Pragmatic Trial Design

To maximize applicability to primary care, the trial was designed to be pragmatic. Eligibility criteria facilitated enrollment of diverse patients from primary care. Interventions were delivered with flexibility in medication selection and dosage. Patients were allowed to participate in nonpharmacological pain therapies outside of the study and were encouraged to complete outcome assessments regardless of their participation in the active interventions.

Participants

Eligible patients had chronic back pain or hip or knee osteoarthritis pain that was moderate to severe despite analgesic use. Chronic pain was defined as pain nearly every day for 6 months or more. Moderate or greater severity was defined by a score of 5 or more on the 3-item pain intensity, interference with enjoyment of life, and interference with general activity (PEG) scale (range, 0-10).

Patients on long-term opioid therapy were excluded. Other reasons for exclusion included contraindications to all drug classes in either group, including class-level opioid contraindications (eg, active substance use disorder), and conditions that could interfere with outcome assessment (eg, life expectancy <12 months). Patients with severe depression or posttraumatic stress disorder symptoms were not excluded because these patients often receive opioids in practice.

Patients were recruited from 62 Minneapolis VA primary care clinicians from June 2013 to December 2015 (Figure). Primary care clinicians were located at multiple clinics affiliated with the Minneapolis VA Health Care System, including clinics in the main medical center building and 4 outpatient clinics in the greater Minneapolis-Saint Paul metropolitan area. Potentially eligible patients were identified by searching the electronic health record (EHR) for back, hip, or knee pain diagnoses at a primary care visit in the prior month. Study personnel screened patients by telephone and then conducted a focused chart review.

Open in a separate window

Figure.

Flow of Participants Through the Study

^aPatients could decline to participate at any point in the screening process, including before the telephone eligibility interview; therefore, patients who declined to participate were not necessarily eligible.

Randomization and Blinding

To ensure balanced numbers of patients with back and osteoarthritis pain in each group, randomization was stratified by primary pain diagnosis. The SAS (SAS Institute), version 9.4, uniform random number generator was used to produce a computerized randomization table. Approximately 1 week after the enrollment visit, patients met with the study clinical pharmacist, who initiated random group assignment using a programmed study application that automatically assigned the next unused position in the randomization table. This process simultaneously informed the pharmacist and patient of group assignment. EHR documentation informed patients’ primary care clinicians of study participation and group assignment. Study medications were visible in the EHR. Outcome assessors were blinded to group assignment.

Intervention Delivery

Medication was delivered using a collaborative pain care model with demonstrated effectiveness. In both groups, patients received structured symptom monitoring and a treat-to-target approach to medication management delivered primarily by a single pharmacist. After randomization, the pharmacist reviewed past medications and identified individual functional goals. The initial medication regimen was determined by the assigned group and considerations such as patient preference and comorbidities. Follow-up visits were monthly until a stable regimen was established, then visits occurred every 1 to 3 months. Visits were in-person at 6 and 12 months when possible and otherwise mostly by telephone.

Both interventions used 3 medication steps. Medications were adjusted within the assigned group to achieve targets of improved PEG scores and progress toward individual goals. Study medications were dispensed from the VA pharmacy.

Opioid Prescribing Strategy

Per protocol, patients in the opioid group started taking immediate-release (IR) opioids. Step 1 was morphine IR, hydrocodone/acetaminophen, and oxycodone IR. Step 2 was morphine sustained-action (SA) and oxycodone SA. Step 3 was transdermal fentanyl. Single-opioid therapy was preferred, but dual therapy with a scheduled SA opioid and as-needed IR opioid was considered based on patient needs and preferences. Opioids were titrated to a maximum daily dosage of 100 morphine-equivalent (ME) mg. If dosages were titrated to 60 ME mg/d without a response, rotation to another opioid was considered before dosage escalation.

Nonopioid Prescribing Strategy

In the nonopioid medication group, step 1 was acetaminophen (paracetamol) and nonsteroidal anti-inflammatory drugs (NSAIDs). Step 2 included adjuvant oral medications (ie, nortriptyline, amitriptyline, gabapentin) and topical analgesics (ie, capsaicin, lidocaine). Step 3 included drugs requiring prior authorization from the VA clinic (ie, pregabalin, duloxetine) and tramadol. Patients were initially prescribed a step 1 medication, unless all were clinically inappropriate. Subsequent changes included titrating, replacing, or adding medications.

Intervention Adherence

Patients were instructed to receive medications for back, hip, or knee pain only from the study. Nonpharmacological therapies were allowed outside of the study. If patients desired discontinuation of all study medications, they were transitioned back to preenrollment pain medications. Medication adherence was monitored by discussion with patients and checking the state prescription monitoring program website.

Descriptive Measures

Before randomization, patients were asked to state their preferred treatment group, perceptions of effectiveness and safety of opioid and nonopioid medications, and expectations for improvement on 0 to 10 scales (higher scores=more favorable). To characterize the study population and provide data required by federal funders, self-identified race/ethnicity was assessed by asking patients to select from 6 categories.

Main Outcomes

The primary outcome was pain-related function, assessed with the 7-item Brief Pain Inventory (BPI) interference scale. Pain intensity, the main secondary outcome, was assessed with the 4-item BPI severity scale. Both BPI scales yield 0 to 10 scores (higher score=worse function or intensity). A prior study of chronic pain in primary care estimated a minimal clinically important difference (MCID) of 0.7 points for both BPI interference and BPI severity. Following consensus guidelines, this trial used a 1-point difference as the MCID for BPI interference and BPI severity, and used a 30% reduction from baseline as MCID for moderate improvement. The primary adverse outcome was a patient-reported checklist of 19 medication-related symptoms, modified from the original version by adding common analgesic adverse effects (eg, memory problems, sweating).

Secondary Health Outcomes

Secondary outcomes were as follows: the Veterans RAND 12-item Health Survey (VR-12) quality-of-life measure (range, 0-100; higher score=better quality of life, standardized to mean of 50), the 11-item Roland-Morris Disability Questionnaire (RMDQ) measure of pain-related physical function (range, 0-11; higher score=worse function, MCID=2.0), the 8-Item Patient Health Questionnaire (PHQ-8) depression measure (range, 0-24; higher score=worse depression, MCID=5), the 7-Item Generalized Anxiety Disorder measure (GAD-7; range, 0-21; higher score=worse anxiety, MCID=5); the Patient-Reported Outcomes Measurement Information System (PROMIS) sleep disturbance short form (range, 8-32; higher score=worse sleep disturbance); the Migraine Disability Assessment (MIDAS) questionnaire (range, 0-270; higher score=worse headache disability), the Arizona Sexual Experience Scale (ASEX; range 5-30; higher score=worse sexual function); and the Multidimensional Fatigue Inventory (MFI) general fatigue, mental fatigue, physical fatigue, reduced activity, and reduced motivation scales (for each scale: range, 4-20; higher score=worse, MCID=2). Additional secondary outcomes not reported here were the global impression of pain change, the Fullerton Advanced Balance scale, 6-m gait speed, chair stand, grip strength tests, cold pain tolerance, free testosterone, and the Indiana University Telephone-Based Assessment of Neuropsychological Status.

Assessment for Adverse Events and Potential Opioid Misuse

At each assessment, patients reported new hospitalizations, emergency department (ED) visits, and falls. VA hospitalizations and ED events were identified by searching EHR databases from enrollment to 13 months after randomization. Two independent raters determined whether events were analgesic-related. Discrepancies were resolved by discussion.

Opioid misuse describes use of prescription opioids in a manner other than as prescribed. This study used multiple approaches to evaluate for potential misuse, including medical record surveillance for evidence of “doctor-shopping” (seeking medication from multiple physicians), diversion, substance use disorder, or death; checking the state prescription monitoring program website at each visit and as needed; and completing the Addiction Behavior Checklist at each intervention visit. The Addiction Behavior Checklist measures aberrant medication-related behaviors that may indicate misuse (range, 0-20; higher score=more aberrant behavior; 3=threshold for opioid misuse). At 6-month and 12-month assessments, patients completed self-report measures and had urine drug testing. Substance use was assessed with the Alcohol Use Disorders Identification Test (AUDIT) and drug use questions from a National Institute on Drug Abuse screening tool.

Assessment of Study Treatment Received and Nonstudy Co-Interventions

Pain medication dispensing data were obtained from EHR databases. Total study visit duration was calculated for each patient as the sum of minutes from clinician-entered Current Procedural Terminology (CPT) codes for all intervention encounters; for CPT codes that include a range of minutes (ie, 5-10, 11-20, 21-30), the highest value was used. Nonstudy co-interventions were obtained from patient report and EHR data.

Pain and Health Outcomes

There was no significant difference in pain-related function between the 2 groups over 12 months (overall P=.58). At 12 months, mean BPI interference was 3.4 in the opioid group (SD, 2.5) vs 3.3 in the nonopioid group (SD, 2.6); difference, 0.1 (95% CI, −0.5 to 0.7). Pain intensity was significantly better in the nonopioid group over 12 months (overall P=.03). At 12 months, mean BPI severity was 4.0 in the opioid group (SD, 2.0) vs 3.5 in the nonopioid group (SD, 1.9); difference, 0.5 (95% CI, 0.0 to 1.0).

Functional response (≥30% improvement in BPI interference) occurred in 69 patients (59.0%) in the opioid group vs 71 patients (60.7%) in the nonopioid group; difference, −1.7% (95% CI, −14.4 to 11.0); P=.79. Pain intensity response (≥30% improvement in BPI severity) occurred in 48 patients (41.0%) in the opioid group vs 63 patients (53.9%) in the nonopioid group; difference, −12.8% (95% CI, −25.6 to 0.0); P=.05.

Health-related quality of life did not significantly differ between the 2 groups (physical health overall: P=.23; difference at 12 months, −1.3 [95% CI, −3.8 to 1.3]; mental health overall: P=.40; difference at 12 months, 0.7 [95% CI, −2.4 to 3.8]). Of the remaining secondary outcomes, only anxiety significantly differed between groups (Table 2; eTables 1-2 in Supplement 2).

Table 2.

Patient-Reported Primary and Secondary Outcomes Among Patients With Chronic Back Pain or Hip or Knee Osteoarthritis Pain Randomized to Opioid vs Nonopioid Medication

Outcome	Opioid Group, Mean (SD) (n=119)	Nonopioid Group, Mean (SD) (n=119)	Between-Group Difference (95% CI)a	Overall P Valueb
Pain-Related Function (Primary Outcome)
BPI interference scale (range, 0-10; higher score=worse)c				.58
Baseline	5.4 (1.8)	5.5 (2.0)	−0.1 (−0.6 to 0.4)
3 mo	3.7 (2.1)	3.7 (2.2)	0.0 (−0.6 to 0.6)
6 mo	3.4 (2.1)	3.6 (2.4)	−0.2 (−0.8 to 0.4)
9 mo	3.6 (2.2)	3.3 (2.4)	0.4 (−0.2 to 1.0)
12 mo	3.4 (2.5)	3.3 (2.6)	0.1 (−0.5 to 0.7)
Pain Intensity (Secondary Outcome)
BPI severity scale (range, 0-10; higher score=worse)d				.03
Baseline	5.4 (1.5)	5.4 (1.2)	0.0 (−0.4 to 0.3)
3 mo	4.3 (1.8)	4.0 (1.7)	0.3 (−0.2 to 0.7)
6 mo	4.1 (1.8)	4.1 (1.9)	0.0 (−0.5 to 0.5)
9 mo	4.2 (1.7)	3.6 (1.7)	0.7 (0.2 to 1.2)
12 mo	4.0 (2.0)	3.5 (1.9)	0.5 (0.0 to 1.0)
Additional Secondary Health Outcomes
VR-12 physical health (range, 0-100; lower score=worse)				.23
Baseline	27.2 (9.0)	27.0 (7.2)	0.2 (−1.9 to 2.2)
3 mo	32.5 (9.8)	33.5 (9.9)	−1.0 (−3.6 to 1.6)
6 mo	33.3 (9.7)	33.6 (10.0)	−0.3 (−2.8 to 2.2)
9 mo	32.0 (10.5)	34.8 (10.9)	−2.9 (−5.8 to 0.0)
12 mo	32.7 (10.1)	33.9 (9.9)	−1.3 (−3.8 to 1.3)
VR-12 mental health (range, 0-100; lower score=worse)				.40
Baseline	47.3 (11.2)	47.8 (13.0)	−0.3 (−3.4 to 2.8)
3 mo	51.8 (10.1)	50.5 (12.0)	1.3 (−1.6 to 4.3)
6 mo	51.6 (9.8)	50.3 (12.5)	1.4 (−1.5 to 4.3)
9 mo	51.8 (10.7)	52.6 (11.5)	−0.8 (−3.8 to 2.2)
12 mo	51.2 (11.6)	50.4 (12.6)	0.7 (−2.4 to 3.8)
RMDQ-11 pain-related physical function (range, 0-11; higher score=worse)e				.47
Baseline	8.0 (2.5)	8.6 (1.9)	−0.5 (−1.1 to 0.0)
6 mo	6.3 (3.3)	7.1 (3.1)	−0.8 (−1.7 to 0.0)
12 mo	5.8 (3.4)	5.9 (3.5)	−0.1 (−1.0 to 0.8)
PHQ-8 depression symptoms (range, 0-24; higher score=worse)f				.13
Baseline	6.3 (4.5)	5.8 (5.0)	0.5 (−0.7 to 1.7)
6 mo	4.4 (3.9)	4.8 (5.2)	−0.4 (−1.6 to 0.8)
12 mo	4.3 (4.0)	4.5 (5.3)	−0.2 (−1.5 to 1.1)
GAD-7 anxiety symptoms (range, 0-21; higher score=worse)g				.02
Baseline	4.0 (3.6)	3.5 (4.0)	0.5 (−0.5 to 1.4)
6 mo	3.0 (3.5)	3.2 (4.5)	−0.2 (−1.3 to 0.8)
12 mo	2.5 (3.3)	2.8 (4.2)	−0.4 (−1.4 to 0.7)
PROMIS sleep disturbance (range, 8-32; higher score=worse)g				.33
Baseline	25.5 (7.8)	24.2 (8.4)	1.2 (−0.8 to 3.3)
6 mo	22.2 (8.8)	22.0 (9.0)	0.2 (−2.2 to 2.5)
12 mo	23.4 (8.2)	21.0 (8.3)	2.3 (0.1 to 4.6)
MIDAS headache disability (range, 0-270; higher score=worse)h				.82
Baseline	6.1 (16.5)	6.1 (16.2)	−0.1 (−4.2 to 4.1)
6 mo	3.8 (12.6)	5.5 (18.8)	−1.7 (−6.0 to 2.5)
12 mo	3.7 (11.6)	3.2 (11.6)	−0.5 (−2.7 to 3.6)
ASEX sexual function (range, 5-30; higher score=worse)i				.49
Baseline	17.4 (5.6)	17.7 (6.0)	−0.3 (−1.8 to 1.3)
12 mo	17.9 (6.0)	19.0 (6.5)	−1.1 (−2.8 to 0.7)
MFI general fatigue (range, 4-20; higher score=worse)j				.68
Baseline	13.8 (3.8)	12.8 (4.1)	1.0 (−0.0 to 2.0)
6 mo	12.7 (3.9)	12.5 (4.3)	0.2 (−0.9 to 1.3)
12 mo	12.5 (3.9)	12.0 (4.4)	0.6 (−0.6 to 1.7)
MFI mental fatigue (range, 4-20; higher score=worse)j				.39
Baseline	10.0 (4.2)	9.6 (4.7)	0.4 (−0.7 to 1.6)
6 mo	9.0 (4.2)	9.3 (4.4)	−0.3 (−1.4 to 0.9)
12 mo	9.2 (3.9)	9.3 (4.3)	0.1 (−1.3 to 1.0)
MFI physical fatigue (range, 4-20; higher score=worse)j				.73
Baseline	13.6 (4.1)	12.9 (4.1)	0.7 (−0.3 to 1.8)
6 mo	12.9 (4.4)	12.5 (4.5)	0.4 (−0.8 to 1.5)
12 mo	12.4 (4.3)	11.8 (4.3)	0.7 (−0.5 to 1.9)
MFI reduced activity (range, 4-20; higher score=worse)j				.74
Baseline	11.4 (4.1)	10.9 (4.6)	0.5 (−0.7 to 1.6)
6 mo	10.6 (4.6)	10.5 (4.5)	0.2 (−1.0 to 1.4)
12 mo	10.6 (4.2)	10.3 (4.5)	0.3 (−1.0 to 1.5)
MFI reduced motivation (range, 4-20; higher score=worse)j				.09
Baseline	9.8 (3.6)	8.8 (3.8)	1.0 (0.0 to 2.0)
6 mo	9.1 (3.6)	8.9 (4.0)	0.2 (−0.8 to 1.2)
12 mo	8.6 (3.2)	8.8 (3.7)	−0.2 (−0.7 to 1.6)

Open in a separate window

Abbreviations: ASEX, Arizona Sexual Experience Scale; BPI, Brief Pain Inventory; GAD-7, 7-Item Generalized Anxiety Disorder Questionnaire; MFI, Multidimensional Fatigue Inventory; MIDAS, Migraine Disability Assessment Scale; PHQ-8, 8-Item Patient Health Questionnaire; PROMIS, Patient Reported Outcomes Measurement Information System; RMDQ-11, 11-Item Roland-Morris Disability Questionnaire; VR-12, Veterans RAND 12-item Health Survey.

^aUnadjusted time-specific between-group comparisons.

^bP values are from mixed models for repeated measures comparing between-group difference during the 12-mo trial, controlling for baseline and including all available time points.

^cMissing data for 1 patient in the opioid group at 9 mo.

^dMissing data for 1 patient in the opioid group at 3 mo.

^eMissing data for 2 patients in the nonopioid group at 12 mo.

^fMissing data for patients: at 6 mo, 3 in the opioid group and 9 in the nonopioid group; at 12 mo, 12 in the opioid group and 15 in the nonopioid group.

^gMissing data for patients: at 6 mo, 2 in the opioid group and 8 in the nonopioid group; at 12 mo, 11 in the opioid group and 12 in the nonopioid group.

^hMissing data for patients: at 6 mo, 3 in the opioid group and 8 in the nonopioid group ; at 12 mo, 13 in the opioid group and 14 in the nonopioid group.

ⁱMissing data for patients: at baseline, 11 in the opioid group and 9 in the nonopioid group; at 12 mo, 19 in the opioid group and 17 in the nonopioid group.

^jMissing data for patients: at baseline, 2 in the opioid group and 3 in the nonopioid group; at 6 mo, 2 in the opioid group and 9 in the nonopioid group; at 12 mo, 14 in the opioid group and 18 in the nonopioid group.

Adverse Outcomes and Potential Misuse

The opioid group had significantly more medication-related symptoms over 12 months than the nonopioid group (overall: P=.03; difference at 12 months, 0.9 [95% CI, 0.3 to 1.5]) (Table 3).

Table 3.

Adverse Outcomes and Measures of Potential Misuse Among Patients With Chronic Back Pain or Hip or Knee Osteoarthritis Pain Randomized to Opioid vs Nonopioid Medication

Outcome	Opioid Group	Nonopioid Group	Between-Group Difference (95% CI)a	P Value
Primary Adverse Outcome
Medication-related symptom checklist (0-19; higher score=worse), mean (SD)b				.03c
Baseline	1.2 (1.9)	1.2 (1.9)	0.0 (−0.5 to 0.5)
3 mo	2.3 (2.5)	1.3 (1.8)	1.0 (0.5 to 1.6)
6 mo	2.1 (2.7)	1.3 (2.3)	0.7 (0.1 to 1.4)
9 mo	1.9 (2.8)	0.9 (1.9)	1.0 (0.4 to 1.6)
12 mo	1.8 (2.6)	0.9 (1.8)	0.9 (0.3 to 1.5)
Secondary Adverse Outcomes
All-cause hospitalization, No.(%)d				.94e
0	99 (83)	99 (83)	0 (−10 to 10)
1	15 (13)	16 (13)	1 (−9 to 8)
≥2	6 (5)	5 (4)	1 (−5 to 6)
All-cause ED visit, No.(%)d				.18e
0	60 (50)	73 (61)	−11 (−24 to 2)
1	34 (28)	30 (25)	3 (−8 to 15)
≥2	26 (22)	17 (14)	8 (−2 to 17)
Number of falls in 12 mo after enrollment, No.(%)f				.19e
0	63 (53)	63 (53)	0 (−13 to 13)
1	26 (22)	17 (14)	8 (−2 to 17)
≥2	29 (25)	39 (33)	−8 (−20 to 3)
Potential Misuse Measures
Patients with ≥1 positive urine drug tests for an illicit drug or unexplained prescription drug, No. (%)g
Illicit drug positive	6 (5)	12 (11)	−5 (−12 to 2)	.13e
Unexplained prescription drug positive	11 (10)	9 (8)	3 (−5 to 10)	.67e
Clinician-assessed behaviors, No.(%)
Significant PMP finding at any visith	6 (5)	4 (3)	2 (−3 to 7)	.75i
Misuse behavior at any visitj	11 (9)	8 (7)	3 (−4 to 9)	.47e
Patient-reported substance use at 12 mo, No.(%)
Hazardous alcohol usek	2 (2)	4 (4)	−2 (−6 to 3)	.44i
Past-year drug usel	17 (16)	13 (13)	3 (−6 to 13)	.56e

Open in a separate window

Abbreviations: ED, emergency department; PMP, Prescription Monitoring Program.

^aUnadjusted time-specific between-group comparison of means or percentages.

^bMissing data for patients: at 3 mo, 1 in the nonopioid group; at 6 mo, 1 in the opioid group and 1 in the nonopioid group; at 12 mo, 3 in the opioid group and 3 in the nonopioid group (n=119 in each group).

^cP value for treatment by time interaction.

^dHospitalization and ED visit events were counted until 13 mo after randomization for all randomized patients (n=120 in each group). Events that started in the ED and resulted in hospitalization were counted as hospitalizations and do not contribute to the ED visit count.

^eP value from χ² test.

^fThe sum of falls reported at each follow-up interview. Missing data for 1 patient in the opioid group.

^gIllicit drugs are illegal substances, including cannabis. Unexplained prescription drugs are potentially prescribed substances for which there was no known prescription. Missing data for patients: 4 in the opioid group and 6 in the nonopioid group.

^hSignificant PMP finding is any prescription that was not disclosed and for which there was no clear acute pain-related indication (n=119 in each group).

ⁱP value for Fisher exact test.

^jMisuse behavior was an Addiction Behavior Checklist score of 3 or more at any visit (n=119 in each group).

^kHazardous alcohol use is Alcohol Use Disorders Identification Test score of 8 or more. Missing data for patients: 4 in the opioid group and 6 in the nonopioid group.

^lPositive result was defined as a patient report of any past-year use of cannabis, cocaine, methamphetamine, inhalants, hallucinogens, street opioids, or prescription medications (opioids, sedatives, or stimulants) for nonmedical purposes. Missing data for 13 opioid patients and 17 nonopioid patients.

There were no significant differences in adverse outcomes or potential misuse measures (Table 3). Two hospitalization or ED visit events were determined analgesic-related: 1 hospitalization in the nonopioid group and 1 ED visit in the opioid group. No deaths, “doctor-shopping,” diversion, or opioid use disorder diagnoses were detected.

Intervention Adherence and Retention

Number and duration of study visits were similar in the 2 groups (Table 4). Twenty-three patients (19%) in the opioid group and 10 patients (8%) in the nonopioid group discontinued study medication (eTable 6 in Supplement 2). Most patients in the opioid group received low or moderate dosage therapy (eTables 7-8 in Supplement 2). In each 90-day follow-up period, fewer than 15% of patients in the opioid group had a mean dispensed dosage of 50 ME mg/d or more. In the nonopioid group, tramadol was dispensed to 4 patients (3%), 6 patients (5%), 8 patients (7%), and 13 patients (11%) in the first, second, third, and fourth 90-day follow-up windows, respectively. eTables 9 to 10 in Supplement 2 show nonstudy pain treatments.

Table 4.

Medications and Visits Over 12 Months From the Electronic Health Records of Patients With Chronic Back Pain or Hip or Knee Osteoarthritis Pain Randomized to Opioid vs Nonopioid Medication

	Opioid Group (n=119)		Nonopioid Group (n=119)
	Mean (SD)	Median (IQR)	Mean (SD)	Median (IQR)
Study drugs, No.a	1.7 (0.8)	2.0 (1.0-2.0)	3.8 (1.7)	4.0 (3.0-5.0)
Study prescribed analgesic, months, No.b
Acetaminophen	0.1 (0.5)	0.0 (0.0-0.0)	2.6 (3.2)	1.0 (0.0-4.0)
Oral NSAID	0.4 (2.0)	0.0 (0.0-0.0)	5.9 (4.9)	5.0 (0.5-10.0)
Analgesic adjunct	0.2 (1.4)	0.0 (0.0-0.0)	3.3 (4.3)	1.0 (0.0-6.2)
Topical	0.0 (0.6)	0.0 (0.0-0.0)	3.5 (3.5)	3.0 (1.0-6.0)
Tramadol	0.1 (0.6)	0.0 (0.0-0.0)	0.4 (1.3)	0.0 (0.0-0.0)
Opioidc	8.1 (4.1)	8.4 (5.6-11.2)	0.0 (0.0)	0.0 (0.0-0.0)
Study visits, No.
In-person visits	2.8 (2.0)	2.0 (2.0-3.0)	2.8 (2.2)	2.0 (2.0-3.0)
Telephone visits	6.2 (2.9)	7.0 (5.0-8.0)	6.2 (2.5)	7.0 (5.0-8.0)
Total study visit duration, mind	231 (95)	230 (159-289)	217 (82)	197 (155-267)
Nonstudy outpatient visits, No.e
Primary care	6.8 (6.5)	5.0 (2.0-8.0)	7.1 (7.1)	4.0 (2.0-9.0)
Specialty	6.7 (12.0)	3.0 (1.0-8.0)	6.3 (6.4)	4.0 (1.0-9.0)
Mental health	4.8 (10.3)	0.0 (0.0-6.0)	7.5 (22.1)	0.0 (0.0-5.0)
Rehabilitation	4.5 (15.8)	1.0 (0.0-3.0)	3.1 (6.1)	1.0 (0.0-4.0)

Open in a separate window

Abbreviations: IQR, interquartile range; NSAIDs, nonsteroidal anti-inflammatory drugs.

^aNumber of unique study-prescribed medication formulations during the intervention, regardless of duration of use.

^bAnalgesic months is the sum of the number of months of medication dispensed from Veterans Affairs outpatient pharmacies for each discrete medication within a category during the 12-mo intervention period. For example, a patient dispensed analgesic A for 6 mo and analgesic B for 12 mo would have 18 analgesic months. Crossover (ie, nonopioid medications in the opioid group and vice versa) is accounted for by patients who desired discontinuation of all medications in their assigned study group. Study clinicians restarted preenrollment medications if requested by these patients, but did not manage or adjust these off-protocol medications.

^cOpioid months do not include tramadol.

^dThe sum of minutes extracted from clinician-entered Current Procedural Terminology codes for all study encounters.

^eOutpatient visits include both in-person and telephone encounters with any type of clinician, including physicians, mental health providers, physical therapists, and nurses. Encounters for diagnostic testing (eg, radiology examinations, endoscopy) and nonmedical ancillary services (eg, social work, nutrition, education) are not included.