Radiostereometric analysis (RSA) is an accurate and precise radiographic method that can be used to measure micromotion of implants and study joint kinematics in vivo. A calibration cage with radiopaque markers is used to calibrate the RSA images;​ however, the thickness (250 mm) of the calibration cage restricts the available area for the patient and equipment during RSA recordings. A thinner calibration cage would increase the recording area, facilitate handling of the cage, and ease integration of the cage with the RSA system. We developed a thinner calibration cage without compromise of accuracy and precision. First, we performed numerical simulations of an RSA system, and showed that the calibration cage thickness could be decreased to 140 mm maintaining accuracy and precision using 40 fiducial and 30 control markers. Second, we constructed a new calibration cage (NRT cage) according to the simulation results. Third, we validated the new calibration cage against two state-of-the-art calibration cages (Umeaa cage and Leiden cage) in a phantom study. All cages performed similar for marker-based analysis, except for y-rotation, where the Umeaa cage (SD = 0.064 mm) was less precise compared to the NRT (SD = 0.038 mm) and Leiden cages (0.042 mm) (p = .01). For model-based analysis the NRT cage had superior precision for translations (SD ≤ 0.054 mm) over the Leiden cage (SD ≤ 0.118 mm) and Umeaa cage (SD ≤ 0.093 mm) (p < .01). The combined study confirmed that the new and thinner calibration cage maintained accuracy and precision at the level of existing thicker calibration cages.