Introduction: Roche Diagnostics runs a quality control scheme named QCS (Quality Control Service). QCS is a combination of internal and external quality control. Both procedures use the same data getting from analyzers. In this scheme Roche considered the problem of matrix effect. Matrix effect is the difference between a human sample and the corresponding control sample. Purpose: The detection of matrix effect and the institution of corrective actions in order to prevent this phenomenon on the future. Material - Method: As experimental materials were used the QCS’s control samples (PCU QCS 10651257/922 and PCTM QCS 10651265/922). These control specimens (PreciControls) were analyzed on the analyzers Roche/Hitachi Elecsys 1010 and 2010. Both analyzers use the technology of electrochemiluminescense and they measure a large variety analytes. The statistical analysis and storage were done with the use of program QCS Easy 4. Results: Statistical important differences were found between the results of human and control samples. These findings reveal the following: Roche determine its analytes on PreciControls as well as on human serum pools. These pools are stored in such conditions that retain their substances steady for a long time. On the other hand Roche determine the consensus mean of PreciControls before their posting to its customers. Although PreciControls are analyzed in referral laboratories, Roche noticed differences when different reagent’s lots were used for the determination of the same PreciControl’s lot. There is no any similar difference between different reagent’s lots and the same human pool. Discusion: The explanation of these differences lies in the matrix effect. It seems that the conservatives and the additives which are added in the serum change the protein structure, especially the antigen chemical affinity. Conclusion: Because of the matrix effect Roche proposes that the interlaboratory comparisons must refer not only in a certain control serum but in a combination of control serum and reagent lot.