Among the biological markers of morbidity and mortality, albumin holds a key place in the range of criteria used by the High Authority for Health (HAS) for the assessment of malnutrition and the coding of information system medicalization program (PMSI). If the principle of quantification methods have not changed in recent years, the dispersion of external evaluations of the quality (EEQ) data shows that the standardization using the certified reference material (CRM) 470 is not optimal. The aim of this multicenter study involving 7 sites, conducted by a working group of the French Society of Clinical Biology (SFBC), was to assess whether the albuminemia values depend on the analytical system used. The albumin from plasma (n=30) and serum (n=8) pools was quantified by 5 different methods [bromocresol green (VBC) and bromocresol purple (PBC) colorimetry, immunoturbidimetry (IT), immunonephelometry (IN) and capillary electrophoresis (CE)] using 12 analyzers. Bland and Altman's test evaluated the difference between the results obtained by the different methods. For example, a difference as high as 13 g/L was observed for the same sample between the methods (p <0.001) in the concentration range of 30 to 35 g/L. The VBC overestimates albumin across the range of values tested compared to PBC (p <0.05). PBC method gives similar results to IN for values lower than 40 g/L. For IT methods, one of the technical/analyzer tandem underestimates the albumin values inducing a difference of performance between the immunoprecipitation methods (IT vs IN, p <0.05). Although, the albumin results are related to the technical/analyzer tandem used. This variability is usually not taken into account by the clinician. Thus, clinicians and biologists have to be aware and have to check, depending on the method used, the albumin thresholds identified as risk factors for complications related to malnutrition and PMSI coding.
Keywords: albumin; analytical methods; commutability; nutrition.