CLIN. CHEM.

30/5, 627-630 (1984)

Direct Spectrophotometry of Serum Hemoglobin: An Allen Correction Compared with a Three-Wavelength Polychromatic Analysis
Dennis A. Noe, Victor Weedn, and William R. Bell2
An Allen correction and a polychromatic analysis are about
equally effective in minimizing effects of interference by bilirubin and triglyceride turbidity in the direct spectrophotometry of serum hemoglobin: interference from bilirubin is nearly eliminated, that from turbidity substantially decreased. The limit of detectability of hemoglobin is 8 mg/L in the
presence of a moderate concentration of bilirubin. A change

in hemoglobin concentration as small as 16 mg/L can be detected in serum having a concentration near the upper limit

of the reference interval, i.e., at the medical decision level.

The polychromatic formula gives concentration estimates approximately 5% greater than those of the Allen correction. The formula for the Allen correction is hemoglobin (mg/L) = 1.68 mA415 0.84 mAsso 0.84 mA450. That for the polychromatic analysis is hemoglobin (mg/L) = 1.65 mA415 0.93 mA3so 0.73 mA470.
-

results by the method of Harboe did not correlate well with those by a benzidine technique (6) or with another Allencorrection method (20) in a small number of hyperbilirubinemic sera. No thorough analytical evaluation or formal comparison study of the method of Harboe has been published. This communication reports the findings of a study in which we compared the performance of the Allen correction recommended by Harboe with a three-wavelength polychromatic assay, a different approach to minimizing analytical interferences in direct spectrophotometry.

Materials and Methods

Reagents
In preparing blood freshly hemoglobin standards we used nonlipemic collected into tubes containing dipotassium ethylenediaminetetraacetate. The plasma was removed and the erythrocytes were washed three times with saline (145 mmoliL NaC1). The packed erythrocytes were re-suspended in saline and lysed by freeze-thawing three times. The resulting hemoglobin concentrates were mixed thoroughly, allowed to equilibrate with the atmosphere, then diluted with saline to give a hemoglobin concentration of 100 g/L as measured in an appropriately calibrated and quality-controlled Co-Oximeter 282 (Instrumentation Laboratory, Lexington, MA 02173). A typical distribution of the various forms of hemoglobin as measured with this instrument was: 91.3% oxyhemoglobin, 7.7% methemoglobin, and 0.4% carboxyhemoglobin. The hemoglobin solution was then further diluted with saline to produce a 2000 mgfL stock solution.
Working standards were made by diluting this stock tion with saline. Bilirubin solutions were prepared from lyophilized solu-

AdditIonal Keyphrases: reference inte,val vascular hemolysis

detection

of intra-

An above-normal hemoglobin concentration in serum or plasma is a sensitive and specific marker of acute intravascular hemolysis (1, 2), one for which there currently is no reliable substitute. Direct spectrophotometry has been used to measure the concentration of hemoglobin in serum and plasma (3,4). Because this method is subject to interference from bilirubin, which commonly is present in increased concentration in hemolytic states, colorimetry of hemoglobin by use of benzidine (5, 6) has been preferred for many years. The Occupational Safety and Health Administration has, however, severely restricted the use of benzidine because of its carcinogenicity (7). As a result, direct spectrophotometry of plasma hemoglobin is again popular. In addition, chromogenic assays are available that involve benzidine substitutes: p-axninobenzoate (8), o-tolidine (9), tetramethylbenzidine (10, 11), leucomalachite green (12, 13), o-dianisidine (13), and 2,2-azino-di-(3-ethylbenzthiazoline sulfonate) (14). Recently, nephelometric techniques have also been introduced (15, 16). The most frequently cited direct spectrophotometric assay is that described by Harboe (17). In this method an Allen correction (18) is used to correct for the presence of interfering substances-that is, by subtracting the background shoulder absorbance. This method demonstrably eliminates interference from substances in normal sera and from mild turbidity (17). In a study reported by Kahn et al. (19), The University of Texas Health Science Center, Houston, TX. 1 Current address: Dept. of Medicine, Blalock 1036, The Johns Hopkins Hospital, Baltimore, MD 21205 (to which address correspondence). 2Hubert E. and Anne E. Rogers Scholar in Academic Medicine, The Johns Hopkins University School of Medicine, Baltimore, MD.
Received November 23, 1983; accepted January 25, 1984.

protein-based reference material (Bilirubin Calibrator, 9000SD; American Monitor Corp., Indianapolis, IN 46268). After reconstitution with saline, the solutions were assayed br total bilirubin in a discrete analyzer (aca II; Du Pont Co.,
Clinical Systems Division, Wilmington, DE 19898).

Turbid micellar triglyceride solutions were obtained by diluting a commercial soybean-oil emulsion used for intravenous hyperalimentation (Intralipid; Cutter Laboratories, Emeryville, CA 94608). The hemoglobin standards and the bilirubin and triglyceride solutions were kept protected from light and refrigerated at 5-10 #{176}C. As the diluent buffer for the procedure we used tris(hydroxymethyl)methylamine, final concentration 62.5 mmolIL, pH adjusted to 8.0 with concentrated HC1. We elected to use an alkaline diluent so as to decrease serum turbidity (21). Hemoglobin standards and serum specimens diluted 10-fold in this diluent proved to be stable; i.e., they gave unchanged values for hemoglobin after a week of refrigerated storage.
CLINICAL CHEMISTRY, Vol. 30, No. 5, 1984 627

Derivation of Polychromatic

Formula

Bilirubin and triglyceride turbidity are known to be the most important sources of interference in the direct spectrophotometry of hemoglobin (4). Polychromatic analysis of an analyte in the presence of two interfering substances requires that absorbance measurements be made at three separate wavelengths (22), in the present case 380,415, and 470 nm. The first wavelength was used as a marker wavelength for triglyceride turbidity because the absorptivities of hemoglobin and bilirubin are small at that wavelength relative to that of turbidity. The second was selected because it is near an absorbance maximum for hemoglobin in alkaline medium. The last was chosen because it is near an absorbance maximum for bilirubin but absorptivity of hemoglobin is low. The polychromatic formula was derived by solving the system of simultaneous equations that characterizes the absorbance readings at the three wavelengths in terms of the concentrations and absorptivities of hemoglobin and the interfering substances (22).

least five separate absorptivity calculations for each substance at each wavelength. For the stock solution of micellar triglyceride the relative apparent absorptivity at 415 urn was arbitrarily set at 10 milliabsorbance units. The relative apparent absorptivities at the other wavelengths were calculated from the ratios of the absorbance of the stock triglyceride solution at those wavelengths to its absorbance at 415 nm. We made eight absorptivity determinations at each wavelength and used the average values in the correction formulas.

Analytical Variables
Linearity. Two sets of hemoglobin standards (100, 500, 1000, 1500, and 2000 mg/L) were measured on separate days. The results were subjected to least squares linear regression analysis (23). Precision. Specimens were not assayed as part of a run so we could not evaluate within-run precision. Between-run precision was evaluated by using hemoglobin standards having concentrations within the reference interval, slightly above it, and markedly above it. During a month we made 20 determinations on each such standard. Limits of detectability were calculated in accordance with the recommendations of the International Federation of Clinical Chemistry (24). A hemoglobin-free bilirubin solution was used as a blank. Low-concentration hemoglobin solutions were made by supplementing separate volumes of the bilirubin solution with a 2000 mg/L hemoglobin standard so as to obtain final concentrations of 20, 40, and 80 mgIL. We also determined the magnitude of a detectable change in hemoglobin concentration at low hemoglobin concentrations. A serum specimen with a hemoglobin concentration of 40 mgfL served as the native (i.e., unaltered) sample. To simulate mild hemoglobinemia, we added hemoglobin standard to portions of this serum to give final hemoglobin concentrations of 80 and 120 mgfL. The mean and standard deviation for the hemoglobin concentrations were calculated from 10 separate determinations on each sample. Analytical recovery. For this, we used 48 serum specimens submitted to the clinical chemistry laboratory for other determinations. Most were selected at random but some turbid and icteric specimens were deliberately included. Hemoglobin concentrations were measured in the original serum and in portions of serum to which we added 1/10 volume of 2000 mgfL hemoglobin stock solution. For each batch of specimens we also determined the analytical recovety in saline. Recovery from serum was calculated as the ratio of the recovery for serum matrix to recovery for saline matrix. Interference. The magnitude of the interference by bilirubin and turbid micellar triglyceride was determined for saline and hemoglobin solutions with concentrations of 200, 600, and 1000 mg/L. Bilirubin and micellar triglyceride

Procedures
Assay. Using plastic spectrophotometry cuvettes with a 1cm lightpath (Evergreen Scientific, Los Angeles, CA 90058),

we measured the absorbance of 0.3 mL of sample diluted with 2.7 mL of Tris buffer at the specified wavelengths (380, 415, and 450 nm for the Allen correction; 380, 415, and 470 nm for the polychromatic formula), with diluent used as the blank. Absorbances were measured at room temperature. For spectrophotometry we used a Coleman 55 Spectrophotometer (Perkin-Elmer Corp., Norwalk, CT 06856). Calculations. We calculated the concentrations of hemoglobin in the samples, in milligrams per liter, using the Allen correction recommended by Harboe and the polychromatic formula. Harboes Allen correction (17) is
2A415 - A
Hb415
-

A
(Z}5()

aff.,3()

dilution

where A, A415, and A0 are the absorbance readings in milliabsorbance units at 380,415, and 450 nm, respectively, and a}0,, a15, and am are the absorptiviti#{233}s of hemoglobin in milliabsorbance units per milligram of hemoglobin per liter at 380, 415, and 450 nm, respectively. The polychromatic formula is
D4a,,M7O.44j5 (D4atur,,i5

D,D4

+ D2a,)A4io D2D3
-

+ D

,,70A3

dilution

where
D2
= = =

D1

(aturM7oa}1M15)
-

(a.Ml5aip7o)

(aturM7oabjlj4l5) (turs8oami7o) (a1,,bOabII47O)

(aturMl5abili47o) (0turM7O(21Th380) ((hurM7O1biIi380)

D3 D4

readings in milliabsorbance units at 380, 415, and 470 nm, respectively; am is the absorptivity of hemoglobin in milliabsorbance units per milligram of hemoglobin per liter at the indicated wavelengths; abII, is the absorptivity of bilirubin in milliabsorbance units per micromole of bilirubin per liter; and at is the relative apparent absorptivity of micellar triglyceride (actually, of scatter) in milliabsorbance units. We calculated absorptivities for hemoglobin and bilirubin from the absorbance readings of hemoglobin and bilirubin solutions of various known concentrations. Absorbances were measured by using the described procedure. The values for absorptivity that we entered into the spectrophotometric formulas were found by taking the average of at 628 CLINICAL CHEMISTRY,Vol.30, No. 5, 1984

A380, A415, and A470 are the absorbance

were final final zero

added to these solutions to give combinations of five concentrations of bilirubin (0-0.5 moIIL) and five concentrations of micellar triglyceride ranging from to the stock concentration.

Comparison of Methods and Preliminary Reference Intervals

The same serum specimens we used in the analytical study were also used to compare the results of the two formulas and to construct preliminary reference intervals. The regression method of Deming was used to define the linear relationship between the results of the two methods (25). Because the imprecision of the two methods can be assumed to be equal, depending as they do upon the
recovery

same

absorbance measurements, the ratio of the error of the two methods was set equal to unity. The correlation coefficient was calculated by the standard method (23).
variances

Interference. The results were essentially identical for both methods. Over the range of bilirubin concentrations we studied, only a slight negative interference was detected when bilirubin was the sole interferent (maximum interference: 4% decrease in 200 mgfL hemoglobin solution). There was appreciable negative interference owing to turbidity, and it increased linearly with the degree of turbidity (maximum interference: 30% decrease in 200 mg/L hemoglobin

Results
Formulas
Table 1 lists the absorptivities of hemoglobin, bilirubin, triglyceride in alkaline diluent. Substitution of these absorptivity values yields:
and micellar

solution). Interference by turbidity was markedly lessened when low concentrations of bilirubin were present but
reappeared rubin nor saline-based at higher bilirubin concentrations. Neither bili-

hemoglobin
and

concn, mgfL

1.68A415

0.84A380

0.84A450

negative concn, mgfL correction

=

alone produced interference in the specimen, but together they caused substantial interference.

turbidity

hemoglobin

1.65A415 0.93A
-

0.73A470 formula,

Comparison-of-Methods

and Preliminary
we calculated

Reference

for the Allen respectively.

and the polychromatic

Interval
The hemoglobin values by using the two
methods were highly correlated (r = 0.991; p < 0.001). The regression equation for the results is: hemoglobin by the polychromatic formula = 4.9 + (1.05 - hemoglobin by Allen correction). Thus hemoglobin concentrations as estimated

Analytical Variables
Linearity. The Allen correction and the polychromatic formula gave values that were linearly related to hemoglobin concentration throughout the range of concentrations studied. The linear regression equation for the data from the Allen correction is: Observed concentration = -9.12 (SD 5.0) + [1.001 (SD 0.007) target concentration]; (r> 0.999, p < 0.001, std. error of the estimate = 15.74 mg/L). For the polychromatic formula the line is: Observed concentration = -6.36 (SD 4.6) + [0.997 (SD 0.007) target concentration]; (r> 0.999, p < 0.001, std. error of the estimate = 14.46 mg/L). Precision. Tables 2 and 3 record the results of the precision study. The two formulas demonstrated equivalent precision. At low hemoglobin concentrations (<100 mg/L) the CVs for the saline-based hemoglobin solutions and the supplemented bilirubin solutions were near 20%. The CVs for serum were much lower. With a one-tailed t-test at p = 0.05, the limits of detectability were 7.6 and 7.7 mgfL for the Allen and polychromatic formulas, respectively. This limit corresponds to the detection of hemoglobin in the presence of moderate bilirubin concentrations-a common and clinically important necessity. The magnitudes of detectable change in serum hemoglobin concentration were 15.8 and 15.7 mg/L for the Allen and polychromatic formulas, respectively. These values apply in the concentration range of transition to hemoglobineinia. Analytical recovery. The mean recovery for both formulas was 103.6%. The frequency distributions were slightly skewed to the left; the inner 94-percentile range (i.e., 45 of 48 samples) for the Allen correction was 81-116%, and 82116% for the polychromatic formula. Recovery was not correlated with either the initial hemoglobin concentration (r = 0.124) or the concentrations of bilirubin (r = 0.234) or turbidity (r = -0.245) as calculated from polychromatic formulas for the interfering substances (22).

by use of the polychromatic formula are about 5% greater than those with the Allen correction. Examination of the frequency histograms of hemoglobin concentration results in the hospital-based population revealed two subpopulations. Three quarters of the specimens studied had hemoglobin concentrations less than 81 mgfL (Allen) or 86 mg/L (polychromatic); the remainder had concentrations between 111 and 686 mg/L (Allen) or 100 and 691 mg/L (polychromatic). The subpopulation with lower hemoglobin concentrations is used as the reference population. The upper limits of the reference intervals are therefore 81 and 86 mgfL for the Allen and polychromatic

Table 2. Precision of Hemoglobin Formulas. 1. Hemoglobin Solutions1

Harboe formula Mean concn, mg/L
mg/L CV, % separate days dunng a

Poiychromatlc formula 20
5 25

20
5 25

447
20 4.5

1355
40 3.0

444
20 4.5

1349
42 3.1

SD,

aSamples were stored refrigerated. Determinations were made on 20

month.

Table 3. PrecisIon of Hemoglobin Formulas. 2. Supplemented Bilirubin Solutions and Sera

Alien formula
Biiirubin solutions Blank 1.5 1.3 87 Polychromatic formula

Supplemented
20 46 3.3 11 17 24 77 16 21

Blank Supplemented
1.5 1.4 22 46 3.3 11 15 24 7.7

Mean concn, mg/L

SD, mg/L
CV, %
Limit of detectability,

77 16

93
Native

21

7.6 Native Supplemented

42 81 2.4 8.8
5.7 11

mg/Lb Sera Supplemented 83

8.7 10

Table 1. Absorptivities of Hemoglobin, Bilirubin, and Micellar Triglyceride in Alkaline Diluent

Hemoglobin Wavelength, nm mA/(mg/L)
Bilirubin Miceilar triglyceride

Mean concn, mg/L

SD, mg/L

123 18
14.6

44
2.6

124
18 14.5

CV, %
Detectable change in concn, mg/L

5.9

15.8

15.7

380
415

2.08
7.47

450
470

0.90
0.54

mAi(pmol/L) 19.9 38.2 61.4

mAI(relatlve concn) 11.8 10.0 7.6

aSamples stored refrigerated and protected from light. Determinations

made on 10 separate days. bLimit of detectability = mean blank result + 1.73 cDetable change in concn. = 1.73 (SDL. + SDOW
+

SDL

==)#{189}.

,)#{189}.

CLINICAL

CHEMISTRY,

Vol. 30. No. 5, 1984

629

formulas,

respectively.

Physiological

considerations

dictate

that the lower limits be set at 0 mgtL.

1910.1010, Title 29, Code of Federal Regulations, Fed Reg, January 30, 1979, pp 3825-3845.
8. Nazario M. Neuvo metodo de dosaje de hemoglobima plasmatica empleando acido para-aniino benzioco. Rev Asoc Bioquim Argent 21, 6-10 (1956). 9. Lewis GP. Method using ortho-tolidine for the quantitative determination of haemoglobin in serum and urine. J ClinPathol 18, 235-239 (1965). 10. Levinson SS, Goldman J. Measuring hemoglobin in plasma by reaction with tetramethyl benzidine. Clin Chem 28, 471-474

Discussion
Bilirubin and turbidity interfere with direct spectrometry of serum hemoglobin (4). The interferences can be minimized by use of an Allen correction (17, 20), by derivative spectrophotometry (26), or by polychromatic analysis. In this study we found that the performance of the polychromatic analysis was comparable to, but no better than, the Allen correction. Hemoglobin concentration estimates found using the polychromatic formula tend to be slightly larger than those found by using the Allen correction. Both methods are simple, rapid, and inexpensive. Concentrations of hemoglobin as low as 8 mg/L can be detected, and linearity is maintained up to 2000 mgJL. At concentrations near the upper limit of the reference interval a change in concentration of 16 mg/L can be detected. This satisfies the clinical needs for analytical precision at this concentration, which is the medical decision level for the detection of intravascular hemolysis. The analytical-recovery study demonstrates that the two methods are accurate even though the interference by turbidity is not completely abolished by either method. The reference intervals were determined for serum specimens because serum is the usual matrix in clinical chemical
assays, and is the preferred matrix in which to measure the

(1982).
11. Standefer JC, Vandeijagt D. Use of tetramethyl benzidine in plasma hemoglobin assay. Clin Chem 23, 749-751 (1977). 12. Slaunwhite D, Clements J, Tussey RL, Reynoso G. Leucomalachite green assay for free hemoglobin in serum. Am J Clin Pat hod 72, 852-855 (1979). 13. Ahlquist DA, Schwartz S. Use of leuco-dyes in the quantitative colorimetric microdetermination of hemoglobin and other heme compounds. Clin Chem 21, 362-369 (1975). 14. Marklund S. Determination of plasma or serum haemoglobin by peroxidase activity employing 2,2-azino-di-(3-ethyl-benzthiazolinsulphonate-6) as chromogen. Scand J Clin Lab Invest 38, 543547 (1978). 15. Engler R, Pointis J, Rondeau Y. Determination immunochimiquede lhemoglobine humaine dan les liquides biologiques. Clin Chim Acta 77, 159-165 (1977). 16. Gadsden R, Ward B, Lepes-Virella MF, et al. Immunochemical determination of free hemoglobin in plasma and urine with a laser nephelometer. Clin Chem 25, 497-499 (1979). 17. Harboe M. A method for determination of hemoglobin in plasma by near ultraviolet spectrophotometry. Scand J Clin Lab Invest 11, 66-70 (1959). 18. Caraway WT. Mathematics in clinical chemistry. In Clinical Biochemistry: Contemporary Theories and Techniques, 1, HE Spie-

concentrations of bilirubin, haptoglobin, and hemopexin, as well as other markers of hemolysis. The intervals found here are only preliminary, because of the small number of samples. Each laboratory should construct definitive intervals based upon a reference population appropriate to the
clinical application. in part by Research Grants from the NIH, NHLBI. HLO 1601,

gel, Ed., Academic Press, New York, NY, 1981, pp 93-94. 19. Kahn SE, Watkins BF, Bermes EW. The spectrophotometric scanning technique for measurement of plasma hemoglobin. Ann
Clin Lab Sci 11, 126-131 (1981).

This work was supported HL24898, and T32HL07525

References
1. Crosby WH, Dameshak W. The significance of hemoglobinemia and associated hemosiderinuria, with particular reference to various types of hemolytic anemia. JLab Gun Med 38, 829-841 (1951). 2. Ham TH. Hemoglobinuria. Am J Med 18, 990-1006 (1955). 3. Flink EB, Watson CJ. A method for the quantitative determination of hemoglobin and related heme pigments in feces, urine and blood plasma. J Bid Chem 146, 171-178 (1942). 4. Hunter Fl, Grove-Rasmussen M, Soutter L. A spectrophotometnc method for quantitating hemoglobin in plasma or serum. Am J Clin Pathol 20, 429-433 (1950). 5. Wu H. Studies on hemoglobin: An ultramicromethod for determination of hemoglobin as a peroxidase. J Biochem 2, 189-194 (1923). 6. Crosby WH, Furth FW. A modification of the benzidine method for measurement of hemoglobin in plasma and urine. Blood 11, 380-383 (1956). 7. Occupational Safety and Health Administration Standards, Part

20. Blakney GB, Dinwoodie AJ. A spectrophotometric scanning technique for the rapid determination of plasma hemoglobin. Gun Biochem 8, 96-102 (1975). 21. Wu H. Studies in hemoglobin I. The advantage of alkaline
solutions for determination of hemoglobin. J Biochem 2, 173-180

(1923). 22. Hahn B, Viastelica DL, Snyder LR, et al. Polychromatic analysis: New applications of an old technique. Gun Chem 25,951959 (1979).
23. Natrella MG. Experimental Statistics, fice, Washington, DC, 1966, pp 5-3-5-23. U.S. Govt. Printing Of-

24. Buttner J, Borth R, Boutwell RI, et al. Approved recommendation (1978) on quality control in clinical chemistry. Part 2. Assessment of analytical methods for routine use. J Clin Chem Gun Biochem 18, 78-88 (1980). 25. Cornbleet PJ, Gochman N. Incorrect least-squares regression coefficients in method-comparison analysis. Clin Chem 25,432-438 (1979). 26. Amazon K, Soloni F, Rywlin AM. Separation of bilirubin from hemoglobin by recording derivative spectrophotometry. Am J Gun Pathot 75, 519-523 (1981).

630

CLINICAL CHEMISTRY, Vol. 30, No. 5, 1984