To the Editor: In clinical work, little attention has been paid to high plasma colloid osmotic pressure (Π_{plasma}). Which method for calculating Π from plasma proteins (1, 2, 5) would be the best one at high protein levels?
Landis and Pappenheimer (1) formulated equations for calculation of Π from total plasma protein (TP), albumin, and globulin. In each equation, a firstorder term represented van't Hoff's law, square and cubic terms deviations from that law caused by Donnan effects, and proteinprotein interactions (1). The albumintoglobulin ratio was not included.
Nitta et al. (2) combined and corrected the equations to include albumin and nonalbumin fractions. Into their equation, I have introduced calculations of the fractions from albumin (alb) and TP (represented by C in equations, both in g/dl), with Π_{Nitta,mmHg} multiplied by 1.36 giving
At moderate and high nonalbumin protein (“globulin”) fractions, Π_{Nitta} is higher than Π_{Yamada} and the difference increases with increases of TP. At 85 g/l TP and 45 g/l albumin, Π_{Nitta} is 43.8 and Π_{Yamada} 42.3 cmH_{2}O, with the difference being 1.5 cmH_{2}O. At the same TP and 30 g/l albumin, the difference is 2.3 cmH_{2}O; in analbuminemia (compare with Ref. 1), the difference is 3.8 cmH_{2}O.
The protein fractions undergo change during the course of many human diseases, for example, rheumatoid arthritis. In phases of mild joint activity, the average level of albumin (TP − globulin) was shown to be 47 g/l and that of globulin was 27 g/l (4). In severe joint activity, albumin dropped by 8 g/l and globulin increased by 10 g/l (4), but the calculated drop of Π_{Nitta} (from 38 cmH_{2}O, by 2 cmH_{2}O) and Π_{Yamada} (from 37.5 cmH_{2}O, by 2.5 cmH_{2}O) remained moderate.
A twopore theory (3) revived my interest in Π_{plasma}. At low filtration rates, the plasma − tissue Π differences (ΔΠ) draw lowprotein fluid into plasma through small (radius of ∼4.5 nm) endothelial pores, whereas hydrostatic pressure differences almost unopposed by ΔΠ drive highprotein fluid through sparse large (radius of 25–30 nm) pores (3). Testing of the proposal that ΔΠcorrelating recirculation increases protein clearance from plasma (3), and presumably protein mass (Π?) in tissues, and testing of the authors' (3) early conjecture that this mechanism might participate in the regulation of Π_{plasma} calls for a reliable and cheap (compare with Ref. 5) method for calculation of Π_{plasma}.
The early globulin equation (1) was misleading (2,5). Lacking a mathematical mind, I kindly ask Yamada et al. (5) to tell me whether a nonalbumin fraction cubic term would improve the usefulness of Eq. 2 at the high globulin levels often met in clinical practice.
 Copyright © 2003 the American Physiological Society
REFERENCES
REPLY
To the Editor: I was delighted when I was asked to reply to Dr. Ahlqvist's letter concerning plasma protein osmotic pressure. First, it reminded my wife and me of our memorable trip to Finland for the International Physiology Congress in 1989. Second, in a world gone mad with molecular biological hype, Dr. Ahlqvist's interest in applied physiology is a ray of sanity.
Dr. Ahlqvist's direct question is whether a cubic term for the nonalbumin portion of Eq. 2 would improve its usefulness. The simple answer is, No. In the 1991 paper by Yamada et al. (12), a cubic equation was a statistically better fit for the albumin osmotic pressure, according to our collaborator, Dr. Licko, a fine mathematician and statistician. However, the same could not be said for the nonalbumin portion of the equation. There was too much variation among the data to require a cubic term; the problem Dr. Ahlqvist poses is essentially insoluble in the real world.
If this answer is not sufficient, readers may want to look up my 1987 review paper (11). In terms of pulmonary liquid and protein exchange, my associates and I have found strong evidence for various types of microvascular inhomogeneity in addition to two or more pores. We obtained evidence over the years for the following inhomogeneities: vertical (top to bottom of lung), longitudinal (arterial, capillary, venous), and parallel (side by side). Each type is discussed with references in the article.
My advice is not to push theoretical calculations too far; they contain many simplifying assumptions, not necessarily made explicit. In my opinion, the best clinical approach is to use data obtained from real osmotic pressure measurements. This is especially true in clinical care, where, for example, access to serial plasma samples and pulmonary edema liquid by deep lung suction is most likely available. Perhaps, Dr. Ahlqvist may consider setting up this endeavor in his hospital.
If the reader must rely on calculations, then choose the one that seems most reasonable. As Dr. Ahlqvist showed by calculation in his letter, the variations are of the order of ±10–15%, which in the real world of whole animal physiology is good agreement. To let the other shoe drop, any calculation one applies to plasma osmotic pressure also applies to interstitial osmotic pressure; therefore, the effective differences may be even less than those calculated using plasma alone.
And so it goes.
Footnotes

10.1152/japplphysiol.00694.2002
 Copyright © 2003 the American Physiological Society
The following is the abstract of the article discussed in the subsequent letter:
Yamada, S., M. K. Grady, V. Licko, and N. C. Staub Plasma protein osmotic pressure equations and nomogram for sheep.J Appl Physiol 71: 481–487, 1991.—The equations developed by Landis and Pappenheimer (Handbook of Physiology. Circulation, 1963, p. 961–1034) for calculating the protein osmotic pressure of human plasma proteins have been frequently used for other animal species without regard to the fractional albumin concentration or correction for proteinprotein interaction. Using an electronic osmometer, we remeasured the protein osmotic pressure of purified sheep albumin and sheep plasma partially depleted of albumin. We measured protein osmotic pressures of serial dilutions over the concentration range 0–180 g/l for albumin and 0–100 g/l for the albumindepleted proteins at room temperature (26°C). Using a nonlinear least squares parameterfitting computer program, we obtained the equation of best fit for purified albumin, and then we used that equation together with the measured albumin fraction to obtain the bestfit equation for the nonalbumin proteins. The equation for albumin is II_{cmH2} _{O,39°C} = 0.382C + 0.0028C^{2} + 0.000013C^{3}, where C is albumin concentration in g/l. The equation for the nonalbumin fraction is II_{cmH2} _{O,39°C} = 0.119C + 0.0016C^{2}. Up to 200 and 100g/l protein concentration, respectively, these equations give the least standard error of the estimate for each of the virial coefficients. The computed numberaverage molecular weight for the nonalbumin proteins is 222,000. Using the new equations, we constructed a nomogram, based on the one of Nitta and coworkers (Tohoku J. Exp. Med. 135: 43–49, 1981). We tested the nomogram using 144 random samples of sheep plasma and lymph from 31 sheep. We obtained a correlation coefficient of 0.99 between the measured and nomogram estimates of protein osmotic pressure.
 Copyright © 2003 the American Physiological Society