The quality of sampling and analytical data is a major determinant of database quality. Sampling foods for inclusion in a compositional database is one of the more demanding and difficult aspects of database preparation and often requires the compilers to make intuitive judgements and compromises. This chapter reviews the objectives of sampling and discusses the various aspects for consideration in making these judgements.
Where the necessary information on the composition of a food is not available (as is often the case in developing countries) or is inadequate (e.g. it is no longer applicable to the current food supply or the analytical values need to be measured using more recent methods), then sampling and analytical protocols need to be devised.
Ideally these should be developed in conjunction with each other because the requirements of the analysts will determine the amounts of foods necessary for the analyses and how the foods should be stored and, if necessary, preserved.
Users of compositional databases require representative values for the composition of the foods consumed by the population for whom the database is being prepared.
The primary objectives in sampling, therefore, are to collect food samples that are representative and then to ensure that changes in composition do not take place between collection and analysis.
All foods are biological materials and exhibit natural variations in composition. A secondary objective may be to document this variability as it relates to factors such as season, geography, cultivar and husbandry. Such variations are to be expected and should not be confused with variations associated with the analytical conditions. The combined protocols – that is, for sampling and analysis – should also ensure that the representative attributes are maintained in the portions taken for analysis.
In the context of the following account, the term sampling is used to describe the activities involved in the selection and collection of items of food defined in terms of number, weight and nature of the material to be analysed. Much of the formal terminology developed for use in sampling was designed for use in the commercial sector for the purposes of surveillance and determination of contamination (Horwitz, 1990). Some of these terms have little relevance for nutrient database work and therefore are not discussed further. Table 5.1 outlines the steps involved in the sampling process and provides definitions of the terms that will be used later in this book. Figure 5.1 illustrates the different stages in sampling and analysis, indicating where sampling errors may arise as distinct from analytical errors.
Because of the variability and heterogeneity of foods, all sampling is associated with some degree of error when the results are extrapolated back to the composition of the whole population of a food. Sampling can merely provide data that define the probability that the values will apply to any one isolated unit of the food.
|
Table 5.1 Definition of terms used in sampling of food for a nutritional database | ||
|
Term |
Definition |
Comments on application in food composition studies |
|
Sample |
A portion selected from a larger quantity of material |
A general term for a unit taken from the total amount (the population) of a food |
|
Sampling protocol |
A predetermined procedure for the selection, withdrawal, preservation and preparation of the sample |
Sometimes called a sampling plan |
|
Characteristic |
The property or constituent that is to be measured or noted |
Description of the food, nutrient and other analyses |
|
Homogeneity |
The extent to which a property or constituent is uniformly distributed |
Foods are usually heterogeneous or must be assumed to be so |
|
Sampling error |
The part of the total error associated with using only a fraction of the total population of food and extrapolating it to the whole population. This arises from the heterogeneity of the population |
Because of the heterogeneous nature of foods, replicate samples must always be taken when estimating the composition of the population of a food |
|
Batch |
A quantity of food that is known, or assumed, to be produced under uniform conditions |
Batch numbers should always be noted when sampling foods |
|
Unit |
Each of the discrete, identifiable units of food that are suitable for removal from the population as samples and that can be individually described, analysed or combined |
These units form the basis of most food analysis work (e.g. an apple, a bunch of bananas, a can of beans, a prepared dish) |
Figure 5.1 The relationship of the operation
involved in sampling and analysis
The lower A of the sampling operations continues with the upper A of the A of
the analytical operations
Source: IUPAC recommendation from Horwitz, 1990.
The selection of a representative sample and the combined protocols for sampling and analysis must be based on a clear understanding of the nature of the foods and the population of food being studied (i.e. all the individual units of the food). A database will be used for a considerable period of time and the values derived from the combined protocols will be used as if they were representative, in both space and time, over the lifetime of the database (and often for much longer). The design of the protocols therefore represents a monumental task and one in which it may be necessary to accept compromises. It is essential that such compromises are based on knowledge of the food in question.
The principal sources of food samples are summarized in Table 5.2. These groupings correspond to the levels at which databases are used.
Compositional data obtained from analyses of bulk commodities have wide-ranging uses. They are commonly used in commerce or for surveillance of imports for contamination with agrochemicals or the misuse of growth stimulants. These data also provide the basis for calculating the nutrient values in food disappearance statistics and sometimes in household and industrial recipes. Standard sampling procedures have been defined for many commodities and these should be followed: International Organization for Standardization (ISO, 2003); Official Methods of the Association of Analytical Communities (AOAC International, 2002, 2003); Codex Alimentarius (FAO, 1994; FAO/WHO, 2003). Care should be taken to ensure that samples are truly representative of the bulk commodity. Several samples may need to be taken from separate sacks, cases, packages or carcasses, and at several points in a silo or container. Random sampling is preferable to the collection of readily accessible units. Collectors should take packages from several randomly identified cases or packages, for example. This level of sampling presents logistical problems that are best overcome by taking samples during the loading or unloading of a consignment. Special probes or triers are required (Horwitz et al., 1978) for sampling finely particulate foods (e.g. sugar, grain), fluids (e.g. milk) or solids (e.g. cheese).
Nutrient analyses at this level are often limited to major components, but generally involve many analysed samples (sometimes in the hundreds), and therefore result in very high-quality values.
|
Table 5.2 Major sources of food samples for analysis for a food composition database | ||
|
Source level |
Examples |
Level of use of compositional data |
|
Bulk commodities Wholesale commodities and foods |
Meat carcasses, bulk consignments of grain, fruit, vegetables, wine, edible fats Meat carcasses, prime cuts, bulk packs of foods, often for institutional use |
Used mainly to assess nutritional value of food supplies and for food disappearance statistics. Also useful for intake assessment |
|
Retail foods |
Foods as sold to the consumer, e.g. meat cuts, vegetables, fruits, wine, processed foods |
Used mainly to assess household and individual food and nutrient intake. Also useful for food supply statistics |
|
Field, garden or wild foods |
Foods grown or gathered, hunted animals |
|
|
Foods as consumed |
Foods at the level of |
Used to assess individual |
|
|
consumption, e.g. cooked dishes (single or multiple ingredients), street foods |
food and nutrient intake |
Sampling of wholesale foods generally follows the principal approaches used with bulk commodities. Randomization of sampling is essential.
These foods constitute the majority of foods included in food composition databases in industrialized countries. For primary products such as meats, fruits or vegetables, the major concern of the sampling protocol is to ensure that the complete range of sales outlets is represented. The primary sample should be made up proportionately of the volumes of food passing through the different outlets. The potential for regional variation also needs to be covered in the design of the sampling protocols.
In non-industrialized countries where food distribution systems may be less developed, regional considerations assume greater importance and variations in composition from one rural market to another may be substantial. Regional stratification (see below) of the sampling may be considered a more useful approach in view of the regional variation in the composition of produce. In many cases presenting data that are representative of a very diverse population may not be acceptable.
Proprietary foods constitute an important range of foods in many countries and their composition should be included in the database. Where a database is prepared by government personnel there is often reluctance to include brand names. In practice, for many proprietary foods, the brand name is essential for identification. In some countries, the range of branded items of a food is very numerous, and covering all the different brands increases the analytical workload. Compositional data supplied by the manufacturer may be acceptable provided that they meet the criteria set for analytical quality, and that the manufacturers can assure the compilers that the samples analysed were representative of products as sold retail. Problems can arise using this approach because many proprietary foods are reformulated at frequent intervals and database values rapidly become out of date. Many compilers prefer to restrict this type of database entry to foods that are stable and well-established. In some cases, pooling the different brands according to market share is considered appropriate.
When collecting samples, care must be taken to ensure that the full range of retail outlets is properly represented. When available, retail sales statistics can be useful. In many cases proprietary products are produced under such strict quality control that limited sampling is satisfactory.
These sources of food are often ignored in industrialized countries, but in many countries food produced by the family constitutes an important component of the diet and should therefore be considered by database compilers. These foods tend to be much more variable – the composition of plant foods is especially dependent on the soils and fertilizer treatments. Such factors therefore need to be taken into account in the design of sampling protocols. Most field or garden produce is eaten seasonally as fresh and then preserved according to traditional methods that can differ substantially from commercial practice.
Many communities, especially those living a “hunter-gatherer” or semi-nomadic style of life, consume substantial quantities of wild plant and animal foods. Such foods account for a significant proportion of daily consumption, and their inclusion in a database can be very useful for those studying the nutrition of such groups. Collecting samples of these foods can pose particular problems. They may be difficult to identify properly and also tend to be variable in composition and maturity (Brand-Miller, James and Maggiore, 1993). Frequently random sampling is virtually impossible and “convenience” sampling, as the opportunity arises, is the only option. Provided that this approach is documented in the database, it is acceptable. Documentation will alert users to the limitations of the data and minimize the possibility of them being used inappropriately.
Many dietary intake studies, especially epidemiological investigations, require the measurement of food and nutrient consumption at the individual level, i.e. foods as directly consumed. These foods – “on the plate”, as they are often called – comprise cooked foods of all kinds, including complex mixed dishes. The latter are often prepared using a variety of recipes and cooking methods, which poses difficulties in selecting representative samples. Simulation of the cooking procedures in the laboratory or dedicated kitchens is often used to prepare samples for analysis. This approach is generally satisfactory, although in the domestic context being simulated, food preparation is not always carried out in a controlled fashion and decisions on when cooking is complete are a matter of individual preference and judgement. Nevertheless, laboratory-based sample preparation allows for detailed documentation of all the relevant conditions (cooking temperature, duration, end-point internal temperature, etc.). Collection of cooked dishes from a randomly selected range of households would provide more representativeness, and is sometimes, therefore, the preferred approach (Greenfield, 1990b). However, this approach also presents its own logistical problems.
Samples of institutionally prepared foods from, for example, hospitals, industrial and public canteens and educational establishments, are more easily obtained. Samples from fast food establishments and of “take-away” foods are also easier to collect. The difficulties in sampling, the enormous range of possible variation among cooked foods and financial constraints have frequently led compilers to use calculations from recipes to estimate the composition of cooked dishes.
Foods are inherently variable in composition, and the approach to sampling and the design of the sampling and analytical protocols need to take account of this factor.
In a single country there may be a wide diversity of soil and climatic conditions, resulting in significant variance in food composition. Variations in food marketing and food preparation within different parts of a country – or among countries in the case of a multicountry database
– may also produce notable variance. For these reasons, geographically-specific data may be presented in the database as a supplement to nationwide and/or regionwide averages. In other countries, the variations may be of similar magnitude to those due to other causes, in which case the national sample could be weighted according to the proportions of the population living in the regions or the proportions of the total consumption of the foods.
Seasonal variations in nutrient composition need to be accommodated in the combined protocols. Plant foods are especially prone to variation, particularly in their water, carbohydrate and vitamin content. Fish also show seasonal variations, especially in fat content, and milk and milk products exhibit variations in vitamin content primarily due to seasonal differences in feeding patterns. The collection of samples needs to be organized, in terms of timing and frequency, to reflect these variations. In some cases, seasonal data need to be given separately in the database. The analytical measurements of the seasonal samples can often be restricted to those nutrients showing variation.
The states of maturity of plants and animal foods cause variation in composition: in the concentrations of sugars, organic acids and vitamins in many plants, and of fats and some minerals in animal foods. Some of these variations are a consequence of seasonal effects.
These may be a significant source of variation for some nutrients and the combined protocols will need to provide for this variation. It is desirable to document this cultivar or breed variation within the database. Some research organizations sample specifically to capture cultivar and breed differences. The significance of the differences attributable to cultivar or breed can only be ascertained by controlling for other factors that can influence variation, and by sampling and analysing individually, not in composite, a large number of samples.
The main sampling methods used for nutrient composition databases are summarized in Table 5.3.
Random samples are collected in such a way as to ensure that every item in the population of the food being sampled has an equal chance of being collected and incorporated into the sample to be analysed. This is difficult to achieve in practice because it is difficult to visualize the entire population of, say, all the cabbages in a country let alone ensure that each one has an equal chance of being selected. It is more usual to set up a stratification (see below) of the food population.
In this method the population of food is classified into strata, taking into account the most important causes of variation.
Stratification by geographical area may be useful even where there are no known significant regional variations (Smits et al., 1998). Stratification according to the distribution of the consuming population, among rural and urban sources, or by type of retail outlet, are other useful examples (Torelm, 1997). The sampling of branded foods can be stratified according to manufacturing plant. Where different brands of the same food are not expected to show significant variation, the sample can be weighted according to market share.
Where this information is not available, extrapolating from similar foods or an intuitive assessment will be required.
Selective sampling is widely used in experimental studies of plant and animal husbandry and in home economics. The resultant data are valuable guides for the design of sampling protocols but since they are not generally representative of the foods available, they require careful documentation when included in the database.
|
Table5.3 Main sampling methods used in nutrient composition studies | ||
|
Method |
Definition and characteristics |
Notes on application |
|
Random sampling |
Samples are taken in a way that ensures that any one unit has an equal chance of being included | The theoretical ideal but rarely practicable when sampling foods for nutritional databases |
|
Stratified sampling |
Units of sampling are taken from defined strata (subparts) of the parent population. Within each stratum the samples are taken randomly |
Often the most suitable method for use in database work. Strata may be regional, seasonal, retail sale point, etc., as defined by knowledge of the food being studied |
|
Selective sampling |
Samples are taken according to a sampling plan that excludes material with certain characteristics or selects only those with defined characteristics |
Most commonly used in the analysis of contaminants. Can be used, with caution, for database work |
|
Convenience sampling |
Samples are taken on the basis of accessibility, expediency, cost or other reason not directly concerned with sampling parameters |
Rarely suitable for database work but may be the only practicable way to sample wild or uncultivated foods or composite dishes from households |
Where, however, it is clear that the methods of husbandry and the storage of the foods are comparable with current practice for the production of food the data may be useful.
This method is often legitimately used in the analysis of contamination, where the objective may be to identify maximal exposure to contaminants. The distribution of contaminants in foods is frequently highly skewed. Random sampling will therefore often include samples in which the concentration of the contaminant is below the level of detection. This is the primary reason why data on the levels of contaminants are often held separately from representative nutrient data in the database.
Samples of foods prepared in a laboratory can be regarded as selective samples. Laboratory preparation may be the only practicable way to obtain data on the composition of certain foods and therefore the derived data may be useful in databases. Generally, however, samples collected from cooks working in domestic or industrial kitchens are to be preferred as they can be regarded as more representative of foods generally available for consumption.
The collection of samples from conveniently accessible points is a very common, and possibly misleading, practice in compositional studies. This method may be acceptable as a preliminary exercise to obtain estimates of variation in composition, but in general data obtained using this method should be regarded as low quality.
Convenience sampling may be the only option in the case of wild or uncultivated foods; provided the sources of the samples are fully documented the values can be used in a database.
In all methods the compositional data obtained can only be an estimate of the composition of the food and are subject to limitations imposed by the variation in the composition of foods.
The objective is to prepare well-documented protocols that provide the basis for those involved in collecting and handling the samples, from their collection in the field through to the laboratory. This process serves to ensure that the data generated meet the objectives of the compilers and the requirements of the database users.
In some countries the database compilers control the sampling and analytical work and are responsible, in collaboration with the analysts, for preparing the written combined protocols. In most countries, however, the sampling and analytical work will be carried out under contract(s); here the compilers' input may be restricted to establishing the broad outlines of the work required. These initial specifications should set out the principles of the database requirements with regard to representativeness and the analytical data quality standards that the reports from the contractors must meet.
Detailed combined protocols are then prepared by the contractors in consultation with the compilers. The sampling may be contracted to local sampling groups (e.g. where the database covers a large country or region); again, it is essential that the subcontractors are fully conversant with the sampling objectives.
Where the analytical work is subcontracted, either for all or selected nutrients, the subcontractors must be aware of the preferred analytical methods and have in place the proper data quality assurance schemes. Where the subcontractors wish to use other methods with which they may be more familiar or experienced, they should provide evidence that these are compatible with the preferred methods.
It is of paramount importance that units and modes of presentation of the results are predefined and written into the contracts. For example, laboratories may use ppm (parts per million, mg/kg) or ppb (parts per billion, microgram/kg) to express the results of trace metal analysis, and others use IU (International Units) for some vitamins. Fatty acids should always be reported as units of mass (mg/100 g) and may additionally be reported as a percentage of total fatty acids. It should also be predetermined whether results should be reported on a dry weight basis or wet weight basis. In either case, water content values must be reported.
Some form of stratified sampling will generally be the method of choice. Even where there is no evidence of regional differences in composition, a stratification based on collecting samples on a regional basis of the population of the food consumed will be included in the sampling. For pragmatic reasons it may be necessary to restrict the extent of sampling and most compilations devote the most extensive sampling to the most important “core foods” or “key foods” and those foods that are major sources of particular nutrients, (Chug-Ahuja et al., 1993; Schubert, et al., 1987; Haytowitz, Pehrsson and Holden, 2002; Pennington and Hernandez, 2002; Perry et al., 2000) where, for example, there are public health concerns. Foods that are relatively minor components of the diet are usually less emphasized in the protocols. Many proprietary or branded foods, which are produced in a few factories, can clearly be sampled more simply than, say, meat products which are often “core foods” and which can show great variability, necessitating much more detailed and extensive protocols. Vegetables and fruits, which show seasonal variations in composition, will need to have a seasonal stratification. Each group of foods must be considered on a case by case basis. The logistics of the analytical work often make it desirable to sample foods on a food group basis because sample handling and the actual methods used will be common across the group.
|
Table 5.4 Summary of stages in sampling and preparation of samples in food composition studies | ||
| Terms |
Description |
Main use in food composition studies |
Primary sample |
The collection of one or more units initially taken from the total population of the food |
The usual starting point in compositional studies. The ideal is the collection of several replicates that are treated separately. Primary samples are often mixed to form composites |
Reduced sample |
A representative part of the primary sample obtained by a division or reduction process |
Frequently used to reduce the primary sample to a more manageable weight |
Composite sample |
Mixtures formed by combining primary samples |
Frequently used in food composition studies. Composites may be samples of the same food or combinations of different brands or cultivars |
Laboratory sample |
The sample sent to or received by the laboratory |
The primary sample (or a reduced sample) often requires further handling in the laboratory (e.g. thawing, cooking, separation of inedible matter). The edible portion may need further reduction or mixing |
|
Analytical sample |
The portion prepared from the laboratory sample from which the portions for analysis are taken |
This is usually the form in which the food samples are prepared for analysis |
|
Analytical portion |
The quantity of food of the proper weight for each analytical measurement |
The analysis of duplicate analytical portions is the minimum acceptable; several replications are preferable |
During the course of describing the sampling process a number of stages are met, each of which uses the terms “sample”. Table 5.4 sets out a summary of the stages and some suggested definitions which may be used to make it clear which type of sample is meant at the different points in sampling and analysis.
Size. The total amount of food required for the different analyses forms the basis for deciding the size of individual samples. In practice, because foods are heterogeneous, taking small portions at the primary sampling stage can lead to error. For many foods the individual items for collection are readily identifiable; in other cases they will need to be defined. In practice, 100–500 g represents a convenient guide to the size of a primary sample, with preference being given to the upper end of this range. Some food items, for example certain cuts of meat, are much larger than this and cannot easily be reduced to a smaller but still representative unit; for the purpose of the primary sample these should be used in their entirety.
Number. In order to calculate the number of samples needed, information is first required on the variability of the composition of the food (Proctor and Muellenet, 1998). This also assumes that the concentration of the nutrient is uniformly distributed in the food, which is a reasonable assumption for many nutrients but often not true for trace elements.
In practice, the required information is often incomplete and one has to proceed intuitively. Furthermore, many nutrients, especially vitamins, show greater variability than, say, protein, so the number of samples required formally will be greater.
An example of how the calculations are performed is provided in Appendix 2.
Most sampling schemes adopt a standard of at least ten units and the United States requires data for nutrition labelling to be based on 12 units. However, strictly speaking the number depends on the variability of the nutrients being measured and thus different numbers of food samples are required for certain nutrients.
The protocols are written documents that describe the sampling process: the identity of the food, the size and weight of units to be collected, the stratification to be used and the distribution of sampling sites. Tables 5.5a–5.5d give the information that is required for preparation of the sampling protocol, commencing with the description of the primary food sample (Greenfield, 1989; McCann et al., 1988).
Table 5.5a deals with the identification of the food. The record of the collection is recorded in Table 5.5b, a detailed description of the food collected in Table 5.5c, and the handling in the laboratory in Table 5.5d.
|
Table 5.5aSuggested food sample record for food composition studies: identification | |
| Common name of food | |
| Sample code number | |
| Date of receipt in laboratory | |
| Food identification |
Examples of record |
|
Alternative names |
Other common names (in language of country of origin) and English equivalent where possible |
|
Scientific name |
Genus, species, variety |
|
Plant food |
Entire plant, or part of plant (root, stem, leaves, flower, fruit, seeds) |
|
Animal food |
Entire animal, or part (leg, head, internal organ) |
|
State of maturity |
Immature, ripe, etc. |
|
Grade |
Where appropriate |
|
Other details |
Any details that the collector thinks may be relevant |
The volume of information recommended in this documentation may seem excessive, but experience suggests that information from different stages is very critical when assessing the quality of sampling and subsequent analyses. Moreover, if the details are not recorded at the appropriate time they cannot be recovered retrospectively.
Table 5.5a sets out the information required. The first section constitutes a label that should be securely and permanently attached to the sample. The laboratory may subsequently add an acquisition number. Most of the information required is self-evident.
Table 5.5b sets out the information to be recorded during sample collection. The items recorded correspond to the sampling plan as set out in the combined protocols. This indicates the designed stratification and the method for achieving randomization within the strata. The use of random number tables is one useful approach. The protocol must also specify the procedure to be followed if the defined sample item is not available for collection. This may be the nomination of a replacement item or the need to choose an alternative sampling point.
Most of the items are self-evident. A record of the purchase price can be useful for auditing purposes and for household budget studies. A photographic record, with a measurement scale and colour standard (e.g. Pantone sheet), if available, is recommended to facilitate the identification of the sample (Burlingame et al., 1995b). If photographic records are not practicable, a simple line drawing may suffice (McCrae and Paul, 1996).
|
Table
5.5b Suggested food sample record for
food composition studies: | |
|
Common name of food |
|
|
Sample code number |
|
|
Date of receipt in laboratory |
|
|
Collection details |
Examples of record |
|
Date and time of collection |
|
|
Name of collector |
|
|
Place of origin |
If known, (village, district, province, map reference) |
|
Sampling point |
Type (field, garden, roadside stall, farm market, shop, warehouse, supermarket, take-away food bar, restaurant, household, deep sea, shoreline) |
|
Address(es) of sampling point(s) |
|
|
Conditions of cultivation |
Where known (altitude, rainfall, fertilizer treatment, irrigation, feed regime) |
|
Season |
Time of year, dry or rainy season |
|
Purchase price |
If relevant |
|
Graphical record |
Visual record with scale; line drawing may be sufficient |
|
Transport conditions |
Details, including mode and conditions of transport and storage |
|
Other details |
Any details that the collector considers relevant |
The combined protocol identifies the arrangements for transporting primary samples from the collection sites to the laboratory. The logistical aspects of handling what may be large amounts of food require careful consideration; the storage procedures, including choice of containers and modes of transport, should be specified in consultation with the analysts. These and all other aspects of the combined protocols need to be rehearsed or at least taken through a “paper exercise” with the participation of all those involved. Secure storage in inert containers, which can be heat-sealed using simple equipment, is preferable. Ideally, the samples should be cooled with crushed ice or solid CO2. If this is not possible, they should be transported to the laboratory with minimum delay. In some cases, the limitations of the sampling and transport arrangements may preclude the analysis of nutrients that are likely to be changed by metabolism (see Table 5.6 on page 80).
Where the distance to the laboratory is short, road or rail transport may be suitable but, where longer distances are involved, air transport may be the only alternative. (This will involve liaison with the airlines to ensure that the storage conditions are compatible with airline safety regulations.) In other cases considerable ingenuity may be required to suit local conditions.
|
Table 5.5c Suggested food sample record for food composition studies: description of samples collected | |
|
Common name of food |
|
|
Sample code number |
|
|
Date of receipt in laboratory |
|
|
Description |
Examples of record |
|
Food type |
Food grouping (legume, fruit juice, milk products, etc.) |
|
Local use of food |
In festivals, famine, etc. |
|
Physical dimensions |
|
|
Physical state |
Shape, form (e.g. liquid, solid, whole, divided, particle size) |
|
Process and preservation method |
Canned, smoked, sun-dried, etc. |
|
Preparation method for consumption |
Cooking method |
|
Extent of preparation |
Raw, uncooked, partially cooked, fully cooked, thawed, reheated |
|
Packing medium |
Brine, oil, syrup, water |
|
Container or wrapping |
Can, glass, paper, foil, leaves |
|
Contact surface |
Glass, type of plastic, foil |
|
Label or list of ingredients |
Retain label, estimated by inspection |
|
Batch number |
For branded foods |
|
For branded or pre-packed food |
|
|
Weight of food collected |
|
|
Number of items |
|
|
Weight of individual items |
|
|
Weight of common measure or portion |
|
|
Other details |
Any details that the recorder considers relevant (e.g. after fresh samples were collected they were vacuum sealed) |
The personal security of the samplers should also be considered, as they often carry relatively large amounts of money to pay for the samples that they are collecting; indeed, the large amounts of food they carry may also be a target for theft. Payment for samples can often be arranged by credit, thereby eliminating one of these concerns.
Most of the information suggested in Table 5.5c may be added once the samples have arrived
|
Table 5.5d Suggested food sample record for food composition studies: record of handling in laboratory | |
|
Common name of food |
|
|
Sample code number |
|
|
Date of receipt in laboratory |
|
|
Handling stage |
Examples of record |
|
Weight and nature of inedible matter |
Prior to further preparation (e.g. head and feet of poultry, outer wilted leaves) |
|
Weight and nature of edible matter |
Prior to further preparation (e.g. remainder of poultry carcass) |
|
Method of preparation |
Preparation of raw sample or cooking method, type, time, temperature and end-point temperature of foodstuff |
|
Weight before cooking |
|
|
Ingredients added, if any |
|
|
Weight after cooking |
|
|
Weight and nature of edible portion of prepared food |
|
|
Weight and nature of inedible material |
Bone, gristle, etc. |
|
Method of mixing and reduction |
Grinding, homogenizing in blender (type of blades) |
|
Details of preparation of composite sample, if applicable |
Simple mixing of equal weights or weighting of primary samples from the designated strata |
|
Type of storage |
Addition of preservatives, temperature of storage, etc. |
|
Method used to take analytical samples |
|
|
Storage of analytical samples or further processing |
|
|
Name and signature of person completing record |
|
|
Date of record |
|
|
Other details |
Any details that the collector thinks may be relevant |
at the laboratory, but the details concerning local use and preparation method may need to be added during sampling.
Labels and lists of ingredients should be retained because they provide key information that may prove useful in explaining analytical discrepancies (e.g. foods where supplementary ingredients have not been added and the labelling is incorrect, differences in formulation of branded foods given the same names).
Table 5.5d provides a record of the early preparation of samples in the laboratory leading up to the preparation of the analytical samples. The laboratory may wish to add its own laboratory acquisition number. Laboratory record-keeping constitutes the first stage of a laboratory quality assurance programme, which will be discussed in detail in Chapters 6, 7 and 8. For this reason it is essential to preserve the linkage between the sample ID number and any laboratory acquisition number.
The primary samples will need to be unpacked and the sample compared with the information recorded in Tables 5.5a, 5.5b and 5.5c.
The protocol will specify whether the primary samples are to be analysed individually or combined in some way. Individual analysis of primary samples provides valuable information on the extent of variations in nutrient content, thus helping to define the confidence limits that can be ascribed to the mean values recorded in most databases. Individual analyses require substantial resources, however, and for many databases composite samples are analysed instead. The composite samples may comprise a simple combination of equal weights of all primary samples, or weighted amounts of primary samples from different strata or sampling points according to information on food consumption or production.
Throughout this handling stage, the principal objectives of the sampling process must remain foremost in the minds of everyone involved, namely to ensure the representativeness of the sample and to protect it from changes in composition and contamination. Table 5.6 summarizes the major effects of sample storage and preparation, the nutrients affected and the precautions to be observed.
The samples should be thawed carefully and handled as quickly as possible. Once again rehearsal of these procedures should always be carried out.
In separating the edible and inedible matter the cultural norms of the population consuming the food need to be considered. Complete documentation is essential for later use in the database.
When cutting, mincing or grinding food samples, protective measures must be taken to exclude the possibility of contamination. The procedures should be tested in advance (Wills, Balmer and Greenfield, 1980). The use of plastic or TeflonŽ coated tools may be necessary. Metal implements should not be used where iron and trace elements are to be analysed; some trace elements may be introduced by the use of stainless steel.
The physical characteristics of the sample are among the important factors to consider in preparing the samples. Lichon and James (1990) have reviewed and evaluated a range of 12 homogenization methods. One should also carry out pilot studies to check on the homogeneity produced by the chosen procedure and that fractionation of the samples has not occurred. Each food will need to be considered case by case.
The logistics of sampling preparation usually mean that it is more convenient to store the analytical samples prior to analysis. At least three sample replicates should be stored. Storage in a frozen state is usually the minimum acceptable with preference given to –40 or even –70 °C, which is current common practice. Storage at –20 or –30 °C is acceptable for fat
|
Table 5.6 Effects of sample storage and preparation on nutrient content and precautions required to minimize them |
|||
|
Effects |
Potential changes |
Nutrients affected |
Precaution |
| Drying out | Loss of water | All nutrients | Design of protocol. Keep samples in sealed containers or covered. Weigh food at start and during preparation |
|
Absorption |
Gain of water |
All nutrients, especially in low-moisture |
Design of protocol. Keep samples in sealed |
|
Microbial activity |
Degradation/autolysis |
Losses of carbohydrates, proteins. |
Storage at low temperature. Pasteurization or |
|
Oxidation |
Destruction of unsaturated Loss of vitamins |
Alterations in profile of fats |
Store at -30 °C in sealed containers under |
|
Acid |
Hydrolysis |
Losses of sucrose and higher |
Store at low temperature. Neutralize acid |
|
Alkaline |
Destruction |
Loss of thiamin |
Avoid alkaline conditions and SO2 |
|
Light |
Photodegradation |
Loss of riboflavin |
Protect from light |
|
Contamination during sampling |
From cooking vessels, soil, dust, etc. |
Increases in inorganic nutrients |
Design protocol to minimize contamination, |
|
Contamination |
Increase in |
Increase in major trace elements |
Select apparatus with care. |
|
Separation |
Separation of fats. |
Changes in composition overall, alteration in fibre content |
Avoid overvigorous mixing and thaw/freeze cycles |
|
Enzymatic and |
Changes in organic nutrients |
Losses of sugars, vitamin C, |
Store at low temperatures. Protect folates with ascorbate |
analyses. The container must be closely sealed with the minimum of headspace. When the samples are taken from storage any sublimed water above the sample must be carefully reincorporated in the mass.
Where freeze-drying is possible, storage of the freeze-dried samples in frozen or chilled conditions is satisfactory. Air-dried samples should be stored in such a way as to prevent uptake of water or contamination with insects or mites.
In producing values for a compositional database a range of analytical procedures will be performed, requiring a number of analytical portions – often over a considerable time period (unless a large number of analytical staff are available). The procedures for taking the portions and their size will usually be defined by the nature of the analytical method to be used. It is imperative that all portions taken are representative and the methods used follow procedures defined by an established quality control programme.
Where analytical portions are repeatedly taken from the stored analytical samples the risks of contamination or taking an unrepresentative portion increase. It is therefore desirable to store a number of identical analytical samples and to minimize the number of staff involved in taking portions from them.
It is impossible to specify the sampling procedures for all methods and nutrients, but some typical procedures are given as examples in Appendixes 3 and 4.
The combined protocols provide a detailed basis for estimating the resources required for the sampling and analytical work. It may be necessary to revise the protocol, either by reducing the number of samples or being selective about the range of analyses to be carried out. This will require a re-examination of the processes used to establish the priorities described in Chapters 3 and 4. Combinations of analyses or extrapolation from related samples may be necessary.
Many compilers adopt the strategy of using a simplified sampling protocol for foods that are minor components of the diet and restricting the complete sampling protocols for core foods, foods that are major sources of nutrients and foods that are of greater importance in terms of public health.
It is essential that all those involved in the sampling process are familiar with the objectives of the work and are clear about their roles. This can be done by rehearsal of the procedures if only as a “paper exercise”. This process will identify aspects that are unclear or impracticable and require modification.
|
Table 5.7 Major sources of error in sampling | ||||
|
Source |
Examples |
Precautions |
||
|
Food sample identification |
Poor labelling of samples |
Maintenance of documentation throughout sampling and analytical process |
||
|
Nature of sample |
Samples do not conform to the defined sampling protocol |
Explicit instructions in sampling protocol, training of sampling staff |
||
|
Transport and handling |
Samples contaminated, degraded or depleted during transport or storage. Loss of samples |
Protocol specifies conditions to be maintained, supervision |
||
|
Analytical sample preparation |
Incorrect mixing or homogenization |
Proper supervision in laboratory. Laboratory quality assurance systems |
||
|
Analytical sample storage |
Incorrect storage of samples |
Proper laboratory techniques and supervision |
||
Table 5.7 summarizes the major sources of error in sampling. These highlight the central importance of documentation, staff training and supervision of the various stages. The sampling stages form the first and critical phases of a fully developed quality assurance programme (see Chapters 6, 7 and 8). Unless the samples are collected and handled correctly the analytical work – however well-executed – will be wasted because the values obtained will not relate to representative samples. It is however a truism that “one cannot inspect-in quality [by supervision], it must be built in”. This depends on adequate staff training so that individuals fully understand their roles in the overall process.
