OBJECTIVE: To study the routine use of the E-test for susceptibility testing of penicillin-resistant Streptococcus pneumoniae. METHODS: A multicenter study of penicillin-resistant S. pneumoniae (PRSP) was carried out in Brittany, France (10 general hospitals, and two university hospitals including a coordinating center). Each hospital detected PRSP by the oxacillin (5- micro g) disk method and determined the MICs of penicillin G, amoxicillin and cefotaxime by the E-test under routine conditions. All the PRSP strains were collected in a coordinating center and the MICs were checked by the agar dilution method. The classifications obtained from the MICs determined by the E-test and by the reference method were compared. RESULTS: Between 1 July 1993 and 30 June 1994, 128 PRSP strains were collected. Agreement within 1 log2 dilution was obtained for only 62% of strains with benzylpenicillin, 72.5% with amoxicillin and 76% with cefotaxime. These data are well below published values. In addition, 52% of the strains found to be penicillin-resistant by the reference technique were of intermediate resistance according to the E-test. There were major differences in the quality of the results obtained by the participating laboratories. CONCLUSIONS: There are problems of standardization in the routine use of the E-test. Microbiologists should therefore take particular care when performing the test and when reading the results, and ensure that reference strains are included in the assay.