Comparison of 3 automated assays for C-reactive protein in end-stage renal disease: clinical and epidemiological implications.

Chronic inflammation has been repeatedly reported in individuals undergoing hemodialysis. C-reactive protein (CRP) is considered a marker of chronic inflammation, as well as a mediator of the atherosclerotic process. Clinical and epidemiologic studies are based on plasma values obtained with the use of various automated methods. Our aim was to test 3 commercially available methods and compare the values obtained with the use of these tests in a population of individuals undergoing hemodialysis. We compared the following methods: immunoturbidimetry (AU2700 biochemistry analyzer; Olympus, Rungis, France) laser nephelometry (Behring Diagnostics, Marburg, Germany), and nephelometry (Beckman Instruments, Fullerton, Calif. The 3 methods were used in 3 different centers: Montpellier, France; and Pisa and Turin, Italy, respectively. We prepared samples for the estimation of imprecision values (ie, coefficient of variation [CV]) from the plasma of normal patients by adding purified C-reactive protein at concentrations ranging from 2.6 to 180 mg/L for intraassay variation and concentrations of 0, 1, 2, 3, 5, 10, 20, 50, 100, 150, and 180 mg/L for interassay variation. Intraassay imprecision was determined with the use of 10 replicate analyses on the same sample of the same day. We assessed interassay imprecision using the same sample, divided into aliquots and measured on 5 consecutive days. Agreement between methods was assessed on predialysis serum samples collected from patients with stable chronic kidney disease who were undergoing long-term hemodialysis at the 3 different centers (Montpellier,192; Pisa, 56; Turin,98). Serum was separated from the red cells and stored in 3 aliquots at -70 degrees C until it could be analyzed. Samples were thawed only once, circulated among the 3 centers, and each tested with all 3 of the methods. The Beckman method yielded the most precise results, with intraassay CVs ranging from 1 to 2 and interassay CVs ranging from 1 to 4. The Behring method was the least precise, with intraassay and interassay CVs ranging from 12 to 15 and 7 to 16, respectively. The results of the Olympus method fell between those of the other 2 methods. Agreement between the results of the Olympus and Behring methods was satisfactorily. The Beckman and Olympus methods yielded, on average, similar results over the entire range of CRP values. We detected significant disagreement between the Beckman method and the other 2 methods, obtaining results 10 to 100 times lower with the Beckman method. This became evident in terms of kappa-statistics. Our findings emphasize the need for careful assessment of the methods used to detect CRP in serum samples. Failure to do so may ultimately have a negative impact on the real relevance of CRP as a marker and on the role of chronic implication particularly in epidemiologic studies.