Stable lyophilized ethylenediaminetetra-acetic acid (EDTA)-blood haemolysates were applied in an external quality assurance programme (SKZL, The Netherlands) for glycohaemoglobin assays in 101 laboratories using 12 methods. The mean intralaboratory day-to-day coefficient of variation (CV), calculated from the assay of 12 unidentified pairs over a period of 1 year, was 5.2% (range: 0.2-28.7). Forty-seven per cent of laboratories did not meet the criterion of CV < 5%, whereas 68% did not meet the clinically more desirable 3.3-3.6%. Linearity, as derived from the analysis of five combinations of two haemolysates with low and high glycohaemoglobin percentages over 6 months, was excellent (mean correlation coefficient 0.9953; range: 0.9188-0.9999). Analysis of two samples with high and low glycohaemoglobin percentages gave mean interlaboratory coefficients of variation of 10% for one method performed by several laboratories and 22% for all methods performed by all laboratories. It is concluded that the majority of laboratories do not meet the clinically desirable intralaboratory precision and that an unacceptably high interlaboratory precision exists.