Human pooled serum from diabetic and non-diabetic subjects, and calibrators containing glycated protein or 1-deoxy-1-morpholinofructose (DMF) in a variety of matrices, were distributed to 10 laboratories. When they used their own assay conditions and calibrators, the inter-laboratory variation was unacceptably high. However, when the pool from diabetic patients was reassayed using a calibrator with an assigned value prepared from freeze-dried human serum, and containing no DMF, inter-laboratory variation was reduced significantly. Inter-laboratory agreement for the pool from non-diabetic subjects remained poor despite recalibration. Recalibration using either serum or albumin based solutions of DMF as calibrator failed to effect any significant reductions in inter-laboratory variation. Secondary calibrators based on a protein matrix with no added DMF are recommended for routine use.