A bias-adjusted evidence synthesis of RCT and observational data: the case of total hip replacement