Aims. To determine the trajectories of patient reported pain and functional disability over five years following total hip arthroplasty (THA) or total knee arthroplasty (TKA). Methods. A prospective, longitudinal cohort sub-study within the National Joint Registry (NJR) was undertaken. In all, 20,089 patients who underwent primary THA and 22,489 who underwent primary TKA between 2009 and 2010 were sent Oxford Hip Score (OHS) and Oxford Knee Score (OKS) questionnaires at six months, and one, three, and five years postoperatively. OHS and OKS were disaggregated into pain and function subscales. A k-means clustering procedure assigned each patient to a longitudinal trajectory group for pain and function. Ordinal regression was used to predict trajectory group membership using baseline OHS and OKS score, age, BMI, index of multiple deprivation, sex, ethnicity, geographical location, and American Society of Anesthesiologists grade. Results. Data described two discrete trajectories for pain and function: ‘level 1’ responders (around 70% of cases) in whom a high level of improvement is sustained over five years, and ‘level 2’ responders who had sustained improvement, but at a lower level. Baseline patient variables were only weak predictors of pain trajectory and modest predictors of function trajectory. Those with worse baseline pain and function tended to show a greater likelihood of following a ‘level 2’ trajectory. Six-month patient-reported outcome measures data reliably predicted the class of five-year outcome trajectory for both pain and function. Conclusion. The available preoperative patient variables were not reliable predictors of postoperative pain and function after THA and TKA. Reviewing patient outcomes at six months postoperatively is a reliable indicator of outcome at five years. Cite this article: Bone Joint J 2021;103-B(6):1111–1118