Globulin correction of the albumin gradient: Correlation with measured serum to ascites colloid osmotic pressure gradients

The albumin difference or gradient between serum ascites is presumed to be an effective estimate of the colloid osmotic pressure gradient, although this has never been directly demonstrated. The colloid osmotic pressure gradient is controlled by the degree of portal hypertension. Thus the albumin gradient is clinically useful in detecting patients with ascites caused by portal hypertension, although some overlap in such patients' albumin gradients exists compared with those of patients without portal hypertension. Part of this overlap is related to the inverse correlation of the albumin gradient with serum globulin; globulins also contribute to colloid osmotic pressure. The ability to calculate colloid osmotic pressure in serum and ascites with albumin and globulin concentration or to correct the albumin gradient for the impact of globulins might improve the clinical usefulness of the ascitic fluid analysis in determining the presence of portal hypertension in ascitic patients with borderline albumin gradients. Thus we developed equations to calculate colloid osmotic pressure from multivariate discriminate analyses of albumin and globulin concentrations in serial dilution samples of pooled serum and subsequently validated these equations, along with older methods of calculating colloid osmotic pressure. In an initial set of dilution experiments, globulin concentration was closely correlated with the colloid osmotic pressure to albumin concentration ratio (r = 0.956; p < 0.001). Multivariate discriminate analysis yielded an equation for calculating colloid osmotic pressure from albumin (A) and globulin (G) concentration with a ratio of colloid osmotic pressure to albumin (calculated colloid osmotic pressure = A(1.058G + 0.163A + 3.11) and two other equations. In a second set of dilution experiments, the colloid osmotic pressure calculated from this equation correlated closely (p < 0.0001) with the measured colloid osmotic pressure (r = 0.997) and outperformed more traditional methods of calculating colloid osmotic pressure. More important, the calculated colloid osmotic pressure correlated closely with measured colloid osmotic pressure in clinical samples (r = 0.988; p < 0.0001) in 75 patients, 54 of whom had ascites. The measured colloid osmotic pressure gradient from serum to ascites correlated with the albumin gradient (r = 0.847; p <0.0001) as expected. Furthermore, the correlation between the measured gradient in colloid osmotic pressure from serum to ascites compared with the calculated gradient (r = 0.949; p < 0.0001) was an improvement over the correlation of the measured colloid osmotic pressure gradient with the albumin gradient. Finally, the relationship between the albumin gradient, serum globulin and colloid osmotic pressure was simulated in a computerized model system to determine whether the albumin gradient could be directly corrected for the impact of serum globulin. A corrected albumin gradient could be calculated (corrected albumin gradient = 0.16 (serum globulin + 2.5) albumin gradient) and correlated nearly as well with the measured colloid osmotic pressure gradient (r = 0.928; p < 0.001) as the calculated colloid osmotic pressure gradient. We conclude that (a) accurate calculation of colloid osmotic pressure is possible in clinical specimens at a level of preciseness not previously demonstrated in serum specimens, (b) the correlation of the albumin gradient with the colloid osmotic pressure gradient is demonstrated for the first time and (c) a calculated colloid osmotic pressure gradient or corrected albumin gradient are better correlated with the measured colloid osmotic pressure gradient than is the albumin gradient. Therefore a calculated colloid osmotic pressure gradient or corrected albumin gradient may be useful in evaluating ascitic patients with albumin gradients between 0.9 and 1.4 gm%, particularly if the globulin concentration is less than 3.0 gm% or more than 5.0 gm%. (H EPATOLOGY 1992;16:396–403.)