Objectives. This systematic review evaluates the rehabilitation interventions for patients who have undergone (or will undergo) total knee arthroplasty (TKA) or total hip arthroplasty (THA) for the treatment of osteoarthritis. We addressed four Key Questions (KQs): comparisons of (1) rehabilitation prior (“prehabilitation”) to TKA versus no prehabilitation, (2) comparative effectiveness of different rehabilitation programs after TKA, (3) prehabilitation prior to THA versus no prehabilitation, (4) comparative effectiveness of different rehabilitation programs after THA. Data sources and review methods. We searched Medline®, PsycINFO®, Embase®, the Cochrane Register of Clinical Trials, CINAHL®, Scopus®, and ClinicalTrials.gov from Jan 1, 2005, to May 3, 2021, to identify randomized controlled trials (RCTs) and adequately adjusted nonrandomized comparative studies (NRCSs). We evaluated clinical outcomes selected with input from a range of stakeholders. We assessed the risk of bias and evaluated the strength of evidence (SoE) using standard methods. Meta-analysis was not feasible, and evidence was synthesized and reported descriptively. The PROSPERO protocol registration number is CRD42020199102. Results. We found 78 RCTs and 5 adjusted NRCSs. Risk of bias was moderate to high for most studies. • KQ 1: Compared with no prehabilitation, prehabilitation prior to TKA may increase strength and reduce length of hospital stay (low SoE) but may lead to comparable results in pain, range of motion (ROM), and activities of daily living (ADL) (low SoE). There was no evidence of an increased risk of harms due to prehabilitation (low SoE). • KQ 2: Various rehabilitation interventions after TKA may lead to comparable improvements in pain, ROM, and ADL (low SoE). Rehabilitation in the acute phase (initiated within 2 weeks of surgery) may lead to increased strength (low SoE) but result in similar strength when delivered in the post-acute phase (low SoE). No studies reported evidence of risk of harms due to rehabilitation delivered in the acute period following TKA. Compared with various controls, post-acute rehabilitation may not increase the risk of harms (low SoE). • KQ 3: For all assessed outcomes, there is insufficient (or no) evidence addressing the comparison between prehabilitation and no prehabilitation prior to THA. • KQ 4: Various rehabilitation interventions after THA may lead to comparable improvements in pain, strength, ADL, and quality of life. There is some evidence of no increased risk of harms due to the intervention (low SoE). • There is insufficient evidence regarding which patients may benefit from (p)rehabilitation for all KQs and insufficient evidence regarding comparisons of different providers and different settings of (p)rehabilitation for all KQs. There is insufficient evidence on costs of (p)rehabilitation and no evidence on cost effectiveness for all KQs. Conclusion. Despite the large number of studies found, the evidence regarding various prehabilitation programs and comparisons of rehabilitation programs for TKA and THA is ultimately sparse. This is a result of the diversity of interventions studied and outcomes reported across studies. As a result, the evidence is largely insufficient or of low SoE. New high-quality research is needed, using standardized intervention terminology and core outcome sets, especially to allow network meta-analyses to explore the impact of intervention attributes on patient-reported, performance-based, and healthcare-utilization outcomes.