To investigate the impact of repair vs replacement of failed restorations on patient related outcome measures, and to explore the clinical factors that influence this decision. Multicentre, prospective practice-based study.Setting Dental practices within Salford, Trafford and East Lancashire in the North West of England. General dental practitioners were asked to participate and to recruit adult patients attending for routine dental treatment.Interventions Repair or replacement of failed restorations. Dental anxiety before treatment using the Corah Dental Anxiety Scale and pain intensity immediately and 24 hours post-operatively using the McGill short form pain questionnaire. Operative outcomes included depth of caries, time taken to complete the procedure, use of local anaesthetic and dental material used. Of the 103 patients diagnosed with a failed restoration, a statistically significantly greater number underwent replacement than repair (p = 0.004). Patients undergoing repairs were significantly less anxious (p = 0.008) and had shorter procedure times (p = 0.044). Repairs were associated with minimal caries depth and less use of local anaesthetic. Failed restorations should be repaired where clinically possible, as they are quick and associated with less patient anxiety. Future research should focus on providing high quality prospective data evaluating the longevity of repaired vs replaced restorations.