Department of Error

who were having surgery for their hip hemiarthroplasty with the opportunity to participate (ie, to deliberately include a sample representative of the hip fracture population as a whole). 5 We agree that patients with, for example, diabetes are at higher risk of surgical site infection.However, any such pre injury predisposition would have been balanced by the random allocation of participants to the two treatment groups.This balance is shown in the baseline characteristics table, which, for example, shows that 16•2% of participants in the single antibiotic group had diabetes and 16•7% of participants in the highdose, dual antibiotic group had diabetes.Regarding the subgroup analyses proposed by Chu and colleagues, given the much smaller numbers of participants in these subgroups, any such analyses would have been underpowered to detect differences in the rate of deep surgical site infection.Therefore, we did not include such analyses in our prespecified statistical analysis plan for the WHiTE 8 trial.We would, however, agree with the authors that further investigation of the interaction between baseline patient characteristics and deep surgical site infection is warranted and we have plans to explore these data further in future work.