Fifty-four pediatric cancer patients with a total of 100 febrile neutropenic episodes treated at China Medical College Hospital were randomized to receive meropenem or ceftazidime plus amikacin from January 2001 to April 2002. The characteristics of 76 assessable febrile episodes (39 with meropenem and 37 with ceftazidime plus amikacin) were compared between the 2 groups. The success rate with unmodified therapy was not significantly different between the meropenem group (72%) and the ceftazidime-plus-amikacin group (57%). The incidence of side effects was similar between the 2 groups and these side effects were reversible. Microbiologically documented infection, clinically documented infection, and unexplained fever accounted for 35%, 37%, and 28% of episodes, respectively. The clinical response rates in subgroups of documented infection and unexplained fever did not significantly differ between the 2 treatment groups. Meropenem was significantly more effective than ceftazidime plus amikacin in children at high risks of developing severe infection who had profound neutropenia (absolute neutrophil count [ANC] < 100/mm3), prolonged neutropenia (ANC < 500/mm3 lasting for > 10 days), or clinically deteriorating shock (p=0.045). As an empirical treatment, meropenem seems to be as effective and safe as ceftazidime plus amikacin for febrile episodes in children with cancer and neutropenia. Meropenem is more effective for pediatric cancer patients at the high risk of severe infection.