P. Scheelings and D. Buick
Dr Pieter Scheelings and Don Buick are, respectively, the Regional Director and Senior Chemist, Australian Government Analytical Laboratories, 338–340 Tapleys Hill Road, Seaton SA 5023.
A nutritional analysis program extends beyond the collection of a trolley of foods at the corner supermarket and analysing the products for a variety of nutritional components. Such a program requires an understanding of the size and complexity of the project and a careful assessment of the resources, skills, courage and discipline required to progress the task to completion (Scheelings 1987). It is as much about program planning as it is about analysis and more about data evaluation than reporting.
This review covers the process of data generation and identifies the critical elements of quality assurance for a nutrient analysis program. It is based on four years involvement in food compositional analysis for the Australian food tables program, as well as for nutritional labelling of packaged foods and nutrient stability studies in processed foods, leading to considerable experience in the associated activities of sampling, transport and storage, analytical strategies, methodology and data reliability.
Data production
Data production can be defined simply as the generation of analytical data by machine, analyst or organisation. The usefulness of the data, however, is a function of its reliability and requires correct program design, planning and execution, and thorough data evaluation.
The provision of reliable nutrient data requires the analytical laboratory to understand the objectives and limitations of the food composition program, the end use and application of the data, and the knowledge and qualifications of the commissioning client. Likewise the client should have some appreciation of the complexities of the analytical task, be able to assess the proficiency of the contract laboratory, have some knowledge of the natural variability of nutrients in foods, understand the statistical basis for sampling procedures and preferably have some knowledge of the principles of analytical quality control and assurance in chemical measurements.
Sampling
Much has been written on the importance of representative sampling and the value of considered sampling plans (Taylor 1987, Garfield 1989). Representative sampling for nutritional analysis for food composition tables poses particular difficulties.
A fundamental difference between regulatory or quality control analysis and food composition analysis is the size of the sample lot or population. In regulatory analysis, the size of the sample population is well defined. The sampling plan merely needs to provide for a representative sample of the population or consignment, taking into consideration the size or volume of the lot and the physical state and homogeneity of the product. In on-going programs such as the National Heart Foundation Food Approval Program (Haddy 1990), the sample population is less defined in terms of size, and analysis is performed on a representative test sample which describes as best as possible the product at the time of sale. In the Food Approval Program representative sampling involves the submission to the laboratory of at least six individual samples taken from different production batches. The laboratory sample is prepared by compositing and homogenising the samples submitted and test samples taken from the homogenate using appropriate sampling procedures. The analytical data so produced are normally valid for an agreed period of time (eg 12 months) after which period the product is resampled and submitted for re-analysis.
Due to the high cost of analyses and the extensive range of nutrients normally involved, data produced for food composition tables, however, demand greater durability. In these analytical programs the size of the sample population is even less definitive and the samples taken for analysis can only represent a snapshot of a changing population. It is important that this limitation is both recognised and accommodated in any sampling plan.
Critical steps in data production
The normal sequence of activities in the production of data by the laboratory is described below.
Commissioning
The client submits a request for an analytical service, the laboratory advises on its capacity to undertake the work and provides a quotation and estimated turn-around time; then a contract is signed if parties are agreeable and the laboratory is advised on sampling protocol, documentation requirements and transport.
Sampling
Samples are taken by the client or the contract laboratory in accordance with the specified sampling plans; the samples are delivered to the laboratory and checked for integrity, documentation and transport duration; sample information and analytical requirements are detailed in a sample register and a laboratory worksheet is produced and sent with the samples to the analyst or supervisor.
Analyses
The priority of the task is assessed taking into consideration client requirements, sample stability and laboratory workload. In nutritional analysis, care must be taken to ensure that sufficient time is available to complete analyses for vitamins with limited stability. Highest priority must be given to moisture and the more labile vitamins (eg vitamin C, carotenes) which are often analysed in individual samples. Then client samples are composited, where necessary, and homogenised to produce a laboratory sample. Test portions are taken from the laboratory sample for analysis. An equal number of laboratory samples are frozen and used as duplicates or retest samples in case of dispute or unusual results. The appropriate methods are selected which provide reliable data for the particular food matrix. Universal methods which are applicable to a wide range of food commodities are preferred. In addition, alternative methods such as bio-assays may be employed as check methods for some vitamins. When analyses are completed, and acceptable recoveries and duplicates are obtained (ie evidence that the analytical system is in statistical control), the data are reported on the worksheet and submitted to the supervisor.
Data checking and reporting
The supervisor confirms the validity of the results by checking calculations, examining chromatograms and other raw data. When all the results are available, the data are again scrutinised and checked against expected or reported values. Outlying data may be confirmed by re-analysis. Final data are entered on the computer, and a report of analysis prepared for the client. A report should provide sampling details where appropriate, analytical results, precision data and methodology and be signed by the supervising analyst. If required, laboratory staff should be available to interpret the data or discuss aspects of the methodology employed with the client.
Quality assurance in nutritional analysis programs
In the context of analytical services, data production is concerned with the output of quality data, ie data which are reliable, timely and cost-competitive. Most clients have a clear understanding of time and an even clearer one of cost but generally lack an appreciation of the statistical concepts of data quality.
Analysts and service laboratories are able to provide a quantitative measure of data quality by reporting the uncertainty limits of the measurements involved and describing the quality assurance parameters employed. This is formalised through the existence and use of a quality assurance plan which is an integral part of any Good Laboratory Practice (GLP) protocol. The important elements of a GLP protocol include the employment of competent and qualified personnel, a modern wellequipped laboratory environment, the use of appropriate and validated methods, the use of experienced analysts, proven proficiency in the analytical methods, routine use of appropriate analytical standards and reference materials, comprehensive sample and data recording and reporting procedures, protocol for sample chain of custody, attention to customer service, attention to staff training and development and occupational safety and health, a professionally stimulating work environment, and, sound technical and administrative management.
Quality control
Quality control can be defined as the overall system of activities whose purpose is to provide a quality service or product which meets the need of the user. Quality control in an analytical laboratory covers personnel, equipment and methodology and promotes procedures such as replicate analyses and use of spiked controls and split samples.
Quality assurance
Quality assurance describes a system of activities whose purpose is to provide assurance that the quality control tasks are being performed effectively. This is normally achieved through regular quality audits, use of blind samples, participation in proficiency studies and external laboratory assessments. The effectiveness of the quality assurance process determines the reliability of the data produced by the laboratory. The term reliability means the data produced by a laboratory are accurate (close to the correct value) and precise (able to be reproduced).
Selection of analytical methods
In selecting analytical methods the following parameters need to be given due consideration (Horwitz & others 1978): specificity, sensitivity, detectability, accuracy and precision. The specificity of a method defines the level of matrix interference with the analyte, whilst method detectability determines the limit at which an analyte can be detected with confidence and sensitivity is the change in response per unit of analyte concentration. Other criteria which influence the usefulness of a method include ease of operation, large dynamic range, speed and cost, portability and ruggedness.
In addition to using official or validated methods with known accuracy and precision, the analyst must employ a number of quality control actions to optimise data quality. These include the use of accurate and clean glassware, calibrated equipment, pure or standard chemicals and reagents, reagent blanks, recoveries, duplicate (or replicate) analyses and control samples and appropriate reference samples.
Certified biological reference materials.
Certified biological reference materials (RMs) are specially prepared well-characterised homogenous stable food (or other biological) materials containing analytes at certified levels measured by highly accurate methods (Wolf 1985). RMs are used to validate measurement systems and analytical results. Most RMs in common use in food composition work are for inorganic nutrients and the lack of RMs for organic nutrients has been recognised as a severe problem for nutrient analysis.
Statistical control
Overall the analyst must ensure that the methods are always in statistical control. Until a measurement operation has attained a state of statistical control, it cannot be regarded in any logical sense of measuring anything at all (Taylor 1987).
In effect this means that in the routine application of the method, the accuracy and the precision of the measurements as described for the method should be maintained at all times. The accuracy of the method is established by determining the difference between the true or known value of the analyte and the measured value. The precision of a method is measured by the standard deviation of a set of results from the mean derived from a number of repeated applications of the method.
A convenient laboratory tool for assessing the performance of the measuring process (method and analyst) is a control chart on which values of control or reference samples analysed over a period of time are recorded (Taylor 1987). By maintaining the control chart in real time, the analyst is able to detect changes in the performance of the method at an early stage. Significant changes in the precision or bias of the process are obvious from eyeballing the chart. Warning limits and control limits based on previously determined precision data are included on the chart to facilitate corrective action. Control charts are not limited to measurements on control and reference samples, but can also be used to monitor the range between duplicate measurements. This is particularly useful when reference samples or suitable controls are unavailable. A range control chart however is limited in that it can only indicate changes in analyst precision as any change in method bias will affect each duplicate result. The control chart illustrates whether the measurements process is in statistical control. As a consequence the analyst can assume confidence in data derived from single determinations. Having established that the analytical process is in control, the analyst derives more useful data from a set number of single determinations than half the set number of duplicate tests, the analytical effort remaining the same. However, this situation is mainly applicable when large numbers of similar samples are being routinely tested.
Application of quality assurance concepts to nutritional analysis
Whilst commitment and adherence to GLP procedures provide the first and most important level of quality assurance, the complexity of nutritional analysis due to the diverse range of foods and the wide range of analytes adds another dimension of difficulty. The limiting factors in assessing a laboratory's performance in Australian nutritional analysis programs include the general unavailability of certified reference food materials and a shortage of suitable collaborative laboratories in Australia. Few (if any) methods employed in nutritional analysis are absolute but generally rely on a comparative measurement between a standard (or reference) substance and the analyte. The unavailability of similar reference substrates increases uncertainties about the accuracy of the measurement. Likewise, whilst within-laboratory or operator precision can be readily determined, the shortage of collaborating laboratories reduces opportunities for comparison of between-laboratory precision.
Conclusion
Because of the diversity of food substrates, the extensive range of analytes involved and the differing stabilities of analytes, in particular vitamins, the application of standard analytical quality assurance procedures in nutritional analysis cannot be fully optimised. There is therefore a greater reliance on the experience and intuition of the analysts, a professional laboratory environment and organisation, effective program management and control, due attention to staff training and continuity of expertise and an appreciation of and commitment to the principles of GLP. In addition there is a greater requirement for analysts to check and evaluate analytical results against published values and to support data with details of sampling, methodology, method precision and quality assurance. Food composition analysis for national food composition tables is and will continue to be a professionally demanding task.
References
Garfield, FM. 1989. Sampling in the analytical scheme. J. Assoc. Off. Anal. Chem. 27: 405–11.
Haddy, B. 1990. Food composition data and the National Heart Foundation's Food Approval Program. Food Aust. 42:8, s 14–15.
Horwitz, W, Cohen, S, Hankin, L, Krett, J, Perrin, CH & Thornburg, W. 1978. Analytical food chemistry. Inhorn, SL (ed). Quality assurance practices for health laboratories. Washington, DC: American Public Health Association. 545–646.
Scheelings, P. 1987. Food composition analysis—getting started. English, R & Lester, I (eds). Proceedings of the first OCEANIAFOODS conference. Canberra: AGPS. 39–44.
Taylor, JK. 1987. Quality assurance of chemical measurements. Chelsea MI: Lewis Publishers Inc.
Wolf, WR (ed). 1985. Biological reference materials: availability, uses and need for validation of nutrient measurement. New York: John Wiley and Sons.
