Patients with knee prosthetic joint infection (PJI) frequently receive one- or two-stage revision. To explore the feasibility of a randomised controlled trial (RCT) comparing methods, we analysed a UK registry, interviewed patients and surgeons, systematically reviewed literature, held a consensus meeting, and assessed progress of an RCT in hip PJI. In 2014, in England and Wales, knee PJI was treated with one- or two-stage procedures in 19% and 71% of patients respectively. Between 2007 and 2014, use of one-stage procedures doubled and, in major centres, up to 42% of treatments were one-stage. We conducted in-depth interviews with 16 patients with knee PJI and 11 surgeons performing one- or two-stage revision. Patients considered randomisation acceptable with appropriate counselling and, depending on infecting organisms and health status, surgeons would randomise treatments. In meta-analysis, two-year re-infection rates in 10 one-stage series (423 patients) and 108 two-stage series (5,129 patients) were 7.6% (95%CI 3.4,13.1) and 8.8% (7.2,10.6) respectively. In a series of patients with knee PJI, surgeons from 2 major centres considered 6/15 patients eligible for either treatment, with 4 more potentially eligible after treatment of soft tissue infection. In an ongoing RCT of surgical treatment of hip PJI, 116 patients have been randomised at 14 centres in 3 years. Randomising patients with PJI is feasible but, as knee PJI is uncommon, a multicentre RCT would be required. Based on WOMAC score outcome and appropriate assumptions on eligibility and acceptability, 170 patients would need to be randomised over 4 years at 14 major centres.