Thank you for reading our paper. We acknowledge the point raised that we should have consistently described the variables as ‘predictors of outcome’ rather than using the term ‘predictors of treatment response’, although we defined outcome as improvement in pain level and symptoms. In order to conserve limited funding resources, research is necessarily an iterative process in which risk factors are first tested in cross-sectional then cohort studies and finally in large RCTs, which in turn provide increasing order of merit in assessing causation. In our conclusion, we state that our study is a first step towards selecting patients most likely to respond. We were clear that we conducted a cohort study which is a reliable design to measure prognosis and outcomes, and we addressed the limitations in the Discussion. We think our cohort data provide a good strength of evidence, are interesting for clinicians, and will be important to inform more definitive trials of treatment effect modifiers in the future.