Results of the second international study on the W.H.O. alpha-foetoprotein standard†

Ten preparations containing various levels of alpha-foetoprotein have been assayed by eight different laboratories in an international collaborative study of the W.H.O. alpha-foetoprotein standard. Included among the preparations, all of which were analyzed by the radioimmunoassay technique, were normal serum, sera from women before and after delivery, sera from teratoblastoma, alcoholic cirrhosis, and primary liver cancer patients, and amniotic fluid. The study establishes the validity of the W.H.O. standard for specimens with a wide range of AFP titer. Results have lower variability when expressed in comparison with the W.H.O. standard than when compared with the local standard used in each laboratory. Within-laboratory variability was some 2·5 times as great using the local standard but both standards show important differences between laboratories; the worst laboratory had more than 10 times greater within-laboratory variance (of the log concentration) than the best after all assays failing a rigorous test of technical validity had already been removed. Agreement between laboratories was better for preparations with more than 45 international units of alpha-foetoprotein/ml such as sera from women in the second half of gestation. For preparations with very low AFP content, such as normal human sera, only one laboratory was capable of providing dose-response curves parallel to that of the standard. A statistical technique (cubic splines) is discussed which allows useful information to be extracted from poor or oddly shaped assay curves. It is also noted, however, that an experienced laboratory worker is often even more effective in recovering results from bad assays.