Continuous improvement of medical test reliability using reference methods and matrix-corrected target values in proficiency testing schemes: application to glucose assay.

The reliability of biological tests is a major issue for patient care in terms of public health that involves high economic stakes. Reference methods, as well as regular external quality assessment schemes (EQAS), are needed to monitor the analytical performance of field methods. However, control material commutability is a major concern to assess method accuracy. To overcome material non-commutability, we investigated the possibility of using lyophilized serum samples together with a limited number of frozen serum samples to assign matrix-corrected target values, taking the example of glucose assays. Trueness of the current glucose assays was first measured against a primary reference method by using human frozen sera. Methods using hexokinase and glucose oxidase with spectroreflectometric detection proved very accurate, with bias ranging between -2.2% and +2.3%. Bias of methods using glucose oxidase with spectrophotometric detection was +4.5%. Matrix-related bias of the lyophilized materials was then determined and ranged from +2.5% to -14.4%. Matrix-corrected target values were assigned and used to assess trueness of 22 sub-peer groups. We demonstrated that matrix-corrected target values can be a valuable tool to assess field method accuracy in large scale surveys where commutable materials are not available in sufficient amount with acceptable costs.

Secular trends in mortality associated with bloodstream infections in 4268 patients hospitalized in a cancer referral center between 1975 and 1989.

OBJECTIVE: To study the trends in mortality over 15 years in hospitalized cancer patients with bloodstream infection. METHODS: The yearly incidence rates and risk of death, by type of microorganism, were calculated for 4268 cancer patients hospitalized between 1975 and 1989 in a French cancer referral center. The relative risk of death (RR) associated with each type of microorganism was estimated using the proportional hazards model, taking into account age, hospital ward, underlying disease, geographical origin and year of the first positive blood culture. RESULTS: The incidence of these infections was five-fold higher in 1989 than in 1975. The largest increases were for coagulase-negative staphylococci (CNS), yeasts and Staphylococcus aureus. For the 3756 patients who had a single-microorganism bloodstream infection, the risk of death compared with that of patients with CNS infection was significantly increased in those with Pseudomonadaceae (RR=5.0), yeasts (RR=3.4), Enterobacteriaceae (RR=3.2), S. aureus (RR=2.8) and streptococci (RR=2.1). The risk of death was not significantly different between patients with a single or several positive blood cultures nor between those with nosocomial or non-nosocomial infections. When the study period was divided in two time periods (1975 to 1982 vs 1983 to 1989), a significant variation (p=0.001) in risk of death associated with the different microorganisms was observed. Most risks were lower from 1983 to 1986 than before 1982. This decrease reached 60% for both S. aureus and Pseudomonadaceae. CONCLUSIONS: These data support of continuing use of aggressive empirical antimicrobial therapy for cancer patients with fever.