As a prelude to developing updated, evidence-based, international consensus recommendations for the management of hip and knee osteoarthritis (OA), the Osteoarthritis Research Society International (OARSI) Treatment Guidelines Committee undertook a critical appraisal of published guidelines and a systematic review (SR) of more recent evidence for relevant therapies. Sixteen experts from four medical disciplines (primary care two, rheumatology 11, orthopaedics one and evidence-based medicine two), two continents and six countries (USA, UK, France, Netherlands, Sweden and Canada) formed the guidelines development team. Three additional experts were invited to take part in the critical appraisal of existing guidelines in languages other than English. MEDLINE, EMBASE, Science Citation Index, CINAHL, AMED, Cochrane Library, seven Guidelines Websites and Google were searched systematically to identify guidelines for the management of hip and/or knee OA. Guidelines which met the inclusion/exclusion criteria were assigned to four groups of four appraisers. The quality of the guidelines was assessed using the AGREE (Appraisal of Guidelines for Research and Evaluation) instrument and standardised percent scores (0-100%) for scope, stakeholder involvement, rigour, clarity, applicability and editorial independence, as well as overall quality, were calculated. Treatment modalities addressed and recommended by the guidelines were summarised. Agreement (%) was estimated and the best level of evidence to support each recommendation was extracted. Evidence for each treatment modality was updated from the date of the last SR in January 2002 to January 2006. The quality of evidence was evaluated using the Oxman and Guyatt, and Jadad scales for SRs and randomised controlled trials (RCTs), respectively. Where possible, effect size (ES), number needed to treat, relative risk (RR) or odds ratio and cost per quality-adjusted life year gained (QALY) were estimated. Twenty-three of 1462 guidelines or consensus statements retrieved from the literature search met the inclusion/exclusion criteria. Six were predominantly based on expert opinion, five were primarily evidence based and 12 were based on both. Overall quality scores were 28%, 41% and 51% for opinion-based, evidence-based and hybrid guidelines, respectively (P=0.001). Scores for aspects of quality varied from 18% for applicability to 67% for scope. Thirteen guidelines had been developed for specific care settings including five for primary care (e.g., Prodigy Guidance), three for rheumatology (e.g., European League against Rheumatism recommendations), three for physiotherapy (e.g., Dutch clinical practice guidelines for physical therapy) and two for orthopaedics (e.g., National Institutes of Health consensus guidelines), whereas 10 did not specify the target users (e.g., Ontario guidelines for optimal therapy). Whilst 14 guidelines did not separate hip and knee, eight were specific for knee but only one for hip. Fifty-one different treatment modalities were addressed by these guidelines, but only 20 were universally recommended. Evidence to support these modalities ranged from Ia (meta-analysis/SR of RCTs) to IV (expert opinion). The efficacy of some modalities of therapy was confirmed by the results of RCTs published between January 2002 and 2006. These included exercise (strengthening ES 0.32, 95% confidence interval (CI) 0.23, 0.42, aerobic ES 0.52, 95% CI 0.34, 0.70 and water-based ES 0.25, 95% CI 0.02, 0.47) and nonsteroidal anti-inflammatory drugs (NSAIDs) (ES 0.32, 95% CI 0.24, 0.39). Examples of other treatment modalities where recent trials failed to confirm efficacy included ultrasound (ES 0.06, 95% CI -0.39, 0.52), massage (ES 0.10, 95% CI -0.23, 0.43) and heat/ice therapy (ES 0.69, 95% CI -0.07, 1.45). The updated evidence on adverse effects also varied from treatment to treatment. For example, while the evidence for gastrointestinal (GI) toxicity of non-selective NSAIDs (RR=5.36, 95% CI 1.79, 16.10) and for increased risk of myocardial infarction associated with rofecoxib (RR=2.24, 95% CI 1.24, 4.02) were reinforced, evidence for other potential drug related adverse events such as GI toxicity with acetaminophen or myocardial infarction with celecoxib remained inconclusive. Twenty-three guidelines have been developed for the treatment of hip and/or knee OA, based on opinion alone, research evidence or both. Twenty of 51 modalities of therapy are universally recommended by these guidelines. Although this suggests that a core set of recommendations for treatment exists, critical appraisal shows that the overall quality of existing guidelines is sub-optimal, and consensus recommendations are not always supported by the best available evidence. Guidelines of optimal quality are most likely to be achieved by combining research evidence with expert consensus and by paying due attention to issues such as editorial independence, stakeholder involvement and applicability. This review of existing guidelines provides support for the development of new guidelines cognisant of the limitations in existing guidelines. Recommendations should be revised regularly following SR of new research evidence as this becomes available.