ObjectiveThe purpose of the present study was to determine the learning curve for a novel seven-axis robot-assisted (RA) total knee arthroplasty (TKA) system and to explore whether it could provide superior short-term clinical and radiological outcomes compared with conventional surgery.MethodsIn the present retrospective study, 90 patients who underwent RA-TKA were included in robot-assisted system (RAS) group and 90 patients who underwent conventional TKA were included in the conventional group. The duration of surgery and robot-related complications were recorded to evaluate the learning curve through cumulative sum and risk-adjusted cumulative sum methods. The demographic data, preoperative clinical data, preoperative imaging data, duration of surgery, alignment of the prosthesis, lower limb force line alignment, Knee Society score, 10-cm visual analog scale pain score and range of motion were compared between the RAS and conventional groups. In addition, the proficiency group was compared with the conventional group using propensity score matching.ResultsRA-TKA was associated with a learning curve of 20 cases for the duration of surgery. There was no significant difference in indicators representing the accuracy of the prosthetic installation between the learning and proficiency phases in RA-TKA group patients. A total of 49 patients in the proficiency group were matched with 49 patients from the conventional group. The number of postoperative hip–knee–ankle (HKA) angle, component femoral coronal angle (CFCA), component tibial coronal angle (CTCA), and sagittal tibial component angle (STCA) outliers in the proficiency phase was lower than that in the conventional group, while deviations of the HKA angle, CFCA, CTCA, and STCA in the proficiency phase were significantly lower than those in the conventional group (P < 0.05).ConclusionIn summary, from the learning curve data, 20 cases are required for a surgeon using a novel seven-axis RA-TKA system to enter the proficiency phase. In the proficiency group, compared with the conventional group using propensity score matching, the RAS was found to be superior to the conventional group in prosthesis and lower limb alignment.