Our objective was to evaluate the effects of three doses of Lactobacillus acidophilus strain NP51 and a combination treatment of strains NP51 and NP45 on prevalence of Escherichia coli O157 in cattle. Three hundred steers were assigned randomly to 60 pens (five steers per pen) and received one of five treatments: (i) control, no added direct-fed microbial; (ii) HNP51, high dose of NP51 at 109 CFU per steer daily; (iii) MNP51, NP51 at 108 CFU per steer daily; (iv) LNP51, low dose of NP51 at 107 CFU per steer daily; and (v) NP51+45, NP51 at 109 CFU per steer daily and NP45 at 106 CFU per steer daily. All direct-fed microbial treatments included Propionibacterium freudenreichii at 109 CFU per steer. Individual rectal fecal samples were collected on arrival and every 28 days throughout the feeding period. Fecal and hide samples were collected on the day of harvest. Samples were analyzed for presence of E. coli O157 using immunomagnetic separation methods. Cattle receiving HNP51, MNP51, and LNP51 had a lower prevalence (P < 0.01) of E. coli O157 throughout the feeding period compared with the controls, and the dose response for NP51 was a linear decrease in prevalence with increasing dose (P < 0.01). No decrease in prevalence for cattle receiving the combination NP51+45 was detected compared with controls (P = 0.15). E. coli O157 prevalences averaged across collection times were 23.9, 10.5, 9.9, 6.8, and 17.3% for cattle in the control, LNP51, MNP51, HNP51, and NP51+45 groups, respectively. Least squares mean estimates of fecal prevalence at harvest of E. coli O157 were 31.7, 12.5, 17.4, 8.2, and 41.6% among cattle in the control, LNP51, MNP51, HNP51, and NP51+45 groups, respectively. Least squares mean estimates of the percentage of positive hide samples at harvest were 8.7, 5.9, 4.8, 3.4, and 8.6% among cattle in the control, LNP51, MNP51, HNP51, and NP51+45 groups, respectively. The greatest decrease in E. coli O157 carriage was achieved using NP51 at 109 CFU per steer.