Objective. To describe associations between serum high-sensitivity C-reactive protein (hsCRP), knee bone marrow lesions (BMLs), and knee pain, cross-sectionally and longitudinally, in patients with knee osteoarthritis (OA). Methods. Patients (n 5 192) with symptomatic knee OA (mean age 63 years, range 50-79, women 53%) were assessed at baseline and after 24 months. Serum hsCRP was measured using enzyme-linked immunosorbent assay. Knee BMLs were scored using the modified Whole-Organ Magnetic Resonance Imaging (MRI) Score from T2-weighted fat-supressed fast spin-echo MRI. Knee pain was assessed using the Western Ontario and McMaster Universities Osteoarthritis Index. Results. Quartiles of baseline serum hsCRP were associated with the presence of knee BMLs (prevalence ratio 1.07 per quartile [95% confidence interval (95% CI) 1.00, 1.15]) and total knee pain scores (b 13.66 per quartile [95% CI 2.26, 25.07]) in multivariable analyses. Longitudinally, higher baseline hsCRP was associated with an increase in BML score (risk ratio 1.37 per quartile [95% CI 1.10, 1.70]), and change in hsCRP was positively associated with change in BML score (b 0.19 [95% CI 0.05, 0.34]) in adjusted analyses. Baseline hsCRP was not associated with change in total knee pain, but change in hsCRP was positively and significantly associated with change in total knee pain (b 4.71 [95% CI 0.48, 8.94]). This became nonsignificant after adjustment for changes in BML score. Conclusion. In patients with knee OA, serum hsCRP is associated with knee BML scores and, to a lesser extent, pain both cross-sectionally and longitudinally, suggesting that inflammation is linked with BMLs and their associated pain.