ANALYST, JULY 1992, VOL. 117 1093 Edible Fats and Oils Reference Materials for Sterols Analysis With Particular Attention to Cholesterol Part 1 Investigation of Some Analytical Aspects by Experienced Laboratories Georges Lognay and Michel Severin Faculty of Agricultural Sciences, Department of General and Organic Chemistry, Belgium Achim Boenke and Peter J. Wagstaffe Community Bureau of Reference, DGXII, Commission of European Communities, Brussels, Be Ig iu m B-5030 Gembloux, 200 Rue de la Loi, B-1049 The work reported here is integrated into a programme organized by the Community Bureau of Reference with the aim of developing edible oil reference materials (RMs) certified for cholesterol content. One vegetable oil (RM 162, a blend of soya and maize oils) and two animal fats (RM 163, a blend of pig and beef fats, and RM 164, an anhydrous milk fat) possessing, respectively, low, medium and high cholesterol contents were chosen for this purpose.The present paper summarizes the analytical conclusions resulting from three interlaboratory trials carried out t o identify and correct for the major sources of random and systematic errors linked t o the protocol and t o the gas-liquid chromatographic analysis. Several improvements t o the methodology, and recommendations, have been proposed for the determination of individual sterols within the certification exercise. The latter will be reported elsewhere. Keywords : Cholesterol determination; edible oils; reference mate rial; collaborative study Edible vegetable oils and animal fats consist mainly of triglycerides, which are responsible for their nutritional and physico-chemical properties.Some other minor lipids, such as sterols and tocopherols, attract the interest of food analysts because they have particular nutritional properties and their composition provides a ‘fingerprint’ of a lipid material. The qualitative and quantitative analysis of sterols is essential for nutrition studies and for identification of lipid mixtures when adulteration is suspected. Reliable determina- tion of cholesterol is particularly important both for food labelling purposes and in dietary studies. A preliminary intercomparison of methods for cholesterol organized within the Community Bureau of Reference (BCR) programme had revealed poor agreement and it was concluded by the participating specialists that there was a need for a series of non-spiked lipid reference materials (RMs) with certified cholesterol contents.’ Several analytical methods, including gravimetric, enzymic and spectrophotometric techniques, have been developed to determine the total sterol content in many foods and biological samples.* Currently, high-performance liquid chroma- tography (HPLC) or gas-liquid chromatography (GLC) is used for the establishment of a sterol profile.The separation of A5 and A7 classes is easily achievable on the analytical and semi-preparative scales by normal-phase HPLC.3 However, reversed-phase columns lead to a more complete resolution of molecular species.4-6 In spite of its simplicity, the technique suffers from insufficient selectivity and sensitivity (the com- mon sterol possesses only isolated double bonds, which exhibit low molar absorptivity, their absorption maxima lying between 200 and 210 nm).Except for the determination of ergosterol, a molecule with a typical cis-diene structure (I,,, = 282 nm), the routine application of HPLC to sterol analysis is still limited.7-9 Capillary GLC is recognized as the method of choice for the qualitative and quantitative analysis of sterols in complex matrices such as fats and oils. The sterols are often present at low concentrations in such samples (from 80 to 1200 mg per 100 g of lipid). Extraction and purification prior to GLC analysis is therefore necessary. The procedure generally adopted can be summarized as follows: the sample is saponified with alcoholic potassium hydroxide, and the unsaponifiable matter (USM) is extracted into diethyl ether or another suitable solvent such as hexane.After washing of the crude extract, the solvent is evaporated and the USM is fractionated into its components by thin-layer chroma- tography (TLC). The sterols are then re-extracted from the gel and their constituents are finally separated by GLC, either underivatized or as their trimethylsilyl ethers. Standardized or official methods based on this principle have been adopted.10-12 On the other hand, a method for the rapid isolation of sterols, involving chromatography on aluminium oxide columns, was developed by Hornberg13 and has recently been standardized. 14 The time-consuming, multi-step proto- col of conventional methods has several potential sources of analytical error due, not only to sample handling, but also to the chromatographic resolution for final quantification.This paper summarizes the analytical conclusions resulting from the initial intercomparison of methods that were carried out to identify and correct for the major sources of random and systematic error. Details of the certification of the content of cholesterol and other sterols in three edible lipid RMs (RM 162, a blend of soya and maize oils; RM 163, a blend of pig and beef fats; and RM 164, an anhydrous milk fat) will be reported elsewhere. Experimental A general scheme of the method used by the participating laboratories for the GLC determination of sterols in fatty material is presented in Table 1.The recovery experiments and the improvement of chromatographic conditions were carried out on RM 162, because this RM possesses a more complex sterolic composition than the two other animal fats (RM 163 and RM 164) selected. In addition, the cholesterol content of RM 162 is relatively low and, therefore, well suited to spiking and recovery experiments.1094 ANALYST, JULY 1992, VOL. 117 Cholesterol was the only sterol for which a standard of high purity was available, and in view of its particular importance for nutritional purposes, the calibration and recovery experi- ments were focused on this molecule. High-purity cholesterol (Sigma; ref. (28667) and betulin (Sigma; B9757) samples, each from the same batch, were sent to the participants in the intercomparison studies to eliminate any variability arising from differences in purity of the compounds obtained from different sources.Intercomparisons First intercomparison The first intercomparison exercise undertaken by seven laboratories, using the method that they considered to be the most accurate, had led to the following observations: (1) almost regardless of the method used, the preliminary tests had highlighted poor agreement between laboratories with respect to the absolute values;' (2) the internal standards (ISTDs) were not added sufficiently early in the procedure to control losses at every stage; and (3) examination of RM 162 by gas chromatography-mass spectrometry (GC-MS) con- firmed that TLC clean-up was necessary to remove interfering compounds such as a-tocopherol (co-eluted with cholesterol) or triterpene alcohols and 4-methylsterols, which could interfere with several phytosterols.15 A TLC step was, therefore, strongly recommended for the analysis of vegetable oils.This conclusion is in line with Homberg and Bielefeld's observations. 16 Second intercomparison The second study, in which seven laboratories participated, was designed to eliminate the more serious of the suspected sources of errors appearing during the extraction and the Table 1 General scheme for the determination of sterols in fatty materials Saponification of the fat Extraction of unsaponifiable (USM) into diethyl ether Water washings of the extract Fractionation of USM (preparative TLC) Isolation of sterols from the gel Derivatization (TMS) Capillary GLC analysis J..1 1 1 .1 J. calibration of the cholesterol determination. The protocol to be followed consisted of three distinct, but related, stages: (1) the determination of response factors for cholesterol trimethylsilyl ether (betulin = 1 .OO) at three concentration ratios (0.5, 1 and 2); (2) a check on the entire sterol determination method by analysis of prepared solutions of cholesterol and ISTDs, taking these solutions into the whole procedure, i.e., 'saponification', extraction, washings, TLC and GLC; and (3) the determination of recovery of cholesterol added to RM 162 unsaponifiable. In the second and third steps of analysis, cholesterol concentrations were calculated from the response factors obtained in step 1. The choice of the methodology for the treatment of the samples was left to the participant.The conditions, summarized in Tables 2 and 3, varied widely. Joint evaluation of the results at a meeting of participants, summarized in Table 4, led to the following observations and conclusions. (1) All the participants used betulin as ISTD. (2) The study of response factors demonstrated adequate linearity over the range of interest for cholesterol in oils and fats. Most laboratories found results very close to unity [mean of means k 1 standard deviation (SD) = 1.028 k 0.056, p = 211, implying near-similarity in GLC and flame-ionization detec- tion behaviour of cholesterol and betulin trimethylsilyl ethers. The results of Laboratory 2 were around 1.1 and although no reason was found for this, it was a consistent value in that laboratory.(3) The recovery of cholesterol from simple solutions was quantitative (mean of means k 1 SD = 100 k 5%, p = 21). Extreme minimum and maximum values were 90% (Laboratory 3) and 108% (Laboratory 19). (4) In general, recoveries close to 100% (mean of means L 1 SD = 100 L 4%, p = 21) were achieved, implying good control of errors following the saponification-extraction process. The lower results from Laboratory 3 suggested a systematic bias; on the other hand, the higher values of Laboratory 19 were explained by a partial TLC separation of the betulin and cholesterol bands, which increased the risk of losses in the re-extraction process. The agreement for low cholesterol contents of unspiked RM 162 remained very poor.Indeed, the relative standard deviation (RSD) within laboratories (repeatability) varied from 2 to 31%, and the RSD between laboratories (reproduci- bility) was 60%. This was probably because the method was calibrated for high levels of sterols. The high results of Laboratory 2 (14 _+ 1 mg per 100 g) were excluded for the statistical evaluation. In view of the good recoveries of cholesterol after saponifi- cation (Table 4), it was concluded that a substantial part of the analytical errors arose during the initial saponification/extrac- tion stages. Further work was, therefore, concentrated on this point. Table 2 Second interlaboratory trials. Analytical conditions for saponification and extraction Laboratory Saponification conditions Extraction of the USM code for 5 g of sample into diethyl ether Eluent for TLC clean-up* 2 t 3 Not specified 1 mol dm-3 ethanolic KOH (50 cm?), 1 h at 80 "C 5 7 11 1st 19 50 cm3 ethanol + 10 cm3 KOH, 10% for 30 min 2 mol dm-3 methanolic KOH (20 cm3), 1 h at 80 "C 2 mol dm-3 methanolic KOH (100cm3), 1hat90"C 1 mol dm-3 ethanolic KOH (50 cm3) for 1 h 20 cm3 ethanol + 5 cm3 KOH (60%), 15 rnin at 80 "C 3 X 100 cm3 (15 rnin each) Chloroform-diethyl ether Not specified Hexane-diethyl ether-acetic acid 1 X 250 cm3 (2 min) 1 X 50 cm3 (30 min) 1 x 100 cm3, 2 x 50 cm3 3 X 100 cm3 (1 rnin each) Chloroform 1 X 100 cm3, 2 X 50 cm3 (90 + 10, v/v) (60 + 40 + 1, v/v/v) Chloroform Chloroform-diethyl ether- Hexane-diethyl ether ammonia (90 + 10 + 0.5, v/v/v) (70 + 30, v/v) (1 min each) Chloroform-diethyl ether- (1 min each) ammonia (90 + 10 + 0.5, v/v/v) Recovery of the sterols from the silica gel Hot chloroform (X 3) Diethyl ether (20 cm3 for 2 min) Diethyl ether Diethyl ether (x2) Chloroform ( x 1) Chloroform-diethyl ether Diethyl ether (7+3,v/v)(x5) Diethyl ether (5 cm3) * All TLC was performed on silica gel Si 60. t Standardized methods were used by Laboratory 2 (AFNOR) and Laboratory 15 (IUPAC).ANALYST, JULY 1992, VOL.117 1095 ~~~~~ Table 3 Second interlaboratory trials. Analytical conditions for GLC Laboratory Injector code Derivatization* Column Temperature OV-1701t (30m x 0.33 mm i.d., dfS = 0.3 pm) Isothermal (280°C) CPSIL-19$ (25 m x 0.2 mm i.d., df = 0.22 pm) Isothermal (250 "C) CPSIL-19 (25 m x 0.32 mm i.d., df = 0.22 pm) Isothermal (260 "C) 2 3 5 7 Pyridine-BSTFA + 1% SE-521(25 m x 0.32 mm i.d., df = 0.15 pm) 70-220°C (20°C min-1) On column BSTFA-TMCS (80 + 20), Trisil, 10 min at room Pyridine-HMDS-TMCS Moving needle Split Split 30 min at 60 "C temperature (9 + 3 + 1) TMCS (1 + 1) TMCS (1 + 1) 22@-300 "C (6.5 "C min-1) 11 Pyridine-BSTFA + 1% SE-52 (25 m x 0.32mm i.d., df = 0.15 pm) Isothermal (260°C) Split 15 Split 19 Pyridine-MSTFA (5 + l ) , OV-1711(25 m X 0.32 mm i.d., df = 0.2 pm) 110-260°C (30 "C min-1) On-column Pyridine-HMDS-TMCS CPSIL-19 (25 m x 0.32 mm i.d., df = 0.22 pm) Isothermal (280 "C) (9+3+1) 90 min at 80 "C * BSTFA = N , 0-Bis(trimethylsily1)trifluoroacetamide; TMCS = trimethylchlorosilane; HMDS = hexamethyldisilazane; and MSTFA = t OV-1701: 5% cyanopropyl-7% phenyl-88% methyl-siloxane.$ df = Film thickness. $ CPSIL-19: 85% dimethyl-7% cyanopropyl-7% phenyl-1% vinyl-siloxane. 1 SE-52: 5% phenyl-95% methyl-siloxane. 11 OV-17: 50% phenyl-50% methyl-siloxane. N-methyl-N-trimethylsilyltrifluoroacetamide. ~~ Table 4 Response factors and recovery check of cholesterol (second study) Laboratory 2 3 n t 5 5 Low cholesterol level, C : B = 5* Mean 1.16 0.975 SD 0.02 0.023 Medium cholesterol level, C : B = 1* Mean 1.16 0.994 SD 0.02 0.03 High cholesterol level, C : B = 2* Mean 1.15 0.983 SD 0.02 0.014 n t 5 4 Low cholesterol level, C : B = 0.5* Mean 102.6 91.3 SD 1 205 Medium cholesterol level, C : B = 1* Mean 104.7 89.7 SD 1.5 2 High cholesterol level, C : B = 2* Mean 100.3 92.1 SD 1.5 4.4 C. Recovery of cholesterol added to RM 162 unsaponifiable (results in YO)- A .Response factors- B. Procedure check; recovery from simple solutions (results in %)- Low level of cholesterol added (5200 mg per 100 g of oil) n = 3t Recovery 97.8 95.2 SD 4.6 4.2 Recovery 106 94.2 SD 1.5 2.9 Unspiked RM 162 cholesterol Mean 13.8$ 9.6 (mg per l00g) n = 3 SD 1 1.3 Medium level of cholesterol added (400 mg per 100 g of oil) n = 3t * C : B = cholesterol : betulin concentration ratio. t n = number of replicates in each laboratory. t. Result excluded for statistical evaluation. 5 3 1.01 0.03 0.997 0.03 1.05 0.02 3 97.3 1.5 97.3 5 97.7 1.5 98 3.3 100.4 1.1 1.9 0.2 7 5 1.01 0.01 1.02 0.02 0.98 0.02 5 101.3 1.5 98.5 1.2 99.6 2 99.8 1.8 98.9 4.1 4.8 0.2 11 3 0.995 0.012 1.006 0.016 0.998 0.014 3 95 2.4 99.8 0.8 101.3 3.3 101.2 1.5 100.2 1.1 4.6 0.1 15 5 1.04 0.01 1.05 0.02 1.01 0.02 5 107.5 8.3 101.5 2.1 99.7 0.5 96.6 2.2 101.5 2.9 5 1.3 19 4 1.012 0.004 1.01 0.004 0.992 0.006 4 108.9 5.2 108.1 4.9 106 7.4 104 1 106 1 1.5 0.4 Third and fourth intercomparisons Participants agreed to apply a common saponificatiodextrac- tion procedure for the third intercomparison.The procedure combined the conditions used in each laboratory and although not necessarily convenient for routine analysis, it was ex- pected to ensure fully quantitative saponification and extrac- tion as requested for eventual RM certification purposes. The procedure was as follows: saponification with 100 cm3 of 2 mol dm-3 methanolic KOH solution for 1 h at 75-80 "C; extraction of the USM with 3 x 100 cm3 of diethyl ether; and washing of the ethereal extract with 3 x 40 cm3 of water.Before use, the method was validated by means of radiolabelled sterols, [3H]cholesterol and [3H]cholesteryl oleate, and radiometric measurements,17 which demonstrated that: (1) the added cholesterol (free or as oleate) was quantitatively recovered regardless of the material tested (sunflower oil or butter oil); (2) cholesteryl oleate was totally saponified; (3) losses by washing did not exceed 1%; and (4) there was no detectable amount of sterol degradation products (TLC plus radiodensitometric scanning). In order to complete the study, six participants studied the recovery of cholesterol when the common protocol was applied. The response factors for silylated cholesterol and betulin standard mixtures were first measured directly without TLC.An aliquot was subjected to the saponification/TLC steps, and1096 ANALYST, JULY 1992, VOL. 117 ~ ~ ~~~ Table 5 Third interlaboratory trials. Response factors for cholesterol to betulin, with and without saponificatiodTLC steps Laboratory Response factors code Mean SD t-Test* 2At 2BT 3A 3B 5A 5B 6A 6B 7A 7B 11A 11B 1.020 1.060 1.030 1.010 1.039 1.019 1.045 1.098 0.980 0.990 0.993 0.970 1.020 1.030 1.044 1.040 1.046 0.944 1.041 1.074 0.990 0.990 0.999 0.973 1.040 1.030 1.046 1.060 1.085 0.828 1.066 1.081 0.990 1.010 0.996 0.967 1.020 1.060 1.023 1.030 - - 1.038 1.010 0.990 0.993 - - * Statistical evaluation: t-Test on paired values (95% confidence limits). t A; Without TLC; B; with TLC. $ NS: No significant difference of the means.0 S: Significant difference of the means. 1.010 1.020 1.080 1.050 - 1.036 - 1.038 - 1.057 - 0.930 1.058 1.049 - 1.084 1.OOO 0.994 1.010 0.998 - 0.995 - 0.970 0.010 NSS 0.020 0.010 NS 0.021 0.025 NS 0.096 0.012 S § 0.013 0.011 NS 0.011 0.003 S 0.030 2 0 15 30 Time/mi n Fig. 1 Chromato ram of the sterols of RM 162. Column, CPSIL- 19CB (Chrompackf, 25 m x 0.32 mm i.d., 0.2 pm film thickness. Cold 'on-column' injector. Temperature programme: 60-285 "C at 30 "C min-1, 40 kPa He, FID detector at 300 "C. 1, Cholesterol; 2, campesterol; 3, campestanol; 4, stigmasterol; 5 , stigmastanol + fucosterol; 6, 6-sitosterol; 7, A5-avenasterol; 8, A7-stigmasterol; 9, A7-avenasterol; and 10, betulin (ISTD) the response factors were measured again. A comparison of the two sets of results presented in Table 5 shows that the calculated response factors, with or without treatment, are virtually identical except for those of Laboratories 4 and 14 for which a significant difference (95% confidence limits) was detected.It was, therefore, necessary for the participants to confirm that the saponification/TLC step had no influence on the calibration procedure. If the results differed by more than +5%, the cause had to be identified and corrected before continuing the study. Conclusions The different interlaboratory trials led to some improvement of the methodology, and several recommendations were proposed for the determination of the individual sterol content in fats and oils within the certification exercise. Saponification. The rigorous saponification procedure dis- cussed above was mandatory.Internal standard. In spite of its particular structure (a lupane triterpenoi'd bearing two hydroxy groups), necessi- tating a careful assessment of the derivatization procedure, 9 8 7 - ._ 6 , " 5 0 4 0 3 s r 2 2 i 1 0) i E f 6 ' 3 ,I 2 - i . 0 2 3 4 5 6 7 8 9 1 1 1 5 1 9 Laboratory code Fig. 2 Comparison of the results obtained in three preliminary intercom arisons and the final certification of the cholesterol content of RM l&. (a) 1st intercomparison, 2.50 k 1.13 mg per 100 g; (b) 2nd intercomparison, 4.85 * 2.90 m er 100 g; (c) 3rd intercom arison, 2.78 k 0.91 mg per 100 g; and tJ certification, 2.49 _+ 0 . d m g per 100 gANALYST, JULY 1992, VOL. 117 1097 the betulin is virtually co-eluted with the sterols on the silica gel plates and does not interfere with any other molecule during the GLC run.On the other hand, cholestane, another potent ISTD, suffers two major drawbacks. ( a ) It is not eluted with the other sterols and, therefore, must be added after the TLC run. For that reason, it does not compensate for any loss of sterols during the extraction of USM and the re-extraction of the sterols from the gel. ( b ) The cholestane is not silylated. According to Homberg and Bielefeld,l6 cholestane is particularly useful for the analysis of animal fats where a TLC fractionation is not strictly obligatory. For sterol determination, it is recommended to add the ISTD not later than the saponification stage and at a level that is appropriate, i.e., producing a peak height similar to that of the compound to be determined.In order to guard against undetected drift during a working day, the calibration and sample solutions should be injected alternately. Derivatization . All the chromatograms furnished by the participating laboratories revealed that the different con- ditions of derivatization used throughout the studies led to complete silylation of both cholesterol and of betulin and that this was not a substantial source of error. GLC stationary phases. The use of chromatographic col- umns with high efficiency and selectivity is recommended in order to obtain the best achievable separation of sterols. Stationary phases of medium polarity, such as poly- cyanopropylphenyl(methy1)siloxane (OV-1701, CP-SIL19 CB, DB-1701, etc.) (see Table 3 ) , fulfil these requirements.p-Sitosterol and A5-avenasterol, unresolved on some apolar columns, were better separated on OV-1701 or equivalent phases. Under such conditions, additional peaks were distin- guishable in the tails of the peaks for campesterol and p-sitosterol (Fig. 1). As shown in Fig. 2, there was a sharp decrease in the standard deviation of the mean when these recommendations were followed and it was believed that the accuracy of the results for low cholesterol levels in soya-maize oil (RM 162) significantly (95% confidence limits) improved during the successive interlaboratory trials. The certification of the cholesterol level in the three BCR RMs will be reported elsewhere. This paper places on record the collaborative work of laboratories from several European countries and to whom the authors give full acknowledgement. The following special- ists contributed to the preliminary studies andor to the final certification exercise.G. Contarini and P. M. Toppino, Istituto Sperimentale Lattiero Caseario, Lodi, Milano, Italy. F. Mordret and F. Lacoste, Institut des Corps Gras, Pessac, France. W. D. Pocklington, J. Pearse and M. Burn, Laboratory of the Government Chemist, Teddington, UK. S. P. Kochar and R. Griffith, Leatherhead Food Research Association, Leatherhead, UK. V. Eckelmans, Ministerie van Economische Zaken, Brussels, I Belgium. S. L. Reynolds, Ministry of Agriculture, Fisheries and Food, London, UK. T. Leth, National Food Institute, Soborg, Denmark. B. Muuse and J. de Jong, Rijks-Kwaliteitsinstituut voor Land en S.Mannino, Universita degli Studi di Milano, Milano, Italy. Tuinbouwprodukten, Wageningen, The Netherlands. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 References Pocklington, W. D., Frezenius’ 2. Anal. Chem., 1988,332,674. Gorog, S. , Quantitative Analysis of Steroids, Elsevier, Amster- dam, 1983, pp. 247-289. Mordret, F., Ajana, H., and Gauchet, C., Rev. Fr. Corps Gras, 1985, 32, 308. Morisaki, M., and Ikekawa, N., Chem. Pharm. Bull., 1984,32, 865. Kesselmeier, J., Eichenberger, W. , and Urban, D., Physiol. Plantarum, 1987, 70, 610. Perrin, J. L., and Raoux, R., Rev. Fr. Corps Gras, 1988, 35, 328. Schwadorf, K., and Muller, H., J. Assoc. Off. Anal. Chem., 1989, 72,457. Cahagnier, B., Ind. Agroaliment., 1988, 1J2, 5. Osswald, W., Holl, W., and Elstner, E., Z . Naturforsch., Teif C, 1986,41, 542. IUPAC, Standard Methods for the Analysis of Oils, Fats and Derivatives, Method 2,403, Blackwell, Oxford, UK, 7th (revised and enlarged) edn., 1987. Determinazione del contenuto di steroli mediante gascromato- grafia con colonna capillare. Norme Grassi e derivati. Metodo NGD C72. Rev. Ital. Sost. Grasse, 1987, C4, 553. Corps Gras d’origine animale et vCgCtale. Dosage des faibles teneurs en CholestCrol. Norme Francaise, 1983, NF T 60-249. Homberg, E., Fat Sci. Technol., 1987,89,215. Arens, M., Fiebig, H., and Homberg, E., Fat Sci. Technol., 1990,92, 189. Homberg, E., and Bielefeld, B., Fat Sci. Technol., 1990, 92, 478. Homberg, E., and Bielefeld, B., Fat Sci. Technol., 1987, 89, 255. Lognay, G., Dreze, P., Wagstaffe, P. J., Marlier, M., and Severin, M., Analyst, 1989, 114, 1287. Paper 1 lO6454A Received December 30, 1991 Accepted March 6, 1992