Production and Certification of Secondary Enzyme Reference

CLIN. CHEM. 32/10, 1906-1913 (1986)

1906 CLINICAL CHEMISTRY, Vol. 32, No. 10, 1986

Production and Certification of Secondary Enzyme Reference Materials(ERMs). Part 2: The Need for Meaningful Protocols That Assure PhotometricAccuracy and a Well-Defined Process for Value AssignmentJ. C. Koedam, H. J. van Dreumel, W. v. d. Ham, G. M. Steentjes, and J. B. A. Terlingen

A collaborative study to assign values to two enzyme refer-ence materials (ERMs) was performed by 18 laboratorieswhose spectrophotometers were checked by us, just beforethe study. We measured the wavelength accuracy andrepeatability, the accuracy and linearity of the absorbancecurves, the cuvette pathlength, equilibration time, equilibriumtemperature, and a few other variables. Five spectrophoto-meters exhibited a marked wavelength-dependent nonlinear-ity. Most instruments were rather slow in bringing the sampleto the correct temperature and the final temperature wasoften too high. In the collaborative study, each participantperformed the same manual, well-described methods on fouroccasions in triplicate, using reagents prepared locally. Therelation between the photometric checks and the analyticalresults is discussed, as well as the treatment of outliers and

the effects on the variances. Suggestions are made aboutvarious facets of collaborative studies. The values assignedto the two ERMs carry a 95% uncertainty interval of ± 1-4%

of the mean.

AdditIonal Keyphrasee: interlaborat ory performance - variation,source of spectrophotometry . aspartate aminotransferase

alanine aminotransferase lactate dehydrogenasecreatine kinase

Having prepared enzyme reference materials (ERMs)’ ofhuman origin (1), here we discuss the testing and calibra-tion of instruments used for assigning activity values andthe results of the collaborative study set up to assign thesevalues.

In assays of enzymes, the accuracy of the results dependsentirely on the accuracy of the absorbance measurements,the accuracy of the temperature control, and the valueassumed for the molar absorptivity of(e.g.) NADH. All havebeen dealt with by Expert Panels of the InternationalFederation of Clinical Chemistry (2,3) and others (4).

Previously a few checks and measurements some timeafter a group of laboratories had participated in the assign-ment of values to a batch of ERM gave somewhat disap-pointing results. We found deviations in photometric accu-racy, optical (= radiation) pathlength, and the temperaturein the reaction cuvette. After correction for these errors inthe enzyme analyses and recalculation, the interlaboratoryresults were closer to each other.

Laboratory of Clinical Chemistry, National Institute of PublicHealth and Environmental Hygiene (R1VM), P.O. Box 1,3720 BABilthoven, The Netherlands.

‘Nonstandard abbreviations: SRM, Standard Reference Materi-al; AP5A, P,P-diadenosine pentaphoaphate; G6PD, glucoee-6-phos-phate dehydrogenase; IlK, hexokinase; YSI, Yellow Springs Instru-ment (Co.); BCR, Bureau Communitaire du Reference; other abbre-viations as in preceding paper (1).

Received April 14, 1986; accepted July 11, 1986.

This prompted us to combine a check on the instrumenta-tion with an evaluation of the actual program to assignvalues to our ERMs, mentioned earlier (1). After discussions

with the clinical chemists in the 18 participating labora-tories,2 we decided that two technologists from our labora-tory would visit to check their spectrophotometers.

The analyses for the value assignment were done a fewweeks after these visits. Although we routinely ask 10laboratories to analyze each component of our referencematerials, we now had the opportunity to cooperate with amuch larger group.

Materials

Reference Materials

Holmium oxide filter and attenuator (no. 9423 179 08501;Pye-Umcam Ltd., Cambridge, U.K.).

Didymium filter, as above.Set of three glass filters (NBS SRM 90c; National Bureau

of Standards, Gaithersburg, MD 20899). These filters were

used at 465 and 546 nm, for which certified values areissued by the NBS, and also at 415 nm (see below).

Yellow glass filter (Corning lIT yellow).Quartz cuvette, 10.00 mm (NBS SRM 932).Gallium melting-point standard (NBS SRM 1968).

Reference Instruments

Spectrophotometers (Pye-Unicam SP8-200 and SP8-250;Philips Nederland N.V., Eindhoven, The Netherlands), fit-ted with the reference cuvette or a 10.00-mm-pathlengthdouble-walled cuvette (Hellma 160 OS), through whichthermostated water was circulated (“Clinicon Precitherm”

waterbath; Boehringer Mannheim Nederland b.v., Almere,

The Netherlands).Electronic thermometer, Model 45 CU, fitted with a Model

4501 probe (Yellow Springs Instrument Co., Yellow Springs,OH 45387). This instrument was calibrated with the galli-um standard cell, SRM 1968, at 29.77 #{176}C.The calibrationwas controlled with a Model HP-2801-A and a Model HP-S-00 thermometer (Hewlett-Packard, Avondale, PA 19311).These two instruments had recently been recalibrated at theVan Swinden Laboratory of the “ljkwezen,” the official

2 analyses were performed in the laboratories for clinical

chemistry of the following hospitals: Assen, Psychiatrisch Zieken-him “Licht en Kracht”; Beverwijk, Rode Kruis Ziekenhuis; Drach-ten, Ziekenhuis “Nij Smelinghe”; Eindhoven, Catharina Zieken-huis; Eindhoven, St. Joseph Ziekenhuis; Ermelo, Ziekenhuis “Sa-lem”; Gouda, Bleuland Ziekenhuis; Groningen, R.K. Ziekenhuis;Helmond, St. Lambertus Ziekenhuis; ‘s Hertogenboach, Groot Zie-kengasthuis; Kampen, Stadsziekenhuis “De Engelenbergstichting”;Leiden, Academisch Ziekenhuis; L.eiden, Diaconessenhuis; Naar-den, Diaconeasenhuis; Nijmegen, St. Radboud Ziekenhuis; Tiel,Ziekenhuis Rivierenland; Utrecht, Wilhelmina Kinderziekenhuis;and Zwolle, Ziekenhuis “De Weezenlanden.”

Code

Ref.ABC0EFG

JKLMN0P0RS

CLINICAL CHEMISTRY, Vol. 32, No. 10, 1986 1907

calibrating facility of the Netherlands. All readingsmatched within 0.02 ‘C.

Reference Solutions

Potassium dichromate. Using NBS SRM 935 and potassi-um dichromate from J. T. Baker Chemical Corp., Deventer,The Netherlands (no. 309), we produced two series ofsolutions with concentrations of 0.00, 40.0, 80.0, 120.0,150.0, and 180.0 mgfL in doubly distilled water containing 1mmol of perchioric acid per liter. We compared the twoseries with each other by measuring their absorbances at350 nm in duplicate during five days, using a single set ofcuvettes. The linear correlation between the results was:ABaker = -0.0002 + 1.0003 ANp� (r = 1.0000). Because NBScertifies the molar absorptivities of potassium dichromatesolutions only up to a concentration of 100 mg/L and wewanted to use higher concentrations, we extrapolated theexisting linear relationship between concentration and ab-sorptivity to 180 mgfL. We checked the calculated valueswith our two spectrophotometers and found no differences

>0.1% between the calculated and measured absorptivities.The absorbances of the five solutions, calculated from theNBS-certifled molar absorptivities or obtained through ex-trapolation, were: 0.428, 0.858, 1.290, 1.614, and 1.940.These were denoted “NBS values.” We found no changes inabsorbances when we increased the bandwidth from 0.8(NBS specification) to 2.0 nm. We stored the two sets ofsolutions in brown borosiicate bottles at room temperature.

Cobalt sulfate. We prepared six solutions of this substance(Merck no. 2556) in 0.05 moIIL sulfuric acid. Their absor-bances were about 0.46, 0.78, 1.02, 1.47, 1.69, and 1.96. Wefiltered the solutions through glass fiber (pore size 0.7 pm;Gelman “Metrigard,” no. 64789) and stored them as de-scribed for the dichromate solutions. Differences in absor-bance did not exceed 0.00 1 A when we varied the bandwidthbetween 1 and 6 nm at a wavelength of 510 nm.

Cresol red. We prepared solutions of cresol red in Tris orbarbital-phosphate buffer as described earlier (5).

Reagents for value-assignment analyses. We used thefollowing specific chemicals: L-aspartic acid (no. 3303), L-

alanine (no. 1893), oxoglutaric acid (no. 3312), and N-acetylcysteine (no. 0460), all from J. T. Baker ChemicalCorp.; PP (no. 108979), creatine phosphate (no. 127574),imidazole (no. 353493), NADH (no. 107727), NADP� (no.128031), AMP (no. 102202), .ADP (no. 236675), AP5A (no.161632), LD (EC 1.1.1.27, no. 127221), malate dehydrogen-ase (EC 1.1.1.37, no. 127248), G6PD (EC 1.1.1.49, no.197762), and HK (EC 2.7.1.1, no. 208540), all from Boeh-ringer Mannheim B.V., Almere, The Netherlands. All otherchemicals were analytical grade.

We purchased single lots of each of the above-mentionedmaterials from a single source, and distributed them to theparticipants, a strategy also followed by the organizers ofthe alkaline phosphatase study (6). We realized that indoing so, we may have introduced lot-dependent bias and, tocheck whether this might play a role, we obtained foursubstances (NADH, PP, LD, and malate dehydrogenase)from each of five companies-British Drug House (BDH),Boehringer Mannheim, Calbiochem, Merck, and SigmaChemical Co.-and measured the ASAT activity five timesin each of five fresh clinical samples, by using reagentmixtures prepared with materials from each of the sourceslisted. The maximum difference was 3.4%; we saw nosystematic deviations between any of the reagents.

Methods and Results

Instrument Checks

Organization. A team of two technologists visited eachlaboratory, bringing with them all the requisite materialsfor the evaluation. They spent 4 to 6 h at each site, but webelieve that the time required for an adequate check can beshortened to 3 to 4 h. All solutions were remeasured at leastonce a week at our laboratory, and no changes could bedetected during five weeks.

Table 1 lists the spectrophotometers tested.Wavelength accuracy and repeatability. Wavelength accu-

racy was checked with the holmium oxide and didymiumifiters at 360.9, 453.3, 572.4, and 585.4 nm. Most measure-ments were performed at a bandwidth of approximately 0.5nm, but in three instruments 4- to 8-nm widths had to beused. Table 2 gives the mean values for the peaks deter-mined by using the techniques specified in the operatingmanuals of the various spectrophotometers.

To check the amount of hysteresis present, we alsomeasured the absorbances at 444 nm (a very steep part ofthe absorbance curve of holmium oxide) by manually scan-

ning upwards and downwards three times. We then trans-

formed the differences in absorbances found into differencesin wavelength. For nine of 19 instruments this value was0.1 nm or less; for the other instruments it ranged between

0.3 and 1.1 nm.Accuracy and linearity of absorbance vs concentration

curves. We measured the absorbances at or near severalwavelengths, covering most of the useful part of the near-ultraviolet and visible spectrum. Included were: 340 nm(frequently used in clinical chemistry to monitor NADH-

Table 1. CharacteristIcs of SpectrophotometersUght Mono- Beam

Model source chromatorb configuration”

Pye Unicam SP8-250 0 G DPye Unicam SP8-200 0 G DPye Unicam SP8-200 0 G 0Pye Unicam SP-1 800 W G DPye Unicam SP-1 700 W G DKontron Uvikon 810 Q G DKontron Uvikon 810 0 G DKontron Uvikon 610 CL 0 G SKontron Uvikon 820 0 G 0Zeiss PM-6 W G SZeiss PMQ-3 W P 5Zeiss PM 0-3 W P SShimadzu UV-190 W G DCaryNanan 118 0 P (2x) DBeckman 36 W G DBeckman25 W G DBeckman 25 W G DBeckman 24 W G DBeckman 24 0 G DBeckman 42 0 G S

= quartz, W wolfram light source.= grating, P = prism monochromator.

= double-; S = single-beam configuration.

Table 2. Waveleng th Accuracy (nm)

NominalMean

measured SD Extremes

360.9453.5572.4

360.9453.6572.7

0.310.350.42

360.2-361.4452.9-454.4572.0-573.4

CJR00

-7.55-4.70-3.75-3.01

-2.73

R

K

0

0NO

0KMfEl

8

6

� A.O.60

2

o iia

6A 0.78

4

2

04

8

SI

L14

EF

-�i.I�Ii 11

UI0z4a

a4

is #{149}415 nm

A #{149}0.56

4

A .0.73

a

6

4

2

08

6

UIC) 4z< 2a0Sna4

UI

C)z4a

a4

In

A.1.02

oPI 1

N LH

DJ SF

(I I [E1IOBAGE I I I

.1.46

OPJ

f�1 ID�l�H�

!�1.�9r�1 I I I

A.1.96 I

R MI

0 PJK

[El . M ILA�SS�Ll,.

:#‘4.1.0’

IsI�I

.#{176},�,I.

I5101�14, IT1IsioI�j..jI,

4.1.73

I I ,R�1. Fl

I

A

4

2.

0 -I

8

6A

4

2

0-�a,

6

4,

2

0 -f..

0 94 96 98 100

�. OF CERTIFIED VALUES

92 94 96 98 100 102

!.OF CERTIFIED VALUES

FIg. 1. Absorbances of (left) potassium dichromate at 350 nm, (middle) three NBS neutral glass and one Coming HT yellow glass filters at 415 nm,and (oghi) six concentrations of cobalt sulfate at 510 nm, expressed as percentages of the certified valuesEach square indicates a particular photometer (see Table 1)

102

1. OF CERTIFIED VALUES


dependent reactions), 415 nm (p-nitrophenol), and 546 nm(hemoglobin, bilirubin). Our team performed almost allmeasurements at 25#{176}C,using a bandwidth of 2 nm and thereference cuvette. We made the readings with the followingsolutions and filters: potassium dichromate at 350 nm atfive concentrations; cobalt sulfate at 510 nm at six concen-trations; three NBS glass filters at 415, 465, and 546 nm;and the yellow glass ifiter at 415 nm.

Figure 1 shows some of the results, the absorbances beingexpressed as a percentage of the NBS- or RIVM-certifiedvalues. Each square indicates a particular photometer. Theresults of the measurements with the NBS ifiters at 465 and546 nm are not shown but followed the same pattern as at415 nm. We also report (Table 3) a measure of the curvatureof the absorbance curves of some photometers at two wave-lengths.

Cuvette pathlength. In each laboratory we compared oneor two locally available and routinely used cuvettes with thereference cuvette, using potassium dichromate and cobaltsulfate solutions at absorbances of 0.8 and 1.5. All readingsfor each cuvette were averaged. The mean of the cuvettemean values was 10.01 mm (SD = 0.027 mm, n = 21, range9.96-10.07 mm).

Thermometers. We compared the readings given by the

8

6

4

2

0I0

8

6

4

2

Table 3. NonlinearIty of Curves for Absorbance vsConcentratIon a

Lab. Potassium Cobaltcod. dichromate sulfate

-0.33-0.04-0.20-0.13-0.25

“The magnitude of C X io�, from the equation y = a + bx + c�, where y =

absorbance and x = concentration.

best locally available laboratory thermometer with those ofthe YSI instrument. The mean was 29.94 #{176}C(SD 0.12 #{176}C,range 29.6-30.2 #{176}C).The corresponding values for 25 #{176}Cwere � = 24.98, SD = 0.12, range 24.7-25.5 #{176}C.

Temperature equilibration of reaction cuvette. We intro-duced a solution of cresol red, the temperature-dependentabsorbance of which had been measured at 25#{176}C,into aphotometer cuvette, and without delay placed this cuvette inthe photometer, which had been previously stabilized at30#{176}C.We then recorded the time for the contents of thecuvette to reach a temperature of 0.5, 0.2, and 0.1 #{176}Cbelow

6

4

2

08

6

Table 4. EquIlIbratIon of Cuvette Temperatures to 30 #{176}C

‘C below the final equilibrium


Lab.

code

temperature Final

equilibriumtamp, #{176}C-0.5 -0.2 -0.1

A 3.68 5.5 7.0 30.0B 3.7 5.3 6.6 30.0C <1.0 <1.0 <1.0 30.10 3.2 5.0 6.2 30.1E 3.7 5,6 7.2 30.1F 4.8 7.0 8.7 30.5G 2.0 4.5 6.0 30.0H 4.2 6.0 7.0 30.8I 3.8 5.4 6.5 30.3J 4.3 6.2 7.3 30.5K 3.8 5.8 7.3 30.0L 5.2 7.5 9.5 30.3M 4.5 6.5 7.7 30.4N 2.8 4.1 5.1 30.1

0 3.7 5.2 6.5 30.1P <1.0 1.3 1.7 30.00 3.5 4.9 6.1 29.8R 3.4 4.4 5.0 30.0S <1.0 <1.0 <1.0 30.0

Median 3.7 5.3 6.5

“Time in minutes necessary to reach the specified difference below the finalequilibrium temperature.

the final equilibration temperature (Table 4). We alsomeasured the final (stable) temperature reached. Because

the thermal mass of the YSI thermistor is extremely small,its introduction into a solution does not cause notableaddition or loss of heat.

In making these cuvette temperature measurements, wechecked for any thermal gradient between the upper andlower part of the solutions in the cuvette. In some cases this

difference was about 1 #{176}C.Finally, we detected differences as great as 0.5 #{176}Cbetween

the solution temperatures of four cuvettes in the cuvette

holders of a few instruments.Timing. We measured the 30-s time interval between

readings and found a 1-s difference in two instruments; inthe others, such errors were negligible.

Voltage fluctuations. Over several hours the voltage of themain line varied by 30 V around the nominal voltage of 220V; however, we never detected any appreciable influence ofthis on the photometric measurements.

Collaborative Study for Value Assignment

Analytical methods and organization. We used the meth-ods according to the recommendation of the Enzyme Com-mittee of the Dutch Society of Clinical Chemistry (NVKC).�

Our overall protocol specifies the following:#{149}one method, which is basically a manual one, is used for

each enzyme.#{149}the method is performed in each of the participating

laboratories.#{149}each laboratory

-prepares its own reagents, using doubly distilled

water,

3The methods used during the collaborative study were describedin great detail by a committee, formed by the NVKC and RIVM,and are in accordance with the norms issued by the Dutch Normal-

ization Institute and the International Organization for Standard-

ization. Similar method descriptions were prepared for about 20serum and urine components. Most of these descriptions are avail-able, in English, from our Institute.

-prepares the standard solutions (if applicable),-checks its instrumentation and calibrates pipettes

and diluters,-performs four runs on separate days or parts of the

day, using a freshly reconstituted sample each time, and-performs three tests with a single reagent blank in

each run.The analytical protocol specifies that:

#{149}each vial must be reconstituted with 2.00 (± 0.01) mL ofwater, to be controlled by weighing

#{149}the sample must be used between 0.5 and 2 h after

reconstitution#{149}photometric readings are always taken at 30-s intervals.

For ASAT and ALAT, six readings are taken, starting 30safter initiation of the reaction; for LD, five readings aretaken, starting 60 s after the start of the reaction; and forCK six readings are taken, starting 300 s after the start Ofthe reaction.

#{149}the lir.earity of the reaction curve is tested by calculatingthe S,.� (residual error of estimate or standard error ofestimate) (7, 8).

#{149}the mean of the three results in one run is calculated.

#{149}each laboratory then calculates the overall laboratorymean after subtraction of the blank value and sends allresults and raw figures to the RJVM.

The organizing laboratory (RIVM):#{149}organizes meetings with the participating clinical chem-

ists and their technologists before and after the study andgives additional guidance during the analyses, if necessary.

#{149}supplies most of the chemicals and checks the purity ofsome of them, such as NADH in this case

#{149}(re)calculates all the results from the raw figures andperforms checks on the reconstitution volume of the sam-ples, excessive noise or a non-linear curve (large S�) etc.

#{149}performs the statistical analyses, leading to the finalresults.

Results. The participants encountered no major problems,and they all returned their raw data and results in time. We(RJVM) recalculated all results and made slight adjust-

ments to them. There were no deviations >0.01 mL in thereconstitution volumes, so we made no corrections for thatstep in the procedure.

An analysis of variance (�ovA) was performed on all theresults, before removal of any suspect data or outliers (Table5).

DIscussIon

Instrument Checks

The object of this study was to assign activity values totwo batches of ERMs and to determine the influence ofpoorly calibrated instruments on the results of such acollaborative exercise. The 18 participants used varioussingle- and double-beam instruments (see Table 1). Theresults showed good control for the following variables:wavelength accuracy and repeatability (absence of hystere-sis), optical pathlength, main line voltage, and, to a lesserextent, accuracy of laboratory thermometers.

The wavelength-dependent nonlinearity of the absor-bance vs concentration curves was quite marked for abouthalf of the instruments. As shown in Table 3, we found thatthe deviations for potassium dichromate at 350 nm weremuch greater than those for cobalt sulfate. Two otherexamples are given in Tables 6 and 7. The absorbances wemeasured with photometer J agreed with those of thereference spectrophotometer (SP8-250) for the NBS filter

DF

17

54144

ASAT

Batch 5, 25.2 U/L Batch 6, 66.4 U/I.

U/Lb %#{176} U/L %

2.010.700.99

541927

6.691.712.15

641620

LD

Between lab.Between runWithin run


ALAT

Batch 5, 15.6 U/I. Batch 6, 94.1 U/i.

U/I. % U/I. %

0.790.590.95

342541

11.07.44.8

483220

Ci(

16 92.351 28.3

136 33.9

Batch 5, 215 U/L Batch 6, 235 U/I. Batch 5, 105 U/I. Batch 6, 221 U/I.

DF U/Lb U/L U/I. U/I.

60 92.518 59.522 31.2

51 14.432 3.617 7.0

Filler/Solution

NBS glassCobalt sulfatePotassium

dichromate

729

19

Potassium dichromate, 350 nm Cobalt sulfats, 510 mm

AbsorbanceSolution

no. Ref. Lab A

Absorbancs

1 4302 8603 12954 16205 1946

58 74.614 9.0

28 19.8

#{149}Mean enzymatic activity, as determined by the participants. b Vanance, expressed In U/I.. ‘Percentage of the total variance. One laboratory did not participate

with the analyses of LD and CK.

readings given by photometer J, for instance (Table 6), are0.2, 0.9, and 2.0% too low at about 0.9, 1.3, and 1.6 A,respectively, which seems to be of minor importance. How-ever, the slope of the absorbance curve between 0.9 and 1.3A will be 1.6% too low; between 1.3 and 1.6 A, this valuebecomes 5.9%, increasing to >10% near 1.75 A. The results

of the measurements of the aminotransferaaes will thereforebe too low to the same extent; turbid samples will beparticularly affected.

__________________________________ For several photometers the temperature equilibrationtime was not in control (Table 4). Several instruments took3-5 mm to reach a temperature 0.5 ‘C below the desiredfinal temperature. Moreover, photometers F, H, I, J, L andM were all relatively slow but the final temperature of thecontents of their cuvette was too high-as if the manufac-turers realized the rather poor equilibration characteristics

___________ __________ and tried to compensate by a small maladjustment of the

fl. Lab A temperature setting. Thus, depending on the circum-stances-the duration of the equilibration period and theinitial temperature of the reagents-the reaction tempera-

1018 ture can either be too low or too high, or the two errors may1451 compensate for each other. The duration of the pre-incuba-1670 tion also is important here.1924 For a nonlinear reaction, e.g., that involving LD, stan-

dardizing the time between the start of the reaction and thefirst reading of the absorbance is vitally important-but ishardly compatible with the variations in behavior of theinstruments.

We also found in several cases a temperature gradient

within the cuvette of as much as 1 ‘C, or temperaturevariations between the cuvettes within the cuvette holder.The former was more frequent in instruments in which heatwas applied in or near the bottom of the cuvette holder.

In conclusion, although a number of errors could bedetected, the majority of instruments tested were well undercontrol. However, a large percentage of the instruments hadbeen checked recently by a technical representative of themanufacturer; under “routine” circumstances, the situationmight be less favorable.

429856

12901616

1945

Solutionno.

1 4612 7793 10244 14665 16936 1961

Table 5. ANOVA of Results for Enzyme ActivItIes


Table 6. ComparIson of Llnearltles of PhotometersAbsorbance x iV

Wavelength,

nm Ref. Lab. J

415 1074 1076510 1466 1464350 430 432

860 8581295 12831620 1588

Table 7. ComparIson of Llnearltles at DlfterentWavelengths

and the cobalt sulfate solution, but not for the dichromatesolution (Table 6). The comparison between two Pye-Uni-cam instruments, however, gave the opposite result: theSP8-200 deviated from the SP8-250 only at 510 nm (Table7).

The latter effect was shown to be due to the occurrence ofstray light. The SPS-200 automatically inserts filters in thelight path at certain wavelengths, but not around 510 nm.After we added an extra interference ifiter, the erroneousreadings were corrected completely. The SP8-250 isequipped with a double monochromator and has very lowstray light at all wavelengths.

Linearity should thus be checked not only at a convenientwavelength (e.g., 465 or 546 nm with NBS glass filters) butalso at or very near the wavelength in actual use (e.g., 340nm). Nonlinearity can play an important role, especially in

the measurement of ASAT and ALAT, for which initialabsorbances may be around 1.3 or more. The absolute

Value Assignment

The literature gives little guidance about how to setup avalue assignment program. We know of only one extensive

Certification of enzyme reference materials

#{149}00

‘5

50

25

U/L

C �O O�0�9

0 0

as #{149} . #{149}#{149}s,.#{149} . .. S

ASAT

Batch 5, Batch 6,24.8 U/i. 65.6 U/L

o�,

- #{149}#{149}#{149}#{149}S

.... -


DF

12- 13

33- 41

88-106Varlance, in U/I.. ‘Percentage of total variance.

‘... .... - .- .... .... ,#{149} ....

�0 ASAT ALAT LD

it A B A B A B

Batch5 25 ±0.3 15.5 ±0.6 215 ±5Batch 6 66 ±1 94 ±1 235 ±5

Table 9. AssIgned ActIvIty Values (U/L)

Fig. 2. Results from four runs performed in each of 18 laboratorIes ofthe analyses of ASAT (#{149})and ALAT (0) in two ERMsResults from laboratories where spectrophotometers showed a marked nonlin-earity are marked with � a statistical outher is marked with x

CK

A B

105 ±2

221 ±5

A = certified value, B = 95% uncertainty interval (t� s/V�n) where t = valueof Student’s tat probability p = 0.05, s = standard deviation, n = number ofresults (13-17).


study (9) that was designed to determine a minimumprogram, based on subjective but well-defined criteria. Mostauthors only make general statements about the minimumnumber of participating laboratories and the maximumallowable uncertainty interval around the assigned values,e.g.:

#{149}“Ideally, objective clinical parameters would be used todefine the allowable uncertainty of the certified value of theEM ... unfortunately there is at present no possibility ofestablishing such relationships” (10).

#{149}“There is general agreement among laboratories experi-enced in interlaboratory testing that the preferable numberof participating laboratories is 15 or more” (11).

#{149}“The number of participating laboratories, number ofunits of EM and number of replicate determinations perunit must be sufficient to give an estimate of the consensusvalue with a precision suitable for most uses” (12).

The lack of agreement on the design of a value assign-ment system has led to an enormous variety of schemes (9,11, 13-23). The NCCLS-approved standard ASC-2, “Cali-bration reference materials and control materials in clinicalchemistry” (16), gave reasonable guidance in this field, butit has been withdrawn and is being replaced by two Guide-lines for calibration and control materials (25, 26) that,unfortunately, do not address the subject.

When we started the production and analyses of referencematerials (“accuracy control” materials) about 1957, four tosix laboratories participated in the value-assignment proc-ess. This number has gradually increased to 10 in 1985.During the present exercise, 18 laboratories were availableto participate. This larger group gave us the opportunity tostudy the effect of variations in the number of participantson the outcome of the analyses.

Performing analysis of variance on all results (Table 5)showed that the between-laboratory component of the totalvariance was the largest of the three. The between-runvariance was always small in comparison with the between-laboratory component and, in fact, in six out of eight caseswas the least of the three variance factors contributing tothe total variance. Therefore, increasing the number of runsdoes not seem necessary, because this would not appreciablydecrease the overall uncertainty in proportion to the addedeffort (17).

Data trimming. We removed some ASAT and ALAT

values for the following reasons (see also Figure 2):

a. Results from laboratories where we could clearly de-tect analytical errors (mostly nonlinearity at higher absor-

bances) during the period of the survey-laboratories C, D,J, 0, and R, as indicated with arrows in Figure 2. The effectof the slow temperature equilibration we observed severaltimes was more difficult to judge; it may have been partlycompensated for by the too-high temperature that wassometimes produced.

b. Results from laboratories that showed a large within-run or between-run error. Within-run differences were quitelarge (10% or more) in laboratories C and J; laboratory Jalso had large between-run errors. No further results were

removed because these laboratories had already failed thephotometric evaluation.

c. Results that failed a statistical outlier test (24)-threeof the four results from laboratory F (but not the results forALAT in batch 5).

One should keep in mind that the five laboratoriesmentioned above under a could be rejected a priori (theyshould not have participated at all in the analyses for theaminotransferases), which leaves a more realistic rejection

rate of only one out of 13. The nonlinearity does not play asignificant role in the analyses of LD and CK, because theyare measured at absorbance values near 1.0, where theabsorbance curves are still almost straight in all 18 photo-meters.

The results of an analysis of variance on the remainingdata for ASAT and ALAT are given in Table 8. All vari-ances have decreased, particularly the between-laboratoryvariance. The between-run variance found for the ALATresults in batch 6 are larger than those for ASAT, whichreflects the somewhat larger vial-to-vial variability and thegreater instability after reconstitution. No CK or LD valueshad to be removed. The mean values were only marginallyaffected by the removal of part of the results because,although most of the data we deleted were lower than theoverall mean, several of these results were unexpectedlyhigh. Apparently, some other factors beyond our controlstrongly influence the results of the analyses.

The final values that we assigned to the two ERMs aregiven in Table 9.

Table 8. ANOVA of Results for ASAT and ALAT afterRejectIon of Some Data

________________ ALAT

Batch 5, Batch 6,________ ________ 15.5 U/I. 93.8 U/I.

U/La %‘ U/I. % U/I. % U/I. %

0.01 1 1.24 33 0.55 32 2.13 26

0.52 48 0.73 19 0.47 30 4.13 500.55 51 1.78 48 0.61 38 2.02 24

12

10

8-

6

4

2

#{149}ASAT

o ALAT

a CK

x LID

0 1 2 3 4 5 6 7 8 9 10 11121314

Number of laboratories

Fig. 3. Effect of increasing numbers of laboratories on the magnitude of the uncertainty interval (see text for explanation)


Number of participants. Because it will seldom be possibleto find as many clinical chemists as are represented by those18 laboratories who are willing and able to cooperate in sucha time-consuming exercise as this, let us consider whathappens if the number of participants is varied. Suppose westart with three participants and increase their number.The uncertainty interval (see Figure 3) is reduced drastical-ly as we add the fourth and fifth participant, but thedecrease gradually becomes smaller as further results areadded. For comparison, the data from the recent alkalinephosphatase transferability study (6) are also shown inFigure 3.

It does not seem realistic to increase the number ofparticipants above 10, but a lower number makes it virtual-ly impossible to apply a rigorous statistical test to check forouthers. Whether the final uncertainty (95% interval) isadequate cannot be said at this moment; we consider itrather high, although the overall variances obtained weregenerally smaller than those of other studies (6,21).

ConclusIons

In accordance with these findings, we believe that evalua-tion of analytical data should primarily depend on both theclinical variables (27) and the analytical state of the art (28).In their paper, Skendzel et al. obtained so-called “medicallyuseful CVs” from physicians’ answers to questions on therelevance of certain changes in laboratory values. Theycalculated a useful CV of 14.3% for a change of ASATactivity at twice the upper value of the reference range andcompared this with the current level of infralaboratoryprecision of between 3.96% and 10.29% (CV) for ASAT atvarious activity concentrations (29). One could argue, how-ever, that the rather large fluctuation that seemed to betolerated by the physicians had to do with the present poorstate of interlaboratory standardization� Furthermore, theerror introduced by a bias in the assigned value of areference material should be small compared with the otheranalytical errors.

The scheme we used in this study and on other occasionsfor other analytes has both strengths and weaknesses. Theuse of well-described methods of proven reliability is superi-or, in our opinion, to the use of consensus values, such as themean of all results from (large) quality-control programs,even after statistical “outlier culling” procedures have beenapplied. Although such mean values for nonenzyme ana-lytes sometimes agree well with the values obtained with a

reference method or definitive method (30), there is noscientific justification for the assumption that the low andthe high results will adequately compensate for each other.Moreover, this approach is not applicable to enzyme analy-sis, systematic bias being inherently part of many enzymemethods. In enzymology one must rely upon recommenda-tions made by a national body or, preferably, by the IFCC,unless calibrated enzyme reference materials become gener-ally available.

Because our ERMs are primarily intended for Dutch use,we followed the NVKC recommendations, although valuesdetermined according to IFCC procedures were very similarin magnitude. As was shown in the previous paper (1), theERMs show good commutability and can therefore be usedto compare and calibrate methods as well as be used inquality-control programs.

Fortunately, international efforts to assign IFCC valuesto ERMs are now well underway, directed by BCR andNRSCIiNCCLSINationa1 Bureau of Standards. A primaryASAT reference material will soon be available from NBSas will a similar material for ganuna-glutamyltransferasefrom BCR. In combination with the stable and commutablesecondary sera we have described, all of them certified interms of the IFCC methods, the standardization of enzymemeasurements will certainly improve during the comingyears.

We gratefully acknowledge the advice, given to us by Drs. G. N.Bowers, Jr., and R B. McComb, Hartford Hospital, Hartford, CTand we are very much indebted to the colleagues and theirtechnologists, who skillfully performed the analyses for us.2

2.t.-�-vn-

e/o of mean o#{149}Alkaline phosphatase 1984 USA

value


References

1. Koedam JC, Steenijes GM, Buitenhuis 5, Schmidt E, Klauke R.Production and certification of secondary enzyme reference materi-als (ERMs). Part 1: Preparation of the sera and some of theirproperties. Gun Chem 1986;32:1901-#{246}.2. IFCC Expert Panel on Instrumentation. Guidelines (1984) forlisting specifications of spectrometers in clinical chemistry. IFCCrecommendations and related documents, Vol. 1. Berlin-New York:Walter de Gruyter, 1984:104-9.3. IFCC Expert Panel on Enzymes. Approved recommendation on�FCC methods for the measurement of catalytic concentrations ofenzymes. Clin Chim Acts 1979;98:163F-74F.4. Vanderlinde RE, Richards AH, Kowalski P. Linearity andaccuracy of ultraviolet and visible wavelength photometers: aninterlaboratory survey. Clin Chim Acta 61;1975:39-47.5. Van Dreumel HJ, Bosman HA, Koedam JC, Terlingen JBA.Kresolrood oplossingen voor temperatuur metingen in (spectro)foto-meters. Berichten uit het R1VM. (ISBN 90 346 0573 6) Bilthoven,The Netherlands: Rijks Instituut voor Vollcsgezondheid en Milieu-hygiene, 1984:178-80.6. Tietz NW, Rinker AD, Burtis C, et al. Transferability study ofthe AACC Reference Method and the IFCC method for measure-ment of alkaline phosphatase activity. Clin Chem 1984;30:704-6.7. Edwards AL. An introduction to linear regression and correla-tion. San Francisco: WH Freeman and Co., 1976:38-9.8. Velapoldi RA, Paule RC, Schaffer B., Mandel J, Moody JR. AReference Method for the determination of sodium in serum. NBSSpecial Publication 260-60, Washington, DC: U.S. Department ofCommerce/National Bureau of Standards, 1978:27.9. Passing H. An optimized design for the establishment of as-signed values in control sera: J Clin Chern Clin Biochem1981;19:1167-79.

10. Hjelm M. Components in a model for the production of certifiedbiochemical reference materials. In: Quality control in clinicalendocrinology. Eighth Tenovus Workshop, Cardifl U.K., 1979.11. Uriano G, ed. ISO Guide 35-1985 (E): Certification of referencematerials-general and statistical principles. P.O. Box 56, Geneva,Switzerland: International Organization for Standardization, 1985.12. Sut.arno B., Steger HF. ISOIREMCO Draft Guide 35. Statisticalprinciples for the preparation and interpretation of reference mate-rial certificates. Chapter 4: Certification of reference materials byinterlaboratory consensus. P.O. Box 56, Geneva, Switzerland: Inter-national Organization for Standardization, 1982.13. Horwita W. Report of the committee on interlaboratory studies.J Assoc Off Anal Chem 1983;66:455-66.14. Marchandise H, Colinet E. Assessment of methods of assigningcertified values to reference materials. Fresenius Z Anal Chem1983;316:669-72.15. Qualitatasicherung in der Laboratoriumsmedizin. Verfahrenzur Sollwertermittlung in Richtichkeits-Kontrollproben. Berlin,F.KG.: Normenausschusz Medizin (NAMed) im DIN (DeutschesInstitut fur Normung e.V.), 1983.

16. NCCLS Approved Standard ASC-2: Calibration reference ma-terials and control materials in clinical chemistry. 771 E LancasterAve., Villanova, PA: National Committee for Clinical LaboratoryStandards, 1975.

17. Hansert E, Stamm D. Determination of assigned values incontrol specimens for internal accuracy control and for interlabora-

tory surveys. J Cliii Chem Chin Biochem 1980;18:461-90.

18. Schumann V. Statistische Modelle zur Erstellung von Sollwer-ten und Sollbereichen in Kontrollproben. Med Lab 1976;29:271-6.

19. Peters T Jr (Chairman). AACC position paper. In: Boutwell JH,ed. A national understanding for the development of referencematerials and methods for clinical chemistry; proceedings of aconference. Washington, DC: American Association for ClinicalChemistry, 1978:47-85.

20. wuo Expert Committee on Biological Standardization. Techni-cal Report Series, 29th rep. 1211 Geneva 27, Switzerland: WorldHealth Organization, 1978:125.

21. Addendum to National Bureau of Standards certificate ofanalysis: SRM 909: human serum. Washington, DC: U.S. Depart-ment of Commerce/National Bureau of Standards, 1980.

22. Merten B. Die EffectivitAt von Ringversuchen als externeQualitatasicherung mi medizinischen Laboratorium, Bewertungs-masstAbe mid ihre Auswfrkung� Sollwert mid Sollbereich. Lab Med1981;6:16-24.

23. Milford Ward A, White PAE, Thompson RA, et al. Preparationof a calibration material for specific protein assay (SPS-01). AnnClin Biochem 1984;21:254-6.

24. Dixon WJ. Processing data for outliers. Biometrics 1953;9:74-88.

25. NCCLS tentative guidelines for calibration materials in clini-cal chemistry. 771 E Lancaster Ave., Villanova, PA: NationalCommittee for Clinical Laboratory Standards, 1982;2:499-526.

26. NCCLS tentative guidelines for control materials in clinicalchemistry. Ibid.,527-54.

27 Skendzel LP, Barnett RN, Platt R Medically useful criteria foranalytic performance of laboratory tests. Am J Cliii Pathol1984;83:200-5.

28. Lott JA, Massion CG. The College of American Pathologistsenzyme survey: a summary, Proc Aspen Conf Cliii Enzymol.Skokie, IL: College of American Pathologists, 1983:233-56.

29. Lohff MR, DiSilvio TV, Ross JW, et a!. Analytic clinicallaboratory precision. State of the art for selected enzymes. Am JClin Pathol 1982;78:634-43.

30. Whitehead TP, Woodford VP. External quality assessment ofclinical laboratories in the United Kingdom. J Cliii Pathol1981;34:947-57.

Documents

Production and Certification of Secondary Enzyme Reference