Abstract Introduction Bone marrow lesions (BMLs) are features detected on MRI that are important in the pathogenesis of knee osteoarthritis. Since BMLs reflect heterogeneous pathologies this prospective cohort study examined whether BMLs detected using different MRI sequences are associated with distinct structural and clinical endpoints. Methods A total of 297 community-based adults without knee pain were examined to identify BMLs visualised using three-dimensional T1-weighted gradient-echo fat-suppressed (T1-weighted sequences) fat-suppressed and fat-saturated FSE T2-weighted MRI sequences (T2-weighted sequences) at baseline. Cartilage volume was measured at baseline and follow-up, while incident knee pain was assessed at follow-up, an average of 2.3 years later. Results At baseline, 46 BMLs were visualised in 39 participants. Of the 45 BMLs visualised on T2-weighted sequences, 34 (74 %) were also seen on T1-weighted sequences. One BML was seen on only T1-weighted sequences. Knees with BMLs visualised on both T1- and T2-weighted sequences had significantly higher medial tibial cartilage volume loss (45 mm 3 /annum, standard error of the mean (SEM) 14) than those with BMLs identified on only T2-weighted sequences (−13 mm 3 /annum SEM 19), after adjustment for age, gender and body mass index ( p  = 0.01). Incident knee pain was more likely in individuals with BMLs in the medial compartment visualised on both T1- and T2-weighted (eight participants, 53 %) compared to those with BMLs on only T2-weighted sequences (0 %) or no BMLs (76 participants, 31 %, p  = 0.02). Conclusions BMLs present on both T1- and T2-weighted MRI sequences were associated with increased medial tibial cartilage loss and incident knee pain compared with those BMLs seen only on T2-weighted sequences. This suggests that combining different MRI sequences may provide more informative targets in the prevention and treatment of knee osteoarthritis.