CLIN.CHEM.

39/9, 1804-1810 (1993)

Characterization and Mathematical Correction of Hemolysis Interference in Selected

Hitachi 717 Assays
Dennis W. Jay”2’3 and Debra Provasek2

The effect of hemolysis on several assays performed with the apparent analyte concentration in the presence of
the Hitachi 717 was quantified by relating the amount of interferent, and A0 is the analyte concentration in the
error to the concentration of hemoglobin. Hemolysis inter- absence of interferent, is often dependent on analyte
ference was judged clinically significant when analyte concentration and may give misleading results when
concentration varied by >10% from the initial value. He- applied to analyte concentrations other than the one(s)
molysis interference was significantfor alkaline phospha- tested. Absolute error, calculated as A1 A0, is often
-

tase, aspartate aminotransferase, cr-amylase, bilirubin, independent of analyte concentration (10-14); conse-
creatine kinase, y-glutamyltransferase, lactate dehydro- quently, it can be applied to a wide range of analyte
genase, lactate dehydrogenase-1, potassium, and theo- values.
phylline assays. Error (expressed in absolute terms) was The dependence of interference on analyte concentra-
linearly dependent on hemoglobin concentration and in- tion can be determined by the model of Kroll et al. (13,
dependent of the initial analyte concentration in each 14), wherein multiple concentrations of analyte are
case, except for bilirubinand theophylline, where multiple tested at each concentration of interferent. This method
regression analysis was required to quantify the effect. offers a rigorous statistical treatment of interference by
Relative error was dependent on the initial analyte con- relatingabsoluteinterferenceto hemoglobin concentra-
centration in all cases. Correction formulas were calcu- tion and provides insight into the mechanism of hemol-
lated from linear regression of absolute error vs hemoglo- ysis interference. However, a statistical program is re-
bin concentration. Clinical application of correction formu- quired for data reduction.
las and mechanisms of hemolysis interference for each Here, we propose a simplified method of interference
assay are discussed. evaluation based on clinical significance. In addition, we
report the correction factors for hemolysis interference
IndexIngTerms: analytical error enzyme activity - bilirubin for selected analytes on the Hitachi 717.
theophylline
MaterIals and Methods
Hemolysis has long been recognized as a source of Reagents. Reagents tested from Boehringer Mann-
error in a variety of chemical analyses (1-9). Major helm Corp. (Indianapolis, IN) included those for assays
sources of error have been attributed to the release of of albumin (bromcresol purple method), alkaline phos-
erythrocyte contents, to spectral interference from he- phatase [in 2-amino-2-methyl-1-propanol (AMP) buffer],
moglobin, and to interference of hemoglobin or its de- alanine aminotransferase (IFCC-recommended meth-
rivatives with chemical reactions. od), a-amylase [with p-nitrophenyl phosphate (PNP)
Quantitative correction for the effect of hemolysis substrate], a-amylase [with 4,6-ethylidene(G7)-p-nitro-
through measurements of serum hemoglobin was first phenyl (G1)-a,D-maltoheptaoside substrate (EPS)],
proposed by Mather and Mackie for potassium and phos- P-amylase (with EPS substrate), aspartate aminotrans-
phate (1). Caraway (2) proposed a correction formula ferase (IFCC-recommended method), bicarbonate, bili-
based on the relative distribution of analyte between rubin (2,5-dichlorophenyldiazonium tetrafluoroborate
serum and erythrocytes. His approach is valid, but re- method), urea nitrogen, calcium (in y-aminobutyric acid
quires prior knowledge of the intracellular erythrocyte buffer), cholesterol, creatine kinase (N-acetylcysteine-
analyte concentration and is applicable only to correc- activated method), creatinine, y-glutamyltransferase,
tion for the dilutional effects of erythrocyte contents. glucose (hexokinase method), lactate dehydrogenase,
Glick et al. (8,9) have published “interferographs” for lactate dehydrogenase-1, sodium, potassium, chloride,
several instrument-specific applications, but their stud- phosphate, total protein, triglycerides (glycerol phos-
ies present error in relative terms. The percentage of phate oxidase method), and uric acid (p-ami-
relative error, expressed as 100(A1 - A0)1A0, where A, is nophenazone method).4 Also tested were bicarbonate
(EM Diagnostic Systems, Inc., Gibbstown, NJ) and theo-
1Depment of Pathology and Laboratory Medicine, Texas
phylline (GDS Diagnostics, Elkhart, IN).
A&M University College of Medicine, and2 Pathology and Labo-
ratory Medicine Service, Olin E. Teague Veterans’ Center,Tem-
ple, TX 76504.
3Mdreas for correspondence: Department of Pathology and Lab- 4Nonstandard abbreviations: AMP, 2-amino-2-methyl-1-pro-
oratory Medicine, Olin E. Teague Veterans’ Center, Temple, TX panol; PNP, p-nitrophenylphosphate;EPS, 4,6-ethylidine(G7)-p-
76504. Fax 817-771-3098. nitrophenyl (G1)-a,D-maltoheptaoside; IFCC, International Feder-
Presented in part at the AACC national meeting, Chicago, IL, ation of Clinical Chemistry; NCCLS, National Committee on Clin-
July 1992 (Cliii Chem 1992;38:1026, abstract 392). ical Laboratory Standards; and HEPES, 4-(2-hydroxyethyl)-1-
Received September 24, 1992; accepted March 26, 1993. piperazuneethanesulfonic acid.

1804 CLINICAL CHEMISTRY, Vol. 39, No. 9, 1993

 Sodium aside (cat. no. S-2002), Tris (T-1503), and 4-(2- Individual reagent components were examined for in-
hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES; teraction with hemolysate by substituting the component
H-3375) were obtained from Sigma Chemical Co. (St. of interest for reagent Ri on the Hitachi 717 and perform-
Louis, MO). AMP (cat. no. 8-A896) was obtained from ing the alkaline phosphatase assay with the R2 volume set
J. T. Baker Chemical Co. (Philipsburg, NJ). to zero and the R2 stirring paddle removed. Absorbances
Adenylate kinase activity was measured by omitting were recalled with the reaction monitor function.
creatine phosphate from the creatine kinase reagent R2. Spectral studies were performed with a Shimadzu
Preparation of hemolysates. Hemolysates were pre- UV-160U spectrophotometer (Shimadzu Corp., Kyoto,
pared by an alteration of the method of Meites (15) as Japan).
follows. Remove a 2-mL aliquot from a well-mixed Temperature equilibration on the Hitachi
717. The
EDTA-treated whole-blood sample <4 h after collection time to achieve temperature equilibration
inside the
and centrifuge at 1500 x g for 5 mm. Remove the cuvette was determined by the method of Bowie et al.
plasma and discard, taking care not to remove cellular (17). Cresol red (10 mg/L in 0.1 mol/L Tris buffer, pH
elements. Wash the cells three times as follows: add 10 7.5) was used as a sample and 0.1 molJL Tris buffer, pH
mL of 0.15 moIJL sodium chloride, mix by inversion, 7.5, was used as Ri and R2 reagents. Absorbance was
centrifuge at 1500 x g for 5 mm, and remove the saline. monitored by using a primary/secondary wavelength
Add type I reagent-grade deionized water (1.0 mL) and pair of 570/660.
vortex-mix for 30 5; after storage at -30 #{176}C for at least
1 h, thaw the mixture and centrifuge for 5 mm at 10 000 Results
x g. Filter the supernate through a 1.2-jun (pore-size) Hemolysis interference expressed as percent relative
ifiter (cat. no. 02360; Schleicher & Schuell, Inc., Keene, error was found to be dependent on the concentrations of
NH), followed by a 0.2-jm (pore-size) ifiter (cat. no. both analyte and hemoglobin in each case. An example
4454; Gelman Sciences, Ann Arbor, MI) to remove cel- is illustrated for a-amylase (PNP method) in Figure 1A.
lular material.
Hemoglobin analysis. On the Hitachi 717, hemoglobin 0
was measured bichromatically with primary/secondary
wavelengths of 5 70/600 nm as reported previously (16).
The hemoglobin assay was standardized with a dilution -20
of 100 jtL of well-mixed, fresh, EDTA-treated whole
blood in 10.00 mL of type I reagent-grade water. Whole-
I-
blood hemoglobin was analyzed by the cyanmethemo-
-40
globin method on the Coulter S+W analyzer (Coulter
Electronics, Hialeah, FL), calibrated with Coulter S-Cal
calibrator.
-60
Hemolysis interference studies. Serum pools were pre-
pared from clear, visibly nonhemolyzed patients’ sam-
ples. We prepared at least 5 mL of specimen for each
-80
analyte, selecting sera to represent low, normal, and high
0 2 4 6 8 10
values in comparison with the reference range for the
analyte of interest. At least three analyte concentrations Hemoobln, g/L
were tested in each case. The amounts of hemolysate
added to serum varied for each analyte, but we have 0
found the following protocol satisfactory. To 0.5-mL ali-
quota of serum we added 2-, 5-, 10-, 20-, and 30-j.L -I
aliquots of hemolysate, to yield hemoglobin concentra- -20
tions of -0.5, 1, 2, 4, and 5 g/L, respectively. Samples
were analyzed for hemoglobin and the analyte of interest.
Error was calculated in both relative and absolute
I-40
terms, according to the aforementioned equations, and
was plotted as a function of hemoglobin concentration.
Clinical significance was judged relevant when absolute -60
error exceeded 10% of the result obtained for the non-
hemolyzed sample. Correction factors were calculated -80
by regression analysis of absolute error on hemoglobin
concentration. Multiple regression analysis was per- 0 2 4 6 8 10

formed with the SPSSIPC+ program (V4.0.1; SPSS Inc., Hemoobln, g/L
Chicago, IL) with the observed analyte concentration as
Fig. 1. Effect of hemolysis on the a-amylase assay with p-nitrophenyl
the dependent variable and with the analyte, hemoglo- phosphate substrate: (A) relative error - I100(A1 - A0)1A0J; (8
bin, and the product of hemoglobin and analyte concen- absolute error, A1 - A.)
trations as independent variables. Initial amylase actIvity: 64 (#{149}),
170 (#{149}),
and 516 (A) U/L

CLINICALCHEMISTRY,Vol.39, No.9, 1993 1805

 30 Table 1. Hitachi 717 Assays with Clinically Significant
-J Hemolysls Interference, and Correction Factors
Factor, per g/L
Analyt. of h.mogloblna r n
110 Alkaline phosphatase -5 6 0.731 28
a-Amylase (EPS) -13 13 0.896 25
= -10 a-Amylase (PNP) -7 6 0.956 30
Aspartate aminotransferase +5 3 0.984 18
Creatine kinase +10 2 0.998 34
Potassium, mmol/L +0.3 0.1 0.991 18
,‘-Glutamyltransferase -3 4 0.914 18
Lactate dehydrogenase +100 15 0.999 21
Lactate dehydrogenase-1 +38 22 0.962 15
a U/L except where noted.
0 2 4 6 8
Hemoglobhi, g/L
Table 2. Multiple Regression Parameters for Hemolysis
Interference in the Bilirubin and Theophylline Assays
2
RegressIon
coefficient
0
Variable Mean SE t P
-2 Dependentvariable:observedbillrubin
(r= 0.995, F= 518, P<0.001)
-4 [Hemoglobin] 4.60 0.58 7.872 <0.001
[Hemoglobin][bilirubinl -0.0592 0.0075 -7.896 <0.001
[Bilirubin] 0.943 0.034 27.991 <0.001
-6
Constant -3.98 2.62 -1.520 0.148
Dependentvariable:observedtheophylilne
-8 (r = 0.998, F = 1869, P <0.001)
[Hemoglobin] -0.4769 0.1233 -3.867 0.001
-10 0 1 2 FI1mrvIInhin11thnhuIlinA1
.,‘.. .-p. -0.0188 0.0054 -3.502 0.002
I [Theophyllinel 0.925 0.024 37.848 <0.001
Hemoglobin, g/L Constant 2.67 0.57 4.674 <0.001

Fig. 2. Effect of hemolysison the assays of bilirubin(top) and

theophylline (bottom) both analyte and hemoglobin (Figure 2), multiple re-
Initial bilirubin concentrations: 10 (#{149}),
29 (U),55 (A),82 (x), and 140 (+)
mol/L Initialtheophytlineconcentrations:7.2 (#{149}),
11.7 (U), 26.8 (A), and gression analysis (14) was performed (Table 2). The
35.2 (x) mg/L resulting equation for predicting the actual bilirubin
concentration was (observed [bilirubin] 4.60[hemoglo- -

Note that 10% relative error occurs at widely different bin] + 3.98)/(0.943 0.0592[hemoglobin]),
- where bili-
hemoglobin concentrations, according to the initial an- rubin is in jLmol/L and hemoglobin is in gIL. The actual
alyte concentration. Because clinical significance could theophyl]ine concentration = (observed [theophyllinel
not be judged by relative error for more than one ana- + 04769[hemoglobin] 2.67)/(0.925
- 0.0i88[hemoglo-
-

lyte concentration, we used absolute error instead. bin]), where theophylline is in mgIL.
Clinically significant negative hemolysis interference We also investigated the mechanism of hemolysis in-
was found for the alkaline phosphatase assay, both terference. For aspartate aminotransferase, lactate de-
a-amylase assays, and the y-glutamyltransferase assay. hydrogenase, lactate dehydrogenase-i, and potassium,
Significant positive hemolysis interference was found the mechanism has been characterized previously and is
for the assays of aspartate aminotransferase, creatine attributed to large differences between intracellular and
kinase, lactate dehydrogenase, lactate dehydroge- extracellular concentrations for these analytes (2). In
nase-i, and potassium. Variable interference was found these cases mathematical correction is the only method
for the bilirubin and theophylline assays. that can be applied for correction of results.
We found that absolute error could be estimated by Hemolysis interference in the creatine kinase assay
linear regression on the hemoglobin concentration (Fig- has been attributed to intracellular adenylate kinase.
ure 1B) for each analyte except bilirubin and theophyl- Correction of creatine kinase for adenylate kinase ac-
line (Figure 2). The amount of error is predicted by tivity can be done by adding inhibitorssuch as adeno-
multiplying the hemoglobin concentration by the slope sine monophosphate and diadenosine pentaphosphate,
obtained from linear regression. Correction factors and or by subtracting the activity measured in the absence
regression parameters are presented in Table 1. of creatine phosphate (18). The Boehringer Maunheim
Because hemolysis interference for bilirubin and reagent contains the aforementioned inhibitors; how-
theophyffine was dependent on the concentrations of ever, omission of creatine phosphate from the R2 re-

1806 CLINICAL CHEMISTRY, Vol. 39, No. 9, 1993

 Table 3. Comparison of Creatine Kinase Interference 4000
and Residual Adenylate Kinase Activity
inItial Added Interference, U/L ResIdual
cr.atlne hemoglobIn, adenylate 3500
klnase, UIL g/L Observed Calculated’ klnase, U/L
46 1.35 14 14 14
2.52 28 25 28 3000
5.01 58 50 56
9.99 117 100 113
97 1.17 12 12 15 2500
2.53 28 25 28
4.91 55 49 54
9.88 116 99 113
2000
348 1.21 13 12 16 0 2 4 6 8 10
2.46 13 25 29
4.84 47 48 57 Thw, mi
9.87 96 99 117 Fig. 3. Time dependence of the absorbance of cresol red in Tns
‘Calculated by usingfactor fromTable 1. buffer, pH 7.5
Arrow at R2 denotes the timeof addition
of thisreagent

agent demonstrates residual adenylate kinase activity,

which is well approximated by mathematical correction 1.2
(Table 3). Given that the method of substrate omission
requires additional reagent, mathematical correction 1.1
provides an economical advantage.
For alkaline phosphatase, a-amylase, y-glutamyl- 1.0
transferase, and theophylline, hemolysis interference is
caused by spectral overlap and by a chemical reaction 0.9
between hemolysate and reaction components. Each
method is a bichromatic rate assay with primary wave- 0.8
lengths near a hemoglobin absorbance peak. The pri-
mary/secondary wavelength pair for each of the en- 0.7
zymes is 4 15/660 nm; theophylline is measured at 546/
800 nm. Hemoglobin absorbance peaks occur at -417, 0.6
540, and 575 nm. The reaction of a reagent component 0 2 4 6 8 10

with hemolysate causes a time-dependent reduction in This,

absorbance at the primary wavelength, resulting in an
Fig. 4. Time dependence of the reaction of hemolysate (final hemo-
apparent reduction in analyte iictivity or concentration. globin concentration, 0.05 g/L) with 0.93 mol/L AMP (-), Boeh-
Before investigating the rate of reaction of each assay nnger-MannheimRi reagent withoutMg (-) and with Mg (“#{149})
component with hemolysate, we wanted to eliminate Absorbanceis normalized to 3-mm readings to account fortemperature Ma-
temperature equilibration as a source of reaction rate bilization
variation. Results of this study are shown in Figure 3.
For a 6-L sample, we determined that at least 3 min is
1.2
necessary for temperature equilibration when the Ri
reagent volume is 250 L; at least 2 mm more is re-
1.0
quired when an additional 50 j.tL of R2 is added. These
conditions are applicable to the alkaline phosphatase 0.8
(AMP) and a-amylase (EPS) assays. The same temper-
ature profile was displayed for the y-glutamyltrans-
ferase assay: sample volume = 8 L, Ri volume = 300
pi, and R2 volume = 60 hl.
The alkaline phosphatase assay contains AMP as an
enzyme activator, hydrochloric acid, zinc sulfate hep- 0.2
tahydrate, and magnesium L-aspartate. When we added
to the hemolysate 0.93 mol/L AMP, with HC1 added to 0.0
adjust the pH to 10.5, a biphasic, time-dependent reduc- 350 400 450 500 550 600
tion in absorbance at 417 nm was evident (Figure 4).
Wavelength, nm
Spectral changes included marked decreases in the ab-
Fig. 5. Absorbance spectra of hemolysate in the absence of AMP
sorbance at 417, 540, and 575 nm and a shift in wave- (-) and -18 h after addition of hemolysate to 0.93 mol/L AMP
length maximum to 395 nm (Figure 5). These changes solution (---)
are in accord with the formation of s1ki1ine hematin Final hemoglobin concentration, 0.26 g/L

CUNICALCHEMISTRY,Vol.39, No.9, 1993 1807

(19) and are in direct contrast to the results of Grosset et 1.02
al. (20), who found no effect on the hemoglobin spectrum
with AMP addition to hemolysate. If the rate of reaction
1.01
between hemoglobin and AMP were constant, rate
blanking could be used, in which the decrease in absor-
bance during the Ri reaction could be subtracted from ‘11.00
the assay rate after the addition of R2. As Figure 4
‘It
demonstrates, the rate of AMP/hemolysate reaction dif- 0.99
fers, depending on the reagent source. A more pro-
nounced biphasic reaction rate is seen with our source of
AMP than with the Boehringer Mannheim Ri reagent. 0.98
This could be caused either by differences in AMP reac-
tivity or by the presence of zinc sulfate heptahydrat.e. 0.97
We did not investigate addition of zinc sulfate heptahy- 0 2 4 6 8 10
drate to our source of AMP because the concentration is
This, nn
not stated in the package insert, nor was it available
Fig. 6. Time dependence of hemolysate absorbance in 105 mmol/L
from Boehringer Mannheim upon request. The presence
HEPES, pH 7.1 (-): 105 mmol/L HEPES, pH 7.1, pIus 15.4
of magnesium L-aspartate did not modify the reaction mmol/L sodium aside (#{149}..Boehnnger-Mannheim Ri reagent with-
#{149});
rate. out substrate(----); and Ri reagent with substrate (-. -)
A “two-test” or Rate-B assay is required to perform Final hemoglobinconcentration, 0.05 g/L; absorbance normalized to 3-mm
rate blanking on the Hitachi 717: the rate after Ri readingsto account fortemperaturestabilization
addition is subtracted from the rate after R2 addition,
dium aside, Ri reagent without substrate, and Ri re-
with each reagent addition defined as a separate test.
agent with substrate. When reaction curves for HEPES
Use of the “two-test” method requires that the Ri/inter-
and HEPES plus aside are compared, at least part of the
ferent reaction rate is constant and remains so after R2
interferenceappears to be due to the interaction of so-
addition. The method also requires that R2 does not
interact with the interferent. In keeping with the tem-
dium aside and hemoglobin. However, the same nega-
tive reaction rate is not observed in the curve for Ri,
perature equilibration restrictions defined in Figure 3,
which also contains azide. The response for the interac-
the measuring points for the rate-blanked assay were
tion of hemolysate with Ri in the presence of substrate
chosen as follows: for Ri, 4.8 - 3.2 min; for R2, 10.0 -
is biphasic. Therefore, rate blanking cannot be used for
8.4 mm. The data in Table 4 demonstrate that rate-
this assay to correct for hemolysis interference; more-
blanking and mathematical correction both provide a
over, there is a possible interaction of hemolysate with
good approximation of alkaline phosphatase activity in
the substrate reagent.
the presence of hemolysis.
Figure 7 demonstrates the interaction of y-glutamyl-
The a-amylase (EPS) Ri reagent assay contains
transferase reagent components with hemolysate. The
HEPES buffer, sodium chloride, magnesium chloride,
Ri reagent contains i22 mmol/L Tris buffer, pH 8.25,
and sodium aside. The Ria reagent tablet contains
and sodium aside. Glycylglycine is the Ria reagent. No
a-glucosidase and substrate. Figure 6 displays the in-
interaction is seen for Tris buffer, but in the presence of
teraction of hemolysate with HEPES, HEPES plus so-

1.02
Table 4. CorrectIon of Hemolysls Interference In the
AlkalIne Phosphatase (ALP) Assay 1.01
Added Rate ALP, U/L
Initial hemoglobin,
ALP, U/L g/L R2’ Rib Rate-blanked Calculatedc ‘ 1.00
61 0.94 277 -21 59 59
2.15 259 -28 57 62
1.00
4.20 184 -112 59 58
8.37 92 -227 63 61
224 1.02 1085 -15 216 220 0.99
2.03 1064 -25 214 220
4.15 954 -143 216 210 0.98
8.41 794 -251 205 200 0 2 4 6 8 10
332 1.01 1664 -6 330 334
2.20 1586 -17 316 326 This, nn
4.13 1504 -130 323 318 Fig. 7. Time dependence of hemolysate absorbance in 122 mmol/L
8.30 1319 -227 305 302 Tris, pH 8.25 (-); 122 mmol/LTris, pH 8.25, plus 15.4 mmol/L
R2 rate
#{149} = 10 - 8.4 mm absorbance. sodium aside (‘.); Boehnnger-Mannheim Ri reagent without gly-
bAl rate = 4.8 - 3.2 mmabsorbance. cylglycine (----); and Al reagent with glycyiglycmne(- . -)
“Calculatedby usingfactorfromTable 1. Anal hemoglobinconcentration, 0.05 g/L absorbance normalized to 3-mm
readingsto account fortemperaturestabilization

1808 CLINICAL CHEMISTRY, Vol. 39, No. 9, 1993

 Table 5. CorrectIon of Hemolysls Interference in the Table 6. Reported Potassium Correction Factors
y-Glutamyltransferase (‘y.GT) Assay Factor, mmol/L potassium
Added Rate per g/L hemoglobin Reference
?GT, ti/L
Initial hemoglobin, 0.30 This study
‘yGT, U/L g/L R2a Rib Rate-blanked Calculatedc 0.33 1
43 0.88 81 -5 49 42 0.29 2
1.67 76 -4 45 41 0.30 3
3.55 66 -24 51 42 0.28 6
7.56 64 -15 45 50 0.25 7
146 0.84 288 0 146 145 0.50 12
1.83 280 -3 144 143
3.52 266 0 135 140
7.16 241 -24 135 139 integrity for any given analyte. The importance of using
473 0.84 933 8 465 466 multiple concentrations of analyte and interferent in
1.68 923 13 457 462 interference evaluation is seen in the following exam-
3.41 888 11 441 450
ple. In the interference screening method proposed by
6.98 829 -34 434 430
the National Committee for Clinical Laboratory Stan-
Footnotes as in Table 4. dards (NCCLS) (22), the analyte concentration is chosen
to be at a medical decision point and the interferent is
sodium aside a negative reaction rate is seen. However, tested at a relatively high concentration. If these crite-
the Ri reagent displays no interaction with hemolysate, ria were used for bilirubin at a concentration of 55
although aside is present in this reagent. A negative moJJL and hemoglobin at a concentration of 5 g/L (Fig-
reaction rate is seen for Ri with glycylglycine added. ure 2), no hemolysis interference would be detected.
We attempted rate blanking with this assay, using the Although the NCCLS Guideline states that “testing at
same measuring points as with the alkaline phospha- more than one decision level may be appropriate de-
tase assay. As Table 5 demonstrates, satisfactory results pending on the analyte,” which analytes require this
were obtained with either rate blanking or mathemat- approach is not known a priori. If accurate character-
ical correction. ization of interference is desired, multiple analyte and
The GDS theophylline assay is a rate assay with cy- interferent concentrations must be used. We have found
tochrome c as the first reagent and theophylline oxidase that, in most cases, a minimum of three analyte and
as the second reagent. The oxidation of theophylline is four interferent concentrations is sufficient for the de-
coupled with the reduction of cytochrome c, for which scription of hemolysis interference.
the absorbance maximum is 550 nm. At lower hemoglo- Because methods for enzyme and bilirubin analysis
bin and theophylline concentrations a slight positive vary greatly, comparing derived correction factors with
interference is present. Negative interference becomes literature values is not feasible in every instance. How-
more prominent as theophylline and hemoglobin con- ever, the factors we found agree well with those of in-
centrations increase (Figure 2). The mechanism of in- vestigators using similar methods (3, 6, 7, 23). Table 6
terference is complex and requires further investiga- shows a list of correction factors for potassium derived
tion. Mathematical correction by multiple regression is from hemolysis interference data in the literature. De-
necessary and provides a good approximation of hemo- spite the facts that interference is method dependent
lysis interference (Table 2). and that multiple analyte and interferent concentra-
The bilirubin assay is an endpoint assay and displays tions were not used in each case, the reported factors
both spectral and chemical interference. The chemical were calculated by assuming that absolute hemolysis
interference of hemoglobin with the Jendrassik-Gr#{243}f interference for potassium is independent of analyte
assay has been characterized previously (21) and is at- concentration. All factors are in agreement except that
tributed to the formation of acid hematin. For the bili- found by Pal and Cyr-Manthey (12). The major differ-
rubin assay this negative interference is more promi- ence appears to be the method used to simulate hemol-
nent at higher bilirubin concentrations (Figure 2). The ysis. Pm and Cyr-Manthey pierced clots with wooden
positive spectral interference is more pronounced at sticks to create hemolysis, whereas all other studies
lower bilirubin concentrations and is due to spectral used physical disruption of washed erythrocytes. The
overlap of hemoglobin with the primary wavelength. former authors also found unpredictable hemolysis in-
Because the interference is both dependent and inde- terference for lactate dehydrogenase, whereas our data
pendent of the analyte concentration, mathematical cor- show a linear response. We are currently investigating
rection by multiple regression analysis (14) is necessary the mechanism(s) for these differences.
and provides an adequate prediction of hemolysis inter- Although we have shown that correction factors can
ference (Table 2). be derived for hemolysis interference, certain precau-
tions should be taken in their application. The source of
Discussion hemolysis plays an important role in the degree of he-
The manner in which interference is presented deter- molysis interference. For example, hemolysis from in
mines what clinical action is taken to maintain sample vivo sources produces less interference in potassium and

CLINICAL CHEMISTRY, Vol. 39, No. 9, 1993 1809

creatine kinase assays than does in vitro hemolysis (24, vitro hemolysis on chemical values for serum. Clin Chem 1978;24:
25). However, the vast majority of hemolyzed samples in 1966-70.
7. Sonntag 0. Hemolysis as an interference factor in clinical
our laboratory are the product of in vitro processes. Of chemistry. J Clin Chem Cliii Biochem 1986;24:127-39.
further concern is interindividual differences in intra- 8. Glick MR. Ryder KW, Jackson SA. Graphical comparisons of
cellular analyte concentrations, which could produce interferences in clinical chemistry instrumentation. CliiiChem
variability in correction factor detennination. This area 1986;32:470-5.
9. Glick MR, Ryder KW, Vroon DH, Masters BE, Sonntag 0.
also requires further investigation. Practical uses of serum indices to reduce errors from lipemia,
The characterization of hemolysis interference is a icterus, and hemolysis [Abstract]. Clin Chem 1990;36:1008.
complex issue and depends on many factors. As we have 10. Karkoski DJ. Hemoglobin interference with the BMD total
shown, the interaction of hemolysate with reagent may bilirubun assay in the Hitachi 705 analyzer, and its relation to the
hemolytic index [Letter]. Cliii Chem 1985;31:791.
differ with different reagent sources. In reagents with 11. Greenson JK, Farber SJ, Dubin SB. The effect of hemolysis on
multiple components, individual components may dis- creatine kinase determination. Arch Pathol Lab Med 1989;113:
play markedly different reaction kinetics with hemoly- 184-5.
sate, compared with situations when all components are 12. Psi SH, Cyr-Manthey M. Effect of hemolysis on chemistry
tests. Lab Med 1991;22:408-10.
present.
13. Kroll MH, Ruddel M, Blank DW, Elm RJ. A model for
Estimates of hemolysis interference can be used in a assessing interference. Cliii Chem 1987;33:1121-3.
variety of ways. We have adopted the following method 14. Kroll MH, Chesler R. Rationale for using multiple regression
in our laboratory. Every visibly hemolyzed sample is analysis with complex interferences. Eur J Clin Chem Clin Bio-
chem 1992;30:415-24.
analyzed for hemoglobin. A computer program is then
15. Meites S. Reproducibly simulating hemolysis, for evaluating
used to list which tests are affected at that hemoglobin its interference with chemical methods [Letter]. Clin Chem 1973;
concentration. The user enters the test result and the 19:1319.
program determines whether correction is necessary on 16. Provasek D, Jay D. Bichromatic determination of hemoglobin
the basis of clinical significance (± iO%). If correction is on the Hitachi 717 [Abstract]. Cliii Chem 1992;38:942.
17. Bowie L, Esters F, Bolin J, Gochman N. Development of an
necessary, the amount of interference is calculated and aqueous temperature-indicating technique and its application to
comments are then generated to accompany results. For clinical laboratory instrumentation. Clin Chem 1976;22:449-55.
example, the potassium result for a hemolyzed sample 18. Szasz G, Gerhardt W, Gruber W, Bernt E. Creatine kunase in
with a hemoglobin concentration of 2.50 g/L would be serum: 2. Interference of adenylate kinase with the assay. Cliii
Chem 1976;22:1806-11.
accompanied by the comment: “If a diagnosis of intra- 19. Hunter Fr. The quantitation of mixtures of hemoglobin deriv-
vascular hemolysis can be excluded, in vitro hemolysis atives by photoelectric spectrophotometry.Springfield, IL: Charles
has caused a false increase in potassium of -0.8 mmoll C Thomas, 1951:124-5.
L.” We have received positive feedback from our clini- 20. Grosset A, Knapp ML, Mayne PD. The effect of hemolysis on
the measurement of plasma alkaline phosphatase activity. Ann
cians with this method, especially in the outpatient set- Clin Biochem 1987;24:513-7.
ting, where repeat samples are not easily attainable. 21. Shull BC, Lees H, Li PK. Mechanism of interference by
hemoglobin in the determination of total bilirubin. II. Method of
References Jendrassik-Grof. Cliii Chem 1980;26:26-9.
1. Mather A, Mackie NR. Effects of hemolysis on serum electrolyte 22. National Committee for Clinical Laboratory Standards. Inter-
values. Clin Chem 1960;6:223-7. ference testing in clinical chemistry; Proposed Guideline. EP7-P.
2. Caraway WT. Chemical and diagnostic specificity of laboratory Villanova, PA NCCLS, 1986.
tests. Am J Cliii Pathol 1962;37:445-64. 23. Randall AG, Garcia-Webb P, Beilby JP. Interference by he-
3. Brydon WG, Roberts LB. The effect of hemolysia on the deter- molysis, icterus and lipemia in assays on the Beckman Synchron
mination of plasma constituents. Clin Chim Acta 1972;4 1:435-S. CX5 and methods for correction. Ann Cliii Biochem 1990;27:345-
4. Schwartz MK. Interferences in diagnostic biochemical proce- 52.
dures. Adv Clin Chem 1973;16:1-45. 24. Blank DW, Kroll MH, Ruddel ME, Elm RJ. Hemoglobin
5. Laessig RH, Jassemer DJ, Paskey TA, Schwartz TH. The effects interference from in vivo hemolysis. Cliii Chem 1985;31:1566-9.
of 0.1 and 1.0 percent erythrocytes and hemolysis on serum 25. McKinney CD, Bruns DE. Predicting the effect of hemolysis on
chemistry values. Am J Clin Pathol 1976;66:639-44. measured creatine kinase: a caveat [Letter]. Arch Pathol Lab Med
6. Frank JJ, Bermes EW, Bickel MJ, Watkins BF. Effect of in 1992;1 16:7.

1810 CLINICAL CHEMISTRY, Vol. 39, No. 9, 1993