Fresenius' Journal of
Fresenius J Anal Chem (1990) 338:430-434
@ Springer-Verlag 1990
Food and nutrition
Development of food reference materials for nutritional analysis Peter C. H. Hollman 1 and Peter J. Wagstaffe 2
State Institute for Quality Control of Agricultural Products, Bornsesteeg 45, NL-6708 PD Wageningen, The Netherlands 2 BCR Programme, Commission of the European Communities, Rue de la Loi 200, B-1049 Brussels, Belgium
Summary. The development of five reference materials for major nutritional properties, whole milk powder, pork muscle, wheat and rye flour, and haricots verts beans is described. Homogeneity and stability of these materials proved to be adequate. A preliminary intercomparison of methods showed that results for total fat and total dietary fibre were method dependent. Evaluation of methods used for available carbohydrates revealed poor solubilisation and hydrolysis of starch in some laboratories. This intercomparison has given valuable information for the final certification of these materials.
Preliminary intcrcomparison of methods
Prior to the certification of selected candidate reference materials an intercomparison of methods was organised involving 15 experienced food laboratories. In this way, valuable information on the state of the art of nutrient analysis by experienced food laboratories was obtained. In addition, the most accurate methods for certification were identified. In this intercomparison participants were requested to make 5 separate determinations of each nutrient. The laboratories were invited to apply the methods that they considered
100
90
Introduction
The growing awareness that dietary factors are important in the development of certain diseases, has stressed the importance of collecting reliable data for the nutritional composition of foods. To this purpose many governments give support to national food tables and data banks. Nutritionists and dieticians rely heavily on these compiled data on the composition of foods. Epidemiologists are frequently involved in international comparisons and consequently need comparable data for the composition of foods. The expanding nutritional labelling of foods, which will be helpful for consumers trying to choose a balanced diet, will also require analyses of nutrients. These developments increase the need for analytical data for nutrients in foods. Major nutrients, protein, fat and carbohydrates have been measured by analysts for more than hundred years. However, little or no data are available on the accuracy and compatibility of results for major nutrients from different laboratories. In 1985 an interlaboratory trial was carried out, showing major discrepancies between laboratories for proximate constituents [1]. These results are summarised in Fig. 1. Thus the need for quality control programmes in nutritional analysis was clearly demonstrated. In this context the BCR Programme started the development of food reference materials for major nutrients which will provide an important part of such quality control programmes. Offprint requests to: P. C. H. Hollman
~
80 70
~>~ ~ -G _~ O ~_ ~_ > O
60 50
$ 40 30 []
20 ÷
1o o 0
÷
•
I
1
I
I
I
I
I
I
10
20
30
40
50
60
70
BO
Mean value ( g / l O 0
90
g)
Fig. 1. Summary of the coefficients of variation (CV) of reproducibility of the results of the first interlaboratory trial [1]. Egg powder, milk powder, wheat and rye flour, biscuits and haricots
verts beans were analyzed by 19 laboratories using its own routine methods. The following nutrients were determined: • protein; + total fat; • available carbohydrates; • total dietary fibre; • ash
431 most reliable in their hands. All results were reported as grams per 100 g dry matter as determined by a prescribed vacuum oven method and, also reported on sample as received. Two materials were used: rye flour and a "chicken dinner" baby food, a mixture of vegetables and chicken in the form of a "roller" dried powder. These materials differ from the candidate reference materials and were chosen because at that time the candidate materials were not available. Homogeneity of these materials was checked by Kjeldahl nitrogen, sodium and potassium analysis and proved to be acceptable. The conclusions drawn from the intercomparison for each nutrient are summarised as follows.
Table 1. Summary of the results of the intercomparison of methods Chicken dinner
Rye flour
Total protein nitrogen Number of labs Mean Range
cv % CV within %
15 15 g protein nitrogen/100 g dry weight 3.938 1,528 3.432-4.306 1,242-1.752 5.3 7,5 0.87 1,4
Total fat Number of labs Mean Range CV % CV within %
14 14 g fat/t00 g dry weight 10.474 1.644 8,008-11.720 1.158-2.804 8.6 27.1 1.7 6.6
Mean Range CV % CV within %
Mean Range CV % CV within %
10 10 g monosaccharides/100 g dry weight 57.615 79.510 41.151-70.847 63.203 -90.579 13.6 9.8 1.7 2.9 11 12 g dietary fibre/100 g dry weight 5.239 12.187 3.020 - 7.519 8.720 - 14.094 31.6 15.1 16.2 3.4
The reproducibility of the fat determination again was rather poor (Table 1), however, repeatability compared favorable to the results of the first interlaboratory trial [1]. In general two types of methods were used: methods according to Schmid-Bondzynski-Ratzlaf (SBR) [2] and Weibull-Stoldt (WS) [3]. Laboratories using the SBR type methods found on average higher results than laboratories using WS type methods (Fig. 2). This was in agreement with the first interlaboratory trial [1]. These differences were significant (t-test, 5% level) for rye flour. Ether-extractable compounds other than fat probably explain these consistent findings in both trials. The International Organization for Standardization norm ISO 8262-1 [4] states that the SBR procedure is not suitable for foods containing more than 5 % starch or dextrin. These carbohydrates give ether-extractable compounds in the digestion with acid resulting in too high values for the fat content.
Methods
Lab nr-15 eye £10ue 9 13 3 12 7 16 II I 4 2 6 14
Total fat
Available carbohydrates were defined as free sugars plus starch. The level of agreement was very poor, with results ranging from 41 to 71 g/100 g dry weight for chicken dinner and 63 to 91 g/100 g dry weight for rye flour (Table 1). The within-laboratory variation was rather low. Methods showed many differences. Laboratories 10, 11, 12, and 15 (Fig. 3) isolated sugars and starch by separate extraction and determined in each extract sugars and starch with various methods. The other laboratories did not separate sugars and starch. Solubilisation of the starch was done by heat treatment, DMSO, alkali, perchloric acid, HC1 and by enzymatic action. Hydrolysis of the liberated starch was mostly done by amyloglucosidase, pancreas or a mixture of amyloglucosidase and e-amylase. The liberated glucose was determined by reductiometric (laboratories 1, 4, 11), colori-
Total dietary fibre Number of labs
The overall coefficient of variation (CV) for rye flour was rather high (Table 1). Methods used were Kjeldahl methods and differed in choice of catalysts (Cu, Hg, TiOz, Se alone and in different combinations) and in procedures for digestion, distillation and determination of the ammonia formed. No significant effects of those procedures were evident.
Available carbohydrates
Available carbohydrates Number of labs
Total protein nitrogen
I
I
I
~
it
8BR 0
I I i
WS SBR SBR 8BR o ws
i&
rll II
9 " -T
t L g~i
ws w8 w8
t~J 1, rAII ,,
r ,
,&n i
I t IlL
II
1,1t5
w8 ws w8
1,35
z o5
'
2,S5
3,TB5
Fig. 2 Total fat (g/100 g dry weight) content in rye flour analysed by different laboratories. Vertical dashed refer to the 5 separate determinations within one laboratory. Black triangles represent the means of these separate determinations
432
Lab nr 18 r~e flou~ 4
|
2
k
@
II 14 13 15 I 6 12
I
h I l•l
Fig. 3 Available carbohydrates (g/100 g dry weight) content in rye flour analysed by different laboratories. Vertical dashes refer to the 5 separate determinations within one laboratory. Black triangles represent the means of these separate determinations
I I I
I
,
601
,
7~]~
r
8~
98 I
lob
Methods
Lab n~ chicken dinne~ 13 L8 1 4 lI 16 12 3 15
'
a' A
i
J
I
I ii~
i
II~l I
I jr
A N
I
I
A E A A
I
E E
14
'
I! v
21
4r
6i
8r
lib
Fig. 4 Total dietary fibre (g/100 g dry weight) content in chicken dinner analysed by different laboratories. Vertical dashes refer to the 5 separate determinations within one laboratory. Black triangles represent the means of these separate determinations
l~.
Table 2. Selected candidate reference materials and preliminary values of nutrient contents
Total protein N
Total fat
Available carbohydrates
Total dietary fibre
Ash
26.4 11.2 1.2 1.2 -
37.2 80.8 81.7 73.1
-
6.0 4.6 0.9 0.8 2.3
g/100 g dry weight Whole milk powder Pork muscle Rye flour Wheat flour Haricots verts beans
4.5 13.5 1.2 2.1 1.1
metric (laboratories 2, 10, 13), and enzymatic (laboratories 6, /5) techniques. Laboratory 14 determined starch with a polar•metric method. Whenever a separate determination o f the free sugars was made, this was done by enzymatic methods (laboratories 6, 14, 15), reversed phase H P L C (laboratories 10, 11), and ion chromatography (laboratory 12). Underestimation of carbohydrates caused by poor solubilisation and or incomplete hydrolysis of the starch, was probably the most important source of error. The low results of laboratory 12 may be due to the solubilisation with an alkali treament (30 rain at 60 ° C), which is not sufficient. The low result of laboratory 6 may be due to hydrolysis of starch with amyloglucosidase at 60 ° C for only 15 min, which is unlikely to lead to completion. To tal dietary fibre There was a large variability in dietary fibre values between the different laboratories (Table 1). This was partly due to the wellknown between-methods differences. The AOAC method [5] was used by most of the participants, laboratory 6 used a related method described
8.2 3.4 11.9
by Asp [6]. In Fig. 4 the letter A refers to these enzymaticgravimetric methods. The lower values reported by laboratories using Englyst methods (E) [7, 8] are to be expected because dietary fibre as determined by this type of method does not include lignin and resistant starch. So, as expected the results were strongly method dependent.
Candidate reference materials
Selection and preparation F o o d reference materials ideally should cover the whole range of matrices relevant to human nutrition. F r o m an analytical point of view, materials with a low and a high level of each analyte are to be preferred. Further, materials should be selected based on expected homogeneity and stability of the nutrients of interest. However, it is not possible to meet all of those criteria, so compromises have to be made. M a n y thoughts were given to the favorable prospects of a mixed diet material representing the average food intake, and consequently containing most of the food components possible. However, difficulties were envisaged in producing
433 a homogeneous batch of a mixed diet with respect to fat and dietary fibre. These considerations led to the selection of the materials indicated in Table 2. Materials were purchased from commercial suppliers, except for the pork muscle. The meat was specifically, selected, minced and commercially freeze-dried. The freezedried meat was finally ground and mixed. The average particle size was measured and found to be less than 100 gm, except for the haricots verts beans where the average particle size was 7 0 - 200 gm. Each material was mixed thoroughly in a 50-kg ribbon blender prior to being packaged. Samples of 100 g of each material (50 g for pork muscle) were weighed into laminated foil pouches, flushed with nitrogen, and then double heat sealed.
[]
c¢ ¢
[]
-FJ
(]) _Q
> ©
OA ,A•
Homogeneity Sample homogeneity was tested as follows. During the packaging, 10 samples of each product were taken at 100package intervals and analysed for the nutrients of interest. To determine the analytical precision, one sample of each product was analysed for each nutrient ten times. All results were calculated on dry matter. For each nutrient and material the standard deviation between the samples was compared with the standard deviation within the sample (analytical precision). The results show that there is no significant difference (F-test, 5% level) between these standard deviations. One exception has to be made for ash in haricots verts beans. The variation between the samples (CV between = 1.5%) differs significantly (1% level) from the analytical variation (CV analytical = 0.7 %). However, analysis of potassium in these samples did not reveal any heterogeneity (CV between = 1.0%, CV analytical = 2.1%). Figure 5 shows a summary of these results; the standard deviations are expressed as coefficients of variation. The between-samples variation was small, and could largely be ascribed to analytical error rather than to true differences between different samples of one material. The reference materials can thus be regarded as homogeneous.
Stability In order to confirm the expected stability of the nutrients in the powdered dry materials, a 24 months stability study is being carried out. All nutrients in the samples stored at - 1 8 , 4, 20 and 37°C are monitored at regular intervals. Results of storage during 12 months do not show evidence for degradation of the nutrients of interest. Figure 6 shows a typical example, fat in wheat flour, the variation in fat content is well within the analytical precision (standard deviation = 0.10 g/100 g).
"1",,,I.
I
0
1
0
CV
analytical
(%)
Fig. 5. Results of the homogeneity study, comparing for each material and nutrient the variation between sachets (CV between) with
the analytical precision (CV analytical). ÷ total protein nitrogen; • total fat; • available carbohydrates; • ash 1.7.
1.5 A
',r--
v
O~
,4-
,,I-
0 0
.i,,
g
1.3
~5
LL
1.1
09.
l
l
l
I
2 Discussion and conclusions
Whole milk powder, freeze-dried pork muscle, rye and wheat flour, and dried haricots verts beans potentially are suitable reference materials for nutritional analysis. Batches of these materials were produced and proved to be homogeneous with respect to major nutrients. These nutrients were stable, "even after storage at 37°C during at least 12 months. A preliminary intercomparison of methods with two materials involving 15 experienced food laboratories generally showed lower variation within laboratories compared to a
2
T
l
l
T
5
l
l
t
f
~
l
8
11
14
Period (months) Fig. 6. Total fat content in wheat flour monitored at regular intervals after storage at different temperatures ( + - 18 ° C; • + 20 ° C; •
+37°C)
previous interlaboratory study [1]. However the laboratories still produced widely different values for the concentrations of fat, carbohydrates and dietary fibre. As for the causes of
434 these discrepancies, differences in methods play an important role. For total fat results were clearly method dependent, with SBR methods giving higher values than WS methods. As both procedures are well established in different countries, two separate values for total fat based on each method will be certified. Evaluation of methods used for the determination of available carbohydrates revealed two critical steps in the procedures: solubilisation and hydrolysis of starch. In the final certification guidelines will be given for these two steps. Results for dietary fibre also were clearly method dependent. Concerning the enzymatic-gravimetric method, 7 out of 8 laboratories had agreed upon the A O A C procedure as described by Prosky [5]. As a result two values for dietary fibre will be certified according to the A O A C method [5] and the Englyst procedures [7, 8]. This intercomparison of methods has given valuable experience for the certification exercise which has been started. Hopefully, the certified reference materials will be ready for release at the end of 1990.
Acknowledgements. The authors are grateful for the cooperation of members of the following laboratories: Unilever Research Colworth Laboratory, Bedford, UK; Bundesforschungsanstalt ffir Ern/ihrung, Karlsruhe, Federal Republic of Germany; Centro Nacional de Alimentacion y Nutricion, Madrid, Spain; Istituto Nazionale della Nutrizione, Roma, Italy; Leatherhead Food RA, Leatherhead, UK; Institute of Food Research, Bristol, UK; Rijkskeuringsdienst van Waren, Maastricht, The Netherlands;
Minist6re de l'Agriculture, Paris, France; Rijks-Kwaliteitinstituut voor Land- en Tuinbouwprodukten (RIKILT), Wageningen, The Netherlands; Agricultural University, Wageningen, The Netherlands; Laboratory of the Government Chemist, London, UK; Ministerio de Sanidad y Consumo, Madrid, Spain; Dublin City Public Analyst, Dublin, Ireland; Nest6 Deutschland AG, Frankfurt, Federal Republic of Germany; CIVO-Instituten TNO, Zeist, The Netherland; AFRC Food Research Institute, Norwich, UK; National Food Institute, Soborg, Denmark; Ministerie van Economische Zaken, Brussels, Belgium; Energie Centrum Nederland, Petten, The Netherlands.
References
1. Hollman P, Katan M (1987) Fresenius J Anal Chem 326:690695 2. Schormfiller J (1968) Handbuch der Lebensmittelchemie III/1. Springer, Berlin Heidelberg New York, p 807 3. Osborne D, Voogt P (1978) The analysis of nutrients in food. Academic Press, New York London 4. ISO 8262-1 (1987) Milk products and milk-based foods - Determination of fat content by the Weibull-Berntrop gravimetric method (Reference method) 5. Prosky L, Asp N, Furda I, De Vries J, Schweizer T, Harland B (1985) J Assoc Off Anal Chem 68:677-679 6. Asp N, Johanssen C, Hallmer H, Siljestr6m H (1983) J Agric Food Chem 31:476-482 7. Englyst H, Cummings J (1984) Analyst 109:937-943 8. Englyst H, Hudson G (1987) Food Chem 24:63-76 Received June 22, 1990