0095-1137/82/070168-06$02.00/0
Evaluation of Two Photometers Used in
Enzyme-Linked
Immunosorbent
Assays
V. M.GENTAtANDJ. H. BOWDRE*
Clinical Microbiology Laboratories, North Carolina MemorialHospital, University ofNorth CarolinaSchool
of
Medicine,Chapel
Hill, North Carolina 27514Received27August 1981/Accepted13April1982
The performance of two commercially available high-speed photometers, designed for through-the-plate reading, was evaluated. Linearity of instrumental reading and reproducibility of same-day and2-day measurementswere assessed by least-squares analysis and analysis of variance, respectively. For both instru-ments,the photometric error was on the order of thousandthsofan absorbance
unit andwasmuch smallerthan theerrorof thecurrentlyavailableenzyme-linked
immunosorbentassays.
High-speed photometers designed for
through-the-plate
reading
insure arapid
turn-over time for
measuring
colorimetric reactionsin microtrays.
Consequently,
thequality
oftheseinstruments is an
important
considerationin the adoption of
enzyme-linked
immunosor-bent assays(ELISA) and similarassays
by
labo-ratories with large numbers of
specimens.
Un-like conventional photometers, these
photometers use a vertical light path for
through-the-plate reading. Oneresultisthat the
length of the light path is notfixed by cuvette
thickness,sothat theabsorbance varies withthe
volumeof solutioncontained in each well. For a prototype ofone through-the-plate photometer (Titertek Multiskan; Flow Laboratories, Inc.,
McLean, Va.), Ruitenberg et al. (5) evaluated
intrarun imprecision and concluded that
photo-metricinaccuracy is probably minor in
compari-son with biological variability and dispensing
errorin theperformance of ELISA.We
expand-ed these studies and evaluated the linearity of
instrumental reading and the reproducibility of
measurementswithinthe samedayandbetween
2consecutive days fortwocommercially
avail-able photometers, Titertek Multiskan (Flow
Laboratories)andMicroelisaAutoreader
(Dyna-techLaboratories, Inc., Alexandria, Va.).
MATERIALS AND METHODS
Photometers. TheTitertek Multiskan was equipped with 405-, 414-, 450-, and 492-nm filters with a speci-fied half-band pass of 8 to 12 nm. We estimated the half-band pass of the 405-nm filter to be 10 nm. The Microelisa Autoreader MR580 was equipped with 410-, 455-, 490-, and 570-nm filters; the half-band pass
ofthesefilters is specified as 10 nm but could not be
t Present address: General Hospital Pathologists, Ltd., General Hospital ofVirginia Beach, Virginia Beach, VA 23454.
assessed byusbecause of thedesign of the instrument. Both photometers make an automatic background absorbance correction by means of a selected cup (Microelisa Autoreader)or aselected column of cups (TitertekMultiskan) filled with anappropriate buffer solution.
Testchromophoresolution. Aconventional chromo-genic solution commonly used in ELISA (8) was
prepared byincubatingtheenzymealkaline phospha-tase (from calfintestine, type VII; Sigma Chemical Co., St.Louis, Mo.) with its substrate, p-nitrophenyl-phosphatedisodium(Sigma104phosphatase substrate tablets) in diethanolamine buffer (pH 9.8) at room
temperaturefor 30 min. Thereactionwasthenstopped by the addition of 3 N NaOH. This chromophore solution was diluted with diethanolamine buffer to
produce test solutions in the range of 0.050to 1.400 absorbance units. The absorbanceof thetestsolution
at405 and 410nm wascheckedby repeated measure-mentswithacalibrated dual-beamspectrophotometer (Beckmanmodel 25; Beckman Instruments, Inc., Palo Alto, Calif.) before and after assays with each of the twophotometers.
Microdispenser. The test solution was dispensed into the cups ofmicroplates with a calibrated micropi-pette(Gilson Pipetman; Rainin Instrument Co. Inc., Woburn, Mass.). The coefficient of variation of this micropipettewaslessthan1%.
Microplates.Flat-bottomed polystyrene microplates (Linbro EIA microplates; Flow Laboratories) with 12 columnsof eight wells each were used.
Experimental designs. (i) Linearity of instrumental
reading. (a) For each microplate, the first column of cups wasfilled withbuffer, and columns 2 through 6 were filled with 200
RId
of test solution containing increasing amounts of chromophore. (b) For each microplate, the first column of cups was filled with buffer, and columns 2 through 6 were filled with 50, 100, 150, 200,and250,u1of a test solution containing aconstantconcentration of chromophore.
(ii) Effect of adding solvent to a determined amount of
chromophore.Toassess the effect ofdispensing error inaddingasolvent to a determined amount of chromo-phore, thefirst column of cupsof a microplate was
168
on February 7, 2020 by guest
http://jcm.asm.org/
EVALUATION OF TWO PHOTOMETERS IN ELISA
0.6[
E
C
It)
0 04
0.2
0.6I
E
C
0
0.41
0.2
0.4 0.8 1.2
A 41On (BECKMAN, MODEL 25)
FIG. 1. Comparison of the Multiskan and Mi-croelisa Autoreader with a dual-beam spectrophoto-meter.Abscissas: Absorbancesat405nm(A) and410 nm(B)determined withaBeckman model 25
spectro-photometer for 200 ,ulofsolutionscontaining
increas-ingamountsofchromophore. Ordinates: Absorbance of thesamesolutions determined at 405nm with the Multiskan(A) and 410nmwith the Microelisa
Auto-reader (B). (-) Least-squares regression lines; (0)
meansofeight independent measurements; (---)95%
predictionintervals.
filled with 50 Fl of buffer, and columns 2 through 5 werefilled with 50,ul of chromophore solution. Then tocolumns 3through5wereadded50,100,and150,ul
of buffersolution, respectively.
(iii) Intrarunimprecision. The intrarunimprecision
of instrumentalreadingswas assessedby performing eight repeated measurements of 200 ,ul of solutions with absorbancereadingsin therangeof 0.050to0.850 absorbance unit.
(iv) Imprecision intradayand betweendays.The first column ofamicroplatewasfilled with 200,ulof buffer solution. Columns 2through12werefilled with 200,u1
oftestsolution.Groupsofindependentmeasurements weretakenthesamedayorover2 consecutivedays.
To minimizeevaporation, theplateswere sealed and
keptinarefrigeratorat4°C.
Statistical analysis. Least-squares regression
analy-sis (3) and one- and two-way analyses of variance
(ANOVA) (3) were employed to study linearity and
reproducibilityof instrumentalmeasurements,
respec-tively. Data were analyzed by using the Statistical
Package for the Social Sciences (4) or the Minitab
Statistical Package (6), which were available to us
through the University of North Carolina and the
NorthCarolinaMemorial Hospitalcomputer systems.
The graphical display ofmeanscomparisonswas
con-structed by the method of Andrewsetal. (1).
RESULTS
Sinceboth of these photometersuse avertical light path for through-the-plate reading, the lin-earity of instrumental reading was verified by regression analysis of the absorbance curves generated by either increasing the amounts of chromophore in afixed volume oftest solution
orincreasing the volume oftestsolution contain-ing a fixed concentration of chromophore.
Fig-ure 1 shows the linearity of readings when increasing amounts of chromophorewere
pres-entin afixed volume oftestsolution. Readings
obtained with a dual-beam spectrophotometer
were used for comparison. For the Titertek Multiskan, the regression equation was y =
0.6
E
C
In)
0
0.41
0.21
0.41
E
C
0ci
0.2
50 100 150 200 VO LU ME (p
I1)
FIG. 2. Linearity ofreadings withincreasing vol-umesof the same chromophore solution. Abscissas: Volumes ofchromophoresolution. Ordinates: Absor-bancesdetermined at405 nmwith theMultiskan(A)
andat410nmwith the Microelisa Autoreader(B). (-)
Least-squares regression lines; (0) means of eight independentmeasurements;(---)95%prediction
inter-vals. A.MULTISKAN FLOW
yz0.002+0.542X/X
r2-0.999
¼~~~~~~~~~~~~~
A405nm
B.MICROELISA AUTOREADER DYNATECH
_ A,
9=O0.004+O0.584X
r2.0. 999 /
I,
/I
X
A.MULTISKAN FLOW 7
y 0.052+0.002X ---I r2 O0.999
I
B.MICROELISAI
AUTOREADER DYNATECH
A ,
y:0.002+0.002X
r2.0.999
I,,
- I
-I | I
B.MIREIS VOL. 16,1982
F
on February 7, 2020 by guest
http://jcm.asm.org/
TABLE 1. Effect ofdilutingaconstantamountofchromophore: ANOVA resultsfor theMicroelisa Autoreader andTitertek Multiskan
Photometer Source df squares
Sum
of Meanof squaressum
Fratio FprobabilityMicroelisa Autoreader Between 3 0.0001361 0.0000454 0.76 P = 0.526 Within 28 0.0016617 0.0000593
Titertek Multiskan Between 3 0.001839 0.000613 P = 0.029
Within 28 0.004916 0.000176 3.49
0.002+0.542x, and the coefficientof determina-tion, r2,was0.999. For the Microelisa Autoread-er, the regression equation was y = 0.004 +
0.584x, and r2was0.999. Thelarge coefficient of
determination indicates a strong linear relation-ship between readingswithadual-beam spectro-photometer and with each of the two
photo-meters.Itwasapparentbyvisualinspectionthat both regression lines were linear; thiswas
con-firmed by thetestfor lack offit(P < 0.05) (3). The dashed lines delimit the 95% prediction interval, whichwas narrowandwasabout 0.010
absorbance unit over the range of
measure-ments. These data indicate that the spread around the regression line was minimal, since for each level ofx,95% ofthe observed values of
yfell withinanintervalof 0.020 absorbance unit. Thelinearity of instrumental readingwasalso checked against volumetric increments of the
same chromophore solution (Fig. 2). For both
photometers, least-squares analysis generated straight regression lines with narrow 95%
pre-diction intervals of about 0.015 absorbance unit and r2 = 0.999. These results confirmed the
linearity of instrumental readings and indicated that,asexpected, these photometers differ from
conventionalphotometers in that the volumetric
errorin dispensing a chromophore solution
af-fects the overallerror.Theslopes of both
regres-sion linesindicate thatavolumetric increment of
1 ,ul determined a photometric increment of
0.002absorbance unit.
Conversely, for a through-the-plate-reading
photometer, the volume in which a given
amount of chromophore is diluted should not
affect themean absorbance measurements. We verified this hypothesis under extreme condi-tions. To four columns of eightwells ofa
micro-plate containing 50,ul of chromophore in dieth-anolamine bufferweadded 50, 100,or150
RI
ofthesamebufferor nobufferatall. The hypothe-sisofequality ofmean absorbancesamong col-umns was then tested with one-way ANOVA. The ANOVA summary tables forthe Titertek
MultiskanandMicroelisaAutoreaderareshown in Table 1. For both photometers, the sums of
squares(estimates of the variance) between and
within groups (columns) were very small. For theTitertekMultiskan, the F ratio is 3.49 (P =
0.029). For the Microelisa Autoreader, the F
ratio is 0.76 (P = 0.526). The results indicated
thatdilution ofagivenamountofchromophore
in anonabsorbing solution did not significantly
affectthe mean absorbance measurements.
The imprecision of instrumental measure-ments was assessedby comparingthe means of groupsofmeasurements taken either in the same
run, in different runs the same day, or over 2
consecutive days. Table 2 shows the mean,
standard deviation, and coefficient of variation
(CV) forgroups ofmeasurementstaken from the
same microplate. The CV for bothinstruments
wasless than
12%,
evenfor the smallestabsor-bance value in therangestudied.TheCVvalues
obtained with the Microelisa Autoreader were
somewhat smaller than those with the Titertek
Multiskan.
The one-way ANOVA for the means of groupsof eightmeasurementstaken on the same
day with the Microelisa Autoreader did not
reveal statistically
significant
differencesbe-tweenthesemeans(P=
0.468).
Similarly,
statis-tically significant differences were not found
betweenmeansofgroupsofmeasurements
per-formedover2consecutive
days (P
= 0.779).Since the light beam for the Titertek
Multi-skan isdivided intoeight
independent
pathways,TABLE 2. Intrarunimprecisionofphotometer
measurementsfromasingle microplate
Photometer Intrarun imprecision
group Mean SD cv
Microelisa Autoreader
1 0.052 0.006 11.5
2 0.106 0.004 3.8
3 0.217 0.002 0.9
4 0.430 0.002 0.5
5 0.844 0.007 0.8
Titertek Multiskan
1 0.061 0.007 11.5
2 0.108 0.010 9.2
3 0.210 0.008 3.8
4 0.413 0.008 1.9
5 0.805 0.025 3.1
on February 7, 2020 by guest
http://jcm.asm.org/
TABLE 3. TitertekMultiskan: imprecision of same-day results two-way ANOVA between columns and rows
(D
Source df Sum of Meansumof Fratio Fprobability
squares squares
Rows 7 0.00083909 0.00011987 13.26 P<0.0001
Columns 10 0.00017861 0.00001786 1.98 P =0.0491
Error 70 0.00063266 0.00000904
which are aligned with corresponding rows of
cups of a microplate, the imprecision was
as-sessed by simultaneouslytesting theequality of
mean measurements for columns and rows by two-way ANOVA (Table 3). There were no
statistically significant differences between
col-umns (P= 0.0491), but the differences between
rows(eightseparatelight paths) were
statistical-ly significant(P < 0.0001). This large F ratio is
explained by the very small mean sum of
squares error (0.00000904). This statistic is an
unbiased estimate of the variance of the entire
population
of
wells andwas one orderofmagni-tude smaller than the mean sum of squares
between rows (0.00011987). Thus, it was the
very precision of the measurements which
al-lowed asmallsystematic error betweenrows to
be detectedasstatistically significant. The
actu-al range of mean readings for rows was very narrow(0.557to 0.565 absorbance unit)(Fig. 3),
andtherewas partial overlapping of95%
confi-dence intervals (bars). Differences onthe order
0.567
-
0.564-E
C
0
0.61
0.558
{-A B C D E F G H
ROWS
FIG. 3. Imprecisionof the Multiskan readings be-tween rows (eight separate light paths). All wells contained 200 Fl of the same chromophore solution. Abscissa: Plate coordinates for the rows. Ordinate: Absorbances at 405 nm. The means of11
measure-mentsareindicatedasdots and their95% confidence intervals asbars.
of thousandths of an absorbance unit are of no practical significance in results of ELISA, in
which the inherent variability is much greater
(7).
The means of 11 groups of eight measure-ments taken on 2 consecutive days were
com-pared (Fig. 4). The differences withindayswere
minimal, and the 95% confidenceintervals(bars)
were largelyoverlapping. The absorbance
read-ings of day 2 were consistently greater than
those ofday1. Theone-wayANOVAgaveanF
ratio of 5.02 (P < 0.0001). However, again the
pooledstandard deviation was very small(0.004
absorbance unit), and the range of the means was narrow(0.550to0.561 absorbanceunit). In
this case also, very small differences were
de-tectable as statistically significant because the
overallvariabilityofrepeated readingswasvery
small. Differences ofthis magnitude in ELISA
results are of minimal practical significance.
Further studies todetermine the relative
impor-tanceof the various componentscontributingto
thesumofsquareerrorsforsame-dayand2-day
assays were not
performed. However,
there-sults of the dilution
experiment
indicated thatevaporation
of the solvent would have atrivialeffect.
i -I I 2
DAY
FIG. 4. ImprecisionoftheMultiskanover2days.
Allwells contained200 ,ul ofthe same chromophore solution. Abscissa: 22groups ofmeasurementstaken
on2consecutivedays. Ordinate: Absorbances deter-minedat405nm.Themeansof five
repeated
measure-ments are represented as dots and their95% confi-denceintervalsasbars.
VOL.16,1982
on February 7, 2020 by guest
http://jcm.asm.org/
0.7 _ 0.7 I~~~~~~~~~~~
0
0.5 0-0.4
Aw
sobc at44100 a -A455nm
wavelngthode o opertion)Well 455,aine 200
-j Y=0.0044-0.584X Y=0.006+0.450X
m r250999rsom O :
length, 410nm) or in thedual-wavelength mode(0.3E
003 0.
legts n 45n) In
E
ah
IEreiaAtradr
samasomirplts ca opr ednsa w ae
0.1 0.1
0.2 06 1.0 1.4
A 'BECKMANMODEL 25 4IOnm
FIG. 5. MicroelisaAutoreader, comparisonof
ab-sorbance at410nmwiththatat410 455nm (dual-wavelengthmodeofoperation). Wells contained 200
pd of solution with increasing amounts of
chromo-phore.Abscissa:Absorbancedeterminedwitha
Beck-man model25 spectrophotometer. Ordinates:
Absor-bance determined with the Microelisa Autoreader
operatingeither in thesingle-wavelengthmode (wave-length,410nm)or inthedual-wavelengthmode
(wave-lengths,410and 455 nm).
The Microelisa Autoreader, as a means of
correctingfor scratchesorotherartifacts onthe
microplates, cancomparereadings at two
wave-lengths selected from the four standard filters
supplied
with the instrument. One of thesewavelengths is selected to be near the
absorp-tionmaximum ofthechromophoreused,andthe other is selected to serve as a reference
wave-length. Ifthereference wavelength is absorbed
to some extent by the chromophore, the
dual-wavelength mode gives absorbance values whichare smaller than those measured with the
one-wavelengthmode. Figure5shows thiseffect
by comparing the regression lines obtained for
the same chromophore solutions at the
wave-length of 410 nm (this is the regression line shown inpB)Fig. and atthe wavelength of410 455 nm (dual-wavelength mode). The two
regression lines were significantly different by
visual inspection, and this difference was
con-firmed by the statistical test for equality of
regressionlines (P< 0.0001) (3).
DISCUSSION
These studies indicated that for these two
commercially available through-the-plate-read-ing photometers, linearity in measuring
absor-bance of solutions with increasing amounts of
chromophoreandsame-dayand2-day
reproduc-ibilitywere satisfactory.
Theimprecision of instrumental readingswas
verysmall,ontheorder of thousandths ofaunit
over arange of 0.050 to 1.400absorbanceunits.
Theseestimatesaresimilartothosereported by
Ruitenberg etal. (5). Asmight be expected, the
CV forboth instruments was smaller for
read-ings ofatleast0.2absorbance unit than formore
dilute solutions. Therefore, the CVcanbe
mini-mized by designing ELISAtests togive
absor-bance values greater than 0.2 for samples of
interest.
The Titertek Multiskan showed statistically
significant differences among mean absorbance
readings of the eight individual
light paths
andbetween mean absorbance readings of the 2
consecutive days. However, these statistically
significant differences were very small (0.003
and 0.004 absorbance unit,respectively) and of
nopractical clinicalimportance for the available
ELISAtests, in which differences of 0.003
ab-sorbance unit are notcritical in
evaluating
testresults.
Forboth
through-the-plate-reading
photome-ters, there was a linear
relationship
betweenvolumes of chromophore solution and
absor-bance
readings,
and theslope
was 0.002absor-bance unitfor both the Titertek Multiskan and
the Microelisa Autoreader(Fig. 2). This clearly
indicates, as do the results ofRuitenberg et al.
(5), that volumetric inaccuracy in dispensing
chromophore is amajorsource oferror.
Thevolume in whichadeterminedamountof
chromophore
was diluted did not affect theinstrumental
readings.
Consequently, in ELISA,volumetric errors inadding buffer and inhibitor
solutions do not affect the overall error of the
assay.
Toobtain comparable results of ELISA
per-formed in laboratories with different
through-the-plate-readingphotometers, the photometers
should be calibrated
against conventional
spec-trophotometers. Furthermore, for the
Mi-croelisa Autoreader, the performance of the
dual-wavelengthmodeshould becompared with
thatof theone-wavelength mode. Itisimportant
tonote thatreadings taken in these two modes
were not comparable, particularly for the larger amountsofchromophore (Fig. 5).
These studies were addressed to aspects of
the evaluation of two commercially available
through-the-plate-reading photometers. The
re-sults of this study confirm and extend those of
Ruitenbergetal. (5),who studied aspects of the
performance ofaprototype oftheTitertek
Mul-tiskan. For anexhaustive evaluation ofthe
per-formance ofspectrometers, thelisting of
specifi-cations proposed by the International
on February 7, 2020 by guest
http://jcm.asm.org/
EVALUATION OF TWO PHOTOMETERS USED IN ELISA 173
Federation of Clinical Chemistry should be
con-sulted (2).
LITERATURE CITED
1. Andrews, H. P., R. D. Snee, and M.H. Sarner. 1980. Graphical display ofmeans.Am.Stat. 34:195-199.
2. Haeckel, R., C.H.Collonbel, T. D.Geary, F. L. Mitchell, R.G. Nedeau,andK.Okuda. 1980. Provisional guidelines (1979) for listing specifications ofspectrometersin clinical chemistry. Clin. Chim. Acta 203:249-258.
3. Neter, J., and W. Wasserman. 1974. Applied linear statisti-calmodels. Richard D.Irwin, Inc., Homewood, Ill. 4. Nie, N. H., C. H.Hull, J. G.Jenkins, K.Steinbrenner,and
D. H. Bent. 1970. Statistical package for the social
sci-ences,2nd ed. McGraw-Hill Book Co., New York.
5. Rultenberg, E. J., V.M.Sekhuis, and B. J. M. Brosi. 1980. Some characteristics ofa newmultiple-channelphotometer forthrough-the-plate reading of microplatestobeused in enzyme-linked immunosorbentassay. J. Clin. Microbiol.
11:132-134.
6. Ryan, T.A., Jr., B. L. Joiner, and B. F. Ryan. 1976. Minitab:astudent handbook.Duxbury Press, North
Scitu-ate,Mass.
7. Vejtorp,M.1978.Enzyme-linked immunosorbentassayfor determination of rubellaIgG antibodies. Acta Pathol. Mi-crobiol.Scand. Sect. B 86:387-392.
8. Voller, A., D. Bidwell, and A. Bartlett. 1980. Enzyme-linked immunosorbent assay, p. 359-371. In N.R. Rose and H.Friedman (ed.), Manual of clinical immunology, 2nd ed. AmericanSociety forMicrobiology, Washington, D.C.
VOL.