Outcomes and Data Extraction

Trials were included (Figure 1) if they reported data for at least 1 of the following outcomes: knee pain, physical function, or joint structure defined as radiological joint space narrowing (JSN).¹⁰ The primary outcome was mean change from baseline to the end point (≥12 months) in knee pain because pain is likely to be the outcome that matters most to patients, physicians, and caregivers. Secondary outcomes were changes in physical function and joint structure.

When pain or physical function outcomes were measured using different scales in the RCTs, the outcomes were prioritized as recommended by Juhl et al¹¹ (additional details appear in eTable 2 in Supplement 2). Using this method of prioritization, the preferred pain outcome measure was the pain subscale¹² of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC); followed by the visual analog scale (VAS) for pain during any activity, pain during walking, or a global measure of knee pain; and then any other pain measures.

For physical function, the preferred outcome measure was the WOMAC physical function subscale (the other types of physical function measures and a full hierarchy for both the pain and physical function outcomes appears in eTable 2 in Supplement 2).¹¹ The WOMAC is a disease-specific questionnaire separately addressing the severity of pain (5 questions) and any limitation in physical function (17 questions) for the activities of daily living during the past 48 hours.¹² In the Likert scale version, each answer is scored on a scale from 0 to 4 and 0 represents “none” and 4 represents “extreme” (score range: 0-20 for pain and 0-68 for physical function). In the VAS version, each answer is scored on a 100-mm VAS and 0 mm represents “none” and 100 mm represents “extreme” (score range: 0-500 mm for pain and 0-1700 mm for physical function). When pain severity is assessed as the answer to a single question (eg, “how much is your pain during walking today?”) on a 100-mm VAS, 0 mm represents “absent” and 100 mm represents “the worst imaginable” (score range: 0-100 mm). Data on JSN were derived from measuring change in the radiological joint space width in millimeters at the medial tibiofemoral joint.

For each outcome, the change from baseline was extracted at the longest available time point after at least 12 months, if reported; otherwise, numerical data for the outcome were extracted at baseline and at the longest available time point after at least 12 months, and the change from baseline was calculated. For graphical information, numerical data were extracted using a standard procedure.¹³

Other extracted data included characteristics of the study design to assess trial quality, baseline demographic characteristics (age, sex, and body mass index), administration route and dose of each treatment, and duration of treatment and follow-up. Only trials with extractable data were included. No additional information was requested from authors.