OBJECTIVES: A Telerehabilitation service following Total Knee Replacement (TKR) has been shown to be not inferior to standard rehabilitation (SR) in recovering the active knee flexion range of motion (ROM) in patients who had not complication related to TKR surgery. However, little is known about its cost-effectiveness. The objective of this economic evaluation was to assess cost-effectiveness of telerehabilitation versus SR. METHODS: A Markov model was employed to simulate the natural progression of TKR, assuming telerehabilitation does not influence this. The revision risk was calculated from patient-level data (multicentre KAT trial) employing a parametric model including established prognostic factors. The others state-transition probabilities and treatments effects were obtained from published literature. Rehabilitation and transportation costs were assessed adopting Ita-NHS perspective employing Italian tariffs. The patients without complications followed the TR-SR programme receiving half of the sessions in SR and half in telerehabilitation regime. Patients with any complication followed the SR receiving face-to-face sessions. Results were adjusted applying an annual discount rate of 3% and half-cycle correction. Probabilistic sensitivity analysis (PSA) described the parameters uncertainty.  RESULTS: A cohort of 1000 patients with the features of KAT trial population (70 years old, 44% male and 19% had any complication related to the surgery) were first assigned to SR then to TR -SR. The mean (SE) lifetime healthcare cost in SR arm were €1,033.81 (€21.97) and €818.44 (€13.82) for telerehabilitation (mean±95%CI difference -215.33±€10.64). Mean (SE) ROM was 16.68 (0.04) in SR programme and 18.84 (0.04) in telerehabilitation programme (mean±95%CI difference 2.16±0.09). The probability that telerehabilitation is cost-saving (WTP:€0) is 92%. CONCLUSIONS: Adopting the healthcare provider perspective, TR-SR seems cost-effective when compared to a SR programme. Further sensitivity analyses are required to relax the model assumptions and to assess the robustness of the model.