Background. Achieving optimal prosthesis alignment during total knee arthroplasty (TKA) is essential. Imageless computer-assisted surgery (CAS) is developed to improve knee prosthesis alignment and with CAS it is possible to perform intraoperative alignment measurements. Lower limb alignment measurements are also performed for preoperative planning and postoperative evaluation. A new stereoradiography system, called EOS, can be used to perform these measurements in 3D and thus measurement errors due to malpositioning can be eliminated. Since both CAS and EOS are based on 3D modeling, measurements should theoretically correlate well. Therefore, objective was to compare intraoperative CAS-TKA measurements with pre- and postoperative EOS 3D measurements. Methods. In a prospective study 56 CAS-TKAs were performed and alignment measurements were recorded two times: before bone cuts were made and after implantation of the prosthesis. Pre- and postoperative coronal alignment measurements were performed using EOS 3D. CAS measurements were compared with EOS 3D reconstructions. Measured angles were: varus/valgus (VV), mechanical lateral distal-femoral (mLDFA) and medial proximal tibial angle (mMPTA). Results. Significantly different VV angles were measured pre- and postoperatively with CAS compared to EOS. For preoperative measurements, mLDFA did not differ significantly, but a significantly larger mMPTA in valgus was measured with CAS. Conclusions. EOS 3D measurements overestimate VV angle in lower limbs with substantial mechanical axis deviation. For lower limbs with minor mechanical axis deviation as well as for mMPTA measurements, CAS measures more valgus compared to EOS. Results of this study indicate that differences in alignment measurements between CAS measurements and pre- and postoperative EOS 3D are mainly due to the difference between weight bearing and non-weight bearing position and potential errors in validity and reliability of the CAS system. Surgeons should be aware of these measurement differences and the pitfalls of both measurement techniques. Level of evidence. IIb. Disclosures. The department of Orthopaedics, University of Groningen, University Medical Center Groningen receives research institutional support from InSpine (Schiedam, NL) and Stryker (Kalamazoo, Mich. USA). One of the authors (ALB) will be and has been paid as a consultant by Zimmer (Warsaw, IN, USA) for purposes of education and training in knee arthroplasty