To determine the extent to which treatment of patients with symptomatic knee osteoarthritis (OA) with non-steroidal anti-inflammatory drugs (NSAIDs) and acetaminophen (ACET) reduces total effusion volume and synovial tissue volume, as quantified by magnetic resonance imaging (MRI). Sequential pilot studies used subjects whose knee OA was treated with NSAIDs (n=10) or with ACET <or=4 g/day (n=20), respectively. After a five half-lives washout of their pain medication, the OA knee with the higher pain score >or=15 of 25 on the Western Ontario and McMaster Universities' pain scale underwent l.5T MRI. Effusion was quantified in axial short tau inversion recovery images; to measure synovial tissue volume, fat-suppressed T1-weighted axial images were obtained 3 min after i.v. injection of gadolinium contrast. After the initial MRI examination, patients resumed their customary pain medications until the severity of knee pain returned to baseline, when pain was again measured and the MRI was repeated. Pain severity after washout was similar in subjects taking ACET and NSAIDs. Reinstitution of ACET resulted in a 50% decrease in the mean of pain scores (P=1.7 x 10(-12)) that was comparable with that seen after the reinstitution of NSAID (49%, P=6.0 x 10(-7)). The mean total effusion volume measured during the flare of knee pain induced by the withdrawal of the two drugs was comparable (ACET 16.9 ml, NSAID 16.2 ml; P=0.884). Significant decreases in mean total effusion volume were observed after reinstitution of both ACET (-4.5 ml, P=0.009) and NSAID (-3.3 ml, P=0.013); the difference between drugs was not significant. Analyses of synovial volume yielded similar results. While uncontrolled and derived from small samples, these data suggest that ACET may have a significant anti-inflammatory effect in patients with knee OA, comparable with that achieved with NSAIDs, possibly through an effect on neurogenic inflammation. Joint pain is the clinical feature of OA that most often leads the affected individual to seek medical attention. Because many patients with OA improve symptomatically with the use of NSAIDs, it has been widely assumed that the pain of OA is due to synovial inflammation. However, the origins of OA pain are numerous and may vary from patient to patient and, within the same subject, from visit to visit. Although the articular cartilage is usually the site of the most obvious pathological changes in this disease, it is aneural and, therefore, is not the source of joint pain. However, in addition to the synovium, the subchondral bone, joint capsule, osteophytes, menisci, ligaments, periarticular tendons, entheses and bursae all contain nociceptive nerve endings, stimulation of which by chemical or physical mediators may be a basis for OA pain.