different nonpharmacologic and pharmacologic modalities that were previously identified by separate expert panels (4-7); note that tramadol was considered separately from opioid analgesics because the central analgesic effect of tramadol is thought to be mediated not only by a weak opioid receptor agonist effect but also through modulation of serotonin and norepinephrine levels. Literature searches were not performed for medications that are not commercially available in the US and Canada. The initial searches were conducted in Medline (1950–2009), Embase (1980–2009), and The Cochrane Library (issue 3, 2009) by applying database subject headings and relevant keywords (see Supplementary Appendix A for the search strategy employed for exercise, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658). Published search filters were modified and used to limit the search to specific high-quality study designs (15-17). The initial searches for pharmacologic modalities were conducted during the third and fourth quarters of 2008, while those for the nonpharmacologic modalities were conducted during the second and third quarters of 2009. The goal of the literature search was to identify the most current systematic review(s) and meta-analysis(es) that would provide reliable estimates of benefits of the intervention for the prespecified clinically relevant outcomes of pain and function as well as providing data on safety of the intervention; clinically relevant safety outcomes differed by type of intervention. If no systematic review or meta-analysis was available, the search results were screened for randomized controlled trials (RCTs). If more than 1 systematic review or RCT was identified for the modality and outcomes of interest, the quality of the systematic review or RCT was assessed in order to select the best-quality evidence. Observational studies on safety of interventions were included if there were no RCTs. Data from the Food and Drug Administration (FDA) Adverse Event Reporting System and unpublished data from product manufacturers or investigators were not solicited, as adequate denominator populations are often not available to allow the calculation of numbers needed to harm. Forming the Technical Expert Panel (TEP) and developing clinical scenarios. While the initial literature review was being completed, a TEP was convened (members of the TEP are shown in Appendix A). The TEP included nationally recognized academic and practicing rheumatologists, primary care physicians, physiatrists, geriatricians, orthopedic surgeons, and occupational and physical therapists. The TEP was asked initially to develop a series of clinical scenarios representing patients with hand, hip, or knee OA who presented for management decisions. The scenarios included a base case of a patient with symptomatic hand, hip, or knee OA; clinical variations of the base case scenario depended on specific joint groups involved for hand OA and the lack of a satisfactory response and/or the presence of comorbidities for hip and knee OA. The scenarios were collated and then submitted to the ACR Board of Directors for review. Once the literature review was completed, the TEP was asked to evaluate the evidence and formulate recommendations for OA treatment in the situations outlined in the clinical scenarios. First, the TEP was provided with summaries of the best available evidence for all interventions, including