Objectives: Is a magnetic resonance imaging (MRI) definition for tibiofemoral osteoarthritis [(TFOA(MRI)) (definite osteophyte and full-thickness cartilage loss (or a combination of these factors with other MRI osteoarthritis (OA) features)] more sensitive to detect structural OA compared with the Kellgren & Lawrence (K&L) grading? And which definition shows the strongest association with (1) knee pain at baseline, (2) persistent knee pain during 2-year follow-up, (3) new onset of knee pain +/- 2 years later, and (4) body mass index (BMI). Design: Of 888 females of the open population Rotterdam Study, radiographs and MRI of both knees were assessed for knee OA defined by K&L >= 2 and TFOA(MRI). Pain in or around the knee is measured at baseline and 2 years later. GEE analyses are used for the associations. Results: Of 1766 knees, 77 knees (4%) were diagnosed with K&L >= 2, whereas 160 knees (9%) met the TFOA(MRI) criteria. Only 43 knees met both definitions (34 knees were graded with K&L >= 2 and no TFOA(MRI) and 117 knees met only the TFOA(MRI) criteria). The association between the definitions and knee pain at baseline was higher when TFOA(MRI) was included [TFOA(MRI) alone: odds ratio (OR) = 2.83 (95% confidence interval (CI): 1.84-436); TFOA(MRI) & K&L >= 2: OR = 6.28 (95% CI: 2.99-13.19)] than for K&L >= 2 alone (OR = 1.83 (95% CI: 0.63-5.32)). This was similar for the association between the definitions and persistent knee pain, and between the definitions and BMI. Conclusions: TFOA(MRI), detects more cases of knee OA than K&L >= 2. Together with a better content validity and at least equal construct validity, we conclude that the TFOA(MRI) definition for knee OA is more sensitive in detecting structural knee OA. (C) 2014 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.