Clinical guidelines recommend non-surgical interventions for knee osteoarthritis (OA) management. There is such a huge body of research on interventions for knee OA that clinical guidelines are used to synthesise the evidence into an accessible and translatable format. At the intervention level, synthesis is also needed, including the evidence for non-surgical interventions, such as gait modifications, that aim to reduce knee joint load indicators since they are associated with OA progression. In addition, evaluating the clinical importance of treatment effects is crucial. Yet, the terminology around this concept makes this opaque in interpreting trial results. The overall aims of this thesis are to: 1. Explore the effectiveness of current non-surgical interventions for people with knee OA, in particular physiotherapy and gait modifications 2. Provide estimates of minimal clinically important change (MCIC) and difference (MCID) of knee OA outcome measures after non-surgical interventions in order to better plan and interpret research. First, a narrative review of effects of physiotherapy interventions was conducted. This review found that land-based exercise could relieve pain and improve physical function. Next, to bring up-to-date research on the efficacy of neuromuscular gait modifications in people with medial knee OA, a systematic review and meta-analysis (using quality-adjusted meta-analytic models) was conducted. We found that ipsilateral trunk lean, toe-out, and toe-in reduced knee joint load indicators. The descriptive synthesis found that medial knee thrust, medial weight transfer and wider-step gait modifications are also effective. These reviews also exposed the lack of reporting of clinically important change values in research studies. Second, we provided MCIC and MCID values using a three-step approach. First, we established known MCIC and MCID values. Second, we used these established MCIC values to infer the MCIC of knee flexion using a meta-analytical approach based on the anchor method concept. Third, we obtained consensus from consumers, clinicians and researchers to determine the MCIC and MCID of knee flexion in a qualitative approach. A systematic review synthesised MCIC, MCID and MDC (minimum detectable change) of knee OA outcome measures (for example, pain measured by the Visual Analog Scale (VAS) and function by the Western Ontario and McMaster Universities Arthritis Index (WOMAC)) after non-surgical interventions from studies of any calculation method: 12 anchor studies, one consensus study and 35 distribution method studies (estimated MDC). These estimates were specific to the measurement tool and calculation method (MCIC values for 13 outcome tools, MCID for 23 tools and MDC for 126 tools were synthesised). Two subsequent studies of this thesis estimated MCIC and MCID of knee flexion. The meta-analytical approach study estimated MCIC of knee flexion after establishing linear relationships between change in (Δ) flexion (mean change between baseline and immediately after the intervention) with Δ pain and Δ function, by conducting a systematic review of randomised clinical trials using Bayesian meta-analytic models. The point estimates of MCIC of knee flexion ranged from 4° to 7°. The consensus method study using nominal group technique discussions sought consensus for MCIC and MCID of knee flexion. The consumer groups estimated MCIC of knee flexion, mode was 5° (group 1: 50% consensus and group 2: 75% consensus); the clinician groups, mode was 10o (≥ 80% agreement by two groups); the researcher group, mode was 10° (55% consensus). The researchers group reached ≥ 80% agreement for MCID as 10°. The thematic analysis of the discussions revealed that patient factors (baseline flexion, flexion-restricted duration, and physical activity level) and measurement factors (instrument error and study protocol) potentially influence MCIC and MCID estimates. In conclusion, this thesis has evaluated the evidence for knee OA outcomes after non-surgical interventions and provided a useful contribution to research methodology in knee OA by providing MCIC and MCID values. The systematic review revealed that neuromuscular gait modifications reduce knee joint load indicators in the laboratory. The synthesised MCIC and MCID of knee OA outcome measures will be useful for evaluating the treatment efficacy and sample size. MDC values are useful for getting some idea about measurement errors. Findings suggested that patient and measurement factors should be considered when applying MCIC and MCID in clinical and research contexts. We recommend that clinicians and researchers use the estimated MCIC and MCID values to interpret research outcomes and design clinical trials.