Background: We investigated the effects of spinal and general anaesthesia and surgical tourniquet on acute pain and early recovery after total knee arthroplasty (TKA). Methods: Patients (n¼413) were randomised to four parallel groups: spinal anaesthesia with or without tourniquet, and general anaesthesia with or without tourniquet. The primary outcome was patient-controlled i.v. oxycodone consumption over 24 postoperative hours. Results: Results from 395 subjects were analysed. Median i.v. oxycodone consumption did not differ between the four groups (spinal anaesthesia without [36.6 mg] and with tourniquet [38.0 mg], general anaesthesia without [42.3 mg] and with tourniquet [42.5 mg], P¼0.42), between spinal (37.7 mg) and general anaesthesia (42.5 mg) groups (median difference e3.1, 95% confidence interval [CI] e7.4 to 1.2, P¼0.15) and between tourniquet and no-tourniquet groups (40.0 vs 40.0 mg, median difference e0.8, CI e5.1 to 3.5, P¼0.72). Vomiting incidence was higher with spinal than with general anaesthesia (21% [42/200] vs 13% [25/194], CI 1.05 to 3.1, P¼0.034). The mean haemoglobin decrease was greater without than with tourniquet (e3.0 vs e2.5 g dl À1 , mean difference e0.48, CI e0.65 to e0.32, P<0.001). No differences were observed in pain, pain management, incidences of blood transfusions, in-hospital complications, or length of hospital stay. Conclusions: For TKA, spinal and general anaesthesia with or without tourniquet did not differ in 24-h postoperative opioid consumption, pain management, blood transfusions, in-hospital complications, and length of hospital stay. Vomiting incidence was higher in the spinal than in the general anaesthesia group. Tourniquet use caused smaller decreases in haemoglobin levels. Clinical trial registration: EudraCT 2016-002035-15.