Three hundred and thirty-four patients with knee osteoarthritis, at risk of poor outcomes (<= 26 on the Oxford Knee Score) and scheduled for a total knee arthroplasty, were randomized to receive 6 weeks of targeted physiotherapist-led rehabilitation or 6 weeks of home exercise. The primary outcome of interest was the Oxford Knee Score at 52 weeks. Secondary outcomes of interest included Oxford Knee scores at 14 and 26 weeks, Timed Up and Go (TUG) test scores at 14 weeks, and worst pain and average pain on a Visual Analog Scale (VAS). Patient satisfaction was also assessed 1 year post-treatment. No statistically significant differences in Oxford Knee Scores were observed between the two groups at 52 weeks post-treatment (p=0.07). Oxford Knee Scores at 14 weeks (p=0.04), but not 26 weeks (p=0.07) were statistically significantly in favour of the physiotherapist-led group. No statistically significant differences in worst pain or average pain scores on a VAS were observed between the two groups at all time-points (p>0.05 for all). Whilst patients were equally satisfied with the operated knee (p=0.75), patients in the physiotherapist-led group were statistically significantly more satisfied with pain relief (p=0.02), the ability to perform regular activities (p=0.02), and the ability to perform heavy work or sporting activities (p=0.04).