C-reactive protein as an outcome predictor for maintenance hemodialysis patients

Abstract

The possible association between inflammatory processes and other outcome measures in ESRD patients led us to measure the blood C-reactive protein (CRP) concentration in a large sample of hemodialysis patients, and to evaluate its statistical relationship with other common laboratory measures and patient survival. This was performed in a prospective, observational analysis with mortality as the principal outcome measure. One thousand fifty-four routine blood samples, collected from as many patients during June and July 1995 (one sample per patient), were randomly selected for measurement of CRP, prealbumin, and other routine laboratory measures. Six months after the initial blood tests, patient survival was determined: Logistic regression analysis was the primary statistical tool used to evaluate laboratory associations with odds of death. Bivariate regression and correlation analyses were performed using all available data. The distribution of CRP values was skewed; approximately 35% of the values exceeded the upper limit of the laboratory's reference range. Serum albumin and prealbumin concentrations both correlated with the serum creatinine concentration (r = 0.378 and r = 0.347, respectively; P's < 0.001), and were inversely associated with the CRP (r = -0.254 and r = -0.354, respectively; P's < 0.001). CRP was also inversely associated with blood hemoglobin concentrations (r = -0.235; P < 0.001). Using multiple regression analysis to further explore these relationships, the serum creatinine concentration was inversely associated with CRP (r = -0.140; P < 0.001). However, after adjustment for the linkage of the serum creatinine with the serum albumin concentration (r = -0.378; P < 0.001), no relationship with creatinine was observed. Before and after adjustment for serum albumin and prealbumin concentration, the ferritin concentration correlated directly with CRP (r = 0.148; P < 0.001). Ferritin was inversely and highly correlated with total iron binding capacity (r = -0.516; P < 0.001). Independent associations of hemoglobin with albumin (t = 7.16; P < 0.001), prealbumin (t = 2.39; P = 0.017), and CRP (t = -4.27; P < 0.001) were observed. Also, the dose of erythropoietin was directly associated with the CRP concentration, before (r = 0.081, P = 0.009) and after (t = 2.03, P = 0.042) adjustment for the serum albumin and iron concentrations. CRP correlated directly with neutrophil (r = 0.318; P < 0.001) and platelet counts (r = 0.180; P < 0.001), but was weakly and inversely correlated with the lymphocyte count (r = -0.071; P = 0.04). A logistic regression analysis performed using the laboratory variables revealed a strong, independent, and inverse relationships between the serum albumin and creatinine concentrations, total lymphocyte count, and the odds risk of death. In this model, no significant relationship was observed between the odds risk of death and CRP. The data presented herein suggest that: (1) strong predictable associations exist among laboratory proxies for malnutrition, anemia, and the acute phase reaction, and (2) the pathobiology implied by these laboratory abnormalities influence patients' mortal risk primarily through depletion of vital body proteins, not inflammation.