Objective: To investigate whether relationships between upper leg muscle strength and activity limitations are non-linear in patients with knee osteoarthritis, and, if so, to determine muscle strength thresholds for limitations in daily activities. Design: Baseline data were used for 562 patients with knee osteoarthritis in the Amsterdam-Osteoarthritis cohort. Upper leg muscle strength (Nm/kg) was measured isokinetically. Activity limitations were measured with the timed Get Up and Go test and timed Stair Climb Test, subdivided into stairascent and stair-descent. Linear and non-linear relationships between muscle strength and activity limitations were evaluated, and thresholds were determined. Results: Non-linear models improved model fit compared with linear models. The improvement in percentage variance accounted for was 5.9, 8.2 and 5.2 percentage points for the timed Get Up and Go, stair-ascent and stair-descent times, respectively. Muscle strength thresholds were 0.93 Nm/kg (95% confidence interval (95% CI) 0.82-1.04), 0.89 Nm/kg (95% CI 0.77-1.02) and 0.97 Nm/kg (95% CI 0.85-1.11) for relationships with timed Get Up and Go, stair-ascent and stair-descent times, respectively. Conclusion: In patients with knee osteoarthritis, relationships between muscle strength and activity limitations are non-linear. Patients with muscle strength below the described thresholds might benefit more from muscle strength training to reduce limitations in daily activities than would patients with muscle strength above the thresholds. Further research is needed to assess the clinical value of the thresholds determined.