Background Total knee arthroplasty (TKA) is associated with varying degrees of pain. A considerable proportion (25-40%) of patients experience severe pain, despite a comprehensive multimodal analgesic regimen. We hypothesized that adductor canal block (ACB) would reduce pain in this patient category compared with placebo. Methods Fifty patients with severe pain, defined as having a visual analogue scale (VAS) pain score of >60 during active flexion of the knee on the first or the second postoperative day after TKA, were included in this randomized, double-blind, placebo-controlled trial. All the patients had received a comprehensive multimodal analgesic regimen. Group A received an ACB with ropivacaine 0.75%, 30 ml at time 0 and isotonic saline after 45 min. Group B received an ACB with isotonic saline at time 0 and ropivacaine 0.75%, 30 ml after 45 min. Results A 32-mm difference in VAS pain score, during active flexion of the knee (primary endpoint), was observed in favour of Group A, 95% confidence interval (CI): 23-42, P<0.0001. At rest, the difference in VAS pain score was 15 mm in favour of Group A, 95% CI: 8-23 mm, P=0.0001. Individual patient analysis revealed that 25% of the patients had no effect during active flexion. At rest, however, only 8% had more than mild pain after ACB compared with 57% at inclusion. Conclusions ACB reduced VAS with 32 mm, during active flexion of the knee, in patients with severe pain after TKA, but a large proportion (78%) still had at least moderate, movement-related pain. Clinical trial registration www.clinicaltrials.gov, NCT01549704.