SANDERSON ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 971
FOOD COMPOSITION AND ADDITIVES
Thermoluminescence Detection of Irradiated Fruits and Vegetables: International Interlaboratory Trial DAVID C.W. SANDERSON, LORNA A. CARMICHAEL, and SAFFRON FISK Scottish Universities Research and Reactor Centre (SURRC), Scottish Enterprise Technology Park, Rankine Ave, East Kilbride, G75 OQF, UK Collaborators: P. Christensen; H. Delincée; K. Hammerton; H. Nootenboom; J. Pfordt; S. Pinnioja; G.A. Schreiber; and M. Toyoda
An international interlaboratory trial was conducted to validate thermoluminescence methods for detecting irradiated fruits and vegetables. Five products were used in this study. This paper presents the results from prestudy material, homogeneity testing, details of sample preparation, and participants’ results. Prestudy results provided a basis for cross comparison of instruments in different laboratories. A wide range of sensitivities, reproducibilities, and signal-to-background ratios were observed. Homogeneity testing showed that the method can distinguish between nonirradiated and irradiated products, including those bleached with 100 J/cm2 artificial daylight, provided that sensitivity rejection criteria are rigorously applied. Blind results were returned by 9 participants in the form of first and second glow integrals and glow ratios for all samples and a qualitative classification for each product. Of the 387 results reported, 327 valid results were obtained from participants. Where valid data were obtained, correct qualitative identifications were made by participants in all cases. Participants’ results and homogeneity testing both confirm the validity of the thermoluminescence method for detecting irradiated fruits and vegetables.
he development of luminescence tests for identifying irradiated foods began at Scottish Universities Research and Reactor Centre (SURRC) in 1987. It has been established that thermoluminescence (TL) of foods is associated with the silicate minerals adhering to the surface of all foodstuffs (1, 2). This forms the basis for the UK-validated and current European standard methods (3–16) and has also been adopted by the Codex Alimentarius Commision. Initial successful work with herbs and spices prompted research into the extension of TL techniques to other products. During 1995
T
and 1996, an international blind trial was conducted on fruits and vegetables and is the subject of this paper. Research had already confirmed (17) that silicates could be extracted from a wide variety of fruits and vegetables and that irradiated samples could be detected under controlled conditions. The effects of optical bleaching were examined with the conclusion that TL signals would be reduced in samples exposed to light but that radiation-induced signals were still detectable. An earlier trial on irradiated fruits and vegetables, conducted in Germany (18) and using herbs and spices detection criteria, resulted in poor detection rates. This raised several procedural questions, particularly in the areas of quality assurance. For the SURRC study, laboratories with TL capabilities were invited to participate. Those accepting were sent protocols that included detailed quality assurance stages; comments were invited prior to sample distribution. Attention was given to definition of laboratory background, minimization of light exposure to samples, and calibration of TL temperature scales. Instructions were included for rejecting data where quality assurance criteria were not met. Five sample types were distributed blind to participants (strawberries, mangos, papayas, mushrooms, and avocados) in 3 conditions: nonirradiated, irradiated (1 kGy), and irradiated and optically bleached (1 kGy + 100 J/cm3 artificial daylight). Bleached samples were included to assess the impact of handling in daylight on detectability. Study materials were sealed individually prior to treatment and handling. Random allocation procedures were used to select samples for treatment and distribution; response to the treatments and assessment of within-laboratory variation (homogeneity testing was performed at SURRC) contemporaneously with the trial. Participants performed silicate extractions and TL analyses following the protocol using individual items of fruit or vegetables and reported data from first glow (G1), second glow (G2), and glow ratios (G1/G2), together with qualitative classifications for each product. Experimental Protocol
Received February 27, 2003. Accepted by SG May 23, 2003. Corresponding author’s e-mail:
[email protected].
The analytical protocol was drafted in international standard format to facilitate adoption as a standard/validated
972 SANDERSON ET AL. : JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 Table 1. Prestudy—Observed LiF peak V and VI temperature ranges and feldspar T1/2 positions LiF temperature range, °C
Feldspar T1/2 positions
Lab
Peak V
Peak VI
F1, 200°C
F2, 250°C
F3, 300°C
1
235–245
305–310
—
—
— 288.5 ± 0.5
2
233–239
295–300
184 ± 1.0
235 ± 2.0
3
291–319
361–387
288 ± 5.0
325 ± 18.0
390 ± 14.0
4
236–241
295–311
179 ± 0.0
211 ± 3.0
280 ± 2.0
7
262–268
326–334
198.5 ± 0.5
236.5 ± 0.5
302.5 ± 1.5
8
238–240
301–312
211 ± 1.0
260 ± 8.0
306 ± 6.0
9
251–288
318–350
189.5 ± 0.5
227 ± 1.0
293 ± 3.0
10
257–259
318–322
200 ± 0.6
249 ± 0.7
293 ± 2.5
11
238–251
259–267
234.5 ± 15.5
261 ± 5.0
320 ± 5.0
13
249–260
—
207.5 ± 4.5
245.3 ± 0.5
307 ± 4.0
233–319
259–387
210 ± 32
250 ± 30
309 ± 31
Overall range
method. It extended the herb and spice validated method V27 (9) by including an additional density separation and adjusting detection criteria in response to earlier work on bleached fruits and vegetables (17). Prestudy Materials To compare analyses of blind samples from different laboratories using different equipment, prestudy materials were distributed in May 1995. To provide a basis for comparison of temperature scales and intensities measured in a variety of units, participants were sent a 14C source, LiF dosimeters with matched sensitivities irradiated to a high precision, and preannealed International Atomic Energy Agency (IAEA) F1 feldspars. Prestudy materials were used to define individual minimum detectable levels for each laboratory and to monitor detection sensitivity throughout the trial itself. Feldspars were included to provide an independent check on temperature by using a material similar to the silicates separated from the blind samples. Choice of Samples and Predistribution Treatment Exotic fruits and vegetables are candidates for irradiation because they are high-value products, easily perishable, and can benefit from shelf-life extension. Some of these foods originate in countries with irradiation facilities, whereas other produce is traded through countries with an active irradiation program. The samples chosen for this interlaboratory trial were purchased from a Glasgow fruit and vegetable wholesaler. They consisted of strawberries from Holland, papayas from Brazil, mushrooms from Ireland, avocados from Kenya, and mangos from Venezuela. Samples were individually bagged immediately after purchase to avoid the possibility of cross-contamination between treatment categories and to enable random allocation. Samples were randomly selected and repacked to receive a commercial irradiation dose of 1 kGy at Isotron (Swindon, UK). Isotron used dosimetry to confirm the distribution of ab-
sorbed dose across the products; to enable cross-reference, dosimeters were also brought back and measured at SURRC. All samples received a mean dose of 0.89 ± 0.03 kGy. A number of irradiated samples were then randomly selected, placed in purpose-built light boxes containing artificial daylight tubes, and bleached. The bleaching period was calculated to deliver a 100 J/cm2 bleach. Nonirradiated control samples were stored under refrigerated conditions (4°C) until the irradiation and bleaching sequences were completed and until all samples were ready to be randomly selected, packed, and dispatched to participants. Participants received each of the 5 fruit and vegetables in each of their 3 blind categories as coded samples. Participants were asked to analyze the contents of each sample as a single aliquot, following the protocol. Homogeneity Testing At SURRC, in addition to a set of blind samples, homogeneity testing was conducted on randomly selected excess material to ensure that the bulk material for the interlaboratory trial was sufficiently homogeneous. Because of the perishable nature of the material, this procedure was performed in paral-
Table 2. Prestudy—Comparison of peak V results from different trial data Laboratory/date
Peak V range
Mean peak V position
BCR Solidus 1991
190–328
270 ± 43
BGA Solidus 1993
196–277
246 ± 29
BGVV Solidus1994
174–312
240 ± 30
MAFF Solidus 1996
233–319
255 ± 19
SANDERSON ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 973
lel with the blind analysis. Twelve-fold replication was conducted on the 5 sample types in their 3 blind categories to assess intrinsic variability. Preparation of Blind Samples
run in parallel with samples were used, together with prestudy results, to determine individuals’ minimum detectable levels (MDL). TL Measurement and Recording of Results
Silicate minerals suitable for TL analysis are present on the surface of fruit and vegetables. After ultrasonic agitation in deionized water, the released minerals are separated from organic constituents with heavy liquid (sodium polytungstate) and cleaned in HCl. Full details of the procedure can be found in EN 1788 (16). Separated silicates are deposited onto stainless steel disks for TL readout. Glassware and process blanks
Participants were asked to measure first glow (G1) TL from room temperature to 400°C at 5°C/s, and then to irradiate the samples, using either a 60Co or a 90Sr source prior to recording the second glow (G2) under identical conditions. Participants were also asked to report their TL data (G1 signal integrated over 20°C bands, G2 signal integrated over 20°C bands, and G1/G2) together with their qualitative classifica-
Figure 1. Glow ratio (G1/G2) for each of the 5 products: homogeneity testing data.
Figure 2. Participants’ raw glow ratio data, prior to rejection of low-sensitivity samples.
974 SANDERSON ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003
tion of each sample. This classification was to be based on both G1/G2 (glow ratio) and G1 shape. Where G2 did not exceed 10 MDL as defined for each laboratory, the sample was to be rejected. Separation into nonirradiated or irradiated was defined by a threshold of G1/G2 = 0.1. Results and Discussion Prestudy Participants’ results from the prestudy showed a wide range of sensitivities, reproducibilities, and signal-to-background ratios. There is also evidence of wide variation in temperature calibration. Table 1 shows the observed temperature values for LiF peaks V and VI and the F1 feldspar standards. Considerable differences are observed in the LiF peak positions. The overall range of temperatures for peak V spans from 233° to 319°C, implying variation in both laboratory calibration and thermal contact. Heating rate does not explain this variation because all laboratories except one used 5°C/s. The 6.5°C/s used by the exception did not produce significant differences in LiF peak temperatures. To investigate whether the temperature shifts observed with LiF dosimeters were due to thermal lags, F1 feldspars (which were dispensed on thin layers on disks) were also measured. Results, presented in Table 1 as T1/2 values (first peak half maximum), show a similar variation. Thermal lag is therefore not the main cause. Variation between the laboratories, however, was greater than within any one laboratory. This prestudy made it possible to compare current temperature variations with those seen in previous trials (18, 19). Table 2 summarizes the LiF peak V position from previous trials. It is notable that very little improvement has been achieved since the BCR trial in 1991 (10, 18, 19). For the general application of a standard method, which uses both glow shape and glow ratio over a specific temperature range in classification of food samples, it is clear that there is a need for confidence in temperature calibration. This study suggests that this has yet to be achieved. The calibration variation must also be taken into account when comparing data for blind samples; on the basis of participants’ prestudy results, temperature shifts were applied as necessary to their data from the blind analyses.
Homogeneity Testing The main purpose of homogeneity testing was to characterize the response of the material within a single laboratory, so that laboratory-to-laboratory variations could be examined independent of the natural variations between and within samples. Figure 1 shows the glow ratios integrated between 220° and 240°C of homogeneity testing results presented for each product. Glow ratios separate the nonirradiated from the irradiated samples, with slightly lower glow ratios from bleached samples. The glow ratios from irradiated samples, whether bleached or not, typically lie between 0.1 and 1. The responses to irradiation and to irradiation and bleaching are highly consistent from product to product. The nonirradiated product data show glow ratios ranging over 1–2 orders of magnitude, a greater spread than for the irradiated samples. Cleaner products (mushrooms and avocados) produced fewer valid results than the other products did as a result of lower mineral yields per disk. On the basis of the homogeneity testing, it was expected that the blind trial would produce adequate separation between irradiated and nonirradiated samples. Participants’ Results Nine laboratories returned results of analyses in the form of G1 and G2 integrals and glow ratios. Rigor of application of rejection criteria varied from laboratory to laboratory. There was evidence that process blanks had changed between the prestudy stage and analysis of the blind samples, and not all laboratories varied their rejections in response to this. Two laboratories did not reject or identify low-sensitivity samples but did include data for blanks to enable rejection at SURRC. A total of 60 results was rejected, either by the participants or by SURRC. Participants also returned qualitative classifications, sometimes commenting on the importance of glow shape in reaching their decisions. Low mineral yield, especially for the mushrooms and to some extent the avocados, was remarked on by the participants and was the main cause of rejection. Where the prestudy had identified a significant temperature shift for a laboratory, raw data were re-analyzed to achieve standardization, but this did not alter the qualitative classification in any instance. Participants’ raw glow ratio data for the 220°–240°C
Table 3. Numbers and percentages of valid results (participants) as a proportion of total analyses undertakena Nonirradiated Product
Irradiated
All samples
No. of analyses
No. valid
No. of analyses
No. valid
No. of analyses
No. valid
Valid, %
Strawberries
25
23
50
50
75
73
97.3
Avocados
26
23
52
41
78
64
82.1
Mushrooms
28
13
56
27
84
40
47.6
Papayas
27
19
54
48
81
67
82.7
Mangoes
29
27
58
56
87
83
95.4
135
105
270
222
405
327
80.7
Total a
All valid results were correctly classified.
SANDERSON ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 86, NO. 5, 2003 975
band (Figure 2) show that the majority of irradiated samples can be distinguished from nonirradiated samples for all 5 samples. Product-to-product variation is not significant, but the importance of applying sensitivity checks can be seen from the greater overlap observed in low-sensitivity samples (mushrooms and avocados) in laboratories where such checks were not applied. Optically bleached samples are less distinct from nonirradiated ones than unbleached samples are, but there is still separation with the level of bleaching used. Participants’ qualitative data were compared with the true status of each sample. There were no misclassifications in the 327 valid data—an identification rate of 100%. The percentages of all samples analyzed leading to valid results, however, varied from product to product as a result of varying mineral yields from each product. The data presented in Table 3 show that successful mineral separations were achieved for only 47.6% of mushrooms, whereas all other products had success rates in excess of 80%. Across all products and classes, 80.7% of the analyses led to successful (and correct) determinations. The results show that, given a successful mineral extraction and rigorous low-sensitivity rejections, TL analysis produces correct determinations. Conclusions The prestudy results show that significant errors in temperature calibration are still a feature in numerous laboratories, but temperature correction applied to data from blind analyses did not alter the qualitative detection performance. Homogeneity testing confirmed that for all 5 products examined, TL analysis separated irradiated and nonirradiated material. The irradiated and bleached samples displayed glow ratios between those characteristic of nonirradiated and irradiated but not bleached samples, but the samples were distinguishable from nonirradiated material. Low mineral yield was associated with failure to meet quality assurance criteria. A total of 387 results was reported by participants, of which 327 remained after the application of sensitivity rejections. All these led to correct classifications of the blind samples, using both glow ratio and glow shape. Participants’ data confirm the results of the homogeneity testing and provide a basis for thorough validation of TL detection methods for fruits and vegetables. Acknowledgments Support for this work by the Ministry of Agriculture Fisheries and Food (MAFF) under contract 1B073 is gratefully acknowledged. We are also grateful to the following participants for the work and time they contributed to the study, without which it would not have been possible: P. Christensen, Risø, Roskilde, Denmark H. Delincée, Federal Research Centre for Nutrition, Karlsruhe, Germany K. Hammerton, ANSTO, Sydney, Australia H. Nootenboom, Food Inspection Service, Nijmegen, The Netherlands
J. Pfordt, Staatliches Lebensmitteluntersuchungsamt, Oldenburg, Germany S. Pinnioja, University of Helsinki, Helsinki, Finland G.A. Schreiber, BGVV, Berlin, Germany M. Toyoda, National Institute of Health Sciences, Tokyo, Japan References (1) Sanderson, D.C.W., Slater, C., & Cairns, K.J. (1989) Nature 340, 23–24 (2) Sanderson, D.C.W., Slater, C., & Cairns, K.J. (1989) Radiat. Phys. Chem. 34, 915–924 (3) Sanderson, D.C.W., Slater, C., & Cairns, K.J. (1989) Int. J. Radiat. Biol. 55, 5 (4) Sanderson, D.C.W. (1990) in Food Irradiation and the Chemist, D.E. Johnston & M.H. Stevenson (Eds), Royal Society of Chemistry, Cambridge, UK, pp 25–56 (5) Sanderson, D.C.W., Carmichael, L.A., Ni Riain, S., Naylor, J., & Spencer, J.Q. (1994) Food Sci. Technol. Today 8, 93–96 (6) Sanderson, D.C.W., Carmichael, L.A., & Naylor, J.D. (1995) Food Sci. Technol. Today 9, 150–154 (7) Sanderson, D.C.W., Carmichael, L.A., & Naylor, J.D. (1996) in Detection Methods for Irradiated Foods, C.H. McMurray, R. Gray, E.M. Stewart, & J. Pearce (Eds), Royal Society of Chemistry, Cambridge, UK, pp 124–138 (8) MAFF (1992) Detection of Irradiated Herbs and Spices, Validated Methods for the Analysis of Foodstuffs, V27, Ministry of Agriculture, Fisheries and Foods, Norwich, UK (9) MAFF (1993) J. Assoc. Public Anal. 29, 187–200 (10) Sanderson, D.C.W., Schreiber, G.A., & Carmichael, L.A. (1991) A European Interlaboratory Trial of TL Detection of Irradiated Herbs and Spices, Phase 1, SURRC report to BCR, European Commission, Brussels, Belgium (11) Raffi, J., Fakirian, A., & Lesgards, G. (1994) Ann. Falsif. Expert. Chim. Toxicol. 87, 125–134 (12) Pinnioja, S. (1993) Radiat. Phys. Chem. 42, 394–400 (13) DelincJe, H. (1993) Radiat. Phys. Chem. 42, 351–357 (14) Roberts, P.B., & Hammerton, K.M. (1994) in Detection Methods for Irradiated Foods, C.H. McMurray, R. Gray, E.M. Stewart, & J. Pearce (Eds), Royal Society of Chemistry, Cambridge, UK, pp 178–181 (15) Hammerton, K.M., & Banos, C. (1994) in Detection Methods for Irradiated Foods, C.H. McMurray, R. Gray, E.M. Stewart, & J. Pearce (Eds), Royal Society of Chemistry, Cambridge, UK, pp 168–171 (16) European Standard Method (1997) Foodstuffs—Detection of Irradiated Food from which Silicate Minerals Can Be Isolated: Method by Thermoluminescence, BS EN 1788, BSI, London, UK (17) Sanderson, D.C.W., Carmichael, L.A., Clark, P.A., & Clark, R.J. (1992) Development of Luminescence Tests To Identify Irradiated Foods, Final report N1701 Identification of irradiated fruits and vegetables, MAFF N1701, London, UK (18) Schreiber, G.A., Helle, N., & Bögl, K.W. (1993) Bundesgesundheitsblatt 36, 355–359 (19) Schreiber, G.A., Ziegelmann, B., Mager, M., & Bögl, K.W. (1993) in Detection of Low Dose Irradiation of Fruits and Vegetables, G.A. Schreiber (Ed.), Bundesgesundheitsamt, Berlin, Germany, pp 66–70