The advantages of a multicenter trial can easily be lost if results from individual centers cannot be safely combined for statistical analysis. One objective of the Kroc Study was to develop methods that would allow valid amalgamation of results from laboratories at the six clinical centers and a central biochemical laboratory at the University of Newcastle upon Tyne. Responsibilities of the local laboratories, in addition to measurement of plasma glucose, creatinine, and glycosylated hemoglobin, were to obtain and prepare samples for measurement of plasma glucose, glycosylated hemoglobin, and serum lipids at the central laboratory, of C-peptide at the University of Chicago, and to collect and prepare samples for measurement of urinary albumin excretion at Guy's Hospital. The central laboratory was additionally to provide a system to ensure the comparability of plasma glucose determinations between and within centers, and to advise on common procedures for sample handling.
Major problems were encountered with sample labeling, dispatch, and transport to the central laboratory. Although central determinations of plasma glucose and serum lipids were still possible in transported specimens, central assay of glycosylated hemoglobin proved inaccurate and useless. Compliance with the plasma glucose quality control program was variable among centers. Although the difference between the centers recording the highest and lowest values was 24.6% of the mean estimate of plasma glucose level, correlation between local and central plasma glucose determinations was good (r = 0.99, see pages 22–26).
From this experience, we suggest that for satisfactory performance of a multicenter trial an individual is given responsibility for coordinating efforts between laboratories, supervising and auditing local procedures of sample preparation and quality control, and ensuring adequate communication between the central laboratory, local centers, and the data collation center. A biochemist at each local laboratory should in turn be personally responsible for the specimens generated by each experiment, and preparation and dispatch of samples to the central laboratory. Duplication of assays of critical variables, namely at local and central laboratories, is recommended. Blood glucose and glycosylated hemoglobin assays should be by a single method using one type of instrument, the methods chosen being selected by their ability to produce reliable and comparable results in different hands.
These measures should minimize the likelihood of loss of critical information due to poor sample management, and the variability inherent in the combined use of biochemical data derived from multiple laboratories.