We report the results of an external quality assessment scheme for serum total cholesterol measurement involving about 100 Italian laboratories participating in an epidemiological study of post myocardial infarction. Two frozen human serum pools with Abell-Kendall assigned values are distributed quarterly at the laboratories (up to now seven events occurred); the obtained results are evaluated and discussed. In one exercise (# 5) duplicated measurements were repeated on three different days. Eighty-five to 98% of the laboratories obtained results within the total error limits (+/- 8.9%). But, while precision (calculated on the six replicates of exercise # 5) is good (90% of the laboratories obtained CV < 3%), inaccuracy problems are evident in every event. Indeed the mean bias from the reference method value ranged from 1.54 and 3.49% in the various events.