Objective: Kellgren-Lawrence grades (KLG) are frequently used for patient selection in clinical trials. The Ahlbäck radiographic grading system has been developed for moderate and severe knee OA. KLG 3 is comparable to Ahlbäck 1 and KLG 4 is subdivided into Ahlbäck 2–5. The objective of this study was to investigate if the Ahlbäck scoring system is able to subdivide patients with moderate to severe knee OA (KLG 3/4) into groups with different sensitivity to change in cartilage thickness. Materials and methods: This study was based on 108 Osteoarthritis Initiative (OAI) participants with KLG 3/4. Baseline KLG scores were available from the OAI database; Ahlbäck scores were performed using the same x-rays. Cartilage thickness change in the weight-bearing femorotibial cartilage was analysed from baseline and year 1 3D FLASH MRI for the entire femorotibial joint (FTJ), the medial (MFTC) and the lateral compartment (LFTC) and for the location-independent ordered values 1 and 16 (OV 1/OV 16) representing the subregions with largest loss (OV 1) and gain (OV 16) within each knee. Results: Of the 108 patients, n = 30/78 had KLG 3/4. The corresponding Ahlbäck scores (1–5) were n = 30/33/36/9/10. Cartilage thickness changes between Ahlbäck groups showed no statistically significant difference for FTJ, MFTC, LFTC and OV 1, but change in OV 16 was significantly higher in Ahlbäck 4 knees (p = 0.03) compared to Ahlbäck 1–3 knees. Conclusion: Radiographic knee OA grading with Ahlbäck scores was not superior to KLG for prediction of cartilage thickness loss over 1 year, in patients with moderate and severe knee OA supporting the continuous use of the easier and more widely used KLG.