Background and purpose - Computer-assisted surgery (CAS) in total knee arthroplasty (TKA) has been used in recent years in the hope of improving the alignment and positioning of the implant, thereby achieving a better functional outcome and durability. However, the role of computer navigation in TKA is still under debate. We used radiostereometric analysis (RSA) in a randomized controlled trial (RCT) to determine whether there are any differences in migration of the tibial component between CAS- and conventionally (CONV-) operated TKA. Patients and methods - 54 patients (CAS, n = 26; CONV, n = 28) with a mean age of 67 (56-78) years and with osteoarthritis or arthritic disease of the knee were recruited from 4 hospitals during the period 2009-2011. To estimate the mechanical stability of the tibial component, the patients were examined with RSA up to 24 months after operation. The following parameters representing tibial component micromotion were measured: 3-D vector of the prosthetic marker that moved the most, representing the magnitude of migration (maximum total point motion, MTPM); the largest negative value for y-translation (subsidence); the largest positive y-translation (lift-off); and prosthetic rotations. The precision of the RSA measurements was evaluated and migration in the 2 groups was compared. Results - Both groups had most migration within the first 3 months, but there was no statistically significant difference in the magnitude of the migration between the CAS group and the CONV group. From 3 to 24 months, the MTPM (in mm) was 0.058 and 0.103 (p = 0.1) for the CAS and CON groups, respectively, and the subsidence (in mm) was 0.005 and 0.011 (p = 0.3). Interpretation - Mean MTPM, subsidence, lift-off, and rotational movement of tibial trays were similar in CAS- and CONV-operated knees.