The present study has involved a sample of 750 medical or medical/surgical intensive care units. The aim was to identify the variations of antibiotic (AB) use, measured in monetary terms, and to sort out explicative variables accounting for the corresponding expense. Activity and expense data have been recorded for 1997. "Second intention" antibiotics have been defined as follows: imipenem, ceftazidime, cefpirome, cefepime, piperacillin/tazobactam, amikacin, isepamicin, vancomycin and teicoplanin. Only 60 evaluable sheets have been sent back. They include data about 28,000 admissions and 183,960 hospital days. The results are presented as means +/- standard deviation (SD) and the variability has been considered to be important if the ratio SD/mean was > 1. Nosocomial infections (NI) surveillance, antibiotic advisory board and restriction of use for some molecules were present in 95%, 67% and 78% of units, respectively. The units usually had 10 beds (range: 6-24) and the mean activity was 468 +/- 184 admissions/year and 3,066 +/- 1,454 hospital days/year. Mean duration of hospitalisation (MDH) was 6.9 +/- 2.7 days and mean omega score 114 +/- 61. Mean age of patients was 56.5 years, IGS II score 35.7 +/- 7; 29 +/- 16% of patients were mechanically ventilated for more than 48 hours and mortality rate was 17 +/- 7%. The mean number of bacterial isolates per unit was 369 +/- 323: Staphylococcus sp. 30% [including 25% of meticillin-resistant Staphylococcus aureus (MRSA)]; enterobacteria 30% (including 14% of cefotaxime-resistant isolates); Pseudomonas sp. 14% (including 40% of ticarcillin R isolates); other 26%. Pharmaceutical expense was 834 +/- 364 FF per day of hospitalisation, including 536 +/- 273 FF for drugs. Antibiotics accounted for 32% of the expense and second intention molecules for nearly 50% of antibiotic expense. More than 80% of antibiotic expense was accounted for by only 10 molecules. The mean cost/hospital day for the most expensive antibiotic, whatever the molecule ranking first, was 27 FF, for the second and third ones 18 and 14 FF. The expense for the tenth molecule was only 3 FF. There was a correlation between antibiotic expense and number of beds, number of hospital days, MDH, omega score, number of patients mechanically ventilated for more than 48 hours, mortality, number of bacterial isolates and incidence of NI. Molecules thought to be active against MRSA and ticarcillin-resistant Pseudomonas sp. accounted for 7 and 20% of total antibiotic expense, as compared to 7.5 and 4.6%, respectively, of bacterial isolates. As a conclusion, in this sample of 60 intensive care units, differences were shown for the cost of antibiotics, but the variability was low, without major discrepancies. Ten molecules accounted for 83% of total antibiotic expense. The financial impact of molecules against ticarcillin-resistant Pseudomonas sp is high.