Laboratory standardization of a large international clinical trial: the DAIS experience. DAIS Project Group. Diabetes...

To implement a quality control program for the standardization and harmonization of lipid and lipoprotein analyses as performed at two core laboratories (St. Paul's Hospital, UBC [Vancouver], and NPHI [Helsinki]) for the Diabetes Atherosclerosis Intervention Study (DAIS).

A DAISSOFT computer program was designed to minimize the occurrence of data and sample management errors during the course of the study. Fresh human serum was used for the provision of an accuracy based external quality control program that monitored the analytical performance of lipid testing at these two laboratories. A separate program was designed for monitoring hemoglobin A1c (HbA1c). At the outset of the study, allowable total error goals were established for each analyte. Ongoing performance was monitored using bimonthly blinded challenges of fresh human serum. The two EQA programs routinely monitored the analysis of total cholesterol, calculated LDL-cholesterol, HDL-cholesterol, net triglycerides, apoprotein A-1, apoprotein B, and HbA1c.

The EQA precision and accuracy data for the measurement of total cholesterol at the two core laboratories over the last 5 years indicated both laboratories operated with good precision, approximately 1% CV over the time period. The accuracy at both laboratories was similar initially. Part way through the study, the accuracy of the cholesterol method at NHPI tended to drift upward with an operating positive bias (+3%) relative to the Abell Kendall reference method. Triglyceride measurements were the most problematic for the study. By EQA cycle 8, the accuracy of the method at UBC had stabilized and was meeting the accuracy goals of the study. NPHI's method was negatively biased relative to the accuracy base of the DAIS study. In spite of recalibrating their method, NPHI found it difficult to maintain consistent accuracy for the measurement of triglycerides during the study. Both laboratories operated their HDL methods with excellent precision. Accuracy at NHPI was well maintained over the course of the study whereas the accuracy of HDL measurements at UBC was more problematic. There was an inconsistent variation in the accuracy of apoprotein A-1 measurements at both laboratories. In most cases, the bias would be corrected by the time of the next EQA challenge. In the case of apo B, one laboratory was standardized to the CDC while the other laboratory was standardized to IFCC/WHO. The discrepancy between these two accuracy bases was >20%. Recalibration to a common accuracy base rectified the problem. Only minor problems were encountered with the precision and accuracy of the DIAMAT assay for hemoglobin A-1c. The two DAIS core laboratories consistently operated within the 9% total error goals of the study for HbA1c.

Through the use of this program, the two DAIS core laboratories were able to maintain their lipid analyses within the limits of allowable total error that had been established for the study.