Abstract The study aim was to address the need for objective markers of pain-modifying interventions by testing the hypothesis that selective gait measures of knee joint loading can distinguish differences between non-steroidal anti-inflammatory (NSAID), analgesic treatment (opioid-receptor agonist), and placebo in patients medial knee osteoarthritis (OA). A randomized, single-blind washout, double-blind treatment, double-dummy cross-over trial using three treatment arms placebo, opioid (Oxycodone), and NSAID (Celecoxib) in medial compartment knee OA patients. Six patients with Kellgren–Lawrence radiographic severity grades of 2 or 3 completed six testing sessions (gait and pain assessment) at 2-week intervals. A significant increase was found in the knee total reaction moment and vertical ground reaction force (GRF) for Celecoxib compared to placebo (p = 0.005, p = 0.003), but not for Oxycodone compared to placebo (p = 0.20, p = 0.27) treatments. Walking speed was significantly higher for the Celecoxib and Oxycodone compared to placebo treatment (p = 0.041 and p = 0.031, respectively). Self-reported function (WOMAC scores) was not different among treatments (p > 0.05). The changes in total reaction moments and GRFs for only the NSAID suggest that greater increases in joint loading occurs when joint inflammation is treated in addition to pain. The total knee reaction moment, representing the magnitude of the extrinsic moment, appears to be a sensitive marker, more so than self-reported metrics, for evaluating knee OA treatment effects.