Serum Creatinine Determination by High Performance Liquid Chromatography and Five Automated Chemistry Analyzers

EARLE W. HOLMES, Ph.D.*, TRUDE H. OESER*, STEPHEN E. KAHN, Ph.D.*, LEONAS BEKERIS, M.D.t, and EDWARD W. BERMES, JR., Ph.D.* *Departments of Pathology, Biochemistry and Biophysics, Loyola University Medical Center, Maywood, IL 60525 and tDepartment of Pathology, Mac Neal Memorial Hospital Berwyn, IL 60402

ABSTRACT A sensitive and specific procedure is described for the determination of serum creatinine by reverse phase high performance liquid chromatog raphy. The method involves sample pretreatment with a cation exchange resin followed by the isocratic separation of creatinine and an external standard which are detected by their absorbances at 254 nm. The analyt ical recovery of creatinine from serum was 99.8 ±4.9 percent using this method. The between-day coefficients of variation for creatinine concen trations of 0.83 and 8.63 mg per dl were 3.07 percent and 3.2 percent, respectively. Results obtained with this reference method were compared to results obtained with five automated chemistry analyzers. The results showed that while the comparison methods correlated well with the ref erence method (r > 0.98 in all cases), those comparison methods based on the Jaffe reaction showed negative biases. In some cases, these biases were highly statistically significant. The difference between paired results obtained with some of the automated Jaffe-based assays exceeded 0.20 mg per dl in a substantial number of specimens with creatinine concentration in the range of 0.5 to 2.3 mg per dl. The measurement of acetoacetate concentrations of specimens sub mitted to our laboratory for serum creatinine assays showed that interfer ence by this substance with certain Jaffe-based assays could result in clin ically significant false elevations in serum creatinine concentrations.

Introduction Jaffe in 1886. Since that time, this color reaction has been used with numerous The reaction of creatinine with alkaline modifications for the determination of sodium picrate was first described by serum creatinine. As might be expected 503 0091-7370/83/1100-0503 $01.20 © Institute for Clinical Science, Inc. 504 HOLMES, OESER, KAHN, BEKERIS, AND BERMES in the case of a well-studied, widely-used Lim et al.9 The first part of the procedure assay, a number of substances that inter is a sample pretreatment designed to iso fere with this reaction have been iden late creatinine from other serum constit tified. 2’3,14,16 uents. The basis of this pretreatment The nonspecificity of the Jaffe reaction step has already been described.1 for serum creatinine is dealt with in Mini-columns containing 100 mg of several ways. The first of these involves Dowex AG 50W X-12 were poured in pretreatment of serum specimens with 2.5M NaOH and washed with 2.0 ml of adsorbents, organic solvents, protein H20 and 4.0 ml of adsorption buffer (0.04 précipitants or dialysis.10 However, these M citric acid, 0.02 M dibasic sodium techniques have met with variable suc phosphate, pH 3.0). Exactly 0.5 ml of cess. With the exception of dialysis, none sample was mixed with 5.0 ml of adsorp is suitable for use with automated instru tion buffer, and 5.0 ml of the mixture was mentation. A second and more popular applied to the minicolumn. Each column approach to increasing specificity in was washed with 4.0 ml H20, and the serum creatinine assays is to use kinetic adsorbed creatinine was then eluted with versions of the Jaffe reaction. These ki 3.0 ml of 0.5 M sodium acetate, pH 8.6. netic assays aim to differentiate chro One hundred jjlI of a solution of 0.58 mM mogen formation by creatinine from that allopurinol in 0.5 M sodium acetate was due to interfering substances by differ added to each eluent as an external stan ences in their reaction rates.7 A third ap dard. 12 proach to a more specific assay for cre Twenty (Jil of each eluent was injected atinine is to abandon the Jaffe reaction onto a 4 mm x 30 cm column of 10 (jiM entirely in favor of creatinine methods silica particles to which a unimolecular based on the use of enzymes as layer of octadecyltrichlorosilane was reagents10 or methods based on high bonded.* High pressure liquid chroma performance liquid chromatography tography was carried out using 0.01 M (HPLC).1,4,9,11,12 ammonium acetate as the mobile phase The existence of a variety of assays for at a flow rate of 1.0 ml per min. Creati serum creatinine as well as the variability nine and allopurinol were detected by among the automated kinetic Jaffe assays their absorbances at 254 nm. The creat with respect to their precise assay inine concentration of a sample was conditions3 suggests the possibility of calculated by comparing its peak height intra-method and inter-instrument bias ration (creatinine peak height per allo in creatinine results. The purpose of this purinol peak height) to those of a series study was: (1) to develop a sensitive, spe of creatinine standards! that were car cific, and accurate HPLC assay for serum ried through the entire procedure. creatinine and (2) to use this assay as a The analytical recovery of creatinine reference method for the evaluation of by this method was assessed by com the creatinine test on five automated paring the peak height ratios of extracted chemistry analyzers. and unextracted aqueous creatinine stan dards and by adding pure creatinine in Materials and Methods increments of 0.3 to 18 mg per dl to pooled human serum. The average re S er u m C r e a t in in e by HPLC coveries are based on results from tests

The following HPLC assay for serum * u-Bondapak C18, Waters Associates, Milford, MA. creatinine was developed by the present t New England Reagent Laboratory, East Provi authors based on the previous work of dence, RI. CREATININE DETERMINATION BY HPLC 505 performed in duplicate on two separate was separated into aliquots for each of occasions. The precision of the method the comparison methods and frozen. was established by assaying commercial Since it was not possible to assay each control sera on 10 occasions over a one specimen by all of the methods, compar month period. ison results for the MCA, ASTRAN and EKTA methods are based on subsets of C o m pa r iso n M e t h o d s the group of 72. Acetoacetate was measured on specimens sent to the lab Creatinine results determined by oratory for creatinine determination. HPLC were compared with the following Specimens from known diabetics were automated methods using a split-sample excluded from this group. protocol: Dupont ACA III (ACA),$ Technicon SMAC (SMAC),§ Beckman I nterference S t u d ie s ASTRA (ASTRA),» Multistat (MCA),! and EKTACHEM 400 (EKTA).** Unless The responses of the reference and specified otherwise, each automated in test methods to several known inter strument was operated according to the fering substances were tested using manufacturer’s instructions using the cal pooled serum spiked with the substances ibrators and reagents that they supplied. of interest in the concentrations indi The ASTRA instrument was also oper cated in the Results section. Acetoace ated using reagents manufactured by tate concentrations in clinical specimens Nobel Scientific (ASTRAN).ft were determined by an enzymatic A series of aqueous standards were as method.8 sayed as samples on each instrument (ex cept EKTA) to check for systematic bias S ta tistic a l E v a lu atio n that might occur as a result of differences between calibrators. Method comparison data were ana lyzed by linear regression techniques.5 C l in ic a l S p e c im e n s Since the differences between paired re sults for individual specimens assayed by The 72 serum specimens used in the various combinations of reference and method comparison study were obtained comparison method were not normally from blood drawn in red-topped vacutain- distributed according to the Kolmo- ers and submitted to the laboratory for gorov-Smirnov Test,15 the significance of biochemical profiling. Specimens with these differences was tested using the glucose values greater than 140 mg per Sign Test. Differences that were signifi dl were excluded from this group. The cant at the 0.05 level by the Sign Test creatinine concentrations of the speci were further evaluated using Wilcoxin’s mens as determined by HPLC ranged Test.15 from 0.4 to 23.2 mg per dl (mean, 4.5 mg per dl, median, 1.0 mg per dl). Serum Results

HPLC A ssay f o r S er u m C r e a t in in e Dupont Company, Wilmington, DE. § Technicon Instruments Corp., Tarrytown, NY. I Beckman Instruments, Inc., Brea, CA. In figure 1 is shown the elution profile II Instrumentation Laboratory Inc., Lexing obtained when a serum specimen with a ton, MA. creatinine concentration of 3 mg per dl ** Eastman Kodak Company, Rochester, NY. ft Nobel Scientific Industries, Inc., Alexan was extracted and chromatographed as dria, VA. described in the Methods section. Cre 506 HOLMES, OESER, KAHN, BEKERIS, AND BERMES aqueous creatinine standards were ana lyzed. The mean recovery of creatinine from the cation exchange mini-columns was 100.3 percent (sd, 6.0 percent; range 94 to 106 percent) for a series of aqueous E standards with creatinine concentrations c of 1.0 to 10.0 mg per dl. The analytical m CM recovery of creatinine from serum spec co imens with concentrations in the range o of 0.3 to 18 mg per dl averaged 99.8 per cent (sd, 4.9 percent; range 95 to 109 Ul percent). Ü Z The regression equation describing a CO< typical standard curve based on duplicate IK 0.5, 1.0, 3.0, 5.0, 7.0, and 10.0 mg per O dl standards was y = 0.351 X + 0.0276, CO r = 0.99 where X = creatinine concen < tration in mg per dl and y = the peak height ratio (creatinine per allopurinol). The overall standard deviations for the assay over a one month period for serum controls with creatinine concentrations of 0.83 and 8.63 mg per dl were 0.025 and 0.267 mg per dl respectively. The reten tion times for creatinine and the internal standard were constant for a period of at least six months, and the HPLC column showed no signs of decreased perfor F i g u r e 1. Analysis of serum creatinine by HPLC. A serum specimen with a creatinine concentration of mance after more than 250 analyses. 3 mg per dl was applied to a cation exchange mini column which was eluted as described in Methods. C o m p a r i s o n o f M e t h o d s Twenty (jlI of the eluent was chromatographed on a 4 mm X 30 cm column of u Bondapak Cj«. Mobile phase: 0.01 M ammonium acetate, flow rate: 1.0 ml No systematic bias between the var per min, detection: absorbance at 254 nm, 0.01 ious automated Jaffe-based methods was AUFS. Peak 1: Creatinine 0.81 n mol), Peak 2: Al lopurinol (0.39 n mol). Specimens with creatinine detected when a series of aqueous concentrations greater than 6 or 1 2 mg per dl were creatinine standards was analyzed. The chromatographed as described except that the de results of the split-sample correlation ex tector sensitivity was decreased to 0.02 AUFS or 0.05 periment in which patients’ sera were as AUFS, respectively. sayed by HPLC and different automated methods are shown in table I. Results atinine and the external standard had re from the automated methods correlated tention times of approximately 4.6 and well with those of the reference method. 9.5 minutes, respectively, and peak However, in all cases, the automated heights of approximately 115 and 97 mm, Jaffe-based methods gave results that respectively. The two peaks eluting be were higher than those obtained by fore creatinine have not been identified. HPLC. The intermethod differences ob However, they were also observed when served in the HPLC/SMAC, HPLC/ CREATININE DETERMINATION BY HPLC 507 TABLE I

Comparison of Automated Serum Creatinine Assays with an HPLC Assay (x)

M ethod R ange Mean Sign Wilcoxin (y) n m g /d l m g /d l Median m (s) * h fs;t S yx$ r§ B ia s" T e s t T e s t

ACA 72 0.2-25.8 4.6 1.0 1.04(0.014) -0.035(0.106) 0.73 0.99 -0.14 0.05tt n.s.** SMAC 72 0.5-28.7 5.2 1.2 1.14(0.016) 0.102(0.122) 0.84 0.99 -0.71 0.01 it ASTRA 72 0.3-25.9 4.7 1.1 1.03(0.014) 0.114(0.109) 0.75 0.99 -0.24 0.01 it MCA 44 0.5-17.0 3.8 1.2 0.90(0.024) 0.386(0.156) 0.83 0.99 -0.01 0.01 it ASTRAN 30 0.6-26.2 6.0 1.1 1.00(0.020) 0.188(0.186) 0.79 0.99 -0.21 0.05 n.s.** EKTA 27 0.3-15.8 4.0 0.6 1.04(0.016) -0.169(0.105) 0.43 0.99 -0.02 0.05 n.s.**

*m(s) - slope of the least squares line (standard deviation of the slope). fb (s) - y intercept of the least squares line (standard deviation of the intercept). jsyx - standard error of the estimate in the y direction. §r - correlation coefficient. "Bias - the mean of HPLC results minus the mean of the results of the comparison method. **n.s. - bias between the reference and comparison methods not significant at the 0.05 significance level. ffsignificance level at which the hypothesis that the two methods are the same is rejected. $$Not done.

ASTRA, and HPLC/MCA comparisons ASTRA/SMAC comparisons were highly were highly statistically significant. statistically significant. These inter To provide a more practical example of method differences frequently exceeded the magnitude of the intermethod differ 0.35 mg per dl for Group 1 specimens ences that might be encountered when and 0.20 mg per dl for Group 2 speci serum creatinine is determined using mens. two different automated analyzers,

ASTRA creatinine results were com Interference pared with AC A and SMAC creatinine results (table II). For both ranges of cre The HPLC assay for serum creatinine atinine concentrations examined (Group was not affected by glucose (100 to 1500 1 and Group 2, table II), the differenc mg per dl), ascorbate (0.5 to 4.0 mg per es observed in the ASTRA/ACA and dl), hemoglobin (27 to 320 mg per dl) or acetoacetate (0.5 to 5.0 mM). These po TABLE II tential interfering substances did not af Comparison of ASTRA Serum Creatinine fect the automated methods except in the Results with ACA and SMAC Results following cases: glucose at 1500 mg per Reference Method;: ASTRA dl caused a 0.1 mg per dl increase in cre No. of Specimens C om parison Bias Sign Bias ^ ± B ia s 5s ± atinine as measured by the ASTRA and Method Mean Range Test 0.35 mg/dl 0.2 mg/dl ASTRA-N methods; acetoacetate caused

Group if severe positive interference in the ACA 0.10 -0.6 to 0.6 0.01 7/72 ---- SMAC -0.47 -2.9 to 0.2 0.01 20/72 ---- ASTRA, ASTRA-N, and SMAC methods. Group 2% In all three methods, acetoacetate in ACA 0.14 -0.1 to 0.4 0.01 ---- 19/45 SMAC -0.10 -0.3 to 0.2 0.01 ---- 20/45 creased the apparent creatinine concen tration in a linear fashion: ASTRA and *Bias = the mean of high power liquid chromatography results minus the mean of the results of the comparison ASTRA-N, 0.24 mg per dl increase per method. fGroup 1: n * 72, creatinine range 0.5 to 23.2 mg per dl mmol per liter acetoacetate; SMAC, 0.18 as determined by high performance liquid chromatography. mg per dl increase per mmol per liter $Group 2: n = 45, creatinine range 0.5 to 2.3 mg per dl, as determined by high performance liquid chromatography. acetoacetate. When acetoacetate was 508 HOLMES, OESER, KAHN, BEKERIS, AND BERMES measured in serum from a random sam sample pretreatment, — suggesting the pling of 30 non-diabetic inpatients on potential for interference by matrix con whom serum creatinines were ordered, stituents in some samples. a mean of 0.42 ± 0.33 mmol per liter The combination of sample pretreat was found. Serum acetoacetate was ment with cation exchange resins fol greater than 0.5 mmol per liter in eight lowed by reverse phase liquid chroma patients and greater than 1.0 mmol per tography was originally used without an liter in three patients. external standard to purify creatinine from biological fluids prior to analysis by Discussion gas chromatography/mass spectrometry. Creatinine isolated from serum using this The increased popularity of HPLC as technique was found to be homoge an analytical tool has led to the establish neous,9 suggesting that the procedure ment of serum creatinine assays based on could serve as a reference method for ion exchange,1,4 paired ion,11 and reverse serum creatinine. phase912 techniques. All of the HPLC The features of the method as pre methods appear to offer adequate sensi sented here include: (1) a high analytical tivity, precision, and specificity for cre recovery of creatinine, (2) the use of an atinine. The methods differ in the com external standard, (3) good precision, (4) plexity of the sample pretreatment and high specificity, and (5) stable chromato the complexity of the analytical system. graphic conditions. Moreover, the sam The ion exchange method of Brown et al4 ple pretreatment and the conditions of does not require sample pretreatment the HPLC separation are such that the prior to assay; however, creatinine in the test can be performed inexpensively column effluent is detected by the Jaffe using the most basic of HPLC equip reaction using a home-made micro ment. While the method as presented analyzer. The ion exchange method of here would not lend itself to routine use Ambrose et al1 requires preliminary de- in the determination of serum creatinine, proteinization of the samples with tri it has been found by the present authors chloroacetic acid followed by gradient to be useful as a reference method. elution of creatinine from the HPLC The results of our method comparison column, with detection by absorbance at showed that serum creatinine results ob 234 nm. The paired ion method of Soldin tained with five automated chemistry an and Hill11 features a simple pretreatment alyzers correlated with those obtained by step consisting of deproteinization with our HPLC reference method. Despite methanol followed by isocratic HPLC the excellent correlations, all of the test with detection by absorbance at 200 nm. methods except the EKTA, were nega This method, although extremely sensi tively biased relative to HPLC. Given tive, is less precise than the others and the high analytical recovery and the yields more complicated chromatograms. specificity of this reference method, it The reverse phase method of Spierto et would appear that the biases might be al12 involves deproteinization by ultrafil due, in part, to interference by non-cre- tration followed by isocratic HPLC with atinine chromogens in the Jaffe-based detection at 236 nm. An internal stan test methods. Differences in the re dard is included in this procedure re sponses of the test methods to creatinine sulting in improved precision. However, is another possible cause of the biases ob the chromatograms obtained using this served by us. Both proportional and con method are complex owing to the limited stant biases were reported in four dif CREATININE DETERMINATION BY HPLC 509 ferent Jaffe-based assays in a previous study, the average blood acetoacetate study in which the reference method concentration in 18 patients upon initial consisted of ion-exchange sample pre presentation with diabetic ketoacidosis treatment prior to creatinine determi was 3.03 ± 1.0 mM (range, 1.5 to 5.4).13 nation by the Jaffe reaction.3 It is notable that two out of 30 specimens The intermethod variability observed selected at random from a hospitalized, by us could be important when more non-diabetic population had acetoacetate than one instrument is used for serum concentrations sufficient to cause a spu creatinine determination in a laboratory. rious increase of 0.2 to 0.4 mg per dl in In a previous study, a decrease in serum the creatinine result as determined by creatinine of 0.35 mg per dl (from an ini the SMAC or the ASTRA. tial value of 2.6 mg per dl) was identified as the average medically significant change by a group of 63 specialists in in References ternal medicine.6 Based on our data, an 1. A m b ro se , R. T., K e tc h u m , D. E, and S m ith , J. W.: Creatinine determined by “high perfor inter-method difference of this magni mance” liquid chromatography. Clin. Chem. tude would not be uncommon. Such dif 29:256-259, 1983. ferences would be especially important 2. B la s s , K . G., Thibert, R., and L am , L . K.: A study of the mechanism of the Jaffe’ reaction. J. when the serum creatinine is in the Clin. Chem. Clin. Biochem. 12:336, 1974. normal range where small changes may 3. Bowers, L. D . and Wong, E. T.: Kinetic serum be indicative of a significant change in a creatinine assays. II. a critical evaluation and re patient’s glomerular filtration rate. view. Clin. Chem. 26:555- 561, 1980. 4. B ro w n , N . D., S in g , H. C., N e e le y , W. L., and The potential for the misinterpretation Koetitz, S. E.: Determination of “true” serum of changes in serum creatinine concen creatinine by high-performance liquid chroma tography combined with a continuous-flow mi trations as a result of intermethod biases croanalyzer. Clin. Chem. 23:1281-1283, 1977. is further emphasized by the differential 5. Davis, R. B., Thompson, J. E., and P a r d u e , effects of interfering substances on var H. L.: Characteristics of statistical parameters used to interpret least squares results. Clin. ious methods. While interference by Chem. 24:611-620, 1978. some of the most common Jaffe interfér 6 . Elion-Gerritzen, W. E .: Analytical precision in ants has been eliminated in the auto clinical chemistry and medical decisions. Am. J. mated methods used by us, acetoacetate Clin. Pathol. 73:183-195, 1980. 7. F a bin y , D. L., and E rtingshausen , G.: Auto severely interfered with two of the mated reaction-rate method for determination of methods. The magnitude of the interfer serum creatinine with the Centrifichem. Clin. ence was such that the increase in the Chem. 17:696-700, 1971. 8 . L i, P. K., Lee, M. H., M acGillinay, P. A., “apparent” creatinine concentration S c h a e f e r , P. A., and S ie g e l, J. H.: Direct, would be small for acetoacetate concen fixed—time kinetic assay for B-hydroxybutyrate and acetoacetate with a centrifugal analyzer and trations in the normal range (<0.2 mM). a computer-backed spectrophotometer. Clin. In physiological and pathological hyper- Chem. 26:1713-1717, 1980. ketonemic states, acetoacetate interfer 9. L im , C. K., R ic h m o n d , W., R o b in s o n , D. P., and Brown, S. S.: Towards a definitive assay of ence may be substantial depending on creatinine in serum and in urine: separation by the “true” creatinine concentration of the high-performance liquid chromatography. J. sample and the ratio of acetoacetate Chromatogr. 245:41-49, 1978. 10. N arayanan, S., and A p p l e t o n , H. D.: Creati (which interferes with the assay) to beta- nine: a review. Clin. Chem. 26:1119-1126, hydroxybutyrate (which does not inter 1980. fere). In extreme hyperketonemic states 11. Soldin, S. J., and H i l l , J. G.: Micromethod for determination of creatinine in biological fluids by such as diabetic ketoacidosis, the serum high-performance liquid chromatography. Clin. acetoacetate concentration may range Chem. 29:256-259, 1983. from 1.5 to 10.0 mM. In a previous 12. S p ie rto , F. W., M acNeil, M. L., C u lb r e th , P., 510 HOLMES, OESER, KAHN, BEKERIS, AND BERMES

D u n c a n , I., and B u r t is, C. A.: Development R. J., and Z ak, B.: Interesting interferences in a and validation of a liquid-chromatographic pro direct serum creatinine reaction, Microchemical cedure for serum creatinine. Clin. Chem. Journal 2i:370-384, 1976. 26:286-290, 1980. 15. W u , G . T ., T w o m ey , S. L., and T h ie r s , R. E.: 13. St e p h e n s , J. M., S ulw ay, M. J., and W a tk in s, Statistical evaluation of method-comparison P. J. : Relationship of blood acetoacetate and 3- data. Clin. Chem. 2i:315-320, 1975. hydroxybutyrate in diabetes. Diabetes 20:485- 16. Y o u n g , D. S., P est a n e r , L. G ., and G ibb er m a n , 489, 1971. V.: Effects of drugs on laboratory tests. Clin. 14. W a t k in s, R. E., F e l d k a m p , C. S ., T h ib e r t , Chem. 21:2860-2889, 1975.

STATEMENT OF OWNERSHIP, MANAGEMENT AND CIRCULATION (Act of October 23, 1962, Section 4369, Title 39, United States Code) Date of Filing—September 28, 1983 Title of Publication—ANNALS OF CLINICAL AND LABORATORY SCIENCE Frequency of Issue—Bimonthly Location of Known Office of Publication—230 N. Broad St., Philadelphia, PA. 19102 Location of the Headquarters or General Business Offices of the Publisher—Same as above Publisher—Institute for Clinical Sciences, Inc. Editor—F. William Sunderman, M.D., Ph.D. Managing Editor—Same as above Owner—Institute for Clinical Sciences, Inc. Known Bondholders, Mortgagees, and Other Security Holders Owing or Holding 1 Percent or More of Total Amount of Bonds, Mortgages or Other Securities—None Average No. Copies Actual Number of Copies Each Issue During of Single Issue Published Preceding 12 Months Nearest to Filing Date

A. Total No. Copies Printed (N et Press Run) 1300 1300 B. Paid Circulation 1. Sales Through Dealers and Carriers, Street Vendors and Counter Sales None None 2. Mail Subscriptions 1210 1188 C. Total Paid Circulation 1210 1188 D. Free Distribution by Mail, Carrier or Other Means, Samples, Complimentary, and Other Free Copies None None

E. Total Distribution (Sum of C and D) 1210 1188 F. Copies not Distributed 1. Office Use, Left-over, Unaccounted Spoiled After Printing 90 112 2. Returns from News Agents None None

G. Total (Sum of E and F — should equal net press run shown in A) 1300 1300 I certify that the statements made by me are correct and complete—F. WILLIAM SUNDERMAN, M.D., Ph.D.