CLIN. CHEM.

30/10, 1646-1651 (1984)

A MathematicalModelof the Hemoglobin-OxygenDissociation Curve of Human Bloodand of the Oxygen PartialPressureas a Functionof Temperature
Ole

Slggaard-Andersen,1Peter D. Wimberley,1 Ivar G#{246}thgen,2 and Mads Slggaard-Andersen The Hemoglobin-Oxygen DIssociation Curve
The function between the partial pressure of oxygen in the blood (p) and the saturation fraction of oxygen in hemoglobin (s) is conveniently represented by a Hill plot (7) with abscissa and ordinate (x, y) given by:
x
=

A mathematical modelis described giving the oxygen saturation fraction (s) as a function of the oxygen partial pressure (p): y- Yo = x- x0 + h.tanh[k.(xQ], wherey= ln[s/ (1 s)] and x = In(pkPa). The parameters are: Yo = 1.875; x0 = 1.946 + a + b; h = 3.5 + a; k = 0.5343; b = 0.055 [TI (K 310.15)]; a = 1.04 (7.4 pH) + 0.005#{149} cI(mmol/L) + 0.07 {[c,jpGI(mmollL)] 5}, where Ooase is the base excess of the blood and cbPG is the concentration of 2,3diphosphoglycerate in the erythrocytes. The Hill slope, n = dyldx, is given by n = 1 + h k{1 tanh2[k. (x )]}. n attains a maximum of 2.87 for x = , and n-+ 1 for x- . The model gives a very good fit to the Severinghaus standard oxygen dissociation curve and the parameters may easy be fitted to other oxygen dissociation curves as well. Applications of the model are described including the solution of the inverse function (p as a function of s) by a Newton-Raphson iteration method. The pD2-temperature coefficient is given by dlnpldT = [Aap + bflS(l s).BjI[a.p + c.1bns#{149}(l s)],whereA = -dlnaIdT=0.012K1; B= (8lnpaT) = 0.073 K for y = yo; a = the solubility coefficient of 02 in blood = 0.0105 mmol V1 kPa1 at 37 #{176}C; -b = concentration of hemoglobin iron in the blood. Approximate equations currently in use do not take the variations of the p02-temperature coefficient with p and q1, into account.
-

ln(plkPa)

ln[s/(1
,

s)]

(1)

According to binding theory, the function is S-shaped with two tangents of slope 1 for x corresponding to the binding of the first through the fourth 02 molecule (Figure 1) (8). As a good approximation, the function is symmetrical and may be characterized by the following parameters:

the the the totes:

point of symmetry: (po, S) or (x0, yo) slope at the point of symmetry: n0 horizontal (or vertical) distance between the asymp2h
in mmHg, substitute 7.5 mmHg for In(p/7.5 mmHg).

If p is measured throughout, e.g., x =

kPa

AddItIonal Keyphrases: blood gas analyzers Bohr effect 2,3-diphosphoglycerate oxygensaturation . pH

.

Most commercial blood gas analyzers measure pH, Pco2, and P02 by electrodes and calculate several derived quantities by means of a built-in programmable calculator (1-3). Among the derived quantities are the oxygen saturation calculated from the Po2 with correction for the pH effect (the Bohr effect) on the oxygen dissociation curve. A temperature correction is generally incorporated, allowing for conversion of values measured at 37 #{176}C to values for other temperatures. The temperature correction of the Po, presents a special problem because the temperature coefficient varies, being low at high P02 and much higher at low P0, (4, 5). The p02-temperature coefficient has also acquired a new importance in relation to the interpretation of measurements of transcutaneous po, when the electrode and the underlying blood are heated to 44#{176}C (6). Here we describe an accurate mathematical model of the hemoglobin-oxygen dissociation curve and the p02-temperature coefficient of human blood, for applications in modern blood gas analyzers. Fig. 1. Hill plot of the oxygen dissociation curve according to the presentmathematical model The curvesarecharacterized bythepoint ofsymrnetiy ( Yo),the maximalskpe
(rb), arid the honzontal(or vertical)distancebetweenthe two asymptotes(2h), and Illustrate the Bohr effect:fromleft to right the pH changesfrom 7.8 to 7.4 to 7.0. The dottedcurveindicatesthe values tabulatedby Seveiinghaus(10)

Department of Clinical Chemistry, Herlev Hospital, and 21)epartment ofAnesthesiology, Rigshospitalet,University of Copenhagen,Herlev, Denmark, DK-2730. Received June 4, 1984; acceptedJuly 18, 1984. 1646 CLINICALCHEMISTRY, Vol. 30, No. 10, 1984

The function may be represented as the sum of two through (x0, yo): a straight line of slope 1, plus a hyperbolic tangent curve with asymptotes at y Yo = h and a maximal slope of n0 1 (9). For convenience we introduce k = (n0 1)/h. A mathematical model representing the hemoglobinoxygen dissociation curve therefore is
functions
-

y-y0=x-x0+htanh[k(x-xo)1

(2)

The Hill slope, n n where x

=
=

dy/dx, is obtained by differentiation:

. . -

1 + h k {1

tanh2[k

x)J}
-

(3)

= n0, and x -#{247} gives n 1. The of the equation, which are explained in more detail below, are summarized here:

x0 gives n

parameters

Yo

= 1.875 (S = 0.867) xo =1.946+a+b (p0=7kPafora=Oandb= 0) no =2.87+ak h =3.5+a k (n0 1)/h = 0.5343 a = 1.04 (7.40 pH) + 0.005 c.,,/(mmol/L) + 0.07 {[cD/(mmol/J)] 5} b = 0.055 [TI(K 310.15)]
. . -

(4) (5) (6)

(7)

Fig.3. The deviation of the calculated partial pressure of oxygen from the tabulated values of the Severinghaus oxygen dissociation curve (lnp) plotted as a function of the oxygen partial pressure (lnp) A deviation of +0.05 for np appmximatelycorrespondsto a deviation of the calculatedoxygen partial pressure of +5% the parameters: Po = 5.85 kPa, s0 = 0.80, no = 2.79, h = 3.15. In both cases the deviations of the calculated s values are less than 0.003. Until consensus is reached concerning the most appropriate standard oxygen dissociation curve, we prefer to use Severinghaus curve as our reference. Attempts to use a different S-shaped function, i.e., tanx rather than tanhx, did not improve the fit, nor did attempts to use y as the independent variable. The parameter a depends on the allosteric effectors: plasma pH, excess concentration of blood base (Cb), and the concentration of 2,3-diphosphoglycerate (CT3pG) in erythrocytes. In the model we assume that these allosteric effectors displace the lower asymptote only, whereas the upper asymptote remains constant (Figure 1) (14, 15). Therefore, not only x0 but also noand h vary with a. The variation of no was chosen so that no approaches 1 in the case where the lower asymptote merges with the upper (i.e., for a -3.5 or pH 10.8). For simplicity we assume that y remains constant. The effectof the hydrogen ion activity (pH) on the parameter a is derived from
-+ -*

(8) (9) (10)

These parameters give the optimal fit to the empirical standard oxygen dissociation curve for normal blood (10), i.e., minimal deviation of the calculated s (Figure 2). The deviation of the calculated p values is then within 3% for the interval between 1.0 and 14.5 kPa. Outside this interval the deviations are greater (Figures 1 and 3). II we want to minimize the deviation for lnp rather than s, the parameters are different: Po = 5.35 kPa, o = 0.76, no = 2.36, h = 3.05. However, in this case the extreme data points (s 0 and S -#{248}1) receive a greater weight than the intermediate data points, despite the fact that the latter are determined considerably more nearly accurately. Thus we prefer to use the parameters that optimize the fit for s. The parameters may easily be fitted to other experimental curves. We have used a visual method, first estimating the parameters on the basis of a plot of y x vs x superimposed on the same plot rotated 180#{176}. Final adjustments of the parameters are by trial and error on the 1asis of the deviations obtained (Figure 2). For example, the standard oxygen dissociation curve of Winslow et al. (11,12) is fitted by the parameters: Po = 5.1 kPa, SO = 0.73, no = 2.72, h = 2.9; and the curve of Zwart et al. (13) is fitted by
-* -

(a1npmpH)2,,,

(lnlO)

(alog10p/apH)8,,.

(lnlO)

. 3

(11)

#{149} +0.004

where p is the Bohr coefficient for constant blood-base excess, taken to be -0.51 for s = 0.5 (16). This gives f3 = -0.45 for s = s0 = 0.867 with the present model. f3 as a function of x is obtained by differentiating equation 2 after substituting the parameters and keeping y, c, CDPG, and T constant:
(3
=

-0.45

{n

tanh[k

(x

xo)]}/n

(12)

Values for (3 calculated with this equation are in good agreement with the experimental data by Hiastala and Woodson (16) (Figure 4). The effect of the blood-base excess concentration (ci,aae) On the parameter a is derived from
=

(alnP/alflPcO2)+,H

(lnpC0J&tbase)s,pH

(13)
0.0 0.5
.

1.0

Fig 2. The deviationof the calculatedoxygen saturationfrom the tabulated valuesof the Sevennghaus oxygen dissociation curve (as), plotted as a function of the oxygen saturation (s) The maximum deviationiss -0.0043 for $ = 0.93.The meandeviation,i.e., theintegral ofthe curve, is -0.0003
=

The first coefficient is 0.11 for normal blood (17). The second is derived by substituting dcuco = dcb (valid for constant pH), hence (alnpcojaci,8++)pH We have used
=

(alflPco
=

ICopH/CHCo

1kHCO3

(14)
CHC0

20 mmoJL

for normal whole blood. 1647

CLINICALCHEMISTRY, Vol. 30, No. 10, 1984

references in the literature for whole blood under physiological conditions. Imai (14) has published values for hemoglobin solutions of 12.6 kJ/mol without 2,3-diphosphoglycerate and 15.2 kJ/mol with a 2,3-diphosphoglycerate concentration of 2 mmol/L. To obtain a realistic value for h, one must have accurate data points for s > 0.98 as well as s <0.02. Applications of the mathematical model are described in the Appendix. Other Mathematical Dissociation Curve Models of the Oxygen

Fig.4. The Bohr coefficient(13) for constant base excess concentration plotted as a function of the oxygensaturation (s) The curve representsthe present mathematicalmodel.The points are experimentalvaluespublished by Hlastala and Woodson (16) The factor for calculating a (equation 9) therefore is alnp/ = 0.11/(20 mmolIL) = 0.005 Lfmmol. The effectof the erythrocyte 2,3-diphosphoglycerate concentration (cp) on a is derived from (alnph3copG)s,pH= (inlO) c1 [aloglQp/a(cDpG/c}p.,)]8H(15)

The latter coefficient is taken to be 0.60 for normal blood (18-20). The erythrocyte concentration of hemoglobin iron is taken to be 20 mmol/L. The factor for calculating a (equation 9) is therefore: alnp/acp0 = (lnlO) 0.60/(20 mmol/L) = 0.07 L/mmol. The reference value for the erythrocyte 2,3-diphosphoglycerate concentration is taken to be 5.0 mmolJL (20). In this model the effects of pH, c, and CDYJ are assumed to be linear and additive. The model does not take into account that /3 varies slightly with pH in the physiological range and that the other coefficients are also functions of pH, c, and temperature (21). A simpler model is obtained if we assume that (3 is independent of s; i.e., a shift of the oxygen dissociation curve is always a simple parallel displacement in the direction of the abscissa. However, for the present purpose (evaluation of the p02-temperature relationship) we were especially interested in the upper part of the curve and therefore preferred the more complicated model. The parameter b reflects the effect of temperature and is based on a temperature coefficient at constant pH (4,22):
(alnp/aT)+H
=

0.055 K

(16)

Temperature changes at constant pH are assumed to cause a parallel displacement of the function in the direction of the abscissa; i.e., x0, but not Yo, h, or no is affected (22, 23). Because a temperature change in a blood sample is accompanied by a pH change, the overall effect of a temperature change at constant s is given by:
(alnp/aT)
= =

(alnp/aT) 0.055 K

pH + (a1np/apH)5#{149} (apHJaT) + (lnlO) f3 (-0.015 K)

(17)

where (3may be obtained from equation 12. (apHJaT)8 is often called the Rosenthal factor. The parameter h reflects the positive, homotropic, allostenc ligand interaction, and the interaction energy (AG) may be calculated as
=

Many different mathematical models of the oxygen dissociation curve have been published (24-26). The classical Adair equation is based on four parameters, namely, the four equilibrium constants (K1, K2, K3, K4) for the successive binding of the four oxygen molecules to hemoglobin (27). Even with an optimal choice of equilibrium constants, the fit to the standard oxygen dissociation curve of Severinghaus is not better than with the present model and, strangely enough, K3 turns out to be negative (10,28). Winslow et al. (11, 12), using the Adair equation as a model of their standard oxygen dissociation curve, also find that the third Adair constant is zero, implying that K3 is zero and K4 is infInity. They caution against assigning any physical meaning to the K values derived from the Adair fit (except for K1). Modified Adair equations with as many as eight parameters have been proposed, without, however, a better fit to the standard oxygen dissociation curve than by our model (24, 29). Severinghaus proposed a simple modified Hill equation, giving s as a function of p (10). The fit to the standard oxygen dissociation curve is similar to the fit obtained with our model, but his equation does not reflect the theoretical S-shape of the Hill plot: s 1 gives n 3 rather than 1; i.e., his model deviates even more from the tabulated values in the upper end than does ours (Figure 1). Severinghaus also suggested a simple equation in which p is a function of s. The latter function, however, does not represent the exact inverse function, which has been recently derived explicitly by Lobdell and turns out to be a somewhat complicated mathematical expression (personal communication). Lobdell also derived his own invertible simple equation (30), which does not, however, give a better fit to the standard oxygen dissociation curve than does our model. The inverse function of our equation 2 cannot be expressed explicitly, and approximative series expressions are rather lengthy. Therefore, we prefer to solve the inverse function by iteration (see Appendix, Procedure B). The present mathematical model may seem unduly complicated. However, once the calculator is programmed, calculating s from p (Procedure A), p from s (Procedure B), p from p and s (Procedure C), or p and s from c2 and c (Procedure D) requires only entering the variables and then calling the function. The latter three functions, although based on Newton-Raphson iterations, require only a few additional program steps once Procedure A has been programmed, and require only a few secondsto run with most programmable calculators. For example, with a HP 9815 calculator, Procedures A through D require 386 program steps and 33 datastorage registers. Plotter instructions for the HP 9862 AXY plotter require a further 389 program steps to generate the Figures printed in this communication.
-

2 h R
.

(18)

The Partial Pressure of Oxygen as a Function of Temperature in Blood

The temperature coefficient for the partial pressure of oxygen in blood, expressed as = dlnp/dT, may be derived theoretically as follows. The concentration of total oxygen in

When h = 3.5 (the Severinghaus curve), the interaction energy is EG = 18.1 kJ/mol, whereas when h = 2.9 (the Winslow curve), the value is 15.0 kJ/mol. There are few 1648 CLINICALCHEMISTRY, Vol. 30, No. 10, 1984

blood (c2) is the sum of the concentrations of free dissolved oxygen (co,) and hemoglobin-bound oxygen (c), where c02 may be derived as the product of p and a (the solubiity coefficient of oxygen) and c02 is the product of s and c (the concentration of hemoglobin iron): Differentiating stant cto, and c, dlnp/dT
=

verylowp values, 4, again falls slightly as previously shown by Reeves (23). Severinghaus (5) derived an approximate equation for 4, as a function of p: 4,. K = 0.013 + 0.058 {[0.243 p1(100 mmHg)]388 + 1)-
.

ap + c#{149}s with respect to temperature

CtO,
=

(19) (T) with con-

which may be transformed 4i K

=

mathematically

intothe follow(27)

and rearranging, -dlnaidT

-

gives: a p1 dsldT (20) 0.042

-

0.029 t.anh[1.90

#{149}

ln(p/19.2 kPa)]

The parameters of equation 27 are easily read off the graph (Figure 5), which may be approximated by a hyperbolic function with an upper and a lower horizontal asymptote. The abscissa and ordinate of the point of symmetry are 19.2 kPa and 0.042, 0.029 is the half.vertical distance between the asymptotes (at 0.013 and 0.07 1, respectively), and 1.90 is ds/dT = (as/aT) + (i3sl#{228}p)T p dlnp/dT (21) the maximal slope at the point of symmetry divided by The terms (as/aT) and (as/ap)r may be expressed in terms of 0.029. By our graph we find parameters slightly different the Hill slope, n, as follows: from those .of Severinghaus (namely, 18.5 kPa, 0.0425, (as/aT) = -n s (1 s) (alrp/a7) and (22) 0.0305, 1.81). The deviation of this approximation from the p (t9sIop)r = ns(1 s) (23) theoretical curve for normal blood is small. However, for pathological blood the deviation may be significant. Combining equations 20-23 gives a very useful expresChanges inp cause a parallel displacement of the function, sion for the temperature coefficient, 4,, where all the variaas do changes in hemoglobin concentration. With very low bles are known or can be calculated: hemoglobin concentrations (e.g., diluted blood or hemoglobin solutions), the shape of the curve also changes so that a 4, = dlnp/dT = [Aap + cpns(1 s)B]/[ap + s(1 s)] (24) typical maximum for 4, is observed at p Pso (Figure 5). Kelman and Nunn (29) derived an approximate equation a, the solubility coefficient for 02 in blood, may be derived for 4, as a function of s but their results deviated significantas follows: ly from the theoretical function for normal blood, their curve being displaced significantly towards the left (35). However, ln[a/(mmol L kPa1)] = ln(0.0105) 11.5 iO TfK a change in p has less influence on 4, as a function of a than + 21 i0 (1TIK)2 (25) in the previous function (with p as the independent variwhere T = T 310.15 K (28,33,34). able), whereas changes in hemoglobin concentration have a (26) similar effect on both functions (Figure 5, right). A -dlnaldT ( 0.012 K). Generally the relationship between hip and T is assumed B = (alnplaT)8, which is given by equation 17. For (3 = to be linear; i.e., dlnp/dT is assumed to equal iMnpIT for -0.51, B = 0.073 K. practical temperature corrections of p. To test this assumpc is taken to be 9.0 mmol/L for normal blood. p. s, and n tion, we calculated the variation of p with temperature for are interrelated through equations 1-3. constant cto,, using Procedure D. The data show (Figure 6) that the relationship between lap and T is virtually linear Application of the Theoretical Equation for very high p values (slope 0.012 K) as well as low p The relationship between 4, and p for normal blood (37 #{176}C, values (slope 0.073 K-) but is slightly curvilinear for p pH = 7.40) was calculated on the basis of equation 24. a, n, values around 19.5 kPa. Here i.lnp/T determined for the and B were all calculated from p by the equations for the temperature interval 27 to 37 #{176}C is slightly lower (0.037 oxygen dissociation curve (equations 1, 2, 3, 12, 17). The K) than the value obtained for the interval 37 to 47#{176}C function between 4, and lnp is S-shaped, falling between A (0.042 K). andB (Figure 5, left). For lowp values, 4, rises towards B; for Comparison of temperature corrections of the 02 partial

A similar

equation for p/(p. IT) was derived by van Stekelenburg (31, 32). The oxygen saturation (s) is a function of both T and p,
.

#{149}

1
2 5 10

3
I

Lltip
I

20

p/kPa

-,

Fig. 5. The p-,-temperature coefficient (dlnpldl) as a function of the oxygen partial pressure (p, left) or of the oxygen saturation (s, right) normal bloodwith = 3.6 kPaandq., =9 mmol/L; abnormal values with normal (2.0 and5.0 kPafromleft to right) abnormal values with normal (0.1.5.0, and 15 mmotfL fromleftto nght)
-,

CLINICALCHEMISTRY, Vol. 30, No. 10, 1984 1649

the inverse function of equation 2 cannot be expressed explicitly; hence, x must be calculated from y by a method of iteration, e.g., a Newton-Raphson iteration:
=

x + (y

y)In,

where yj and n are calculated from x by equations 2 and 3. The iteration starts with x, = x0 for i = 0 and ends when ly y,i a, where a is an arbitrary small number depending upon the accuracy required, e.g., a = 0.0001. Procedure B is used when, for example, Pso is to be calculated from a = 0.5. With the present model, p50 3.56 kPa at standard conditions (37 #{176}C, pH = 7.40, c = 0 mmol/L, cDpG = 5.0 mmol/L). Procedure C. Partial pressure of oxygen at half saturation (p50)to be calculated from a set of corresponding values for partial pressure of oxygen (p) and saturation (s). Initially calculate the parameter b from T (equation 10). Calculate x from p. andy from s (equation 1). Express parameter a as a function of x andy by using equation 2 after substituting the parameters from equations 4, 5, 6, and 8. However, a must be calculated by a Newton-Raphson iteration:
-

a1.1.1 = a1 + {(y Fig. 6. The partial pressure of oxygen (p) as a function of temperature for three differentpartial pressures (at 37#{176}C): 7.4, 20, and 55 kPa
Forall three cases p = 3.6 kPa and =9 mmol/L The curves,whichwere calculatedby Procedure 0 (seeApoendi*). arevirtually linear, although thereisa slightcurvatureforp (37 #{176}C) = 20 kPa (mkkile curve)

y1)I-n1 + tanh [k (x

x0)]}

pressure performed with the theoretical p02-temperature coefficient (equation 24) and the iterative procedure (Procedure D) shows complete agreement, thus providing a cross check of the mathematical derivations. Experimental confirmation of the present calculations is lacking. Van Stekelenburg (31) confirms the expected effect of the hemoglobin concentration, but his data do not allow definite conclusions concerning the linearity of the hip vs T relationship or the effect of p on the p02-temperature coefficient (we have recalculated all the data on the basis of his graphs). Recent data by Gethgen (35) fully confirm the shape of the function as illustrated in Figure 5, but the effects of the hemoglobin concentration or p remain to be studied systematically. None of the previously published approximating equations (10, 29) are fully satisfactory for abnormal blood, i.e., when P50 and or hemoglobin concentration are abnormal; in fact, they do not represent a great simplification as compared with the theoretical equation (equation 24). Most of the commercial blood-gas instruments are programmed to calculate s from the measured p and pH, and some of the instruments also measure the hemoglobin concentration. Therefore it would be possible at little extra costto program the instrument to calculate the p02-temperature coefficient from the more nearly accurate theoretical equation derived here.

where y1, n1, and x0 are calculated from a1 by equations 2,3, and 5. The iteration starts with a1 = 0 for i = 0 and ends when ly yI = a. Once a is calculated, continue calculating p from s = 0.5 as described in Procedure B. The value of a calculated fromp andsshould be compared with the value calculated from equation 9. A difference between the two indicates a variation in allosteric effectors other than H, C02, or 2,3-diphosphoglycerate, or the presence of an abnormal hemoglobin. Often this iteration procedure is unnecessary; e.g., when 0.2 < s < 0.8, the Hill slope may be taken as constant (ii = 2.7). Therefore p50 can be calculated from p and s by using the simple Hill equation:
-

ln[s/(1

s)]

2.7 ln(pIp50).
.

Procedure D. Partial pressure and saturation of O2(P, s) to be calculated from the concentration of total 02 and the concentration of hemoglobin iron (cHb). Calculate the partial pressure of oxygen (p) by a Newton-Raphson iteration:

pi+i

p1 + (c2

cto)l[a

+ c,m n

(1

S1)/p1]

Appendix: Procedures for Typical Calculations

Procedure A. Oxygen saturation (s) to be calculated from the partial pressure of oxygen (p). First calculate the parameters of the oxygen dissociation curve from pH, c, CD1G, and temperature (equations 4-10). Then calculate x from p (equation 1), y from x (equation 2), and finally s from y (equation 1): s = (e + 1Y. Procedure B. Partial pressure of oxygen (p) to be calculated from the oxygen saturation (s). First calculate the parameters of the oxygen dissociation curve from pH, c, CDPG, and temperature (equations 4-10). Calculate y from s (Equation 1). x is to be calculated from y, but unfortunately 1650 CLINICALCHEMISTRY, Vol. 30, No. 10, 1984

where s, n1, and cto,, are calculated from p, by using Procedure A and equations 3 and 19. The iteration starts with pi = 3 kPa for i = 0 and ends when Ico2 c,2) a mmol/L. Iterative procedures for calculating p from c, by somewhat different algorithms have previously been described (10, 23, 24). Procedure D may also be used to correct po2 for temperature; e.g., p measured at T, is to be corrected to T2, assuming that cto2remains constant: first calculates (Procedure A) and then c (equation 19) with the parameters (e.g., pH and a) applying at T,. p for T2 may now be calculated with this c2 and Procedure D by using the parameters applying at T2; i.e., pH (T1) must be corrected to pH (T2) by means of the Rosenthal factor (equation 17), and a(T2) must be calculated (equation 25).
-

References
1. Ashwood ER, Kost G, Kenny M. Temperature corrections of blood gas and pH measurements. Clin Chem 29, 1877-1885 (1983). 2. Burnett RW. Erroneous temperaturecorrection for blood pH and gas measurements.Clin Chem 24, 1850 (1978). Letter. 3. Kofstad J. Calculation of baseexcesson four different blood gas analyzing instruments (IL 613, Corning 175, Radiometer ABL 2, and AVL 940). Proc IFCC Expert Panel on pH and Blood Gases 5, 83-88 (1981).

4. Astrup P, Engel K, Severinghaus JW, Munson E. The influence of temperature and pH on the dissociation curve of oxyhemoglobin of human blood. Scand J Clin Lab Invest 17, 515-523 (1965). 5. SeveringhausJW. Blood gascalculator. JApplPhysiol 21,11081116 (1966).

6. Gothgen I, Wimberley PD. Algorithmsfor calculating the determinants of heated skin surface p7. Acta Anaesthesiol Scand (in
press).

7. Hill AV. The possible effects of the aggregation of the molecules ofhemoglobinon its dissociation curves.JPhysiol (London) 40, ivvii (1910). 8. Perutz MF. Hemoglobinstructureand respiratorytransport.Sci Am 239 (6), 68-86 (1978). 9. Siggaard-Andersen 0. Determination and presentationof acidbase data. Contrib Nephrol 21, 128-136 (1980). 10. Severinghaus JW. Simple accurate equations for human blood 02dissociation computations. JAppl Physiol: Respir Environ Exerc Phy8iol 46, 599-602 (1979). 11. Winslow RM, Swenberg M-L, Berger RL, et al. Oxygen equilibrium curve of normal human blood and its evaluation by AdaWs equation. J Biol Chem 252, 2331-2337 (1977). 12. Winslow RM, Morrissey JM, Berger RL, et al. Variability of oxygen affinity of normal blood: An automated method of measurement. J Appi Physiol: Respir Environ Exerc Physiol 45, 289-297 (1978). 13. Zwart A, Kwant G, Oeseburg B, Zijlstra WG.Oxygen dissociation curves for whole blood, recorded with an instrument that continuouslymeasurespc,2 and sc, independentlyat constant T, Pco2,and pH. Clin Chem 28, 1287-1292 (1982). 14. Imai K. Allosteric Effects in Hemoglobin, Univ. Press, Cambridge, U.K., 1982, pp 114, 149. 15. Tyuma I, Imai K, Schimizu K. Organic phosphates and the oxygenequilibrium function of somehuman hemoglobins. In Oxygen Affinity of Hemoglobin and Red Cell Acid-Base Status, M Rbrth, P Astrup, Eds., Academic Press, New York, NY, 1972, pp 131-146. 16. Hlastala MP, WoodsonRD. Bohr effect data for blood gas calculations.J Appi Physiol: Respir Environ Exert Physiol 55, 1002-1007 (1983). 17. Cathy L, Robert M, Zaar B. Proton- and carbamino-linked oxygen affinity of normal human blood. Acta Physiol Scand 82, 482-492 (1972). 18. Duhm J. The effectof2,3-DPO and other organic phosphateson the Donnan equilibrium and the oxygenaffinity of human blood. Op. cit. (ref 15), pp 583-594. 19. Samaja M, Winslow RM. The separate effects of H and 2,3-

DPG on the oxygen equilibrium curve of human blood. Br J 41, 373-381 (1979). 20. Wimberley PD. Fetal hemoglobin,2,3-diphosphoglycerate and oxygentransport in the newborn premature infant. Scand J Clin Lab Invest 42 (Suppl 160), 23-27 (1982). 21. Siggaard-Andersen 0, Berth M, NOrgaard-Pedersen B, et al. Oxygen-linkedhydrogenion binding of human hemoglobin.Effects of carbon dioxide and 2,3-diphosphoglycerate. IV. Thermodynamical relationshipbetween the variables. Scand J Clin Lab Invest 29, 303-320 (1972). 22. Zwart A, Kwant G, Oeseburg B, Zijlstra WG.Human whole bloodoxygenaffinity: Effectof temperature. J Appi Physiol:Respir Environ Exerc Physiol (in press). 23. Reeves RB. The effect of temperature on the oxygenequilibrium curve of human blood.Respir Physiol 42, 317-328 (1980). 24. Thomas U Jr. Algorithms for selected blood acid-base and bloodgas calculations. J Appi Physiol 33, 154-158 (1972). 25. Gabel RA. Algorithms for calculating and correcting blood-gas and acid-base variables. Respir Physiol 42, 211-232 (1980). 26. NordbyGL, Ellis JH. Efficientcomputersubroutinesfor interconversion ofoxygentensionand saturation. JAppi Physiol: Respir Environ Exert Physiol 51, 1080-1085 (1981). 27. Adair CS. The hemoglobinsystem.VI. The oxygendissociation curve of hemoglobin. J Biol Chem 63, 529-545 (1925). 28. Roughton FJW, Severinghaus JW. Accurate determination of 02 dissociation curve of human blood above 98.7% saturation with data on 02 solubilityin unmodified human bloodfrom 0 to 37#{176}C. J Appi Physiol 35, 861-869 (1973). 29. Kelman OR., Nunn JF. Nomogramsfor correctionof bloodp, p, pH and baseexcessfor time and temperature. JAppI Physiol 21, 1484-1490 (1966). 30. Lobdell DD. An invertible simple equationfor computationof blood 02 dissociationrelations. J Appi Physiol: Respir Environ Exert Physiol 50, 971-973 (1981). 31. van StekelenburgGJ. The influence of temperature changeson the p of normal anaerobicbloodsamples.Respir Physiol 8, 245259 (1970). 32. van Stekelenburg GJ. Temperature correction of blood pc,2 measurements. Clin Chem 30,595-596(1984). Letter. 33. Christoforides C, Hedley-Whyte J. Effect of temperature and hemoglobin concentration on solubility of 02 in blood. J Appl Physiol 27, 592-596 (1969). 34. Siggaard-Andersen 0. The Acid-Base Status of the Blood, Munksgaard, Copenhagen, 1974, p 22. 35. G#{246}thgen I. Heat-inducedchanges in Po2 and Pco of blood. Acta Anaesthesiol Scand 28, 413-417 (1984).
Haematol

CLINICALCHEMISTRY,Vol. 30, No. 10, 1984

1651