BACKGROUND Exercise is widely accepted as a first-line treatment for chronic non-specific low back pain (CNSLBP). However, the benefits of exercise diminish over time, as does adherence to exercise. It is unclear whether greater exercise adherence is associated with improvements in pain intensity (PI) and functional limitation (FL). We explored the relationship between exercise adherence and patient-reported outcomes in people with CNSLBP. METHODS We conducted a secondary analysis of the Cochrane systematic review, ‘Exercise therapy for chronic low back pain’, using a subset of 24 trials that measured exercise adherence compared to usual care. Random-effects meta-analysis was performed in R for PI and FL at the closest time point post-intervention. We used predefined subgroups of exercise adherence of ‘Good’ (90-100%), ‘Moderate’ (70-89%), or ‘Poor’ (14-69%) adherence. We used the risk of bias judgements provided by Cochrane. RESULTS All trials included were deemed low risk of bias. Compared to usual care, ‘Good’ adherence was associated with reduced PI by 17.83 points on a 100-point scale (95% CI -26.23 to -9.43; I2 = 81.7%) and FL by 9.69 points on a 100-point scale (95% CI -12.64 to -6.74; I2 = 18.9%). ‘Moderate’ adherence was associated with reduced PI by 6.93 points (95% CI -10.43 to -3.44; I2 = 18.3%) and FL by 3.80 points (95% CI -6.10 to -1.49; I2 = 0%). ‘Low’ adherence was associated with reduced PI by 7.50 points (95% CI -19.83 to -4.84; I2 = 89.7%) and FL by 3.35 points (95% CI -10.45 to -3.74; I2 = 82.7%). CONCLUSIONS Greater adherence to exercise is associated with greater improvements in PI and FL in adults with CNSLBP. Further research is needed to understand the causal effect of adherence on patient-reported outcomes. Better reporting of this potentially important exercise parameter in randomised trials is also needed.