An evaluation of the reproducibility and accuracy of the NMR human blood test for cancer described by Fossel, E. T., Carr, J. M. and McDonagh, J., (New England Journal of Medicine 315, 1369-1376) in 1986 has been conducted jointly at the National Cancer Institute-Frederick Cancer Research Facility, Frederick, MD (NCI-FCRF) and the National Research Council, Ottawa, Canada (NRC). The influences on the test of the following were studied: (a) subject fasting; (b) sample collection, storage and handling; (c) use of plasma or serum; (d) variations of test results from the same individual with time; (e) NMR observation parameters including field strength and temperature; and (f) variations in obtaining the Fossel Index (FI) (a number defined by Fossel and co-workers as the average of the widths at half height of the regions in the NMR spectrum of human plasma at 1.3 and 0.88 ppm) by different people from the same plotted spectrum. This test was found to be reproducible but not accurate for screening a general asymptomatic population. The accuracy is defined in terms of the sensitivity, specificity, and predictive values of the test. The accuracy of the test results from our laboratories is compared with the accuracies from other laboratories including Fossel's. The correlation of the Fossel Index with total triglyceride content in the serum has been confirmed by analysing blood components using the following technologies: KBr density gradient centrifugation, high resolution agarose gel electrophoresis, high performance gel permeation chromatography, and chemical analysis.