Background Although the severity of knee osteoarthritis (OA) usually is assessed using different measures of joint structure, function, and pain, the relationships between these measures are unclear. Purpose Therefore, we: (1) examined the relationships between the measures of knee structure (flexion-extension range of motion, radiographic tibiofemoral angle, and medial joint space), function (Knee Osteoarthritis Outcome Scores [KOOS], peak adduction angle, and moment), and pain (visual analog scale [VAS]); and (2) identified variables that best predicted knee pain. Methods We assessed 15 patients with medial knee OA using VAS pain, KOOS questionnaire, 3-D gait analysis, and radiographic examination. Parameter relationships were assessed using Pearson correlation, and variables most predictive of knee pain were determined using a stepwise multiple regression. Results Subjective measurements correlated (|r| C 0.54) with one another, as did most of the objective measurements (|r| C 0.56) except for adduction moment which did not correlate with any variable. All variables correlated (|r| [ 0.54) with VAS knee pain except peak adduction moment. Medial joint space and peak adduction angle best predicted knee pain, accounting for approximately threequarters of the model variance (r 2 = 0.73).