715.full

CLINICAL CHEMISTRY, Vol. 17, No. 8, 1971 715

Feasibilityof Multiple Simultaneous Enzyme Assays,for

Diagnostic Purposes,with the GeMSAEC Fast Analyzer

Thomas 0. Tiffany,1 George F. Johnson,2 and Max E. Chilcote

The GeMSAEC fast analyzer provides the clinical chemistry laboratory withan analytical instrument that can be used to perform large numbers ofkinetic enzyme analyses. Precise enzyme-rate analyses can be doneroutinely, on a large scale, and at a decreased cost per test. Improvedprecision in analyses of enzymes should provide more reliable data be-cause analytical variation is lessened. We have asked how the fast analyzermight provide more useful diagnostic information to the clinician. We haveselected the ratio of SGOT to SGPT activity in serum as an example, andexamined instrumental precision. The coefficients of variation of the ratio,determined in the range of 50 and 140 Karmen units (which represents slightlyelevated to clearly elevated values), are 4.8% and 2.2%, respectively. Weexamined the feasibility of measuring two or more enzyme activities simul.taneously in one sample, to produce a diagnostic enzyme profile. Deter-mination of SGOT, SGPT, and GLDH in parallel is presented as an example.In addition, we illustrate spectrophotometric linearity at 340 nm and discussinstrumental noise and an experimental approach to determining it by useof a premix experiment.

Additional Keyphrases diagnostic aids . liver-function assessment #{149} in-strumental noise #{149} SOOT, SGPT, GLDH, CPK #{149} parallel (profile) assays ofseveral enzymes #{149} CENTRIFICHEM

The introduction of the GeMsAEc3 high-speed,parallel-sample, fast analyzer by Anderson (1) hasprovided research laboratories in general and theclinical laboratory in particular with an analyticalinstrument capable of performing high-volumeenzyme-rate analyses. The ability of this instru-

From the Department of Clinical Chemistry, H. J. MeyerHospital Division, Erie County Laboratories, 462 Grider St.,Buffalo,N. Y. 14215.

NIH ClinicalChemistry postdoctoral Fellow, Departmentof Biochemistry, State University of New York at Buffalo,Buffalo,N. Y. 14214. Present address: MAN Program,Oak Ridge National Laboratories,Oak Ridge, Tenn. 37830.Send reprintrequeststo this address.

2 NIH ClinicalChemistry PostdoctoralFellow,Departmentof Biochemistry,State University of New York at Buffalo,Buffalo,N. Y. 14215.

Nonstandard abbreviationsused: GeMSAEC, an acronym forGeneral Medical Sciences Atomic Energy Commission. KU,Karmen unite (the oxidationof NADH can be measured interms of the decreasein absorbance at 340 nm. A decrea.seof0.001 is equal to 1 icu/ml/min); sooT, serum glutamic-oxalo--acetic transaminase (i-aspartate: 2-oxoglutarate amino trans-ferase, EC 2.6.1.1.);SGPT, serum glutamic-pyruvic transaminase(o-alanine: 2-oxoglutarate aminotransferase, EC 2.6.1.2); GLDH,glutamate dehydrogenase [1.-glutamate: NADP oxidoreductase(deaminating)]; CPK, creatine phosphokinase (ATP: creatinephosphotransferase,EC 2.7.3.2); mA, milliabsorbance units.

ment to perform a large number of tests in parallelallows its operator to run the reaction long enoughto obtain a sufficient absorbance change to assurereasonable precision in the normal range and stillmake a large number of enzyme determinationsper hour. The small sample and reagent volumerequired by the fast analyzer makes it possible toperform kinetic rate enzyme analyses at a greatlydecreased cost per test.

Pragmatically, the fast analyzer offers a solutionto a problem facing the clinical chemistry labora-tory in that it provides precision data at a lowercost per sample and is capable of handling anever-expanding enzyme workload. Furthermore,multiple-sample, parallel, fast analysis should makeit feasible to perform several different enzymeassays simultaneously on one or several samples,with the advantage that the resulting equality oftemperature and sample handling should lead tomore reliable comparisons of the various activitiesof the enzymes in the serum sample. In this man-ner, a panel of enzymes can be assayed as a pro-file for a certain disease state and provide moreinformation for differential diagnosis. If the fastanalyzer can perform enzyme assays with suffi-cient precision and provide a series of enzyme

716 CLINICAL CHEMISTRY, Vol. 17, No. 8, 1971

assays rapidly, at a reduced cost to the patient,and yield useful diagnostic information to theclinician, it will have made a significant contribu-tion to clinical enzymology.

In our initial evaluation of a commercial proto-type of the GCMSAEC fast analyzer we have ex-plored the potential of the fast analyzer to produceprecision enzyme data by measuring the coefficientof variation of the ratio of SOOT activity to SGPTactivity. Also we have investigated the feasibilityof running different enzyme assays simultaneouslyon several serum samples, and present data for sucha profile (for liver-function assessment), con-sisting of SGOT, SGPT, and GLDH. Information con-cerning the instrument evaluation pertinent tothis study is also presented.

Methods and Materials

ApparatusThe fast analyzer used in this work is a CEN-

TRIFICHEM (Union Carbide Research Institute,Tarrytown, N.Y. 10591), a prototype model pro-vided to us for initial evaluation. The transferdisc on this instrument has a capacity of 29 sam-ples. Samples were pipetted satisfactorily by usingvarious sizes of Eppendorf micropipets. Reagentwas added to each of the 29 sample compartmentsor sectors on the transfer disc with a Biopetteautomatic pipet (Schwarz/Mann, Division ofBecton, Dickinson & Co., Orangeburg, N.Y.10962).

Materials

Enzyme kits and reagents used in this studywere purchased from the Boehringer MannheimCorp., New York, N.Y. 10017.

Methods

Enzyme determinations performed in this studywere done in accordance with the manufacturersrecommendations except for temperature, time ofreaction, and volume of reagents used. The tem-perature used was 30#{176}C.The results for severalSGOT and SGPT determinations on samples andcontrols in the normal and moderately enhancedrange indicated the rate to be linear over a A340change of 1.0 absorbance unit. We thus felt justi-fied in running our SGOT, SGPT, and GLDH for 10mm, to obtain better precision in the normal andmoderately elevated ranges. A 1.0-unit absorbancechange would correspond to a sample activity of147 mU/ml or 237 KU (at 25#{176}C).Under the above-described conditions most samples would not haveto be diluted and 75 to 100 samples could be as-sayed per hour, in addition to controls. A samplevolume of 50 d was used for SGOT, and SGPT, 100jl for GLDH. Reagent volume was 400

Results

Initial Concerns

Most of the ultraviolet spectrophotometricenzyme assays now in use are optimized eitheraccording to Henry et al. () or Bergmeyer andBernt (3). For enzyme determinations involvingthe conversion of NADH to NAD the optimized pro-cedures require an initial NADH concentrationwithin the range 0.17 to 0.20 mmol/liter. SGOT,SGPT, and GLDH procedures require larger serumvolumes in the reaction mixture, because activityis less in normal sera. The additive absorbance ofserum and NADH produces initial starting absor-bances of 1.6 to 2.0 absorbance units. In systemswhere reference blanking with an absorbing solu-tion is not practicable, good spectrophotometriclinearity is required. The linearity obtained forNADH solutions in the CENTRIFICHEM (Figure 1) isquite good to 2.0 absorbance units. The linearityof SOOT and SGPT with respect to time is shown inFigure 2.

Contribution of Instrumental Noiseto Total Analytical Variation

In the kinetic methods we used, a zero-orderreaction rate is considered to be proportional toactivity of a particular serum enzyme. Determina-tion of this zero-order rate frequently requires thata small absorbance change with time be measuredat moderately high absorbanees. In a measurementmade under these circumstances, instrumentalnoise can contribute substantially to the observedvariation between replicate samples.

If the standard deviations involved in measur-ing two different absorbances are known, then thestandard deviation of the measurement of the differ-ence between these two absorbance values can becalculated. This is shown in the following expres-sion: $42 = S2A, + S242, where S4, and 84 arethe standard deviations in the separately measuredabsorbance values and &A is the standard devia-tion in the measurement of the difference betweenthese values. The coefficient of variation owing tonoise that would be found upon replicate deter-minations of A can be expressed as &A X 100/A.In the case where a small absorbance change athigh absorbance values is measured, a useful ap-proximation is that $4, = S42. The coefficient ofvariation due to noise can then be expressed asj2(S4) /A. Because most kinetic enzyme assaysof clinical interest exhibit zero-order kinetics over aconsiderable absorbance change, the contributionof noise to the observed analytical variation isminimized by maximizing the measured iSA.

A useful method for studying instrumental noisewith CENTRIFICHEM involves a premixed type ofexperiment. In these experiments, enzyme reac-tions were started in batch by adding serum to

EC0

Ce

UiC,z4

0C,,

4

Fig. 1. Spectrophotometric linearity at 340 nm

1 2 3 4 5 6 7 8 9 10TIME Mm

Fig. 2. Linearity of SGOT and SOPT assay with respect totime (. SOOT, 143 U/liter; #{149}sopT, 90 U/liter)

Table 1. CPK Premix Experiment to AssessInstrumental Noise

LV, mm * x 10 SAA X 10 CV, % SA//2 - 10

Av 1.0

CLINICAL CHEMISTRY, Vol. 17, No. 8, 1971 717

reagents, mixing well, and then pipetting the react-ing mixture into the transfer disc. Data weretaken over a period of time after the reacting mix-ture was spun into the individual cuvets. The re-sults of such an experiment with CPK are shown inTable 1. Absorbance at 340 nm is increasing withtime. As the reaction is measured over longertimes, the coefficient of variation decreases, withthe standard deviation in the measured EA remain-ing constant. Since absorbance is changing at the

2 15.2 1.3 8.8 0.94 30.9 1.3 4.2 0.96 45.2 1.6 3.5 1.18 60.3 1.6 2.7 1.1

10 75.7 2.1 2.7 1.512 88.5 0.9 1.0 0.614 103.4 1.6 1.6 1.116 119.7 1.6 1.3 1.118 134.3 1.6 1.2 1.1

same rate in each cuvet, differences between theEA values measured at the same time for eachcuvet reflect only the random noise of the instru-ment. If there were significant differences betweenthe reaction rates in the cuvets, the standard de-viation in A would increase with A, which in apremix experiment would imply contamination ofthe cuvets. The data from premix experiments forvarious assays allow a time interval to be set foreach assay so that acceptable precision can beobtained in the normal range. A coefficient ofvariation attributable to noise of 1% to 2% at theupper end of the normal range for the enzymes inthis study was considered satisfactory. This allowsthe obtainable precision within a run to be limitedmainly by pipetting errors.

The de Ritis Quotient

Our first approach to our attempt to obtainmore useful diagnostic information and diagnosticenzyme profiles was to see whether or not the in-strument can perform with sufficient precision tomake current tests more reliable. SGOT and SGPTform the basis of a very simple diagnostic enzymeprofile. De Ritis et al. (4) demonstrated a largeincrease of the two serum transaminases in patientssuffering from acute hepatitis, with the increase ofSGPT activity being more pronounced than that inSOOT. This resulted in their suggestion that the useof the ratio of SGOT to SGPT is diagnostically usefulfor differentiating predominately inflammatorylesions from necrotic processes in the liver. Theutility and limitations of this approach have re-cently been reviewed by Zimmerman and Seeff(5). Wilkinson (6), in a recent review of theclinical significance of enzyme activity measure-ments, pointed out that the data obtained in thismanner can be equivocal because of the lack ofprecision in determining the activity of theseserum enzymes, and thus the ratio has not beenextensively applied in diagnosis.

The experimental approach was to select apooled serum control and a commercial enzymecontrol that would provide enzyme activities inthree ranges of interest: normal, slightly supra-normal, and supranormal. We examined in-discvariation and the variation during two weeks.Both SOOT and SGPT were run in parallel on onedisk. Table 2 summarizes these data. In-run varia-tion is 4.6% (1 SD) in the normal range, and1.6% (1 SD) in the abnormally high range.

For a within-disc run, where temperature andreagent variations should be the same, the majorcontribution to the overall error for large per-minute changes in absorbance is due to pipetting.The supranormal control absorbance changes at0.0603 absorbance units per minute (90 U/liter) ora total change in absorbance of 0.603/A per 10mm. The coefficient of variation, 1.6%, obtained

718 CLINICAL CHEMISTRY, Vol. 17, No. 8, 1971

Table 2. In-Disc and Two-Week AnalyticalVariation in Results for SGOT and SGPT

In.dlsc variation: (N - jj)a

A. SGOT(Pooled serum) 48KUX = 20.7 0.7mA/mm; CV = 3.41%

B. SGPT (Pooled serum) 18WLUX = 8.1 0.3 mA/mm; CV = 4.58%

C. SGPT (Control A) 143 WLUX = 60.3 0.9 mA/mm; CV = 1.57%

Two-week variatIon

A. SGOT(Pooled serum) (N = 108)X = 20.9 1.1 mA/mm; CV = 5.20%

B. SGPT(Pooledserum)(N =94)6X = 8.1 0.6 mA/mm; CV = 7.71%

C. SGPT (Control A) (N = 85)6X = 60.1 1.3 mA/mm; CV = 2.12%

N = no.ofdiscsdetermined.

= no.of samples.

for this range is representative of the combinedpipetting error for sample reagent.

The propagation of error for a simple functionu = x/y (where the ratio of SGOT/SGPT is an ex-ample) is presented in the appendix. The result isan expression for the coefficient of variation of theratio, and is given by the equation cv = (cv2 +cv,2) 1/2 If the coefficients of variation for SOOT(x) and SOPT (y) are assumed to be the same (whichis reasonable for an in-disc run of this type), thenthe coefficient of variation for the ratio will becomecv,, = cv,V. With the coefficients of variationlisted in Table 2, the coefficient of variation for theratio in the range of 50 KU at 25#{176}C(30 U/liter) is4.8% and decreases to 2.2% in the range of 140KU at 25#{176}C(90 U/liter) where differential diag-nostic significance has been attributed to the ratio.Thus, as a first approximation, we can cite 2.2%(1 SD) as the analytical variation that we wouldexpect for the use of the ratio in this range.

We have compared the activities of severalSOOT and SGPT samples, as measured on the CEN-TRIFICHEM in parallel, with results for automatedprocedures currently being used in the laboratoryfor SGOT and SOPT (Table 3). The automated SOOTmethod is an adaptation of the method describedby Morgenstern et al. (7), and is designed to cor-relate with spectrophotometric methods. Correla-tion coefficients of 0.92 and 0.99 were obtained forthe SOOT comparison. The standard error of esti-mation (s,,) is lower in both cases for the sootdata than for the SGPT data, and indicates lessscatter about the linear regression line. The SGPTdata are a correlation with a modified automatedReitman-Frankel procedure (unpublished) per-formed on a discrete analyzer. The correlation

coefficients are still good but the larger standarderror of estimation indicates greater scatter aboutthe linear regression line. Giusti et al. (8) have re-ported a correlation study between the spectro-photometric and Reitman-Frankel methods. Theyhave reported that correlation is poorer for theSOPT methods, but that the variation coefficientis of about the same magnitude.

Feasibility of PerformingDiagnostic Enzyme Profiles

The necessity of collating the results of two ormore enzyme determinations to form enzymepanels or diagnostic profiles for differentialdiagnosis and for determining patterns of ab-normality has been established (5, 9, 10). rFheCENTRIFICHEM can perform several differentenzyme determinations in parallel. The largesample-per-disc capacity, in conjunction withsmall reagent and sample volumes, makes this areasonable approach. Most enzyme assays ofclinical interest involve either coupled NAD/NADHassay systems or NAD or NADH as a substrate, sothat wavelength is not a limiting factor.

Henley et al. (11) have suggested that the mito-chondrial enzyme ourn, found in high concentra-tion in the liver, is of value in assessment of liver-cell necrosis. To test the feasibility of the diagnosticenzyme profile approach, GLDH was set up anddetermined in parallel with SOOT and SGPT. Sampleswere preincubated in the rotor in the presence ofNADH and ammonium ion for 20 mm to removeendogenous substrates. Nine sets of samples can bedetermined simultaneously on the CENTRIFICHEMwhen performing a profile that requires threedifferent enzyme determinations.

A replicate analysis of SOOT, SGPT, and GLDHsamples, run in parallel, are shown in Table 4.GLDH activity was determined at 30#{176}C.The GLDHassay is less precise in the normal range because ofits very low activity in the serum.

The rate curves obtained for SOOT, SOPT, GLDH,run in parallel on a sample from a patient withacute hepatitis, are shown in Figure 3. The SGPTspectrophotometric method used in this study wasessentially that of Wroblewski-LaDue (1P2)and wasnot optimal (2). A higher sGoT/sGPr ratio than 1 isfound in this range by some authors (5) and proba-bly reflects the use of suboptimal assay systems(8).

Discussion

We have attempted to show that the multiple-sample, parallel, fast analyzer is becoming a practi-cal reality. We have demonstrated some acceptableprecision data for SOOT and SGPT run in parallel,and have established that of confidence limit for the

In

0.95

0.84

0.60

0.89

Syx

3.4 0.92 0.86

10.9 0.99 0.98

11.7 0.91 0.83

-1.1 0.88 0.78

32631.2

14.8

65.2

375

44 50

223 252

I I

SAMPLR a / sCOOTII 205SK.U.// 75511.1.

55 OPT///7 7131/ #{163}OLOH -

RSI.U.

Sample 3 5SGOT(XU) SGPT(WLU) GLDHU/ml

3 5 3 S

PS 40 40 18 20 5 43 121 122 83 896 26 24

38 46 42 52 57 6 558 43 44 69 77 5 419 25 23 17 17 5 332 17 17 9 10 4 334 29 28 17 166 6 39 25 23 16 17 9 6

41 85 86 49 476 3 4

Range of errorfor the replicate samples:0%-2%.2%-4%.4%-6%.PS, pooled serum samples.

CLINICAL CHEMISTRY, Vol. 17, No. 8, 1971 719

Table 3. Correlation of Data on SGOT and SGPT Activities with Automated Laboratory Data

SGOTN = 111SGOTN = 31SGPTN = 111SGPT

Y = mX + b; Syx = standard errorof estimation.Samples lessthan 100 KU.Samples greaterthan 100 KU.

1 X (Sy-x/V)X1006.1 41 42 14.5%

9.6%

35.2%

29.2%

0

4I

S GOT. SGPT-GLDH(PARALLI!L mUN)

0.I

0.7

0.3

4 5 S 10 15 14

TI MS (MIt..)

Fig. 3. Simultaneous determination of sooT, SGPT, andOLDH on a serum sample from a patient with acutehepatitis

ratio of SGOT to SOPT. We have attempted to showthat the CENTRIFICHEM fast analyzer can produceprecision data and that the use of a ratio deter-mined under carefully defined conditions shouldnot be equivocal because of analytical variation.

Furthermore, we have demonstrated that threedifferent assays for enzyme activities can be per-formed on several different samples simultane-ously, and that the results can be used for a specificdiagnostic profile. Thus, enzymes of interest canbe assayed in one sample, at the same time, at thesame temperature, and with the same (and verylittle) handling of the sample. This approach mayassume a role in special chemistry or as an emer-gency function, where simultaneous information onCPK, LDH, and SOOT activity could be obtainedvery rapidly. The usefulness of this approach iscurrently being evaluated in this laboratory.

We wish to express appreciation for technical assistance pro-vided by Mary Howe, Colleen Heigl, Sandy Casey, and SandyMarszalek. This study was supported in part by TrainingGrant in Clinical Chemistry, 5 TOl GM 01960 (NIGMS).

Table 4. SGOT-SGPT-GLDH Parallel Run. Replicate Analysis of Samples

References

720 CLINICAL CHEMISTRY, Vol. 17, No. 8, 1971

1. Anderson, N. (1., A multiple-cuvette rotor for a new micro-analytical system. Anal. Biochem. 32, 59 (1969).2. Henry, H. J.,Chiamori, N., Golub, 0. J.,and Berkman, S.,Revised spectrophotometric determination of glutamic oxalo-acetic transaminase, glutamic pyruvic transaminase and lacticacid dehydrogenase. Amer. J. Clin. Pathol. 34, 381 (1960).3. Bergmeyer, H. U., and Bernt, E., Glutamic-oxaloacetic trans-aminase. Determination with malic dehydrogenase as indicatorenzyme with optimum conditions for measurements. In Methodsof Enzymatic Analysis, Academic Press, New York, N.Y., 1963,p 837.

4. De Ritis, it., Coltori, M., and Giusti, G., 1)iagnostic valueand pathogenic significance of transaminase activity changes inviral hepatitis, Minerva Med. 47, 167 (1956).5. Zimmerman, H. J.,and Seeff, L. B., Enzymes in HepaticDisease. In Diagnostic Ens ymology, Coodley, E. L., Ed. Lea andFebiger, Philadelphia, Pa., 1970, p 1.6. Wilkinson, J. H. Clinical significance of enzyme activitymeasurements. CLIN. CHEM. 16, 882 (1970).7. Morgenstern, S., Okiander, M. Auerbach, J.,Kaufman, J.,and Klein, B., Automated determination of serum glutamicoxaloacetic transaminase. CL1N. CHEM. 12, 95 (1966).8. Giusti, G. Ruggiero, G., and Cacciatore, L., A comparativestudy of some spectrophotometric and colorimetric proceduresfor the determination of serum glutamic oxaloacetic trans-aminase and glutamic pyruvic transaminase in hepatic disease.Enzyrn. Biol. Clin. 10,17 (1969).9. Schmidt, E.,Schmidt, F. W., Horn, H. D., and Gerlach, U.,The importance of measurement of enzyme activity in medicine.In Methods of Enzymatic Analysis, Bergmeyer, H. U., Ed. Aca-demic Press, New York, N.Y. 1963, p 651.10. Bat.sakis, J. G., Briere, H. 0.,Enzymatic profile of myo-cardial infarct. Amer. Heart J. 72, 274 (1966).

11. Henley, K. S.,Schmidt, E.,and Schmidt, F. W., Enzymesin Serum. Their Use in Diagnosis. Charles C Thomas Co., Spring-field, Ill. 1966, p 54.12.Wroblewski, F.,and LaDue, J.S.,Serum glutamic pyruvictransaminase in cardiac and hepatic disease. Proc. Soc. Exp.Biot. Med. 91,569 (1956).13.Parratt,G.,Probabilityand Experimental Errors in Science.John Wiley and Sons, New York, N.Y., 1961, p 114.

AppendixThe equation for the propagation of error for a function

u = f(x,y) is given by

#{246}u2 #{246}u2S,,2 = (_) 3,,2 + () 52 (1)where S,,2, and 5,2 are the variance (13). For the equationu = x/y, #{246}u/#{246}x= l/y and #{246}u/y= -x/y2 so that the ap-plication of Equation 1 after substitution and rearrangement,gives

S,.2/(x2/y)2 = + _ix2 y2 (2)

Given the following identities, x2/y2 u2, S,,2/,,2 = cv,,2-= cv2 and S,2/,2 = cv,2 (where cv,,, cv,, and cv

represent the coefficient of variation of u, x, and y), theresulting equation for the propagation of error for a ratiox/y is

cv,, = (cv2 X CV,2)h/2 (3)