Isotope Dilution Mass Spectrometry as a Candidate Definitive Method for Determining Total Glycerides and Triglycerides in Serum

Transcription

1 CLIN. CHEM. 41/3, (1995) #{149} Lipids and Lipoproteins Isotope Dilution Mass Spectrometry as a Candidate Definitive Method for Determining Total Glycerides and Triglycerides in Serum Pofly Ellerbe, Lorna T. Sniegoski, and Michael J. Welch A new isotope dilution mass spectrometric method for total glycerides and triglycerides in human serum is described. Total glycerides are defined as the sum of tn-, di-, and monoglycerides plus free glycerol; triglycerides are defined as the pure triglyceride species. In both determinations, serum samples are supplemented by addition of [13C3]tripalmitin, processed, derivatized, and the abundance ratios of selected ions are determined. Bias is investigated by quantifying the analyte in the same samples under different chromatographic and ionization conditions. The analytes were determined in two human serum pools. The CV for a single measurement ranged from 0.35% to 0.72%, and the relative SEM ranged from 0.10% to 0.34%; there was no significant bias in the measurements. The combination of high precision and absence of significant bias in the results qualify this method for consideration as a Definitive Method as defined by the National Committee for Clinical Laboratory Standards. Indexing Terms: glycerol/cholesterol/gas chromatography-mass spectrometry The measurement of triglycerides has long been an important clinical measurement because of the correlation between high concentrations of triglycerides and the risk of heart attack. However, the accuracy and precision needed for that particular clinical purpose were relatively low. More recent data now indicate that, clinically, the total amount of cholesterol is not as important as the amounts of cholesterol associated with the individual lipoproteins, low-density lipoprotein (LDL) and high-density lipoprotein (HDL).2 The lipoproteins may be separated directly by ultracentrifugation, but this is not a practical procedure for the high number of serum specimens received by clinical laboratories. HDL-cholesterol may be quantified by a simple precipitation procedure that is easily run in high-volume testing (1), but the LDL is calculated from a formula (2) that uses the measured values for total and HDL-cholesterol and triglycerides. Thus, accurate National Institute of Standards and Technology, Gaithersburg, MD Author for correspondence. Fax 30i Nonsdard abbreviations: ID-MS, isotope dilution mass spectrometry; GC-MS, gas chromatography-mass spectrometry; NIST, National Institute of Standards and Technology; TMS, trimethylsilyl; SPE, solid-phase extraction; El, electron ionization; CI, chemical ionization; SRM, Standard Reference Material; LDL, low-density lipoprotein; HDL, high-density lipoprotein; and BSA, bis(trimethylsilyl)acetamide. Received September 28, 1994; accepted December 8, calculation of LDL depends on accurate determination of triglycerides. At present, only Reference Methods or field (routine) methods are available for triglycerides assay; there is no true accuracy base. The present Reference Method for measuring triglycerides (3, 4) does not measure the true triglyceride concentration alone, but rather a combination of triglycerides and all or part of the diglycerides and monoglycerides. Some field methods correct for free glycerol; others do not. A need thus exists for a method of demonstrated accuracy and precision, i.e., a Definitive Method. A Definitive Method is not meant for use on clinical samples; rather, it provides an accuracy base, through certified reference materials, by which field methods can be judged. The Definitive Method should be able to differentiate between triglyceride and other glycerides. According to the published guidelines of the National Committee for Clinical Laboratory Standards for Definitive Methods (5), isotope dilution mass spectrometry (ID-MS) is a suitable technique for a Definitive Method, because it does not depend on sample recovery, shows high precision, and can be tested for bias and unknown interferences. ID-MS methods for organic analytes involve adding to a sample a labeled version of the analyte as an internal standard, processing the sample, and then measuring the ratio of unlabeled to labeled analyte by gas chromatography-mass spectrometry (GC-MS). Assuming complete equilibration, less than complete recovery of the unlabeled analyte does not affect the concentrations measured (unless there is a significant isotope effect), because it is the ratio of unlabeled to labeled analyte that is measured. Generally, deuteriuxn is the only stable isotope routinely used in labeling organic compounds for which significant isotope effects can be seen. For this work, therefore, we have used 13C to label the glycerol moiety. Although the probability of a significant measurement interference is low when an isolation procedure is followed by capillary column GC-MS, it is still possible that a substance could coelute with the measured species, contribute to the ion intensity measured for either the unlabeled or labeled form, and thus interfere with the measurement. To test for such an interference, we selected a subset of sera from samples already measured at the principal ion and remeasured them in two ways: with a GC column of different polarity and monitoring another pair of ions, and monitoring a different pair of ions generated with a different ionization method. These sets of measurements are called confirmatory measurements. If an interference were CLINICAL CHEMISTRY, Vol.41, No. 3,

2 present that would bias the principal ion measurements for a given sample, the results for one or both of the confirmatory measurements should be different. Therefore, an interference would go undetected only if it had the same retention time as the measured species on each GC column, and yielded the same ions in the same proportions by each method of ionization. Such a situation is unlikely. Even interferences of unknown nature can be detected by this method. Our group at the National Institute of Standards and Technology (NIST) has developed methods that are, according to the above-mentioned guidelines (5), Definitive Methods for cholesterol (6, 7), glucose (8), uric acid (9)3, urea (10), and creatinine (11) in human serum. Other laboratories have published methods they describe as definitive for cortisol (12-14), cholesterol (15), creatinine (16), glucose (17), and uric acid (18). We have now developed, in cooperation with the College of American Pathologists, two ID-MS methods that fulfill the stringent requirements of a Definitive Method: one for total glycerides, which we define as the sum of triglycerides, diglycerides, monoglycerides, and free glycerol; and another for triglycerides only. This dual approach was based on expert advice (Herbert Naito, Pathology and Lab Medicine Service, VA Medical Center, Cleveland, OH 44106; personal communication). Total glycerides were determined because this value represents all of the related species. We included glycerol in the definition of total glycerides because many clinical laboratories do not blank for the free glycerol in analyzing for triglycerides (common lipase-based tests must be run twice to measure triglycerides without including free glycerol, which is expensive in time and money). Triglycerides were determined because this is the actual species for which a value is desired. Principles The Definitive Method for determining total glycerides is based on the addition of a known weight of ti13c]tripalmitin to a known weight of serum. The serum is hydrolyzed in base to convert the glyceride species to glycerol without the hydrolysis of phosphoglycerides and is then passed over a deionizing resin. The eluates are taken to dryness, and the glycerol is converted to a butylboronic ester; the remaining hydroxyl group is converted to the trimethylsilyl (TMS) ether. For measurement, the derivative is injected into a nonpolar fused-silica capillary GC column that has been inserted directly into the ion source of a modified magnetic mass spectrometer. The principal isotope ratio measurements are made from the ion abundances of the [M- 15] ions atm /z 215 and 218. Standards are made by combining and derivatizing known amounts of pure unlabeled tripalmitin and [13C3]tripalmitin. Standards with weight ratios slightly higher and slightly 3Cohen A, Hertz HS, Schaffer R, Sniegoski LT, Welch MJ. Presented at the 27th Annual Conference on Mass Spectrometry and Allied Topics, Seattle, WA, June 3-8, lower than that of each sample are measured immediately before and after the sample. This measurement technique, known as bracketing, produces results of high precision (6). Confirmatory measurements made on the same samples with use of different ions, different chromatographic conditions, or different ionization techniques provide evidence that there was no bias in the measurement process. The Definitive Method for determining triglycerides involves adding a known weight of [13C3ltripalmitin to a known weight of serum. The sera are then extracted with chloroform:methanol overnight, evaporated to dryness, dissolved in hexane, and passed through silica solid-phase extraction (SPE) cartridges. The triglycerides are eluted with hexane:ether, and the eluates are taken to dryness. Samples are then hydrolyzed in base and processed as described for total glycerides. Standards are made by combining and derivatizing known amounts of pure unlabeled tripalmitin and [13C3]- tripalmitin and then processing as described for sera. Materials and Methods Materials N-Methyl-N-trimethylsilyltrifluoroacetamide and bis(trimethylsilyl)acetamide (BSA) were purchased from Pierce Chemical Co. (Rockford, IL). 1-Butaneboronic acid was purchased from Aldrich Chemical Co. (Milwaukee, WI). Standard Reference Material (SRM) 1595 Tripalmitin, with a certified purity of 99.5%, was obtained from the Standard Reference Materials Program at NIST. Amberlite MB-i resin was obtained from Rohm and Haas (Philadelphia, PA). Reacti-vials were purchased from Pierce. Silica SPE cartridges (3-mL Supelclean LC-Si) were from Supelco (Bellefonte, PA). [ 3C3]Tripalmitin was synthesized in this laboratory by a procedure described elsewhere (19). Egg phosphatidyicholine was obtained from Sigma Chemical Co. (St. Louis, MO). All other chemicals were ACS-reagent grade. The serum pool LP 10 was obtained from the College of American Pathologists (Northfield, IL). The fresh-frozen serum poo1 was prepared from blood donated by volunteers at NIST. The procedures followed were in accordance with the Helsinki Declaration of 1975, as revised in Radioactive triglycerides. [a-14c]monopalmitin was prepared, according to the method of Hartman (20), from [14Clglycerol and palmitoyl chloride in a chloroform solution made homogeneous with the aid of N,Ndimethylformamide; potassium thiocyanate was included as a complexing agent to reduce the reactivity of one primary hydroxyl group. We purified the monopalmitin by Silica Gel G thin-layer chromatography, with hexane:ether:glacial acetic acid (30:70:1 by vol) as solvent. The synthesis was carriedout on a semimicro scale. [ 4C}Dipalmitin (mostly a,a ) was prepared by the same method. [ 4ClTripalmitin was prepared according to Dauben (19) and purified by thin-layer chromatography on Silica Gel G, with hexane:ether: acetic acid (50:50:1 by vol) as solvent. 398 CLINICAL CHEMISTRY, Vol. 41, No. 3, 1995

3 Standards. For each set of serum samples a set of standards was prepared. Stock solutions of SRM 1595 tripalmitin and [ 3C3]tripalmitin in toluene were prepared by weight (-P24 mg of tripalmitin, accurately weighed). The [ 3C3ltripalmitin solution was the same solution as that added to the serum samples. Weighed portions of each solution, sampled by the syringeweighing technique (6), were combined to provide a series of standard mixtures with unlabeled/labeled weight ratios of 0.7 to 1.4. These mixtures were processed in the same manner as sera: Standards for measurement of total glycerides were processed in the same manner as total-glycerides serum samples, and standards for measurement of triglycerides only were processed in the same manner as the triglycerides-only serum samples. Each set contained 7 to 11 standards. Sample Preparation For all sets of samples except set 1, two aliquots were analyzed from each vial of serum; in set 1, three aliquots from one vial and two aliquots from another vial were analyzed. The sera in sets 1-7 were freeze-dried sera. These sera were reconstituted by weight (21), except that when a vial of serum was weighed and water added, the amount of water added was determined by weight rather than by volume. The serain sets 8-11 were frozen sera, and were allowed to thaw. The solution of [ 3C3]tripalniitin used to prepare standards was added to sera as follows: Aliquots of the labeled tripahnitin solution were placed in test tubes. Aliquots of sera were then placed in those test tubes, and the tubes were gently swirled. Each tube contained accurately known quantities of -1 mg of [ 3C3ltripalmitin and sufficient serum (usually 0.5 ml) such that the weight ratio of endogenous and labeled triglyceride was -Wi. All aliquot preparation was done with the syringe-weighing technique (6). Serum densities were measured by a previously described procedure (22). To prepare a set for total glycerides determination, we then added to each of the test tubes containing serum standard 5 ml of 2.5 g/l ethanolic KOH (5 ml of 50 g/l KOH diluted to 100 ml with ethanol) and heated the sera at 70#{176}C for h. Sera and standards were deionized on the day of hydrolysis by use of Amberlite MB-i resin. For each solution to be treated, a column of 20 ml of resin was poured and then washed with 80 ml of deionized water. The basic hydrolysate was loaded, the sample was eluted with 40 ml of deionized water into a mL round-bottomed flask, and the eluate was dried on a rotary evaporator. The residue was transferred in -5 ml of methanol to a screw-capped test tube, and the methanol was removed under a stream of nitrogen at room temperature. The residue was redissolved in 0.2 ml of 1-butaneboronic acid in pyridine(22 mg/5 ml), and the samples were heated at 95#{176}C for 1 h. The contents of each tube were transferred to a 0.3-mL Reacti-vialand evaporated to dryness under nitrogen at room temperature; we then added 50 L of N- methyl-n-trimethylsilyltrifluoroacetainide and heated at 60#{176}C for mm (or let them stand overnight at room temperature). Recovery of total glycerides in serum up to the derivitization step averaged 79.0% (range %). To prepare serum for the determination of triglycerides only, one must separate the triglyceride species from the other glyceride species before hydrolysis. After reconstituting the sera and aliquoting sera and standards as described above, we extracted each serum and standard in 25 ml of chloroform:methanol (2:1 by vol) for 1 h with occasional shaking. We then added 4 ml of saturated sodium chloride solution, and stored the extracts in a refrigerator for several hours or overnight. The top phase was then removed and discarded, and the bottom phase was evaporated to dryness under nitrogen and redissolved in 2 ml of hexane. The hexane extracts were brought to room temperature and then loaded onto individual silica SPE cartridges (3-mL Supelclean LC-Si) preconditioned with 3 ml of hexane. Each cartridge was then rinsed with 0.5 ml of hexane and the triglycerides were eluted with 5 ml of hexane:diethyl ether (85:15 by vol) into a small screw-capped test tube, where the samples were dried under nitrogen. Recovery of glycerides from the SPE was 99.5% (four trials; range %). Sera and standards were then hydrolyzed, deionized, and derivatized as for total glycerides, except that the hydrolysis time was 1 h. Recovery of triglycerides in serum up to the point of hydrolysis was 94.5% (94.4% and 94.6%). GC-MS Instrumentation and Measurement Conditions The instrumentation used for measurements consisted of a gas chromatograph combined with a single focusing magnetic mass spectrometer, controlled by a data acquisition system designed for isotope ratio measurements (10). Electric switching was used to switch between the masses (9). The principal measurements were made at the masses of the EM - CH3] fragment ions at m/z 215 and 218 from electron ionization (El) with a nonpolar GC column. For confirmatory measurements we used the fragment ions [M - C2H5O] at m/z 185 and 187 with a moderate polarity GC column, and the EM + H1 ions at m/z 231 and 234 from methane chemical ionization (CI) with the nonpolar column. For measurement under El conditions, the mass spectrometer was operated at 72 ev with an emission current of 0.5 ma and an ion source temperature of 200#{176}C. For measurement under methane CI conditions, the emission current was 1 ma, the source manifold pressure (ionization gauge) was 1.5 x i03 Pa, the analyzer pressure was 1 X i0 Pa, and the source temperature was 200#{176}C. An adjustable splitter was located at the front of the column, and the end of the column was placed directly into the source. For the principal measurements and the confirmatory methane CI measurements, the GC was equipped with a 30 m X 0.25 mm (i.d.) nonpolar [95% dimethyl-5% diphenylpolysioxane] fused-silica capillary column with 0.25-pmthick film, especially made for mass spectrometry (DB- Sms; J & W Scientific, Folsom, CA). The splitter was set CLINICAL CHEMISTRY, Vol. 41, No. 3,

4 to a vent-to-column ratio of 40:1, and the GC was operated at a temperature of 130#{176}C with a helium flow rate of 3 inlimin. The injection port and the interface to the mass spectrometer were maintained at 200#{176}C. The usual injection was 1 y.l of sample or standard. Under these conditions, the retention time for the glycerol derivative was -6 mm, and the GC peak was usually -10 s wide. Confirmatory measurements were also done on masses at m/z 185 and 187 on a 60 m X 0.25 mm(i.d.) intermediate polarity (50% methylphenylpolysioxane) fused-silica colunin of 0.5-pm-thick film (DB-1701; J & W Scientific). The column temperature was 200#{176}C, and the other conditions were as described above; the retention time of the glycerol derivative under these conditions was -7 mm. The number of data acquisition sweeps per measurement cycle was set at 17, and the number of cycles across each chromatographic peak was 30-40(9). For some of the method development work, a triplequadrupole mass spectrometer (TSQ-70; Finnigan MAT, San Jose, CA) interfaced with a gas chromatograph was used. The injector and transfer line were set to 200#{176}C. Helium head pressure was set to -100 kpa. The nonpolar column (DB-5ms) was programmed for 1 mm at 130#{176}C, then at io#{176}c/min to 280#{176}C. The mass spectrometer was operated in the electron ionization mode, scanning the first quadrupole. The ion source was set to 200#{176}C. Measurement Protocol For the measurement of each sample, two standards were chosen: one of ion intensity ratio lower than that of the sample, and one of slightly higher ion intensity ratio. Each standard and sample was measured twice in succession. The two observed intensity ratios were acceptable only if they agreed within 0.5%; if not, a thirdmeasurement was made, which had to agree with one of the other two within 0.5%, and the three were averaged. For further use of a standard again in any given half-day, only a single measurement was made at each use if the new ion intensity ratio was within 0.5% of the previous value for that standard. Measurements were made in either this order or its reverse: lower weight-ratio standard, sample, higher weight-ratio standard. Thus, each measurement of sample was immediately bracketed both in time and ratio by measurements of standards. On a second day the order of standards was reversed, and the measurement process repeated. If the weight ratios for each sample from both days did not agree within 0.5%, a third day s measurement was done, which had to agree with one of the other days measurements within 0.5%, or all three measurements would be discarded. (In fact, no measurements were discarded in this work.) The average of these separate analyses constituted one valid measurement. The quantity of analyte in the sample was calculated by linear interpolation of the measured ratio of the sample between the measured ratios of the standards, whose weight ratios were known. Other Procedures Total glycerides equilibration and hydrolysis. To test for complete hydrolysis of labeled tripalmitin and equilibration with endogenous glycerol in serum, we added a toluene solution of labeled tripalmitin to a serum sample. After mixing, we incubated the solution at 70#{176}C, removed aliquots at 15 and 30 mm and 1, 2, 4, 6, and 18 h, and then processed the aliquots as described above for total glycerides. Triglycerides equilibration and hydrolysis. To test for complete equilibration of labeled tripalinitin with serum extracted with 2:1 (by vol) chloroform:methanol, we added 2.5 ml of serum and mg of labeled tripalmitin to 125 ml of the chloroform:methanol, and shook the mixture by hand about every 15 mm. Auquots (25 ml) of the extract were removed at 0.5, 1, 2, 4, and 18 h. To each aliquot was added 4.5 ml of saturated NaC1 solution. Each mixture was inverted several times (not shaken) and then stored in the refrigerator overnight. The aqueous layer was removed and discarded, and the bottom layer was evaporated to dryness and then hydrolyzed, deionized, and derivatized as described above. To test for complete triglyceride hydrolysis, we prepared a serum sample in the same manner as for the measurement of triglycerides, up to the point of hydrolysis. Labeled glycerol was then added to the dry extract, followed by 20 ml of the 2.5 g/l alcoholic KOH, and the mixtures were incubated at 70#{176}C for 15 and 30 mm and 1, 2, 4, 6, and 18 h. Sera were then deionized and derivatized as described above. Phospholipid hydrolysis. Egg phosphatidyicholine (5.85 mg) was incubated with 2.5 g/l alcoholic KOH at 70#{176}C for 1 and 24 h. Aliquots were neutralized, dried, derivatized with BSA, and examined in the TSQ-70 mass spectrometer for the presence of glycerol-tms derivative; as little as 3.2 ng of glycerol-tms derivative could be easily detected. Egg phosphatidyicholine (2.34 mg) was derivatized with 100 p.l of BSA, standing overnight unheated; 1-p.L samples of the derivatives were then injected into the TSQ-70. Separation of triglycerides. A chloroform:methanol (2:1 by vol) extract of 2.5 ml of serum was prepared in the same manner as for triglyceride samples. Radioactive triglycerides, diglycerides, monoglycerides, free glycerol, and phosphatidyicholine separately were added to 20-mL aliquots of the extract. The mixtures were then dried, dissolved in hexane, and applied to individual silica SPE cartridges that had been washed with hexane. The solvent elution steps of 5 ml for each cartridge sequentially were as follows: hexane:ether in volume ratios of 90:10, 85:15, 80:20, 75:25, 70:30, 60:40, and 50:50, then pure ether, and then pure methanol. An aliquot of the eluate from each step of the elution was placed in a counting vial and allowed to dry overnight in the hood. The next day we added 10 ml of scintillation fluid and counted the radioactivity in the vial. 400 CLINICAL CHEMISTRY, Vol. 41, No. 3, 1995

5 Results and Discussion The ultimate goal of our work, as set by the Standards Committee of the College of American Pathologists, was to develop a Definitive Method for measuring triglycerides in the forms useful for physicians. The following measurement needs were defined: total glycerides (the sum of tri-, di-, and monoglycerides plus free glycerol) and pure triglycerides (only the actual triglyceride fraction but no monoglycerides, diglycerides, or free glycerol). It was not considered useful to measure the free glycerol alone. Assay Evaluation Choice of label. Tripalmitin was chosen as the species to be isotopically labeled because of the availability of an SRM (No Tripalmitin) for use as the unlabeled primary standard material; no other triglycerides were available as such completely characterized materials. The SRM is a solid and can be easily weighed. 3C was chosen as the isotope because the isotope effects with it are negligible, and because E 3C]glycerol from which to make the labeled tripalmitin was commercially available. With three 13C labels on glycerol, the contribution of the labeled material to the unlabeled material, and vice versa, is very small; over the short bracketing range used, the effect on the results is negligible (23). Phospholipid hydrolysis. If phospholipids were to be hydrolyzed during the serum processing, unlabeled glycerol would be produced, generating a bias that could not be detected with the confirmatory measurement protocol. Thus it is most important to detect whether phospholipid hydrolysis could occur during sample preparation. The most common phospholipid type in serum is phosphatidylcholine, both saturated and unsaturated fatty acids are present, and the most common fatty acid chain lengths are C16 and C18 (24). Therefore, egg phosphatidylcholine, with one palmitic and one oleic acid, was chosen as the test phospholipid. When egg phosphatidylcholine was incubated with 2.5 g/l alcoholic KOH at 70#{176}C for 1 and 24 h, processed as described above, and an aliquot injected into the mass spectrometer, the amount of hydrolysis at 1 h was -0.2% and at 24 h -1.2%. Assuming usual concentrations of phospholipids in serum, and the amounts of hydrolysis and concentrations of triglycerides measured in our samples, the maximum amount of interference in any sample would be -0.2%. This interference is possible only in the samples analyzed for total glycerides, because, in the triglyceride processing, samples are passed over an SPE, which retains the phospholipid, before hydrolysis. (Egg phosphatidylcholine tested under acid hydrolysis conditions did substantially hydrolyze to glycerol.) Although theoretically phospholipid could hydrolyze to glycerol after base hydrolysis, e.g., in the injector of the mass spectrometer, when we treated samples of egg phosphatidylcholine with BSA and injected them into the instrument, no glycerol-tms derivative was detected. Choice of derivative. To introduce glycerol into the G(, it is necessary to convert it into a suitable derivative. Fourteen different derivatives of glycerol were prepared and examined. As described elsewhere (Ellerbe P. Sniegoski LT, Welch MJ, manuscript in preparation), there were three possibilities for a useful derivative: the cyclohexylidene TMS ether, the phenylboronate TMS ether, and the butylboronic TMS ether. However, the cylcohexylidene TMS ether derivative is prepared under acid conditions, which could hydrolyze the phospholipids, causing a bias in the results. The phenylboronate derivative, when prepared from a serum sample, proved to have an unresolvable interference at the main measurement masses. Thus, the butylboronic TMS ether was chosen for use in these methods. Its El spectrum has been published (25). The base peak, m /z 103, is not useful for measurement because the mass is relatively low, and it has lost two of the labeled carbons. The main measurement peaks (unlabeled/labeled) are at m /z 215/218, for which the intensity is --66% of that of the base peak. The El confirmatory peaks are at m/z 185/187 (intensity -42% of that of the base peak). The base peak of the methane CI spectrum, which was used for confirmatory measurements, is m/z 231. Memory effects. We tested the derivative for column memory effects, although none were expected. If a memory effect is present, injections of a sample or standard of a given unlabeled/labeled weight ratio will affect the intensity ratio measured for subsequent injections of sample or standard. We injected a particular standard with its given unlabeled/labeled intensity ratio, then the pure labeled derivative, and finally the particular standard again. The intensity ratio for the last injection of the particular standard was not significantly different from the intensity ratios of the first two injections of that standard. The lack of effect on the measured intensity ratios, even when the weight ratio difference between consecutive measurements was extreme and thus should make any memory effect clearly evident, provides strong evidence of the absence of column memory effects. Standards cross-check. The accuracy of results for serum samples is limited by the accuracy of the standard mixtures for calibration. In these Definitive Method determinations, standards were prepared for each set with the same solution of labeled tripalmitin; thus, any bias due to an error in weighing labeled tripalmitin would not matter because standards and sera were based on the same labeled solution. The preparation accuracy of each set of standards was determined by bracketing with standards from the same set. The weight ratio determined by the ID-MS measurements was then compared with the weighed-in weight ratio for that standard. The agreement between these values for each member of a set allows determination of the presence or absence of a bias of a particular standard. If a particular standard differed from its CLINICAL CHEMISTRY. Vol. 41. No. 3,

6 Table 1. Consistence of standards within a set of standards. Weight ratio, unlabeled/ labeled Stan- Measured dard Bracketed by by ID-MS au as, ax ax au, ar ar ax,ay ay ar,av Weighed-in Average a Data reported are from standards for serum set 4. b [(Measured value - weighed value) x 1001/weighed value. 01ff, %b weighed-in value by >0.5%, that standard was discarded. The average bias for all sets of standards was 0.10% (range %). Table 1 shows the results of a typical standards cross-check. Total glycerides equilibration and hydrolysis. The addition of an isotopically labeled material to a serum matrix may not immediately result in the complete equilibration of the labeled form with the endogenous form. Both complete hydrolysis and equilibration are necessary for accurate measurement. The time required for equilibration depends on the matrix and on the nature of the particular analyte, and affects the results obtained for that analyte. We studied the equilibration of endogenous triglycerides with hydrolyzed labeled tripalmitin and found the equilibration to be complete in 1 h; the ratio remained unchanged for at least 18 h. We chose 2 h as the time for hydrolysis. Triglycerides equilibration and hydrolysis. In contrast to the method for total glycerides, it was necessary to first equilibrate labeled tripalmitin with endogenous triglyceride, and then hydrolyze both tripalmitin and endogenous triglyceride in sera and standards. We studied the equilibration of endogenous triglyceride with labeled tripalmitin, and found that the equilibration was complete in 30 mm. For convenience, we chose an overnight extraction time. Because an extracted serum is not the same matrix as serum itself, we also had to check for complete hydrolysis of endogenous triglyceride and labeled tripalmitin in extracted serum. We studied the hydrolysis of endogenous triglyceride with labeled tripalmitin and found that the hydrolysis was complete in 30 mm. A hydrolysis time of 1 h was chosen. Separation of triglycerides. If one is to measure only the triglycerides in a serum sample, the triglycerides must be separated from the other glyceride species in serum. A silica SPE was used for each species labeled with 4C, and elution recoveries were calculated as counts in each eluate divided by the total counts recovered. When pure species were passed through the SPEs, 98% of the triglycerides eluted in the 85:15 fraction or earlier; 99% of the diglycerides eluted in the 75:25 fraction or after; all eluted either in the pure ether or methanol fractions; and phospholipid was retained on the SPE. A solution of 85:15 hexane:ether was chosen to elute the triglycerides from the SPE. Serum Results The results of the principal measurements of total glycerides and for triglycerides in the freeze-dried material LP-10 are given in Table 2. The results of the principal measurements for total glycerides and triglycerides in the fresh-frozen material are given in Table 3. In the manufacturing process for the freeze-dried material, lyophilized serum is dispensed into vials, with some inevitable variation in fill weight from vial to vial. To separate this variation from variation due solely to the method, we divided the concentrations that were determined for each vial by the fill weight of lyophilized serum (measured after lyophilization) in that vial. Therefore, the units of concentration in Table 2 are millimoles of glycerol per liter of serum, divided by the fill weight. The fresh-frozen material presumably has no variation from this source, so the concentration units in Table 3 are given only in mg/l, ex- Table 2. Results for total glycerides and triglycerides in LP 10 [mmol/(l - Total glyceddes Triglycerides Vial Set I Set 2 Set 3 Set 4 Set 5 Set 6 Set Overall mean CV of single measurement, % Relative SEM, % The concentration in mmol/l is divided by the fill weight of lyophilized serum to correct for vial-to-vial variability in fit weight. 402 CLINICAL CHEMISTRY, Vol. 41, No. 3, 1995

7 Table 3. Results for total glycendes and triglycerides in Table 4. Confirmatory measurements (mgil). fresh frozen serum. El Total glycerides, mg/i. 5et9 Seth Mean 1925 CV of single measurement,% 0.35 Relative SEM, % 0.10 #{149} Expressedas milligrams of tnpalmitin per liter. Triglycerides, mg/i Set8 SetlO pressed as tripalmitin. To express mg/l tripalmitin concentrations as mgfl of other triglycerides, multiply the molar concentration by the appropriate ratio of molecular weights. The mean value for the total glycerides for the freeze-dried material was mmol/(l - g), with the relative SEM of the mean of 0.34% and the CV for a single measurement of 0.72%. The mean value for the triglycerides for the freeze-dried material was mmol/(l g), with the relative SEM of 0.11% and the CV for a single measurement of 0.42%. These mean concentrations for the total glycerides and triglycerides only (expressed as mg/l) are 1661 and 1391, respectively. The mean value for total glycerides for the freshfrozen material was 1925 mg/l, with a relative SEM of 0.10% and the CV for a single measurement of 0.35%. The mean value for triglycerides for the fresh-frozen material was 1794 mg L, with the relative SEM of 0.34% and the CV for a single measurement of 0.57%. For the freeze-dried material, the difference between the total glycerides and triglycerides values is 270 mgfl, which is about 16% of the total glycerides value. For the fresh-frozen material, the difference between the total glycerides and triglycerides values is 131 mg/l, which is about 6.8% of the total glycerides value. Possibly the process of freeze-drying affects the proportion of triglycerides in the material. The results from the confirmatory measurements are given in Table 4. The El confirmatory measurements show no consistent direction of bias, and differences from the principal measurements are all quite small. The CI confirmatory measurements show a small positive bias, but still so small as to be negligible. Overall, the confirmatory measurements demonstrate that there is no significant measurement bias. Error Analysis Although the imprecision of the method is small, and no evidence of significant bias in the measurement process was found, we still made an analysis of possible sources of bias and imprecision. Errors in the standards would contribute to errors in Material Set/sample Princlpar Conf l Conf r Total glycendes LP 10 LP 10 Mean 2/8 4/38 % difference from principala Frozen 9/78 Frozen 11/98 Mean % difference from principar Triglycerides LP1O 5/39 LP 10 7/56 Mean % difference from pnncipald Frozen 8/71 Frozen 10/79 Mean % difference from principald #{149} Principal El measurements made at m/z 215/218 with a DB-5ms GC column. b Confirmatory El measurements made at m/z 185/187 with a DB-1701 GC column. C Confirmatory measurementswith methane Cl made at m/z 231/234 with a DB-5ms GC column. Cl [(Confirmatory value - pnncipal value)/pnncipal value] x 100. determining of glyceride concentrations in serum. The cross-checking with standard sets provided evidence for the absence of significant error. Because the results for independently prepared sets are in good agreement, the effect of random weighing errors is very low. We expect that making large numbers of measurements on each serum pool, using standards from different sets, would further reduce the effect of random error. Isotope effects in the derivatization reaction could lead to bias and imprecision. With the [ 3C3]tripalmitin labeled material, the small differences in results observed between independently prepared serum sets provide evidence for the absence of significant isotope effects in the derivatization, and the coelution of the unlabeled and labeled derivatives from the GC column provides evidence for the absence of a significant isotope effect on the column. The standards and samples were derivatized and measured under as nearly identical conditions as we can achieve; thus, residual isotope effects, if any, should cancel out. Error due to phospholipid hydrolysis has been discussed above. In all previous Definitive Methods, the labeled species has been identical to the measured species except for an isotopic substitution. In this Definitive Method, that identity was not possible, there being so many glyceride species in serum. However, for this nonidentity to be a problem, a glyceride species in serum would have to behave differently from the labeled tripalmitin in processing at any point up to hydrolysis, when all glyceride species are hydrolyzed to glycerol. In either CLINICAL CHEMISTRY, Vol. 41, No. 3,

8 determination, the two procedures used up to that point are chloroform:methanol extraction and SPE separation. The chloroform:methanol extraction is quantitative for lipids (25, 26); consequently, which glyceride species was being examined would be irrelevant. If a particular glyceride species were not recovered from the SPE in the same proportion as the labeled tripalmitin, an error would be introduced in the triglyceride results. However, the average recovery of glycerides from the SPE was 99.5%, calculated as the total counts recovered from the SPE divided by the total counts applied to the SPE. Therefore, because essentially all the glycerides were recovered from the SPE, there is no error due to differential recovery. The triglyceride value could also be affected if the dior monoglyceride were not completely separated from the triglycerides on the SPE. When pure radioactively labeled diglyceride was passed through the SPE in two separate trials, 0.4% and 1.2% of the diglyceride counts eluted in the triglyceride elution. We assume the worst case: that this was all true diglyceride leaking over into the triglyceride elution (and not just a radioactive triglyceride impurity in the radioactive diglyceride itself). Because the amount of diglyceride in human serum is <2% of the triglyceride value [and often is not detectable (27)], and because the crossover from di- to triglyceride is at most 1.2%, this crossover would affect the values by at most -0.02% and thus is a negligible source of error in the triglyceride values. In conclusion, the combination of high precision and evidence for the lack of significant bias qualifies this method as a candidate Definitive Method for total glycerides and triglycerides. Use of this method to certify concentrations of total glycerides and triglycerides in reference materials will permit evaluation of the accuracy of both reference and routine clinical methods. We have used this method to certify the concentration of total glycerides and triglycerides in a freeze-dried material and in a fresh-frozen material. We gratefully acknowledge the assistance of Susannah Schiller with the statistical analysis and thank Jess Edwards and Edward White V for donating blood for the fresh-frozen serum pool. We also gratefully acknowledge the support of P.E. by the College of American Pathologists. Certain commercial equipment, instruments, or materials are identified in thisreport to specify adequately the experimental procedure. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the equipment, instruments, or materials named are necessarily the best available for the purpose. References 1. Bachorik PS, Albers JJ. Precipitation methods for quantification of lipoproteins. Methods Enzymol 1986;129: Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentrations of low-density lipoprotein cholesterol in plasma without use of the preparative ultracentrifuge. Clin Chem 1972; 8: The Lipid ResearchClinics population studies data book, Vol. 1. Prevalence study. DHHS, PHS(NIH) NHLBI Publ. No Bethesda, MD: 1980: Centers for Disease Control in-house triglycerides Reference Method. Atlanta, GA: Special Activities Branch, Div. of Environmental Health Lab Sciences, Center for Environmental Health, CDC, NRSCC1-T. Tentative guidelines for the development of Definitive Methods in clinical chemistry for the National Reference System in Clinical Chemistry. Villanova, PA: National Committee for Clinical Laboratory Standards, Cohen A, Hertz HS, Mandel J, Paule RC, Schaffer R, Smegoski LT, et al. Total serum cholesterol by isotope dilution mass spectrometry: a candidate definitive method. Clin Chem 1980;26: Ellerbe P, Meiselman S, Sniegoski LT, Welch MJ, White E V. Determination of serum cholesterol by a modification of the isotope dilution mass spectrometric definitive method. Anal Chem 1989;61: S. White E V, Welch MJ, Sun T, Schaffer R, Smegoski LT, Hertz HS, Cohen A. The accurate determination of serum glucose by isotope dilution mass spectrometry. Biomed Mass Spectrom 1982; 9: Ellerbe P, Cohen A, Welch MJ, White E V. Determination of serum uric acid by isotope dilution mass spectrometry as a new candidate definitive method. Anal Chem 1990;62: Welch MJ, Cohen A, Hertz H, Ruegg FC, Schaffer R, Smegoski LT, White E V. Determination of serum urea by isotope dilution mass spectrometry as a candidate definitive method. Anal Chem 1984;56: Welch MJ, Cohen A, Hertz H, Ng KJ, Schaffer R, Van der Lijn P, White E V. Determination of serum creatinine by isotope dilution mass spectrometry as a candidate definitive method. Anal Chem 1986;58: Siekmann L, Breuer HJ. Determination of cortisol in human serum by isotope dilution mass spectrometry. J Clin Chem Clin Biochem 1982;20: Jonckheere JA, De Leenheer AP. Candidate definitive method for the determination of cortisol in human serum. Biomed Mass Spectrum 1983;10: Patterson DO, Patterson MB, Cuibreth PH, Fast DM, Holler JS, Sampson EJ, Bayse DD. Candidate definitive method for the determination of cortisol in human serum. Clin Chem 1984;30: Pelletier 0, Wright LA, Breckinridge WC. Isotope dilution mass spectrometry of serum cholesterol. Clin Chem 1987;33: Sieckmann L. Determination of creatinine in human serum by isotope dilution mass speetrometry. J Clin Chem Clin Biochem 1985;23: Pelletier 0, Arratoon C. Precision of glucose measurements in control sera. Chin Chem 1987;33: Sieckmann L. Determination of uric acid in human serum by isotope dilution mass spectrometry. J Clin Chem Clin Biochem 1985;23: Dauben WG. The synthesis of palmitic acid and tripalmitin labeled with carbon fourteen. J Am Chem Soc 1948;70: Hartman L. Glycerides synthesis by direct esterification. J Chem Soc 1957: NBS Certificate of Analysis for SRM 909. Available from the Standard Reference Materials Program, NIST (formerly NBS), Gaithersburg, MD, Sniegoski LT, Moody JR. The determination of serum densities. Anal Chem 1979;51: Yap WT, Schaffer R, Hertz HS, White E V, Welch MJ. On the difference between using linear and nonlinear models in bracketing procedures in isotope dilution mass spectrometry. Biomed Mass Spectrum 1983;10: Bloom RJ, Elwood JC. Quantitation of lipid proffles from isolated serum lipoproteins using small volumes of serum. Clin Biochem 1981;14: Nelson GJ. Blood lipids and lipoproteins. New York: Wiley- Interscience, 1972: Litchfleld C. Analysis of triglycerides. New York: Academic Press, 1972: Hirsch J, Ahrens EH. Separation of complex lipid mixtures by the use of silicic acid chromatography. J Biol Chem 1958;233: CLINICAL CHEMISTRY, Vol. 41, No. 3, 1995