The objective indicators which reflect the past results of end-stage knee osteoarthritis (OA) patients who have already received total knee arthroplasty (TKA) could be helpful for physicians to discuss with patients who are considering TKA. The aim of this prospective cohort study was to examine whether we could predict the knee OA patients who would receive TKA in advance based on baseline data, and to set cut-off points for receiving TKA. The two-hundred and forty end-stage medial-type knee OA patients were enrolled and followed up for 6 months while performing therapeutic exercises. Radiographic findings, visual analog scale for pain and a patient-oriented outcome measure, the Japanese Knee Osteoarthritis Measure (JKOM), were recorded at baseline. Relative risks (RRs) using the area under the curve (AUC) for a receiver operating characteristic (ROC) curve were calculated to evaluate several scores for receiving TKA. While 119 patients (55.3 %) did not undergo TKA, the remaining 96 patients (44.7 %) underwent TKA during this period. The AUCs of the ROC curve for the JKOM total score [0.71 (95 % CI 0.64-0.79)] were higher than those for radiographic parameters. Among the JKOM subcategories, JKOM category III, which indicates the condition in daily life, showed the highest AUC of 0.72 (0.65-0.80). The JKOM total score (65/100) and JKOM category III score (17/40) showed RRs of 2.20 (1.33-3.63) and 1.95 (1.18-3.22) for receiving TKA, respectively. The presence of disability in daily living was found to be an important factor determining whether the patient should undergo TKA.