In a prospective randomized study, ceftazidime monotherapy was compared with a combination of ceftazidime plus cephalothin in 102 febrile neutropenic patients. Thirty bacteriologically documented infections, of which 23 were bacteremias, in 48 clinically assessable patients were treated with ceftazidime alone. Twenty-four bacteriologically proven infections, of which 18 were bacteremias, in 42 clinically assessable patients were treated with a combination of ceftazidime and cephalothin. The clinical response rates in assessable patients were 77% for ceftazidime monotherapy and 88% for the combination. The bacteriological clearance rate was 70% for ceftazidime monotherapy and 79% for the combination. Efficacy against gram-negative pathogens appeared to be excellent, with 93% clearance for ceftazidime monotherapy and 100% clearance for the combination. The bacteriological clearance of gram-positive infections was only 60% for both regimens, with failures mainly due to Streptococcus faecalis and Streptococcus sanguis, which are primarily resistant to both ceftazidime and cephalothin. After addition of vancomycin to those infections which did not respond to empiric therapy, bacteriological clearance rates of 94% (ceftazidime plus vancomycin) and 90% (ceftazidime and cephalothin plus vancomycin) were achieved. Three superinfections were registered in the ceftazidime group and two were seen in the combination group. Other adverse effects of ceftazidime were minimal and were not enhanced by combination with cephalothin. It is concluded that ceftazidime is an effective drug for the empiric treatment of febrile neutropenic patients, especially if one is prepared to modify therapy if resistant gram-positive strains or mycotic infections are encountered. Neither the clinical nor bacteriological cure rates could be substantially improved by adding cephalothin to ceftazidime in initial empiric treatment of febrile neutropenic patients.