To compare the accuracy rates of intra-articular hyaluronic acid (HA) injections for osteoarthritis (OA) of the knee between the modified Waddell approach (an anteromedial approach with manipulative ankle traction at 30 degrees of knee flexion), an anteromedial approach with the subjects seated (hereinafter the seated anteromedial approach) and a lateral patellar approach based on the Kellgren-Lawrence (K-L) radiographic grade (II, III and IV). Fifty patients with knee OA received HA injections through the three approaches. The accuracy rates were confirmed with a single radiograph after injections of a mixture of radiographic contrast medium. In the K-L grade IV cases (n=11), the accuracy rates through the modified Waddell approach (11 out of 11, 100%) were significantly higher than those through the seated anteromedial approach (six out of 11, 55%) and the lateral patellar approach (six out of 11, 55%) (P=0.035). No significant differences were detected in the accuracy rates of the participants classified as grade II (n=21) or III (n=18) cases between the modified Waddell approach (86% and 78%, respectively), the seated anteromedial approach (71% and 56%, respectively) and the lateral patellar approach (86% and 61%, respectively) (P>0.05). Although previous studies have been conducted on the accuracy of needle placement into the intra-articular space of the knee, no evaluations were performed with the results categorized by radiographic severity. This study highlighted the need for clinicians to change the approach employed for HA injections, according to the severity of knee OA.