Background and purpose Cup failure is a recognized problem in total trapeziometacarpal (TM) joint prostheses; it may be related to poor fixation, which can be revealed by radiostereometric analysis (RSA). We compared the early implant migration of cemented trapezium cups to that of uncemented screw cups.Patients and methods In a prospective, parallel-group, randomized patient-blinded clinical trial, we included 32 hands in 28 patients (5 males) with a mean age of 58 (40–77) years and with Eaton stage-2 or -3 osteoarthritis of the trapeziometacarpal joint. Patients were randomized to surgery with a cemented DLC all-polyethylene cup (C) (n = 16) or an uncemented hydroxyapatite-coated chrome-cobalt Elektra screw cup (UC) (n = 16). Uncemented cups were inserted without threading of the bone. Stereoradiographs for evaluation of cup migration (primary effect size) and DASH and pain scores were obtained during 2 years of follow-up.Results The 2-year total translation (TT) was similar (p = 0.2): 0.24 mm (SD 0.10) for the C (n = 11) and 0.19 mm (SD 0.16) for the UC (n = 11). Variances were similar (p = 0.4). Judged by RSA, 2 UC cups and 1 C cup became loose (TT > 1 mm). Both UC cups were found to be loose at revision. Grip strength, pain, and DASH scores were similar between groups at all measurement points.Interpretation Early implant fixation and clinical outcome were equally good with both cup designs. This is the first clinical RSA study on trapezium cups, and the method appears to be clinically useful for detection of loose implants.