Multianalyte analysis of volatile compounds in virgin olive oils using SPME-GC with FID or MS detection: results of an international interlaboratory validation

Received: 18 April 2023 Revised: 12 September 2023 DOI: 10.1002/ejlt.202300079 RESEARCH ARTICLE Multianalyte analysis of volatile compounds in virgin olive oils using SPME-GC with FID or MS detection: results of an international interlaboratory validation Diego L. García-González1Enrico Casadei2Ramón Aparicio-Ruiz1,3 Clemente Ortiz Romero1Enrico Valli2Paul Brereton4Anastasios Koidis4 Martyna Korytkowska4Maurizio Servili5Roberto Selvaggini5Florence Lacoste6 Julien Escobessa6Stefania Vichi7Beatriz Quintanilla-Casas7Alba Tres7 Pierre-Alain Golay8Paolo Lucci9Erica Moret10 Alessandra Bendini2 Tullia Gallina Toschi2 1Instituto de la Grasa (CSIC), Sevilla, Spain 2Department of Agricultural and Food Sciences, Alma Mater Studiorum - Università di Bologna, Bologna and Cesena, Italy 3Department of Analytical Chemistry, Universidad de Sevilla, Sevilla, Spain 4Institute for Global Food Security, Queen’s University Belfast, Belfast, UK 5Department of Agricultural, Food and Environmental Sciences, Università degli Studi di Perugia, Perugia, Italy 6ITERG (Institut des Corps Gras), Canejan, France 7Departament de Nutrició, Ciències de l’Alimentació i Gastronomia, Campus de l’Alimentació de Torribera, Universitat de Barcelona, Santa Coloma de Gramenet, Spain 8Nestlé Research Center, Lausanne, Switzerland 9Department of Agricultural, Food and Environmental Sciences, Università Politecnica delle Marche, Ancona, Italy 10Department of Agri-Food, Animal and Environmental Science Università degli Studi di Udine, Udine, Italy Correspondence Prof. Enrico Valli, Piazza Goidanich, 60, 47521 Cesena (FC), Italy. Email: enrico.[email protected] Funding information Horizon 2020 European Research; European Commission within the Horizon 2020 Programme (2014–2020), Grant/Award Number: 635690 Abstract Theorganolepticassessment(Paneltest)istheonlyprocedurewithintheofficialmethods for determining the quality of virgin olive oils that involves an expert panel. There is an urgent need for analytical methodology that can reliably measure volatile compounds in virgin olive oils that is capable of supporting and anticipating the official Panel test. For this reason, a new method based on solid-phase microextraction–gas chromatography with the choice of two possible detectors (FID or MS) was subjected to a large international interlaboratory validation study. The study involved a twostage process: first, a pretrial phase in which 7 participants were exposed to the method for the first time to identify any initial problems with the methodology; then, Abbreviations: CAR, Carboxen adsorbent; DVB, divinylbenzene; EV, extra virgin olive oil; FID, flame ionization detector; IS, internal standard; L, lampante olive oil; LOD, limit of detection; LOQ, limit of quantification; LRI, linear retention indexes; PEG, polyethylene glycol; SMs, standard mixtures; SOP, Standard Operating Procedure; SPME, solid-phase microextraction; V, virgin olive oil; VOC, volatile organic compounds; VOO, virgin olive oil. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2023 The Authors. European Journal of Lipid Science and Technology published by Wiley-VCH GmbH. Eur. J. Lipid Sci. Technol. 2023;2300079. wileyonlinelibrary.com/journal/ejlt 1of13 https://doi.org/10.1002/ejlt.202300079 2of13 GARCÍA-GONZÁLEZ ET AL. a formal validation stage (trial proper), which involved 20 laboratories from Europe, USA, Japan and China. The performance of the different detectors was investigated. While both methods have advantages, the method using FID provided better results for 11 compounds, in terms of reproducibility, compared to MS. This information will allow to implement the method with accurate information of the method performance depending on the detector used. Practical applications: This study provides information from an interlaboratory validation of a method for measuring volatile compounds in virgin olive oils conducted with laboratories (from industry and academia) working in the olive oil sector. The information on the expected analytical errors in the determination of each volatile compound is necessary to apply this method for supporting the official Panel test (sensory analysis). The SPME-GC-MS/FID methods proposed in this work can be used for the internal quality control of a company/distributor/quality control laboratory and could also be used in cases of difficult/contradictory organoleptic assessment, or to confirm results from sensory panels in cases of disputes/disagreement (Reg. EU 2022/2105). KEYWORDS collaborative trial validation, sensory analysis, SPME-GC-FID/MS, virgin olive oil, volatile compounds 1INTRODUCTION The sensory quality of virgin olive oil (VOO) is one of the most distinctive properties of this oil compared to other edible oils, and for that reason quality characteristics are carefully controlled.[1,2] by international regulations and trade standards that facilitate international trade and attempt to ensure the quality of the oil from production to the consumer.[3–7] The trade standards are specified by the International Olive Council (IOC),[8] which establishes the description of the corresponding methods. In addition other national and international regulatory frameworks also establish specific provisions for VOO quality control (e.g., EU, USDA, CODEX).[3,9–11] Among the methods for controlling quality, the organoleptic assessment (sensory analysis) is one of the most crucial.[12] An oil will be formally attributed as VOO due to the Panel test identifying a sensory defect even though the rest of the chemical and physical–chemical parameters are within extra virgin (EVOO) category.[4,13,14] In this context, a recent report from the European Union (EU) highlighted that the marketing of VOO sold as EVOO is the most widespread type of non-compliance identified with respect to the organoleptic characteristics of olive oils that are declared as “extra virgin olive oil.”[15] Furthermore, this report points out some key issues in the organization and performance of organoleptic assessment, which has led to the search for alternative analytical approaches to support the official Panel test. As a result, the EU H2020 OLEUM research project developed a new method for measuring volatiles based on solid-phase microextraction-gas chromatography (SPME-GC) with the choice of two alternative detectors: flame ionization detector (FID) or mass spectrometry (MS).[16–18] The method resulted from a previous investigation into sources of analytical errors that focused on reducing errors resulting from the quantification procedure.[16,17] This procedure consisted of developing calibration curves by using two mixtures of standards. The composition and concentration of the volatile compounds included in these mixtures were optimized to take into account separation/coelution and concentration of the compounds in VOO. The method underwent an initial peer study to obtain initial information about the performance parameters of the method, such as linearity, repeatability, reproducibility, recovery, and limits of detection (LOD) and quantification (LOQ).[16,17] Thepeer study revealedthatthenormalization of the chromatographic areas by the selected internal standard (IS) (4-methyl-2-pentanol) introduced in the calibration curves produced similar (e.g., 23.04% vs. 25.83%forhexanaldeterminedbyMS)orbetterreproducibilityresults (e.g., 19.18% vs. 30.57% for (Z)−3-hexenyl acetate) compared with the calibration of the chromatographic areas conducted without IS normalization.[16,17] Following the peer study, the method was then subjected to a formal international validation (collaborative trial) with laboratories from industry or other areas working in olive oil sector, some with limited experience ofconductingamultianalyteanalysisof organic compounds (VOCs) as it is currently not a requirement in the EU regulation; that is, there no methods for the headspace analysis of VOO for assessing sensoryquality,[8] and thereforethisanalytical technique is considered as new procedure for the laboratories in the study.[19] Akeypartof the method focuses on avoiding evaporation of VOCs or contamination with other volatile molecules present in the environment. In addition, the method includes a GC system adaptation: new injection configuration, new column, new procedure, and procedures for identifying the VOCs on the chromatograms, that are all relatively novel for analysts 14389312, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejlt.202300079 by Universidad De Sevilla, Wiley Online Library on [23/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License GARCÍA-GONZÁLEZ ET AL.3of13 less experienced in the analysis of VOCs.[20–22] The lack of experience may play a relevant role in introducing more errors so it is import to address this factor when designing the validation scheme while ensuring that an internationally recognized process is followed.[23] In this research work, the study and statistical processes were carried out in compliance with ISO 5725,[24–27] a well-established internationally agreed method validation procedure.[25] With this purpose, an international validation was launched in 2020 following the Protocol for the Design, Conduct and Interpretation of Method-Performance Studies, which is compliant with ISO 5725.[25] The study ran between June 2020 and January 2021. In total, 20 laboratories from Europe, the United States, Japan, and China took part in the study. The method was supplied to these labs in an internationally accepted form (ISO format).[28] A large proportion of participants were involved in olive oil analysis, as such the results of the validation study reflect the results from real end-users. 2MATERIALS AND METHODS 2.1 Chemicals Pure standards of 18 VOCs analyzed were purchased from Merck (Darmstadt, Germany). These were used to prepare two standard mixtures and to prepare the calibration curves. The IS (4-methyl-2- pentanol, purity ≥98%) and the mixture of n-alkanes from 8 to 20 carbon atoms (∼40 mg L–1 each, in n-hexane) for the calculation of the linear retention indexes (LRI) were purchased by each participating laboratory. 2.2 Samples Agroup of 10 “blind samples,” comprising fivepairedsamplesrandomly numbered, were used in the trial proper (see Section 2.6). These samples were selected to ensure 18 VOCs were included in the study and at relevant concentrations. This selection took place after analysis of a range of commercial filtered samples obtained from different producers in order to check for their volatile composition and their suitability for the validation study. In addition, the sensory analysis (Panel test) was carried out by the professional committee of the University of Bologna and all samples were classified according to the commercial category (extra virgin, virgin, and lampante). Those samples in which many VOCs were absent (not detected) were not selected as samples for the trial proper (Table 1). 2.3 IS solution The IS solution was prepared as described by Casadei et al.[16] The IS (4-methyl-2-pentanol) was diluted in freshly refined olive oil in order to have an approximate concentration of 50 mg kg–1. 2.4 Gas chromatographic analysis 2.4.1 SPME-GC-FID analysis SPME-GC-FID analysis was carried out according to Casadei et al.[16] An Excel file was sent to the participant labs (see Supplementary Information) in which the calculations were carried out automatically, just requiring the laboratories to input the exact weights and chromatographic areas (see Supplementary Information), thus reducing the likelihood of error in such calculations. The vial was closed with a septum (polytetrafluoroethylene) and was left for 10 min at 40◦C under agitation to allow for equilibration of the VOCs in the headspace. The SPME fiber was exposed to the sample headspace for 40 min at 40◦C. The volatiles adsorbed by the fiber were thermally desorbed in the hot injection port of a GC for 5 min at 250◦C with the purge valve off (splitless mode) and injected into a capillary column of a gas chromatograph with FID detector. The analysis allowed the use of an auto sampler or manual injection. The capillary column was of a polar phase based on polyethylene glycol(PEG),length60m, internal diameter 0.25 mm,andcoating0.25– 0.50 µm. The transfer line temperature was set at 260◦C. The carrier gas used was open to helium or hydrogen if the lab facility was configured for that. Although hydrogen is the most efficient and economical carrier gas for GC, carbon-carbon double bonds may be hydrogenated in the hot GC injector if introduced by SPME coated with divinylbenzene (DVB) polymer or Carboxen (CAR) porous particles. During this validation process, parallel analyzes were made (both with hydrogen and helium as carrier gas) and no artifacts were detected.[29,30] The oven temperature was held at 40◦C for 10 min and then programmed to increase by 3◦Cmin –1 to a final temperature of 200◦C. A cleaning step was added at the end of the oven programmed temperature by all participants (20◦Cmin –1 to 250◦C for 5 min) to ensure that the column was ready for the next analysis.[16] 2.4.2 SPME-GC-MS analysis SPME-GC-MS analysis was carried out according to Aparicio-Ruiz et al.[17] The calculation of the concentrations was also made with the same Excel file mentioned in the previous section (see Supplementary Information). 2.5 Identification and quantification of VOCs Linear Retention Index (LRI) and standards were used for identification[16] in addition to mass spectrometry.[17] The quantification procedure was based on the calibration curves of the 18 selected volatile compounds built as linear regression (intercept equal to 0) corrected by the chromatographic areas of the IS as described by Casadei et al.[16] and Aparicio-Ruiz et al.[17] Thus, the following equation was used: AAnalyte/AIS =m⋅CAnalyte,whereAAnalyte is the area 14389312, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejlt.202300079 by Universidad De Sevilla, Wiley Online Library on [23/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 4of13 GARCÍA-GONZÁLEZ ET AL. TABLE 1 Information and sensory characteristics of the samples (VOO) considered for the selection in the pretrial and trial proper studies. Characteristics Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7 Sample 8 Sample 9 Pretrial code M4A007 M4A008 – – – – – – Trial proper code TP3, TP11 –TP2, TP4 –TP10, TP12 TP6, TP7 –TP1, TP8 – Quality grade Virgin olive oilaLampante olive oil Extra virgin olive oil Lampante olive oil Lampante olive oilbVirgin olive oila Virgin olive oil Virgin olive oil Lampante olive oilc Sensory characteristics Fusty/muddy sediment, fermentative defects and slightly rancid Winey/vinegary, fusty/muddy sediment and brine No sensory defect detected Fusty/muddy sediment, winey/vinegary, musty/humid/earthy Fusty/muddy sediment, winey/vinegary, musty/humid/earthy, slightly rancid Rancid Slight rancid Winey/vinegary Fusty/muddy sediment, winey/vinegary, musty/humid/earthy, rancid No. of nondetected compounds 2233 0 0431 aNear lampante olive oil. bNear virgin olive oil. cOil resulted from the mixture of 10 virgin olive oils from the three quality grades (extra virgin, virgin, and lampante). corresponding to the analyte, AIS is the area corresponding to the IS, and mis the slope of the calibration curve (built for the selected analyte). The calibration curves were prepared using standard mixtures (SMs), as reported by Casadei et al.[16] The two mixtures, coded as SM- A and SM-B, were prepared to have a concentration of 10 000 mg kg–1 for each VOC and were supplied to participants for s subsequent dilutions, coded as SM1 (200 mg kg–1), SM2 (20 mg kg–1), and SM3 (2 mg kg–1). Thus, the calibration started from the same materials avoiding need for each participant to purchase 18 standards of the volatile compounds. The calibration solutions were prepared by the participants to have 0.05, 0.10, 0.15, 0.20, 0.25, 0.5, 1.00, 1.50, 2.00, 2.50, 5.00, and 10.00 mg kg–1 for the compounds included in SM-A and from 0.20 to 10 ppm (adding three points at 15.00, 20.00, and 25.00 mg kg–1) for the compounds included in SM-B. The concentration values were higher in the latter since some of the compounds in this mixture are present at higher amounts in VOO. 2.6 Collaborative study design (pretrial and trial proper) The design of the collaborative study involved two consecutive phases, namely, the initial (“pretrial”) and the formal validation stage (“trial proper”). Figure S1 shows the general validation scheme used in the OLEUM project. The initial pretrial phase took place in 2019. The objective of the pretrial was a two-way knowledge transfer: (1) to allow the participants to familiarize with the protocol (called Standard Operating Procedure or SOP) as well as with the reporting procedure (Excel file shown in Supplementary Information) and to provide important information regarding deviations from the provided protocol by the participantsand(2) to obtain end-user feedback toimprovethewritten procedure (SOP). A key part of the pretrial process was the two-way hands-on knowledge transfer workshops where participants shared their experiences with the developers and provided feedback on the method so that any errors could be corrected, and changes could be made to make the method easier to understand and use. The resulting workshop in Bologna (Italy)[31] was well attended, with over 70 laboratories and stakeholders in the olive oil sector participating in the event. A timeframe of 3 months from the launch date of each phase was granted for submitting the results of the requested analyses. The formal collaborative trial (trial proper) ran between June 2020 and January 2021. At the start of each phase of the collaborative trial, the participant laboratories were provided with Excel electronic spreadsheets for reporting the results (Supplementary Information), as well as with the appropriate SOP. Any deviation from the protocol had to be reported to the organizers of the study. Twenty laboratories from 9 countries (Europe, the United Kingdom, the United States, China, and Japan) took part in the trial proper and received 10 test materials comprising 5 sets of individually numbered blind duplicates. Twelve laboratories (4 from Italy, 3 from Spain and 1 each from France, Slovenia, China and Japan) submitted valid (compliant) results for the MS procedure and 8 laboratories for FID 14389312, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejlt.202300079 by Universidad De Sevilla, Wiley Online Library on [23/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License GARCÍA-GONZÁLEZ ET AL.5of13 (5 from Italy, 2 from Spain, and 1 from Slovenia). The collaborative study was designed according to the harmonized protocol for method validation of the International Union of Pure and Applied Chemistry (IUPAC).[25] The criteria that must be met for the protocol are as follows: (i) a minimum number of 8 laboratories that deliver valid results; (ii)use a minimumnumber of 5test materials thatare different in terms of analyte and/or matrix concentration; and (iii) finally undertake the statistical analysis to calculate method performance parameters after rejection of outliers. 2.7 Data processing and statistical analysis Prior to the statistical analysis of the results, the reported data and metadata were assessed for compliance, that is, whether the laboratories strictly followed the designated SOP. The assessment of noncompliance was based on (1) comments in the Excel reporting spreadsheet (Supplementary Information) and/or subsequent dialogue with participants and (2) clear gross errors in the analysis/noncompliance assessed. After exclusion of noncompliant results, the data were processed according to the harmonized protocol, which includes stepwise assessment of statistical outliers using Cochrans and Grubbs tests respectively.[25] Method performance parameters were then calculated: (i) relative standard deviation for repeatability (RSDr%), which is a primary estimate of precision when the method is applied within a laboratory; (ii) relative standard deviation for reproducibility (RSDR %), which is an estimate of the precision when the method is applied between laboratories; and (iii) Horrat value (HoR), defined as the ratio between the reproducibility standard deviation and the Horwitz function (SR/H). Horrat values in the interval 0.5 <HoR <2.0 indicate that the precision is within expected values.[32] Additionally, the repeatability and reproducibility limits were calculated. The terms “repeatabilitylimit” and “reproducibilitylimit” were applied specifically to a probability of 95% and they were taken as 2.8 ×Sr(Sr, standard deviation of the repeatability) and 2.8 ×SR(SR, standard deviation of the reproducibility), respectively as specified in the harmonized protocol.[24] Data processing, calculations, and Student’s t-test of paired samples (p<0.05) were carried out with Microsoft spreadsheet program 2016 (Microsoft Corp., Redmond, WA). Analysis of variance (p<0.05) was carried out with Statistica (StatSoft, Tulsa, OK). 3RESULTS AND DISCUSSION 3.1 Sample selection The previous evaluation of the SPME-GC-FID/MS[16,17] method showed that the selection of the validation samples is critical since these oils should contain all the relevant 18 VOCs as they are markers of different sensory defects or fruitiness, as positive attributes. Furthermore, a natural oil (real VOO sample) could not present all the sensory defects at the same time. For that reason, 9 samples were selected that represented either complex aromas or different sensory defects, and, therefore, they were likely to show many if not most of 18 VOCs. The VOCs were analyzed by SPME-GC-FID and the number of nondetected compounds were evaluated to select those 5 samples to be used in the pretrial and trial proper phases. Table 1 shows those 5 samples selected, which presented most of the volatile compounds studied. This selection was also made in order to balance the samples set not only in terms of commercial category (EV, V, L) but also in terms of sensory defects. Sample 9 only presented one nondetected VOC (3-methyl-1-butanol). This sample was a lampante olive oil resulting from the mixture of 10 VOOs from the three quality grades—extra virgin, virgin, and lampante—with different sensory defects. However, this sample was finally not selected because its aroma was considered nonrepresentative of a real VOO. 3.2 Pretrial phase The pretrial phase allowed for the identification of potential unforeseen problems. One of the critical issues was the possible alteration of the volatile fraction during transport and distribution to the labs, since external variables (e.g., light and temperature) may alter the volatile profile,[2,33] thereby introducing errors that are not associated with the method performance. In order to investigate stability during transportation a pilot exercise was carried out by shipping the two pretrial samples from Seville (Spain)—York (UK)—Seville (Spain) and analyzing the VOCs before and after the transportation to identify possible stability issues. No significant differences were found between the concentrations obtained before and after the shipping. Table 2 shows the results of the pretrial, which were computed from 7 labs (2 with FID and 5 with MS). Since the number of labs using FID and MS detectors were low for a separate analysis for each detector the results of all the labs were studied together regardless of type of detector used. The analyses of outliers determined that one of the labs presented the half of the values (9 out of 18) as outliers and as a consequence it was removed from the data set. It was established that a technical problem had occurred in the sample analysis which explained the anomalous values. The other outliers removed from the data set corresponded to cases of wrong identification or a clear problem in the chromatogram integration that resulted in anomalously high/low chromatographic areas. No errors were identified in the calibration curves, provided by participants in their Excel data sheets. As a result of the pretrial, and the following knowledge transfer workshop, a revision of the SOP was made that provided a more detailed description of the method with an example chromatogram for each product category to more easily facilitate identification. Furthermore, a statement about the importance of applying the same integration procedure, in both calibration and sample chromatograms, was added to the written method. The resulting relative standard deviations (RSD%) were in accordance with those published regarding the peer study of the method carried out within OLEUM project[16,17] for some compounds: octane, ethyl acetate, ethanol, 14389312, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejlt.202300079 by Universidad De Sevilla, Wiley Online Library on [23/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 6of13 GARCÍA-GONZÁLEZ ET AL. TABLE 2 Mean concentrations (mg kg–1) and standard deviations (SD) resulted from the pretrial study carried out with two samples (coded as M4007 and M4008). Code Volatile compounds M4007 M4008 Previous interlab study1 Mean ±SD RSDR%Mean±SD RSDR%RSD R%FID RSD R%MS 1 Octane 2.812 ±0.313 25.9 2.874 ±0.349 28.0 12.0 38.5 2Ethyl acetate 2.568 ±0.664 23.0 5.723 ±1.600 25.9 18.2 28.2 3 Ethanol 0.730 ±0.168 24.7 18.220 ±4.715 25.4 35.7 32.3 4Ethyl propanoate 9.590 ±2.368 112.3 30.007 ±7.607 49.7 122.0 39.0 5 Hexanal 0.012 ±0.014 23.1 0.382 ±0.190 33.7 28.0 23.0 63-Methyl-1-butanol 2.428 ±0.560 51.6 1.236 ±0.417 15.0 23.1 26.0 7(E)−2-hexenal 0.199 ±0.103 14.5 7.774 ±1.169 128.0 30.1 19.6 8 (Z)-3-hexenyl acetate 8.855 ±1.282 22.9 0.151 ±0.194 64.8 32.8 19.2 9(E)-2-Heptenal 1.094 ±0.251 37.9 0.209 ±0.136 48.7 26.0 24.9 10 6-methyl-5-hepten-2-one 0.423 ±0.160 33.9 0.195 ±0.095 24.2 47.8 43.2 11 1-hexanol 0.047 ±0.016 65.6 0.239 ±0.058 37.1 48.1 13.3 12 Nonanal 1.322 ±0.867 16.3 2.341 ±0.867 41.4 44.2 46.1 13 1-octen-3-ol 3.693 ±0.602 15.8 8.272 ±3.427 55.7 37.2 31.5 14 (E,E)−2,4-hexadienal 0.038 ±0.006 44.7 0.188 ±0.105 104.9 39.3 63.5 15 Acetic acid 0.051 ±0.023 20.1 0.189 ±0.198 21.7 44.8 17.5 16 Propanoic acid 3.815 ±0.768 36.6 17.257 ±3.735 34.2 21.4 26.7 17 (E)−2-decenal 0.126 ±0.046 74.0 1.072 ±0.367 55.8 57.8 36.7 18 Pentanoic acid 0.474 ±0.350 30.0 1.391 ±0.777 39.0 29.7 27.1 Note: The relative standard deviation for reproducibility (RSDR%) resulted from this study (regardless the detector) and from two previous studies carried out with labs with experience in VOC analysis (FID and MS separated) are also shown. hexanal, 6-methyl-5-hepten-2-one, acetic acid, propanoic acid, and (E)−2-decenal. Other compounds showed higher RSD% although only in one of the two samples, highlighting a probable effect of the sample (their whole volatile profile) or concentration as was the case for 3-methyl-1-butanol, (E)−2-hexenal, (Z)−3-hexenyl acetate, 1-hexanol, 1-octen-3-ol, (E,E)−2,4-hexadienal, and (E)−2-decenal. The fact that the volatile profile of the sample and the concentration of volatile compounds apparently showed some effect highlights the importance of integration procedure, which may be affected by the signal around the analyte peak. Alternatively, the low number of participants (7) in the pretrial, and the fact that the data comprised results from both FID and MS detector systems could explain some of the variability in results. The differences observed with respect to the previous evaluation of the method[16,17] could be explained by incorrect identification and/or integration of the peaks, which supports the idea of introducing changes in the protocol to guide the identification procedure and to give importance to the chromatographic integrators and data systems.[34] It has been reported, that a manual integration carried out on the same chromatogram by 4 different analysts may lead to a maximum variation (RSD%) of 7% in the determined areas.[16] According to the results and information acquired in the pretrial, there were two key sources of errors: (1) calibration and (2) identification/integration. One of them arose from the calibration procedure (from the sample analysis, integration, or the calibration curve). When calibration was the source of error for the concentration data of one VOC, atypically high or low values were observed but the magnitude in relation to other samples was similar to the results from other laboratories, for example, if one sample contained higher concentration than another, in that laboratory their results were still in a similar proportion to those of from other laboratories. The other type of error was due to failings in the identification/integration, not surprising given the complexity of VOO aroma.[21,22,35] In this case, the order of the samples with respect to the magnitude of the concentration was dissimilar to other laboratories with incorrect peaks being quantified. This error was expected to be more frequent when FID detector was used due to its lack of selectivity. However, some errors in identification were also found in some laboratories working with MS, although usually in compounds at low concentration, or in the cases of coeluting peaks. For example, some identification/integration problems were detected for octane with one laboratory reporting no value for this analyte. Similarly, 3-methyl-1-butanol was apparently problematic, probably due to the overlapping with (E)−2-hexenal, and also because it was observed that the order of elution of this compound could be altered depending on the brand of the capillary column used. Some laboratories did not report values for (E)−2-hexenal and (Z)−3-hexenyl acetate, suggesting a problem in their identification. The conclusion of the pretrial phase and the resulting feedback from participants led to some improvements to the SOP and accompanying documentation that could be summarized as follows: 14389312, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejlt.202300079 by Universidad De Sevilla, Wiley Online Library on [23/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License GARCÍA-GONZÁLEZ ET AL.7of13 1. The Excel data sheet for reporting the results in the trial proper included fields to be completed with all the information regarding the weights for building the calibration curves (Supplementary Information). Thus, introducing all the weights and the chromatographic areas, Excel provided the output in mg kg–1. 2. The information to be introduced in the Excel file allowed the traceability of the quantification method to identify the source of gross errors (e.g., anomalous chromatographic area). 3. The information about the injector liner or sleeve to be used was included in the section of the apparatus in the updated SOP. This information was identified as missing in the previous version used in the Pretrial and it was considered that it should be included. 4. The importance of equilibrating the samples to room temperature before their preparation was also highlighted. Samples must look homogeneous, and no waxes or solid particles should be observed. This new input in the SOP comes from a discussion about how to handle the samples with respect to the room temperature. 5. For the method using SPME-GC-FID, the detector temperature was setat260 ◦C and the desorption temperature in the injector was reduced from 260◦C to 250◦C. 6. The need for applying an adequate and uniform integration method for both calibration solutions and virgin olive oil samples was specified in the SOP since it was discussed that it was one of the error sources in the pretrial. In addition to these improvements, it was decided to provide the labs with a table with the exact concentrations of each volatile compound present in the two standard mixtures (SM-A and SM-B) considering the exact measured weights during the preparation of these mixtures. Thus, using the Excel file, participants could build the calibration curves taking into account the exact concentrations of the specific mixtures that they received. Four lots of SM-A and SM-B (10 vials in each lot) were produced and each lab received one SM-A and SM-B of one of these lots. The actual concentrations are indicated in Table S1. Special attention was also paid to the reproducibility of the preparation of SM-A and SM-B to provide standard mixtures with the lowest variation as possible. The relative standard deviations (RSD%) of the concentrations were in the range of 0.69–8.50, and except for 5 compounds, the RSD% were always lower than 2.81% (Table S1). On the otherhand,whenthequantification was carried out with the exactconcentrations (Table S1) and the generic concentration of 10 000 mg kg–1 for all the compounds (concentrations without considering the exact weights in the preparation), the two resulting concentrations differed in only 1.16% (mean value for the 18 volatile compounds). Thus, the error coming from the differences in the preparation of the standard mixtures was considered to be acceptable. 3.3 Trial proper The trial proper was informed by the results and experience acquired in the pretrial, not only in the method itself, but also in the sample selection,homogenization, and the materials (flasks and vials)used. For example, in thetrial proper,theuse of plastic bottles for refined oliveoil (sent for building the calibration curves) was avoided since some peaks derived from plastic were observed during the analysis of refined olive oil in the pretrial. Table 3 shows the mean concentration values for the validation process of FID and MS methods, which were calculated from 20participants (8 for FID and 12 for MS). The comparisonofconcentrationsobtained between the FIDand MS methods for thedetermination of the 18 VOCs in VOO revealed that, for 13 VOCs, no significant differences in concentration were found (p≤0.05, two tails by Student’s t-test of paired samples) while 5 VOCs showed significant differences in concentration between the two method variants. The 5 compounds were 3-methyl-1-butanol, (E)−2-heptenal, 6-methyl-5-hepten-2-one, propanoic acid, and pentanoic acid, with the concentrations obtained with the FID method being higher than the MS method for these analytes (Table 3). These differences could be related to the different sensitivity of the detectors at low concentrations, for example, (E)−2-heptenal. Interms of repeatability,Table4 shows the RSDr%valuesfortheFID and MS methods from the trial proper. The RSDr%valueswerelower than 10% in all cases except for ethyl propanoate, (E)−2-heptenal, 6- methyl-5-hepten-2-one, nonanal, (E,E)−2,4-hexadienal, (E)−2-decenal, pentanoic acid. These compounds presented RDSr% lower than 20% except for ethyl propanoate, (E)−2-decenal, and pentanoic acid in two samples (Table 4). In general, the RDSr% values were similar between FID and MS detectors, pointing out that the sensitivity of these detectors did not have a large impact on repeatability. No significant differences were found in the RSDr%values(p≤0.05, two tails by Student’s t-test of paired samples) for 17 VOCs (94.4% of the compounds) except for ethyl acetate. In this case, the RSDr% value was slightly higher for FID compared to MS. The mean RSDr% values for the 5 paired samples were 6.7% and 5.0% for FID and MS detectors, respectively. Table 5 shows the reproducibility in terms of RSDR% for FID and MS methods. In this case, there were differences in reproducibility for 50% of the VOCs between FID and MS methods. The reproducibility of these compounds (octane, ethyl acetate, ethanol, 6-methyl-5-hepten- 2-one, acetic acid, and pentanoic acid) were lower for FID method compared to MS while 1-hexanol showed a higher reproducibility value for FID compared to MS. For FID, the compounds with the lowest RSDR%(<30%) were ethyl acetate, ethanol, acetic acid, and octane, while the compounds with high RSDR(>60%) were (E)−2- heptenal, pentanoic acid, (E)−2-decenal, (E,E)−2,4-hexadienal, ethyl propanoate, and 1-octen-3-ol, with the last three compounds having a RSDR%>100%. The results could be related to the low concentration of these compounds in the samples, the limits of quantification, and/or the low recovery for these compounds in FID.[16,17] The values of RSDR% were not associated to the concentration value apparently: the highest RSDR% was not necessarily found in those samples with lower concentration, and a regression analysis between RSDR% values in each compound and the concentration values resulted in R2 below 0.6 in all cases, except for (Z)-hexenyl acetate (0.70), 1-octen- 3-ol (0.91), propanoic acid (0.63), and (E)−2-decenal (0.7) in the case of the method with FID detector. For MS detector, octane (0.98), 14389312, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejlt.202300079 by Universidad De Sevilla, Wiley Online Library on [23/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 8of13 GARCÍA-GONZÁLEZ ET AL. TABLE 3 Mean concentrations and standard deviation (mg kg–1) from the trial proper performance study of the method (SPME-GC-FID/MS) carried out with five paired samples. Volatile compounds Samples 1 and 8 Samples 2 and 4 Samples 3 and 11 Samples 6 and 7 Samples 10 and 12 FID MS FID MS FID MS FID MS FID MS Octane 1.541 ±0.190 1.552 ±0.068 0.162 ±0.019 0.161 ±0.012 0.244 ±0.015 0.219 ±0.018 0.450 ±0.034 0.377 ±0.015 1.249 ±0.084 1.438 ±0.040 Ethyl acetate 2.892 ±0.191 3.421 ±0.152 0.542 ±0.032 0.678 ±0.038 0.448 ±0.019 0.593 ±0.016 0.193 ±0.019 0.232 ±0.016 1.035 ±0.068 1.299 ±0.039 Ethanol 51.519 ±5.745 45.892 ±4.080 0.345 ±0.154 0.460 ±0.058 10.249 ±0.518 13.182 ±0.715 4.558 ±0.215 7.102 ±0.446 13.663 ±0.752 17.138 ±0.409 Ethyl propanoate 0.011 ±0.004 0.009 ±0.001 0.028 ±0.008 0.031 ±0.139 0.014 ±0.003 0.004 ±0.014 0.013 ±0.002 0.005 ±0.013 0.038 ±0.008 0.020 ±0.039 Hexanal 0.457 ±0.049 0.584 ±0.035 0.954 ±0.024 0.584 ±0.042 2.265 ±0.091 1.379 ±0.088 1.786 ±0.092 1.494 ±0.034 1.939 ±0.107 1.784 ±0.047 3-Methyl-1- butanol* 0.507 ±0.016 0.425 ±0.019 0.286 ±0.014 0.215 ±0.010 0.242 ±0.010 0.150 ±0.005 0.128 ±0.010 0.076 ±0.007 1.048 ±0.041 0.877 ±0.032 (E)−2-hexenal 0.111 ±0.011 0.137 ±0.007 2.619 ±0.127 2.746 ±0.150 12.092 ±0.482 10.140 ±0.424 3.222 ±0.104 3.211 ±0.137 3.875 ±0.109 4.714 ±0.109 (Z)−3-hexenyl acetate 0.645 ±0.042 0.810 ±0.078 1.753 ±0.063 2.079 ±0.179 0.641 ±0.042 0.520 ±0.038 0.598 ±0.036 0.666 ±0.038 0.722 ±0.065 0.795 ±0.025 (E)−2-heptenal*0.231 ±0.024 0.194 ±0.015 0.103 ±0.020 0.075 ±0.008 0.296 ±0.016 0.161 ±0.021 0.442 ±0.029 0.298 ±0.018 0.577 ±0.037 0.376 ±0.011 6-methyl-5- hepten-2- one* 0.059 ±0.006 0.051 ±0.003 0.059 ±0.010 0.048 ±0.007 0.080 ±0.008 0.068 ±0.007 0.068 ±0.007 0.058 ±0.007 0.069 ±0.006 0.047 ±0.002 1-hexanol 0.514 ±0.030 0.530 ±0.010 1.727 ±0.055 1.763 ±0.094 1.403 ±0.050 1.298 ±0.058 0.579 ±0.020 0.532 ±0.032 1.121 ±0.057 1.103 ±0.022 Nonanal 2.533 ±0.283 2.839 ±0.405 0.546 ±0.056 0.911 ±0.063 1.064 ±0.107 1.022 ±0.177 0.606 ±0.054 0.648 ±0.046 2.886 ±0.378 2.908 ±0.138 1-octen-3-ol 0.029 ±0.009 0.025 ±0.003 0.017 ±0.005 0.018 ±0.001 0.026 ±0.002 0.023 ±0.004 0.112 ±0.002 0.030 ±0.004 0.040 ±0.006 0.039 ±0.003 (E,E)−2,4- hexadienal 0.202 ±0.048 0.020 ±0.007 1.039 ±0.148 0.112 ±0.012 1.883 ±0.075 0.219 ±0.028 0.395 ±0.027 0.101 ±0.012 0.500 ±0.032 0.294 ±0.055 Acetic acid 10.893 ±1.049 10.402 ±0.906 8.487 ±0.179 7.935 ±0.376 3.521 ±0.179 3.141 ±0.142 1.790 ±0.074 1.515 ±0.036 3.215 ±0.095 3.755 ±0.118 Propanoic acid*0.366 ±0.013 0.310 ±0.019 0.229 ±0.022 0.165 ±0.023 0.162 ±0.007 0.010 ±0.006 0.221 ±0.019 0.124 ±0.005 0.537 ±0.024 0.476 ±0.017 (E)−2-decenal 0.651 ±0.173 0.994 ±0.390 0.292 ±0.112 0.295 ±0.092 0.664 ±0.083 0.522 ±0.072 0.929 ±0.246 0.669 ±0.071 0.829 ±0.150 0.858 ±0.088 Pentanoic acid*0.084 ±0.017 0.049 ±0.012 0.071 ±0.005 0.038 ±0.002 0.051 ±0.010 0.038 ±0.008 0.073 ±0.022 0.041 ±0.003 0.080 ±0.004 0.062 ±0.003 *Significant differences between results from FID and MS at p≤0.05, two tails by Student’s t-test. 14389312, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejlt.202300079 by Universidad De Sevilla, Wiley Online Library on [23/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License GARCÍA-GONZÁLEZ ET AL.9of13 TABLE 4 Results of relative standard deviation of repeatability (RSDr%) for the five paired samples used in the trial proper. Volatile compounds Mean RSDr% pairs RSDr% 1 and 8 RSDr% 2 and 4 RSDr% 3 and 11 RSDr% 6 and 7 RSDr% 10 and 12 FID MS FID MS FID MS FID MS FID MS FID MS Octane 9.5 5.9 12.3 4.4 11.9 7.2 6.0 8.2 7.6 3.9 6.8 2.8 Ethyl acetate*6.7 5.0 6.6 4.4 5.9 5.6 4.3 2.8 9.8 7.0 6.6 3.0 Ethanol 16.4 8.3 11.2 8.9 44.6 12.7 5.1 5.4 4.7 6.3 5.5 2.4 Ethyl propanoate 24.6 27.3 34.9 14.2 28.6 13.5 20.2 36.5 14.7 44.9 22.1 6.6 Hexanal 5.6 5.5 10.8 6.0 2.5 7.2 4.0 6.4 5.1 2.3 5.5 2.6 3-Methyl-1-butanol 5.1 5.4 3.1 4.6 5.0 4.7 4.3 3.0 8.1 9.3 4.0 3.7 (E)−2-hexenal 5.4 4.7 9.6 4.9 4.8 5.5 4.0 4.2 3.2 4.3 2.8 2.3 (Z)−3-hexenyl acetate 5.7 7.7 6.5 9.6 3.6 8.3 6.6 7.3 6.0 5.7 9.1 3.2 (E)−2-heptenal 10.6 9.3 10.6 7.8 19.8 10.7 5.5 12.8 6.6 6.0 6.5 3.0 6-methyl-5-hepten-2-one 11.8 10.4 10.6 6.6 17.0 13.5 10.0 10.2 9.6 11.4 8.0 5.2 1-hexanol 4.1 4.4 5.9 1.8 3.2 5.3 3.6 4.5 3.5 5.9 5.1 2.0 Nonanal 10.1 11.4 11.2 14.3 10.2 6.9 10.0 17.3 8.8 7.2 13.1 4.7 1-octen-3-ol 17.8 12.2 32.6 11.9 27.8 7.5 9.1 15.4 1.7 13.9 14.3 6.5 (E,E)−2,4-hexadienal 12.1 17.9 23.0 35.4 14.4 11.0 4.0 12.9 6.8 12.1 6.4 18.6 Acetic acid 5.2 5.1 9.6 8.7 2.1 4.7 5.1 4.5 4.1 2.4 3.0 3.1 Propanoic acid 6.5 7.4 3.7 6.0 9.4 13.6 4.4 6.2 8.6 3.7 4.4 3.6 (E)−2-decenal 26.1 23.7 26.6 39.2 38.5 31.1 12.6 13.8 26.5 10.6 18.1 10.3 Pentanoic acid 19.3 14.1 20.2 23.7 6.3 5.7 20.0 20.8 30.6 6.2 5.4 4.3 *Significant differences between results from FID and MS at p≤0.05, two tails by Student’s t-test. TABLE 5 Results of relative standard deviation of reproducibility (RSDR%) for the five paired samples used in the trial proper. Volatile compounds Mean RSDR%pairs RSD R%1and8 RSD R%2and4 RSD R%3and11 RSD R%6and7 RSD R%10and12 FID MS FID MS FID MS FID MS FID MS FID MS Octane*27.7 39.1 21.9 32.0 30.5 44.9 23.6 42.2 28.1 41.4 34.6 35.1 Ethyl acetate*15.9 29.1 12.4 27.1 8.5 29.9 12.1 29.5 23.9 32.1 22.6 26.7 Ethanol*23.8 45.4 27.1 45.2 53.5 53.3 8.0 41.0 13.0 46.2 17.2 41.5 Ethyl propanoate 111.3 107.2 150.3 82.7 117.8 158.4 98.8 121.8 92.1 102.7 97.4 70.3 Hexanal 39.7 28.0 20.3 17.3 54.5 33.9 56.7 27.4 35.6 30.1 31.6 31.1 3-Methyl-1-butanol 33.5 57.0 12.9 37.4 23.2 63.0 54.6 70.8 48.5 76.1 28.1 37.8 (E)−2-hexenal 38.7 37.6 61.9 64.7 31.6 31.2 21.5 30.3 29.9 31.3 48.7 30.7 (Z)−3-hexenyl acetate 44.7 38.9 49.9 42.9 35.3 43.4 43.5 32.7 52.0 39.5 42.7 36.1 (E)−2-heptenal 68.5 70.4 58.3 67.2 64.8 74.2 100.0 76.0 58.4 67.3 61.2 67.4 6-methyl-5-hepten-2-one*34.0 74.2 27.1 78.2 29.8 66.2 36.3 82.7 30.9 74.4 45.9 69.3 1-hexanol*54.7 43.1 53.1 44.7 39.6 38.0 59.9 44.8 68.2 47.4 52.6 40.8 Nonanal 57.2 59.4 49.6 66.9 61.5 68.1 55.5 60.0 55.5 55.9 63.8 46.0 1-octen-3-ol 114.2 52.4 108.9 43.0 87.4 45.8 79.9 61.6 208.6 56.1 86.0 55.7 (E,E)−2,4-hexadienal 108.7 103.4 113.2 142.4 126.6 93.3 125.7 79.3 99.0 80.2 79.0 121.6 Acetic acid*25.1 40.0 26.4 43.5 20.3 39.2 23.0 35.7 23.4 35.4 32.6 46.0 Propanoic acid 39.4 74.5 25.5 76.8 38.3 68.4 61.1 80.4 43.6 70.3 28.3 76.8 (E)−2-decenal 88.9 92.3 66.8 94.6 145.2 100.1 78.8 105.5 76.6 71.5 77.3 90.0 Pentanoic acid*74.7 86.5 80.9 82.8 70.3 91.1 77.5 87.1 76.8 94.6 68.0 77.0 *Significant differences between results from FID and MS at p≤0.05, two tails by Student’s t-test. 14389312, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/ejlt.202300079 by Universidad De Sevilla, Wiley Online Library on [23/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License