Results of urinalysis and culture of 2181 urine specimens obtained by catheter from febrile children aged less than 24 months were analyzed to determine the following: (1) an optimal cutoff point in considering a bacterial colony count clinically "significant," (2) the accuracy of leukocyte esterase and nitrite tests in identification of pyuria and bacteriuria, and (3) the utility of pyuria (defined as > or = 10 leukocytes/mm3) in the discrimination of urinary tract infection from asymptomatic bacteriuria. Among 110 urine cultures with > or = 10,000 colony-forming units per milliliter, 92 (84%) had > or = 100,000 CFU/ml, 10 (9%) had 50,000 to 99,000 CFU/CFU/ml and 8 (7%) had 10,000 to 49,000 CFU/ml. Urine specimens with 1000 to 49,000 CFU/ml were more likely than specimens with > or = 50,000 CFU/ml to yield Gram-positive or mixed organisms (36/60 vs 7/109; p < 0.001). A count of < 10 leukocytes/mm3 was almost invariably associated with a sterile culture; a count of > or = 10 leukocytes/mm3 was found in 93 of 102 patients with > or = 50,000 CFU/ml. The dipstick leukocyte esterase test had sensitivities of 52.9% and 66.7% in detecting > or = 10 leukocytes/mm3 and > or = 20 leukocytes/mm3, respectively. The dipstick nitrite test had a sensitivity of 31.4% in detecting bacteriuria (> or = 50,000 CFU/ml). Acute pyelonephritis was diagnosed by a renal scan with dimercaptosuccinic acid labeled with technetium 99m in 50 (77%) of 65 patients with > or = 10 leukocytes/mm3 but in none of five patients with < 10 leukocytes/mm3 (p < 0.01). The findings in these five patients were consistent with colonization of the urinary tract rather than infection. For urine specimens obtained by catheter, we believe that urinary tract infection is best defined by both a leukocyte count > or 10/mm3 and a CFU count > or = 50,000/ml. This definition almost always discriminates among true urinary tract infection, bacteriuria resulting from contamination of the urine specimen, and asymptomatic bacteriuria.