Since 1992, there has been an increase in the incidence of Enterobacter sepsis in the neonatal intensive care unit (NICU) of the authors' hospital. From 1995 to 1997, a prospective molecular epidemiological survey of the colonizing and infecting strains isolated from neonates was conducted. Enterobacter cloacae was the most frequent cause of neonatal sepsis, accounting for 19.2% of all neonatal infections, reaching a peak incidence of 2.2/1000 during 1996. Fifty isolates from the NICU and four epidemiologically unrelated strains were characterized by pulse-field gel electrophoresis (PFGE), ribotyping, enterobacterial repetitive intergenic consensus (ERIC)-PCR and plasmid profiling. PFGE was the most discriminatory technique and identified 13 types (two of them classified into two and three subtypes) compared with ERIC-PCR, plasmid profiling and ribotyping that identified 11, 11 and seven types, respectively. A good correlation was found between all techniques. Five different clones caused 15 cases of sepsis. Clones A and B were prevalent in 1995 and 1996, but they were not isolated in 1997. An outbreak caused by clone G in 1997 was controlled by cohort nursing and hygienic measures, without changing the antibiotic policy. Strains were characterized by their antibiotic resistance pattern and divided into three groups. Group I correlated with PFGE types A, B1 and B2, which hyperproduced Bush type 1 chromosomal beta-lactamase and expressed extended-spectrum ?-lactamases (ESBLs). Group II only hyperproduced Bush type 1 chromosomal beta-lactamase and correlated with PFGE-types D1, D2, D3 and I. Finally, Group III, with inducible beta-lactamases, correlated with the rest of PFGE types. The sudden disappearance of E. cloacae after reinforcement of hygienic measures confirms the importance of patient-to-patient transmission.