Use of an efficient measurement uncertainty approach to compare

Subscriber access provided by READING UNIV

Article

Use of an efficient measurement uncertainty approach to compare room temperature and cryogenic sample processing in the analysis of chemical contaminants in foods Lijun Han, Steven J. Lehotay, and Yelena Sapozhnikova J. Agric. Food Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jafc.7b04359 • Publication Date (Web): 14 Nov 2017 Downloaded from http://pubs.acs.org on November 15, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Agricultural and Food Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 27

Journal of Agricultural and Food Chemistry

Use of an efficient measurement uncertainty approach to compare room temperature and cryogenic sample processing in the analysis of chemical contaminants in foods Lijun Han,1 Steven J. Lehotay,*,2 and Yelena Sapozhnikova 2 1

2

College of Science, China Agricultural University; Beijing, 100193; China U.S. Department of Agriculture, Agricultural Research Service, Eastern Regional Research Center; 600 East Mermaid Lane; Wyndmoor, PA 19038; USA

*Corresponding author. Email: [email protected]; Phone:

1-215-233-6433.

Disclaimer: The use of trade, firm, or corporation names does not constitute an official endorsement or approval by the USDA of any product or service to the exclusion of others that may be suitable. Table of contents graphic image:

1

Abstract

2

In this study, analytical results were compared when using different approaches to bulk food

3

sample comminution, consisting of a vertical chopper (Blixer®) at room temperature and at dry

4

ice cryogenic conditions, followed by further subsample processing (20 g) using liquid nitrogen

5

cryogenic conditions (cryomill).

6

in a food mixture consisting of equal parts orange, apple, kale, salmon, and croaker involved

7

QuEChERS with automated mini-column solid-phase extraction (known as ITSP) cleanup

Analysis of the 43 targeted spiked and incurred contaminants

1

ACS Paragon Plus Environment


Page 2 of 27

8

followed by low pressure gas chromatography - tandem mass spectrometry (LPGC-MS/MS).

9

Different ambient Blixer® test portion sizes of 20, 10, 5, 2, and 1 g were assessed, and for

10

cryogenic Blixer® conditions, a 0.5 g test portion was also tested.

11

portions were 2, 1, and 0.5 g, and all subsamples in all cases entailed 5 replicates.

12

concentrations and precisions (CV) of the analytes were compared to assess possible differences

13

in systematic and random forms of error.

14

the procedures to isolate that individual step in the uncertainty measurements using the error

15

propagation sum of squares approach.

16

preparation and LPGC-MS/MS analysis steps were 2-7% and 11% CV, respectively, while

17

uncertainties of sample processing ranged from 6% CV for the cryomill to 12% CV for the

18

ambient Blixer® conditions.

19

uncertainty from 12-15% to 7-10% CV.

20

this study, the minimal test sample weight that gave satisfactory recoveries and precision was

21

found to be 1 g in all cases.

In case of the cryomill, test Determined

A quality control spike was made before each step in

Results indicated that the uncertainty of the sample

The common use of internal standards reduced overall method For the analytes, matrix, conditions, and tools used in

22 23

Keywords: Sample Processing; Comminution; Measurement Uncertainty, Pesticides and

24

Environmental Contaminants; Food Analysis

25 26 27

■ INTRODUCTION Food safety concerns by consumers, governments, and industry, as well as increasing

28

international food trade, have contributed to the greater importance of high-throughput

29

multiresidue analysis of pesticides and environmental contaminants in monitoring laboratories all

30

over the world.

31

process is as critical as any other, but the sample processing (comminution) step is often

32

ignored.1

33

preparation (usually including extraction and cleanup); 3) analysis (e.g. chromatography and

34

detection); and 4) data handling and reporting.

35

quality control (QC) measures usually start from the sample preparation step, in which

36

pre-weighed test portions are spiked with standard solutions in the extraction vessels.

37

Consequently, the systematic and random errors inherent in the sample processing step are

38

eliminated from the measurement uncertainty in the full validated method.

Just as a chain is only as strong as its weakest link, each step in the analytical

The series of laboratory steps typically entail: 1) sample processing; 2) sample In routine laboratories, method validation and

2


Page 3 of 27

39


Relatively few studies address sample processing techniques compared to tens of thousands

40

of publications on sample preparation and instrumental analysis of food contaminants.

41

demonstrate this, a simple search of “food and analysis” was done using Web of Science in

42

August of 2017, which yielded a list 105,414 publications.

43

18,866 papers mentioning “detection or determination or quantitation or quantification,” 13,441

44

on “spectrometry or spectroscopy,” 11,985 on chromatography, 8,987 on “extraction or ‘sample

45

preparation’ or clean[-]up,” but “comminution or ‘sample processing’ or homogenization”

46

yielded only 296 publications.

47

field of food analysis, but that is a different focus of work.2

48

To

Refined searches led to subsets of

“Data processing” is another under-reported area of study in the

In practice, sample processing techniques and the devices used have been much the same for

49

generations.

Out of curiosity, we randomly perused two analytical journals from 1909 and 1962,

50

and we saw little difference in how sample processing was described in food analysis then vs.

51

now.3,4

52

each thoroughly sampled by grinding them in a chopper.”3

53

been placed on comminution as the other steps in analytical methods over time, valuable

54

advancements would have been made in sample processing leading to even better practical

55

benefits and analytical performance than realized today.

For example, Emmett and Grindley wrote, “The portions of the resulting lean beef were Perhaps if as much attention had

56

We believe that establishing a new practice to require that method validation start from the

57

comminution step would force analytical chemists to pay more attention to this long-neglected

58

factor, and it would surely lead to improvements in sample processing technology and techniques.

59

Currently, the vast majority of analytical chemists take sample processing for granted, or they

60

and journal editors/reviewers think the subject is too mundane, but due to great advances

61

recently in sample preparation, analytical separations, detection, and software, sample processing

62

has become the major limitation in sample throughput and overall quality of results in real-world

63

analyses.

64

That is not to say that sample processing has been completely neglected.

CODEX

65

sampling guidelines require that bulk samples for analysis consist of at least 10 or 5 units

66

(depending on type of fruit or vegetable) and must weigh >1 or >2 kg, respectively.5

67

sample is then comminuted to generate a representative test sample portion, typically 10-50 g,

68

for analysis.6,7

69

most routine laboratories is conducted at room temperature.

This bulk

Due to resource limitations and sample throughput needs, sample processing in However, it has been demonstrated 3



Page 4 of 27

70

that cryogenic processing using dry ice (solid CO2) or liquid nitrogen often produces more

71

reliable results.8-11

72

that the same degree of analyte variability could be achieved for a 5 g test sample when using dry

73

ice as 110 g at room temperature, whereas for orange, subsample size of 5 g provided sufficient

74

homogeneity for both sample processing procedures.8

75

fish filets, 2 g subsamples were sufficiently representative when using cryogenic conditions.11

76

For chlorpyrifos in tomato using a specific chopper, Fussell et al. showed

Another study showed that in the case of

In the case of miniaturized high-throughput methods employing 96-well plates, very small

77

sample sizes are a necessity, and at least one report indicates that test portions as small as 100 mg

78

can be satisfactorily representative of the bulk sample.12

However, as exemplified in the

8

79

comparison between tomatoes and oranges, the miniaturized applications cannot be extended to

80

other analytes and sample types without being specifically validated to meet method

81

acceptability criteria.

82

during validation and proficiency testing experiments, but become lax during routine sample

83

analyses.

84

QC practices for every sample analysis,1 especially as test portion sizes are reduced below

85

proven standard practices.

Even then, analysts tend to take extreme care in their analytical practices

Thus, we emphasize that the comminution step should be assessed as part of ongoing

86

To routinely estimate measurement uncertainties, a QC spike can be added to the sample

87

before each step, including sample processing, to isolate the error contribution of each step in the

88

overall uncertainty of the method.13,14

89

and troubleshoots which step is the source of problems when poor results are obtained.

90

simple sum of squares set of equations are used to make the uncertainty assessments:

This simple practice empirically measures uncertainties A

CV2overall = CV2proc + CV2prep + CV2anal ,

91 92

in which CV is the coefficient of variation of the overall method and for each step (proc =

93

sample processing, prep = sample preparation, and anal = analysis).

94

standard deviation (RSD) are mathematically synonymous, in this work we use CV for the

95

calculated measurement uncertainty isolated for that step in the method, and RSD refers to the

96

actual empirically measured precision of the QC spiking compound (e.g. QCproc, QCprep, and

97

QCanal) in the experiment.

98

reflects all of its steps, CVoverall = RSDproc.

99

analytical step, thus CVanal = RSDanal.

100

Although CV and relative

Thus, since QCproc is added at the start of the overall method and Similarly, QCanal depends only on the final

After substituting and re-ordering these terms, the

equation becomes: 4


Page 5 of 27


CV2proc = RSD2proc – CV2prep – RSD2anal ;

101 102

and since, RSD2prep = CV2prep + CV2anal , (or CV2prep = RSD2prep – RSD2anal),

103 104

then, CV2proc = RSD2proc – (RSD2prep – RSD2anal) – RSD2anal,

105 106

which negates RSD2anal and leads to, CVproc = √(RSD2proc – RSD2prep).

107 108

We use these equations to calculate the CV results reported in this study, further splitting CVprep

109

into its component CVextraction and CVcleanup steps.

110

The purpose of this study was to assess and compare the performances of different sample

111

comminution methods including a relatively new type of vertical cutting chopper (Blixer®) at

112

ambient and cryogenic conditions (with or without dry ice) followed by (or not) use of a cryomill

113

apparatus employing liquid nitrogen.

114

portion for the methods using diverse spiked and incurred analytes.

115

uncertainty of each analytical step was also estimated by spiking different QC standards between

116

each step in the methods, and comparing these results with those from the common practice of

117

using internal standards.

Additionally, we investigated the minimal sample test The measurement

118 119

■ MATERIALS AND METHODS

120 121 122

Chemicals and materials. HPLC-grade acetonitrile (MeCN) was from Fisher Scientific (Pittsburgh, PA; USA).

123

Deionized water (18.2 MΩ-cm) was prepared with a Barnstead/Thermolyne (Dubuque, IA; USA)

124

E-Pure Model D4641.

125

containing 45 mg anh. MgSO4/primary secondary amine (PSA)/C18/CarbonX (20/12/12/1,

126

w/w/w/w) were purchased from ITSP Solutions (Hartwell, GA; USA).

127

anhydrous magnesium sulfate (anh. MgSO4), D-sorbitol, ethylglycerol

128

(3-ethoxy-1,2-propanediol), shikimic acid, and formic acid came from Sigma-Aldrich (Saint

129

Louis, MO; USA).

130

Agency’s National Pesticide Repository (Fort Meade, MD; USA), ChemService (West Chester,

131

PA; USA), or Dr. Ehrenstorfer GmbH (Augsburg; Germany).

For the solid-phase extraction (SPE) cleanup procedure, mini-cartridges Sodium chloride (NaCl),

Pesticide standards were obtained from the Environmental Protection Polychlorinated biphenyls 5



Page 6 of 27

132

(PCBs), polybrominated diphenyl ethers (PBDEs), p-terphenyl-d14,

133

5'-fluoro-3,3',4,4',5-pentabromodiphenyl ether (FBDE 126), and benzo(a)pyrene were from

134

AccuStandard (New Haven, CT; USA).

135

Isotopes (Pointe-Claire, Quebec; Canada), and 13C12-p,p'-DDE and 13C12-PCB 153 originated

136

from Cambridge Isotope Laboratories (Andover, MA; USA).

Atrazine-d5 and fenthion-d6 came from C/D/N

137 138

Quality control standards. Figure 1 displays the experimental design of the study.

139

The targeted analytes included 19

140

incurred contaminants at various concentrations plus 12 analytes spiked at 100 ng/g in the 500 g

141

mixed food sample portions prior to comminution with the Blixer.

142

the QCBlixA, QCBlixC, and QCBlixC+cryo experiments, in which BlixA, BlixC, and BlixC+cryo refer

143

to the ambient Blixer, cryogenic (dry ice) Blixer, and cryogenic Blixer+cryomill sample

144

processing conditions, respectively.

145

glob that did not break into smaller pieces when the cryomill was used, thus the cryomill could

146

only be used with the 22 g cryogenic Blixer portions when they were still frozen.

147

processing with the cryomill only (cryo), the QCproc subset of QCcryo consisted of 3 pesticides

148

(diazinon, pyriproxyfen, and tetraconazole) spiked at 100 ng/g.

These served as QCproc for

The 22 g ambient Blixer subsamples formed a large frozen For sample

Sample preparation involved two steps, QuEChERS extraction (extr) and mini-SPE cleanup

149 150

(ITSP).

151

13

152

tubes also served as QCextr, and 3 pesticides (piperonyl butoxide, carbophenothion, and

153

procymidone) added prior to the cleanup step constituted QCITSP at 100 ng/g.

154

consisted of 100 ng/g p-terphenyl-d14, which was included at 0.88 ng/µL (to yield 100 ng/g in

155

final equivalent sample) in the analyte protectants solution (25 mg/mL ethylglycerol, 2.5 mg/mL

156

gulonolactone, 2.5 mg/mL D-sorbitol, and 1.25 mg/mL shikimic acid) in 2/1 (v/v) MeCN/water

157

containing 0.88% formic acid.15

158

extracts just prior to injection in low-pressure gas chromatography – tandem mass spectrometry

159

(LPGC-MS/MS).

160

To isolate these steps, the 5 internal standards (int. stds), atrazine-d5, fenthion-d6,

C12-p,p'-DDE, 13C12-PCB 153, and FBDE 126, added to test sample portions in the extraction Lastly, QCanal

This solution was added to all calibration standards and final

Table 1 summarizes the analyte types, retention times, and MRM conditions for the 43

161

analytes altogether.

For preparation of calibration standards, a series of working standard

162

solutions containing all 43 analytes were prepared in MeCN. 6


Page 7 of 27


163 164 165

Sample processing. Previously analyzed foods found to contain 19 incurred pesticides and environmental

166

contaminants altogether were selected for this study.

These previously comminuted samples

167

consisted of orange (incurred with imazalil and thiabendazole), kale (p,p'-DDE), apple

168

(diphenylamine and thiabendazole), salmon (p,p'-DDE, hexachlorobenzene, and PCBs), and

169

croaker (DDTs, PCBs, and PBDEs).

170

evaluated:

171

6870D Freezer/Mill® cryomill.

172

spatula, the total volume of the stainless steel Blixer container was 2.8 L, and the

173

polycarbonate-walled cryomill vessel including the stainless steel impactor was 165 mL total

174

volume.

175

working ranges of the food processing devices, which is an important practical matter.

176

sample processing, each of the 5 individual food samples were weighed in equal 100 g portions

177

into the 2 L stainless steel Blixer container, and the 500 g uneven composite was spiked with 1

178

mL QCBlix solution prior to processing for 1 min each with and without addition of dry ice.

Two types of sample processing equipment were

a Robot Coupe (Ridgeland, MS; USA) Blixer® 2 and a Spex (Metuchen, NJ; USA) With the s-shaped serrated chopping blade and lid fitted with a

The 500 and 20 g comminuted sample sizes, respectively, fell well within the practical For

179

When using the cryogenic conditions, the food samples were partially frozen in a -20°C

180

freezer before the dry ice pellets were added; otherwise, the dry ice would sublime too quickly

181

and create an eruption of CO2 out the top hole in the container’s lid.

182

were transferred to clear glass jars, which in the case of the cryogenic Blixing procedure were

183

loosely capped and placed in the freezer for 30 min to allow complete sublimation of the dry ice

184

before subsamples were taken for further cryomilling and/or analysis.

185

cryomilling, a 22 g portion of the frozen sample was weighed into a cryomill vessel, and QCcryo

186

solution was added, being careful not to expose the polycarbonate cylinder directly to MeCN.

187

Cryomilling was conducted using liquid N2 for 3 cycles of 1 min each at 10 beats/s of the

188

impactor.

189

the ambient and cryogenic Blixer samples of different test portion sizes.

The comminuted samples

In the case of

Then, the cryomilled sample was transferred into a glass jar for analysis, along with

190 191 192 193

Sample preparation. Sample preparation was conducted using QuEChERS extraction and automated ITSP cleanup as described previously.15

Different test portions from the three sample processing 7



Page 8 of 27

194

approaches were weighed (5 replicates each) into polypropylene centrifuge tubes.

For the

195

ambient Blixer composite, 20, 10, and 5 g subsamples were weighed into 50 mL tubes, and 15

196

mL tubes were used for 2 and 1 g subsamples.

197

subsamples, plus 0.5 g test portions were weighed into 2 mL mini-centrifuge tubes (this test

198

sample weight was not evaluated for the ambient Blixer conditions).

199

portions consisted of 2 and 1 g in 15 mL tubes and 0.5 g in 2 mL mini-tubes.

200

volumes of QCextr solution was added to yield 100 ng/g of the 5 QCextr analytes.

201

For extraction, 1 mL MeCN per g sample was added to the sample tubes.

The same was done for the cryogenic Blixer Cryomill test samples Appropriate All tubes were

202

then shaken for 10 min at room temperature on a Glas-Col (Terre Haute, IN; USA) platform

203

pulsed-vortex shaker at the 80% pulsation setting.

204

(w/w) anh. MgSO4/NaCl (0.5 g per 1 g sample) were added to each tube, which were then

205

shaken another 3 min.

206

(4150 rpm) using a Kendro (Osterode, Germany) Sorvall Legend RT centrifuge, and the

207

mini-tubes were centrifuged at 8000 rcf (10,000 rpm) using a Hill Scientific (Derby, CT; USA)

208

mv 13, each for 3 min at room temperature.

209

Afterwards, pre-weighed amounts of 4/1

Then, the 50 and 15 mL centrifuge tubes were centrifuged at 3711 rcf

After centrifugation, 0.6 mL extract (supernatant) for the 20, 10, 5, 2, and 1 g test samples,

210

and 0.3 mL for the 0.5 g subsamples, were transferred to brown-glass autosampler vials, and

211

QCITSP was added prior to automated ITSP cleanup as described before.15

212

volumes added to the cartridges at 2 µL/s flow rate yielded ≈220 µL volumes in the receiving

213

autosampler vials containing microvial inserts, to which 25 µL QCanal solution with analyte

214

protectants was added.

215

final volumes (and ng/g equivalents) as the 7 calibration standards.

216

was not used in this study due to the presence of incurred pesticides and a lack of blank matrix.

217

To encompass the expected analyte concentrations, calibration standards were 0, 1, 4, 16, 64, 256,

218

and 1024 ng/g for the 100 ng/g spiked analytes, and 40-fold less at each concentration for PCBs,

219

PBDEs, diphenylamine, and hexachlorobenzene, 4-fold less for DDDs/DDTs, half as much for

220

imazalil and thiabendazole, and twice as high for the DDEs.

221

p-terphenyl-d14 were fixed at 100 ng/g equivalent concentrations in all calibration standards.

The 300 µL extract

Another 25 µL MeCN was added to all samples to yield very similar Matrix-matched calibration

The int. std analytes and

222 223 224

LPGC-MS/MS analysis. All final extracts and calibration standards were analyzed in the same analytical sequence 8


Page 9 of 27


225

using an Agilent (Little Falls, DE; USA) 7890A/7010A gas chromatograph / triple quadrupole

226

tandem mass spectrometer for LPGC-MS/MS analysis.15,16

227

15 m × 0.53 mm i.d. × 1 µm film thickness Phenomenex (Torrance, CA; USA) ZB-5MSi

228

analytical column connected using an Agilent Ultimate union to a 5 m × 0.18 mm i.d. uncoated

229

restrictor/guard Hydroguard column from Restek (Bellefonte, PA, USA).

230

parameters were the same as previously published,15 and Table 1 gives ion transition conditions

231

for the specific analytes.

The separation was achieved on a

The LPGC-MS/MS

232 233

■ RESULTS AND DISCUSSION

234 235 236

Practical aspects The purpose of sample comminution is to obtain an acceptably homogeneous analytical test

237

sample portion that accurately represents the bulk sample that needs to be analyzed for the

238

decision-making purpose.

239

and more ease in sample preparation) is achieved by minimizing the test portion weight as much

240

as feasible.

241

less precision) unless technically advanced equipment and techniques are used with proper

242

procedures.

243

sample sizes than is saved in the subsequent sample preparation and analysis procedures, or the

244

smaller sample size makes sample preparation more problematic, then implementation of the

245

advanced comminution method is not worthwhile.

246

considered as well as performance aspects in the results.

247

Greater laboratory efficiency (lower costs, higher sample throughput,

However, reduced subsample amounts tends to decrease accuracy (more bias and If high-quality sample processing takes more time and labor to minimize test

Thus, practical matters have to be

For example, the easiest sample processing method is to quickly cut bulk food into chunks

248

using a knife, which are then chopped in one step using one processing device at room

249

temperature.

250

≈1 kg sample per person from start to finish (cut bulk sample by knife, add to container, chop for

251

1 min, transfer ≈200 g to a storage jar, and wash the container, lid, and blade between samples).

252

At least two sets of containers should be available per chopping motor to increase efficiency and

253

to allow the containers to fully dry between samples (washing should be done with soap, water,

254

and acetone/isopropanol).

255

to appropriately adjust the QC spiking volume depending on the sample weight, determined by

The Blixer® was evaluated in this study for that purpose, which takes ≈10 min per

When routinely using a QCproc standard, it is much easier and faster

9



Page 10 of 27

256

placing the container on a tared balance, than it is to weigh an exact bulk sample amount for a

257

fixed spiking volume.

258

In the case of cryogenic sample processing using dry ice in the Blixer, the food samples must

259

first be cut into reasonably small chunks (requiring more care than needed for room temperature

260

operations), placed into a large beaker, plastic bag, or wrapped in aluminum foil, and put into a

261

freezer for >30 min prior to comminution.

262

large batch of samples.

263

containers, spiked with the QCproc, and chopped for 1 min using about 25 pellets (≈1 cm3 each)

264

of dry ice (which increases reagent needs and cost).

265

same amount of time as the room temperature protocol, but an additional amount of time is

266

needed for the dry ice to sublime.

267

that ≈30 min is needed before the test portions should be weighed.

268

per person is needed in this case for a batch of 20 samples, whereas the batch could be prepared

269

in

Use of an efficient measurement uncertainty approach to compare

Recommend Documents