The weekly mean 24-hour average pain score was calculated from the participant's daily 24-hour average pain ratings using an 11-point numeric rating scale, with scores from 0 (indicating "no pain") to 10 (indicating "the worst possible pain"). The Least Squares Mean estimates were adjusted for baseline, treatment, investigator (pooled), week, treatment*week, and baseline*week.

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline weekly mean 24-hour average pain value were included in the analysis.

A scale that measures the participant's perception of improvement at the time of assessment compared with the start of treatment. The score ranges from 1 (very much better) to 7 (very much worse). The Least Squares Mean estimates were adjusted for baseline value of Patient Global Impression of Severity (PGI-S), treatment, investigator (pooled), visit, and treatment*visit. The PGI-S measures participant's perception of severity of illness at the time of assessment. Scores range from 1 (normal, not at all ill) to 7 (extremely ill).

Modified intent to treat (mITT) population: ITT participants with a baseline PGI-S rating and at least 1 post-baseline PGI-I rating were included in the analysis.

Self-administered questionnaire captures elements of pain, stiffness, and physical disability in participants with osteoarthritis of the knee and/or hip. Index has 24 questions (5 on pain, 2 on stiffness, 17 on physical function). Each question uses a 5-point numeric rating scale ranging from 0 (none) to 4 (extreme). Pain scores range: 0 to 20. Stiffness scores range: 0 to 8. Physical function scores range: 0 to 68. Higher scores=greater impairment. Least Squares Mean estimates were adjusted for baseline value, treatment, investigator (pooled), visit, treatment*visit, and baseline value*visit.

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline WOMAC value were included in the analysis.

Weekly mean 24-hour night pain and worst pain values are calculated from the participant's daily assessments of pain at night and worst pain during the previous 24 hours on an 11-point numeric rating scale, with scores from 0 (indicating "no pain") to 10 (indicating "the worst possible pain"). The Least Squares Mean estimates were adjusted for baseline value, treatment, investigator (pooled), week, treatment*week, and baseline*week.

Modified intent to treat (mITT) population: ITT participants with a baseline night/worst pain value and at least 1 post-baseline weekly mean 24-hour night/worst pain value were included in the analysis.

Measures pain severity and pain interference with function. Severity scores: 0 (no pain) to 10 (severe pain) on each question. Interference scores: 0 (does not interfere) to 10 (completely interferes) on each question assessing interference of pain in past 24 hours for general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life. Mean interference is the average across the 7 interference items. The Least Squares Mean estimates were adjusted for baseline value, treatment, investigator (pooled), visit, treatment*visit, and baseline*visit.

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline BPI-S/BPI-I value were included in the analysis.

The CGI-S scale evaluates the severity of illness at the time of assessment. The scores range from 1 (normal, not at all ill) to 7 (among the most extremely ill participants). The CGI-S must be administered by a study physician in the presence of the participant or after having been in the presence of the participant. The Least Squares Mean estimates were adjusted for baseline, treatment, investigator (pooled), visit, treatment*visit, and baseline*visit.

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline CGI-S value were included in the analysis.

The PGAI is a participant-rated measure of the severity of osteoarthritis (OA) of the knee the participant has experienced in the past week as indicated on an 11-point numeric rating scale, with scores ranging from 0 to 10, where greater numbers reflect greater severity. The Least Squares Mean estimates were adjusted for baseline value, treatment, investigator (pooled), visit, treatment*visit, and baseline*visit.

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline PGAI value were included in the analysis.

30-item BPOMS measures positive and negative aspects of mood states (item score: 0=not at all to 4=extremely). 5 negative factors: tension-anxiety, depression-dejection, anger-hostility, fatigue-inertia, confusion-bewilderment; 1 positive factor: vigor-activity. Factor scores range: 0 to 20; high scores=negative mood (positive mood for vigor). Total score=sum of 5 negative factor scores minus vigor score; range: -20=least disturbed to 100=most disturbed. Least Squares Mean estimates adjusted for baseline value, treatment, investigator (pooled), visit, treatment*visit, and baseline value*visit.

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline BPOMS value were included in the analysis.

The Least Squares (LS) Mean percentage estimates of participants using acetaminophen was determined during each week individually over the full 10-week treatment period based on participant's daily Yes/No assessments for the use of acetaminophen. The LS Mean estimates for the main effect of treatment (average weekly use) were adjusted for baseline value, treatment, investigator (pooled), week, and treatment*week.

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline weekly acetaminophen use value were included in the analysis.

OARSI response is composite Yes/No response assessed at 8 weeks based on decrease in 24-hour average pain ratings, range: 0 ("no pain") to 10 ("worst possible pain"), improvement in functioning (using WOMAC physical function scores, range: 0 [no difficulty] to 68 [extreme difficulty]), and improvement in participant's impression of illness (using PGAI scores, range: 0 to 10; 10=greatest severity). OARSI responder=large response in pain or function components (50% relative and 20% absolute improvement), or moderate response (20% relative and 10% absolute improvement) in 2 of 3 components.

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline OARSI response value, last-observation-carried forward (LOCF) based on values of each of the 3 components listed in the outcome measure description were included in the analysis.

Response is a dichotomous outcome (Yes/No) indicating at least 30% (or 50%) reduction from baseline to endpoint for the weekly mean of the 24-hour average pain ratings. The weekly mean 24-hour average pain score was calculated from the participant's daily 24-hour average pain rating assessed on an 11-point numeric rating scale, with scores from 0 ("no pain") to 10 ("worst possible pain").

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline weekly mean of the 24-hour average pain score value, last-observation-carried forward (LOCF) were included in the analysis.

Response is a dichotomous outcome (Yes/No) indicating at least 30% (or 50%) reduction from baseline to endpoint for BPI-S average pain rating. The BPI-S self-reported scale that measures the severity of pain based on the average pain experienced over the past 24 hours. The severity scores range from 0 (no pain) to 10 (pain as severe as you can imagine).

Modified intent to treat (mITT) population: ITT participants with a baseline and at least 1 post-baseline BPI-S average pain value, last-observation-carried forward (LOCF) were included in the analysis.

Abnormal high HbA1c is defined as a post-baseline HbA1c > 6.1% if baseline HbA1c ≤ 6.1% for lab samples obtained before November 17, 2010 and post-baseline HbA1c > 6.4% if baseline HbA1c ≤ 6.4% for lab samples obtained November 17, 2010 and beyond.

Number of participants with a normal baseline and at least 1 post-baseline abnormal HbA1C value, last-observation-carried forward (LOCF) were included in the analysis.

Abnormal weight gain (potentially clinically significant [PCS] weight gain) is defined as weight gain at last visit ≥ 7% of the baseline weight.

Abnormal weight loss (PCS weight loss) is defined as weight loss at last visit ≥ 7% of the baseline weight.

Participants with a baseline and at least 1 post-baseline weight value, last-observation-carried forward (LOCF) were included in the analysis.

All randomized participants were included in the analysis.

Abnormal DPB (diastolic hypertension) is defined as sitting DBP ≥ 90 mm Hg that is also ≥ 10 mm Hg increase from baseline that is observed at last visit if highest baseline DBP < 90 mm Hg.

Abnormal SBP (systolic hypertension) is defined as sitting SBP ≥ 140 mm Hg that is also ≥ 10 mm Hg increase from baseline that is observed at last visit if highest baseline SBP < 140 mm Hg.

Participants with a normal baseline and at least 1 post-baseline DBP and SBP value, last-observation-carried forward (LOCF) were included in the analysis. Participants with a normal baseline value and a nonmissing endpoint value for the variable of interest were included in the analysis.

Abnormal pulse rate (tachycardia) is defined as a sitting heart rate (HR) ≥ 100 beats per minute (bpm) that is also ≥ 10 bpm compared to baseline, at last visit if highest baseline HR < 100 bpm.

Participants with a normal baseline and at least 1 post-baseline pulse rate value, last-observation-carried forward (LOCF) were included in the analysis.

REU captures information regarding the participant's work status and/or health care utilization. Investigators gather information from medical records, psychiatric history, and direct questioning of the participant and his or her family to complete the questionnaire. Responses to each item, comparing baseline to endpoint, are characterized as "Better," "Same," or "Worse." Better: an increase in time spent working/volunteering/holding a job, decrease in number of health care visits; Same: no change in time spent working/volunteering/holding a job, no change in number of health care visits; Worse: decrease in time spent working/volunteering/holding a job, increase in number of health care visits.

Intent-to-treat (ITT) participants with a baseline and at least 1 post-baseline REU value, last-observation-carried forward (LOCF) were included in the analysis.