples such as trimming and draining should also be described. When the food is cooked, the cooking processes should be described. References to published cooking procedures should be given whenever possible.
Compositing and Homogenizing
Compositing is the process of preparing a single representative composite sample from several food samples. Homogenization is the process of reducing a food sample to equally distributed particles of uniform size. Compositing and homogenization are the invisible components of food assay systems. Homogenization is almost always necessary to transform the large bulk of heterogeneous foods and diets to homogenous representative material suitable for sub-sampling. The process has been described as transforming a 5 pound meat roast into an analytic sample which can be introduced by a 5 mL syringe into a chromatograph. The procedures used for compositing/homogenizing can have a significant impact on the accuracy and precision of the final results. The issues inherent in compositing and homogenizing are crucial to the production of accurate, representative, and precise food composition data. It is important that adequate documentation of these processes be given in food composition papers.
In these homogenization processes there is potential for analyte loss due to non-enzymatic oxidation, to various enzymatic actions and to other destructive reactions. Many homogenizing techniques do not yield homogenous material with mixed foods and diets. In such cases, representative sub-sampling is difficult and the precision of the results deteriorates. Much more work is needed on the techniques for validation of the appropriateness of homogenizing techniques.
Even when many samples are used to make a composite, once they have been composited, the analyst has only one test sample. Thus a primary feature of compositing is that all information on the variation between lots is lost once the individual samples are composited. There are several other issues that need resolution when doing compositing. For example, given the purposes of the assays: How many units should be used in a composite? How much information on real variance is lost when a given compositing procedure is used?
A generalized diagram of an ideal food composition assay system is shown in Figure 1. The key feature of this schema is that a food composition assay should be viewed as a whole. Each part of the assay system must fit with the rest of the assay system. Inappropriate use of any one technique can invalidate the accuracy and/or precision of the entire assay. Basically what is needed is a holistic approach to food composition assays.
Figure 1. An ideal assay system
Published Assay Methods
In most cases published assay methodologies are used for food composition assays. The authors should provide complete references to published assays. In those cases where methods manuals are used, e.g., an AOAC methods manual, the edition of the book and the assay number should be given. Some indication should be given as to how the selected methods were determined to be appropriate for the assays at hand. Most published methods are not appropriate for the assay of every food matrix. If there are known potential interferences in a given matrix, then the authors should use the method of standard additions to verify that the assay is appropriate for the food matrix. Failure to demonstrate quantitative recovery should raise serious concerns of the appropriateness of the choice of assay.
New Assay Methods
When new food assay methodologies are presented, it is important that these methods be validated for their use in obtaining food composition data. What follows is a general description of what is necessary to validate the use of a new assay methodology for a given matrix. The underlying premise is that the validation of an assay method is a process by which the assay method is demonstrated to be capable of producing the desired analytical results when used with the matrix of interest (i.e., assay methodology validations are matrix specific). Usually such a validation is done in some authoritative manner so that the burden of proof will be on those who challenge the assay method or the data from such a method. For most assay methods, the desired results include acceptable accuracy, precision, and sensitivity. The definition of acceptable accuracy, precision, and sensitivity of an assay is a function of the end use of the assay results. Different end uses will change the perception of what is acceptable. The following criteria are presented as an idealized list for assay validation. While it is understood that not every criterion will be met in every new method validation, each individual criterion should be considered when doing new assay method validations. The idealized assay validation criteria are:
the entire assay method must be consistent with the chemical properties of the analyte and the matrix
it should be demonstrated that there are no obvious matrix interferences for any stage of the assay method, or that subsequent or previous stages have eliminated the interferences
the method should give quantitative recovery of pure standards carried through total assay method
the method should give quantitative recovery of analytes to matrices of concern
the method should give acceptable results on composition of standard reference materials of matrices similar to the foods to be assayed
the method should give acceptable precision of replicate assays (five replicates) of the analyte in matrices of concern
the limit of detection (LOD) for analyte determination in matrices should be defined and acceptable
the linear range for analyte determination in matrices should be defined and acceptable
the method should give the same results as a validated (or accepted) method or there should be an adequate explanation for the observed differences
more than one analyst and more than one laboratory should be able to produce the same results with the method
the individual data set validation processes should be made an integral part of any method description.
Recent research studies in analytical chemistry have focused on the development of new instrumentation for the separation and measurement of analytes. Similar significant advances have been made in analytical biochemistry and molecular biology in the development of highly selective and sensitive probes such as immune-reagents and DNA and RNA probes. Significant improvements have been made in the use of enzymes as reagents. Almost all the advanced techniques described above require clean extracts free of interfering compounds. For the most part, it is best if the analytes are dissolved in solutions which are themselves compatible with the separation and measurement components of the assays. These advanced techniques are perfect for the assay of pure standards or the assay of mixtures of pure standards. Unfortunately, most foods do not come in tidy packages free of assay interferences. Rather, most foods are complex mixtures of multi-phase materials with extremely complex chemical matrices. Analyte levels are often low and assay interferences are common. If analysts are to properly use today's marvelous array of analytical tools for the assay of most food components, they first should isolate the analyte from the food matrix.
In an ideal extraction procedure the analyte is quantitatively removed from the food, no analyte remains with the residue, and no analyte is altered by the extraction procedure or by the inherent biochemical and chemical activities of the matrix. The extract should not contain compounds that would interfere with the separation and measurement components of the assays. For example, if chromatographic separations are used, the extract should not contain components which coelute with the analytes or those which alter the chromatographic behavior of the analytes. If immune reactions and/or enzyme measurement systems are used, components which alter those enzymatic and/or immune reactions are unacceptable. Given the complexity and variability of food matrices, quality control procedures for the extraction steps should be required parts of most assays. Certainly, the analyte isolation procedures should be carefully documented and critically reviewed before they are used or published in reputable scientific journals.
Analytical Separation of the Analyte
The current state of the art in analytical separation techniques such as gas liquid chromatography (GLC), and liquid chromatography (LC), super-critical fluid chromatography, and capillary electrophoresis is impressive. These sophisticated separation tools have the capability of separating very complex mixtures in relatively short time periods and as such each of these techniques can be very useful for the food analyst. Accompanying the power and sophistication of these techniques is their complexity. Reproducing assays using these techniques requires detailed information on the entire analytical system including the manufacturer and model of the instrumentation, the column used, the solvents or carrier gases used, the flow rates of solvents or carrier gases, and the temperatures used at the injection port, the column oven, and the detector. Adequate evaluation of any given analytical separation system requires extensive documentation of the system. Each sub-discipline in analytical separations has developed its own shorthand mode of presenting the necessary documentation for system evaluation.
Chromatography systems have several common problems including drift, difficulty in confirming peak identifications, and the difficulty of obtaining reproducible sample injections. The use of internal standards helps to reduce the problems of drift and sample injections. The use of internal standards is now considered to be almost mandatory and the lack of their use is considered to be a serious flaw in the methodology and often leads to rejection of a manuscript. Peak identification can be difficult in some food matrices and care should be taken to document the proper identification of the analyte peaks convincingly.
Analyte Identification, and Quantification
Today's analyst has an enormous array of detection techniques for analyte quantification including atomic absorption and plasma emission spectrophotometry, mass spectrometry, diode array spectrophotometry, various electrochemical detectors, fluorescence detectors, the highly selective and sensitive probes such as immune-reagents, DNA and RNA probes, various chemical detection systems, the use of enzymes as reagents and of enzyme amplifier systems, and enzyme-linkedimmunosorbant-assays. Many of these assay systems have been automated through the use of continuous flow systems, flow injection systems, and robotics. Recently, there have been significant advances in the use of hyphenated techniques such as GLC-mass spectrometry or other combinations such as enzymelinked-immunosorbant-assays (ELISA) systems using electrochemical detection automated through the use of flow injection analysis. All of these systems are powerful and the analyst has an impressive array of quantification tools to draw upon. However, as in the case of analytical separations, the sophistication is usually accompanied by significant increases in complexity. Reproducing given assays using these techniques requires detailed information on the entire analytical system. The analyst needs to provide significant detailed documentation on the quantification system. Many detection systems do not provide unique identification; verification of analyte identification is often necessary. Even the highly selective assay systems such as immune-reagents and enzymes do not give totally unique identifications and are often quite sensitive to interferences in the quantification reactions. Verification of the appropriate use of these systems is usually necessary.
Computation of Results
Unfortunately, computational errors are still one of the most common sources of errors in food composition data (or any other assays for that matter). The introduction of computerized computation systems, such as spreadsheets and black box analytical instrumentation, has not alleviated these problems. Thus it remains important for analysts to check the process by which they do their computations. Since computational errors are still so common, it is prudent for authors to document their quality control procedures for the computation of the analytical results.
Authors should report the means and standard deviation of the composition values and the number of lots assayed. Replicate assays on one test sample or one composite yield a single value and as such are usually not sufficient for journal publication. The significant digit convention (reporting of only all digits known with certainty and the first digit of uncertainty) should be used in reporting all data. Most food assays yield results with no more than three significant digits.
The moisture contents of individual foods are highly variable and thus most composition data should be presented on a dry weight basis. The composition data for beverages are obvious exceptions. Many believe that data presented on a dry weight basis should be accompanied by a moisture value to enable calculation to “as consumed” basis.
Data Set Validation
It is our observation as editors, that mistakes in composition data are a relatively frequent occurrence. Their frequency should be substantially reduced. The challenges in the production of good data are that while there are a very large number of useful assays, their implementation is often complex and mistakes are relatively easy to make. Even experts can get incorrect results and generate incorrect data when using “good” methods. Therefore, to produce good data that are credible, the food composition community should develop protocols for data set validation. Given the complexity of the problems of assay and the wide variety of methodologies which are available today, we believe that the validation of individual data sets is necessary and that all food composition data sets should be individually validated. The concept that each individual data set be validated specifically implies that some type of quality control sample was assayed along with the samples that were assayed. Furthermore, it also implies that the data set results underwent an internal quality assurance check prior to the acceptance of the results. There are many ways to validate data sets including the use of common sense-consistency observation, standard laboratories, standard instruments, certified analysts, certified algorithms, standard reference materials, internal standards, audit trails, and in-house reference samples, (i.e. pool samples). The choice of the data validation procedure depends upon the laboratory, the food samples, and the component being measured. The addition of the concept of validated individual data sets will be of significant help in efforts to provide “good” food composition data that are also perceived to be “good data”.
Comparison of the New Composition Data with Existing Information on the Composition of Foods
One aspect of almost all data quality control operations is a comparison of the new data with the existing body of knowledge. There are very few totally unique food composition data and reviewers will normally evaluate a new set of data by comparing it to the existing knowledge on food composition. Extreme departures from existing knowledge are usually rejected by reviewers unless significant justifications are presented for the acceptance of the new data. Authors are well advised to make such comparisons within the manuscript and provide a rationale for those data which appear to conflict with previously published data.
• Future Actions
While significant improvements in the quality control of food composition data published in journals have been accomplished in the recent years, a great deal of work still needs to be done. A comment by Jorhem and Sundström in a recent paper (11) made the point clearly:
During the last decade the application of analytical quality control measures has gradually been intensified. However, since analytical quality control activities are not yet in general use or standardized, it is often still difficult to compare results from different studies.
Significant efforts by journal editors, reviewers and authors are needed if we are to improve the comparability of data between studies.
More composition data need to be published in refereed journals. The current practice of directly publishing the results of food composition studies in databases rather than referred journals means that the documentation behind those new composition data are usually not placed in the public domain. Thus the end users can not evaluate the appropriateness of the analytical quality control used in those studies. The failure to publish composition data in refereed journals prior to placement in a database is a worst case scenario. The data are available but the user has no idea of their quality. Ignorance is not bliss in such cases.
More analysts need to incorporate more analytical control into their assays and to better document those quality control procedures. These actions can be accomplished by both the editors and reviewers having an absolute requirement for documentation of good quality control procedures in all manuscripts accepted for publication.
Currently, almost all textbooks and courses on analytical chemistry, analytical biochemistry, food analysis and nutritional biochemistry contain little, if any, discussion of or instruction in assay quality control. (An exception is the recent book by Greenfield and Southgate (12)). This is a fundamental failure in our training of future analysts and it should be corrected. We strongly advocate that all analytical courses and text books in these areas contain a thoughtful section on the basics of assay quality control.
Adoption of these actions will have several benefits. Authors will increase the documentation of quality control procedures already in use in their laboratories. Authors will increase the use of acceptable assay quality control procedures in their studies. The existence of published papers with appropriate assay quality control will be useful as good examples to those in the field who wish to improve the quality of their own composition studies. Finally, the existence of papers with good quality control procedures will permit the users of food composition data to better evaluate the appropriateness of each food composition data set for the purpose at hand.
(1) Stewart, K. K. (1987) J. Food Comp. Anal. 1, 1
(2) Stewart, K. K. (1988) J. Food Comp. Anal. 1, 291–292
(3) Stewart, K. K. (1989) J. Food Comp. Anal. 2, 91–92
(4) Stewart, K. K. (1990) J. Food Comp. Anal. 3, 103–104
(5) Stewart, K. K. (1992) J. Food Comp. Anal. 5, 1
(6) Stewart, K. K. (1992) J. Food Comp. Anal. 5, 99
(7) Stewart, K. K. (1992) J. Food Comp. Anal. 5, 183
(8) Rand, W.M. (1992) J. Food Comp. Anal. 5, 267
(9) Stewart, K. K. (1993) J. Food Comp. Anal. 6, 105–106
(10) Stewart, K. K. (1993) J. Food Comp. Anal. 6, 201–202
(11) Jorhem L., & Sundström, B. (1993) J. Food Comp. Anal. 6, 223–241
(35) Greenfield, H., & Southgate, D.A.T. (1992) Food Composition Data: Production, Management and Use, Elsevier Applied Science, London, pp. 127–138
• Additional Reading
Beecher, G.R., & Mathews, R.H. (1990) in Present Knowledge in Nutrition, 6th Ed., M.L. Brown (Eds.), ILSI-Nutrition Foundation, Washington, DC, pp. 430–443
IUPAC (1978) Compendium of Analytical Nomenclature, H.M.N.H. Irving, H. Freiser, & T.S. West (Eds.), Pergamon Press, Oxford.
Klensin, J.C., Feskanich, D., Lin, V., Truswell, A.S., & Southgate, D.A.T. (1989) Identification of Food Components for INFOODS Data Interchange, UNU Press, Tokyo
Official Methods of Analysis (1995) 16th Ed., AOAC INTERNATIONAL, Arlington, VA
Rand, W.M., Pennington, J.A.T., Murphy, S.P., & Klensin, J.C. (1991) Compiling Data for Food Composition Data Bases, UNU Press, Tokyo
Rand, W.M., Windham, C.T., Wyse, B.W., & Young, V.T. (1987) Food Composition Data: A User's Perspective, UNU Press, Tokyo
Stewart, K. K. (Ed.) (1980) Nutrient Analysis of Foods — The State of the Art for Routine Analysis, AOAC, Washington, DC
Stewart, K.K., & Whitaker, J.R. (Eds.) (1984) Modern Methods of Food Analysis, AVI Publ. Co., Westport, CT
Stewart, K.K. (1985) in Methods of Vitamin Assay, 4th Ed., J. Augustin, B. Klein, D.R. Becker, P.B. Venugopal, P.B. (Eds.), Wiley, NY, pp. 1–15
Wernimont, G.T. (1985) Use of Statistics to Develop and Evaluate Analytical Methods, W. Spendley (Ed.), AOAC, Arlington, VA
Wolf, W.R. (Ed.) (1985) Biological Reference Materials, Wiley, NY
Garfield, F.M. (Ed.) (1980) Optimizing Chemical Laboratory Performance Through the Application of Quality Assurance Principles, Proceedings of a Symposium, AOAC, Arlington, VA
Garfield, F.M. (1991) Quality Assurance Principles for Analytical Laboratories, AOAC, Arlington, VA
Taylor, J.K. (1987) Quality Assurance of Chemical Measurements, Lewis Publ., Chelsea, MI
Modern Assay Techniques
Becker, J.M., Caldwell, G.A., & Zachgo, E.A. (1990) Biotechnology, A Laboratory Course, Academic Press, San Diego
Boehringer Mannheim, GmbH (1987) Methods of Biochemical Analysis and Food Analysis, Mannheim, Germany
Borman, S.A. (Ed.) (1982) Instrumentation in Analytical Chemistry, Vol. 2, ACS, Washington, DC
Harlow, E., & Lane, D. (1988) Antibodies: A Laboratory Manual, Cold Spring Harbor Laboratory, New York, NY
Strobel, H.A., & Heineman, W.R. (1989) Chemical Instrumentation: A Systematic Approach, 3rd Ed., Wiley, New York, NY