Serial prealbumin levels as predictors of outcomes in a retrospective cohort of peritoneal and hemodialysis patients

Objective. Although earlier research has suggested that baseline prealbumin level is an independent predictor of outcome among dialysis patients, the prognostic importance of serial prealbumin levels is less clear. The present study had 3 objectives: first, to determine if prealbumin (a marker of visceral protein stores with a relatively short half-life) predicts subsequent albumin levels taken at least 1 month later; second, to examine the association between serial prealbumin levels and clinical outcome; and third, to examine the association between changes in prealbumin level and outcome. Design: The prognostic value of serial prealbumin levels was examined by linear regression analysis and Cox hazard models in an observational cohort study using a repeated measures design and time-dependent covariates. Setting: Patients were followed by a tertiary care center, receiving hemodialysis (HD; at either an in-center dialysis unit or one of several satellite units operated by the hospital) or home peritoneal dialysis (PD). Patients: A retrospective cohort was identified consisting of 268 incident and prevalent chronic HD and PD patients receiving dialysis from June 1998 to September 1999. Main outcome: The study examined the association between serial prealbumin measurements and future laboratory and clinical outcomes (albumin, hospitalization, and death). Results: Serial prealbumin values were independent predictors of future albumin levels among HD patients (P = .04), but not PD patients. Independent predictors of hospitalization included diabetes for PD patients (P = .0012) and advanced age for HD patients (P = .0008). Advanced age and diabetes were independent predictors of death for both HD (P = .0001 and P = .0368) and PD patients (P = .0014 and P = .0164). Serial prealbumin values, measured as time-dependent covariates, did not predict hospitalization or death. Further analyses examined the prognostic value of changes in prealbumin and albumin values as time-dependent covariates. The final multivariate analysis identified low baseline albumin level as an independent predictor of hospitalization among HD patients (P = .0282), whereas low baseline prealbumin was an independent predictor of death for HD patients (P = .0001). Interestingly, negative changes in serial prealbumin measurements were also independent predictors of death among HD patients (P = .0025). Conclusion: Serial prealbumin measurements predict subsequent albumin values among HD patients. As well, low baseline prealbumin level is an independent predictor of adverse outcome among HD patients. Although repeated prealbumin measurements in and of themselves were of no added prognostic value, failing prealbumin values identified by repeated measurements were additional independent predictors of death. These results support the clinical utility of regular prealbumin monitoring among HID patients. (C) 2001 by the National Kidney Foundation, Inc.