OBJECTIVE: To determine the impact of three needleless intravenous systems on needlestick injury rates. DESIGN: Randomized controlled trial. SETTING: 1,000-bed tertiary-care Midwestern hospital. PARTICIPANTS: Nursing personnel from general medical, general surgical, and intensive-care units. INTERVENTIONS: From June 1992 through March 1994, a metal blunt cannula (MBC), two-way valve (2-way), and plastic blunt cannula (PBC) were introduced into three study areas, and needlestick injury rates were compared to three control areas using traditional needled devices. RESULTS: 24 and 29 needlestick injuries were reported in study and control areas. Intravenous-therapy-related injuries comprised 45.8% and 57.1% of injuries in each area. Thirty-seven percent and 20.7% of study and control area needlestick injuries were considered to pose a high risk of bloodborne infection. The 2-way group had similar rates of total and intravenous-related needlestick injuries compared to control groups. The PBC group had lower rates of total and intravenous-related needlestick injuries per 1,000 patient-days (rate ratios [RR], 0.32 and 0.24; 95% confidence intervals [CI95], 0.12-0.81 and 0.09-0.61; P=.02 and P=.003, respectively) and per 1,000 productive hours worked (RR, 0.11 and 0.08; CI95, 0.01-0.92 and 0.01-0.69; P=.03 and P=.005, respectively) compared to controls. CONCLUSIONS: Needlestick injuries continued in study areas despite the introduction of needleless devices, and risks of bloodborne pathogen transmission were similar to control areas. The PBC device group noted lower rates of needlestick injuries compared to controls, but there were problems with product acceptance, correct product use, and continued traditional device use in study areas. Low needlestick injury rates make interpretations difficult. Further studies of safety devices are needed and should attempt greater control of worker behavior to aid interpretation.