Objectives. The study was designed to investigate the analgesic effects and mechanisms of acetaminophen (paracetamol) in symptomatic osteoarthritis (OA) of the knee. Methods. Twenty patients with symptomatic OA were randomly allocated to two groups treated with either acetaminophen or rofecoxib for 3 months. Visits and measurements were scheduled upon entry (T0), at month 1 (T1) and at month 3 (T3). The intensity of joint pain was evaluated with a 100-mm visual analogue scale (VAS). The physical function of the affected knee was evaluated with a questionnaire comparable to the Western Ontario McMaster Universities Osteoarthritis Index (WOMAC). Levels of serotonin, substance P (SP) and β-endorphin (BEND) were determined with commercial enzyme-linked immunoassay kits. The expression of κ opioid receptor (KOR) in peripheral mononuclear blood cells (PBMCs) was quantified by real-time PCR. Results. Both acetaminophen and rofecoxib relieved pain considerably but with different kinetics, and affected different biomarkers. Rofecoxib appeared to be more efficient, reducing pain intensity by 56% at T1 (P<0.01), whereas acetaminophen reduced it by only 29%. Physical function improved in both groups by T3. Correlated with the pain relief, acetaminophen significantly reduced plasma BEND levels, whereas rofecoxib did not do so. In both groups plasma SP levels were elevated compared with T0. A reduction in serum serotonin was detected in the rofecoxib group at T1 ( P=0.004) but had recovered at T3. No changes in KOR mRNA in PBMCs were observed in either group. Conclusions. There is a correlation between reduction in circulating BEND and OA pain relief in patients treated with acetaminophen.