The analysis population included enrolled participants who took part in all follow-up interviews.

Significant variables included in the final model were age over 65, physician gender and specialty, use of opioids (excluding tramadol), use of non-steroidal anti-inflammatory drugs (NSAIDs), and number of medications.

The analysis population included enrolled participants who had non-missing covariates used in logistic regression.

Significant variables included in the final model were pregabalin use, NSAID use, and GAD-7 (7-item, self-reported measurement of GAD severity [not at all severe, severe for several days, severe for more than half the days, severe nearly every day]; Total score=sum of all 7 items; ranges from 0 to 21; Higher score=greater level of anxiety). Odds ratios were based on medication possession ratio (MPR) ≥0.8 for pregabalin and NSAID use among duloxetine initiators at Baseline; MPR=total supply days/total number of days in 12-month study period.

The analysis population included enrolled participants who had non-missing covariates used in logistic regression.

Participants were allowed to select multiple reasons for discontinuing treatment. During the 12-month period post-baseline, a participant could possibly discontinue more than 1 medication, or discontinue the same medication more than once. A reason for discontinuation was given each time a participant stopped taking a medication.

The analysis population included participants who reported discontinuing medication at least once during the 12-month period post-baseline.

The analysis population included participants who were using opioids at Baseline.

Factors significantly associated with discontinuation of opioid use were participant age, Brief Pain Inventory-Interference (BPI-I) score, and Patient Health Questionnaire for somatic symptoms (PHQ-15) score.

The analysis population included participants who were using opioids at Baseline.

Average BPI-I measures self-reported degree of pain interference on function. Interference scores range from 0 (does not interfere) to 10 (completely interferes) for each question assessing interference of pain within past 24 hours for general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life. Average interference = average of non-missing scores of individual interference items.

The analysis population included participants who had non-missing BPI-I scores at Baseline and 12 months.

BPI-S measures self-reported severity of pain. Severity scores range from 0 (no pain) to 10 (severe pain) for each question assessing worst pain, least pain, and average pain in past 24 hours, and current pain.

The analysis population included participants who had non-missing BPI-S scores at Baseline and 12 months.

FIQ measures self-reported fibromyalgia (FM) status, progress, and outcomes over past week. FIQ comprises 20 items: Items 1-11 measure physical functioning (each rated on 4-point Likert-type scale); Items 12 + 13 measure number (no.) of days participant felt well and no. of days participant felt unable to work due to FM symptoms. Items 14-20 are numerical, 11-point Likert-type scales (marked in 10-millimeter [mm] increments) rating work difficulty, pain intensity, fatigue, morning tiredness, stiffness, anxiety, and depression. Total scores range from 0-80; Higher score=greater negative impact.

The analysis population included enrolled participants who had non-missing FIQ scores at Baseline and 12 months.

The SDS is completed by the participant and assesses the effect of the participant's symptoms on work/social/family life. Total scores range from 0 to 30; Higher score=greater disruption in the participant's work/social/family life.

The analysis population included enrolled participants who had non-missing SDS scores at Baseline and 12 months.

The analysis population included enrolled participants who had at least 1 follow-up visit. The number of participants actually used in analyses were less than 1700 because some had missing outpatient visit values.

The analysis population included enrolled participants who had at least 1 follow-up visit. The number of participants actually used in analyses were less than 1700 because some had missing ER visit values.

Partial (day or night) care included day care, day nursing home, and partial hospitalization.

The analysis population included enrolled participants who had at least 1 follow-up visit.

Caregiver burden was described as the number of days family members missed paid work and used a paid caregiver (for example, home healthcare nurse) to care for the participant with FM.

The analysis population included enrolled participants who had at least 1 follow-up visit.

Caregiver burden was described as the number of days family members used an unpaid caregiver (family and friends) to care for the participant with FM.

The analysis population included enrolled participants who had at least 1 follow-up visit.

Participant productivity was described as the number of days the participant stayed in bed, had to reduce normal activity by half, and received disability income due to FM.

The analysis population included enrolled participants who had at least 1 follow-up visit.

Participant productivity was described as the number of days a participant missed work due to FM.

The analysis population included enrolled participants who had at least 1 follow-up visit.