Background: Patient-specific instrumentation in TKA has the proposed benefits of improving coronal and sagittal alignment and rotation of the components. In contrast, the literature is inconsistent if the use of patient-specific instrumentation improves alignment in comparison to conventional instrumentation. Depending on the manufacturer, patient-specific instrumentation is based on either MRI or CT scans. However, it is unknown whether one patient-specific instrumentation approach is more accurate than the other and if there is a potential benefit in terms of reduction of duration of surgery., Questions/purposes: We compared the accuracy of MRI- and CT-based patient-specific instrumentation with conventional instrumentation and with each other in TKAs. The three approaches also were compared with respect to validated outcomes scores and duration of surgery., Methods: A randomized clinical trial was conducted in which 90 patients were enrolled and divided into three groups: CT-based, MRI-based patient-specific instrumentation, and conventional instrumentation. The groups were not different regarding age, male/female sex distribution, and BMI. In all groups, coronal and sagittal alignments were measured on postoperative standing long-leg and lateral radiographs. Component rotation was measured on CT scans. Clinical outcomes (Knee Society and WOMAC scores) were evaluated preoperatively and at a mean of 3 months postoperatively and the duration of surgery was analyzed for each patient. MRI- and CT-based patient-specific instrumentation groups were first compared with conventional instrumentation, the patient-specific instrumentation groups were compared with each other, and all three approaches were compared for clinical outcome measures and duration of surgery., Results: Compared with conventional instrumentation MRI- and CT-based patient-specific instrumentation showed higher accuracy regarding the coronal limb axis (MRI versus conventional, 1.0° [range, 0°-4°] versus 4.5° [range, 0°-8°], p < 0.001; CT versus conventional, 3.0° [range, 0°-5°] versus 4.5° [range, 0°-8°], p = 0.02), femoral rotation (MRI versus conventional, 1.0° [range, 0°-2°] versus 4.0° [range, 1°-7°], p < 0.001; CT versus conventional, 1.0° [range, 0°-2°] versus 4.0° [range, 1°-7°], p < 0.001), and tibial slope (MRI versus conventional, 1.0° [range, 0°-2°] versus 3.5° [range, 1°-7°], p < 0.001; CT versus conventional, 1.0° [range, 0°-2°] versus 3.5° [range, 1°-7°], p < 0.001), but the differences were small. Furthermore, MRI-based patient-specific instrumentation showed a smaller deviation in the postoperative coronal mechanical limb axis compared with CT-based patient-specific instrumentation (MRI versus CT, 1.0° [range, 0°-4°] versus 3.0° [range, 0°-5°], p = 0.03), while there was no difference in femoral rotation or tibial slope. Although there was a significant reduction of the duration of surgery in both patient-specific instrumentation groups in comparison to conventional instrumentation (MRI versus conventional, 58 minutes [range, 53-67 minutes] versus 76 minutes [range, 57-83 minutes], p < 0.001; CT versus conventional, 63 minutes [range, 59-69 minutes] versus 76 minutes [range, 57-83 minutes], p < .001), there were no differences in the postoperative Knee Society pain and function and WOMAC scores among the groups., Conclusions: Although this study supports that patient-specific instrumentation increased accuracy compared with conventional instrumentation and that MRI-based patient-specific instrumentation is more accurate compared with CT-based patient-specific instrumentation regarding coronal mechanical limb axis, differences are only subtle and of questionable clinical relevance. Because there are no differences in the long-term clinical outcome or survivorship yet available, the widespread use of this technique cannot be recommended.