Background: Serum albumin concentration predicts mortality in hemodialysis (HD) patients. While serum albumin concentration correlates with serum concentration of C-reactive protein (CRP) and is dependent upon CRP in multiple regression models in cross sectional studies, CRP does not predict future albumin levels, possibly because CRP changes rapidly, yielding large month-to-month variability in CRP. If inflammation causes rather than is simply associated with hypoalbuminemia, then changes in the levels of acute phase proteins should precede changes in serum albumin concentration.
Methods: The levels of long-lived positive and negative acute-phase proteins (APPs) (C-reactive protein, ceruloplasmin, alpha1 acid glycoprotein, transferrin and albumin) were measured longitudinally in 64 HD patients and a regression model was constructed to predict future albumin levels. Normalized protein catabolic rate (nPCR) was measured monthly. The number of repeated measurements ranged from 9 to 39 in each patient (median 22 and a mean of 23 measurements). To construct a model that would predict serum albumin concentration at any time j, values of all longitudinally measured APPs, positive and negative at any time j - 1, approximately 30 days prior to time j, were used. Other demographic factors (such as, race, access type, and cause of renal failure) also were incorporated into the model.
Results: The model with the best fit for predicting serum albumin at time j included albumin, ceruloplasmin, and alpha1 acid glycoprotein measured at time j - 1. The only demographic variable with subsequent predictive value was diabetes.
Conclusions: The finding that changes in the concentration of the long lived APPs measured one month earlier are associated with predictable changes in the future concentration of serum albumin suggest that changes in inflammation are likely to be causal in determining serum albumin concentration in hemodialysis patients.