Web-Based Cognitive Behavior Therapy for Chronic Pain Patients with Aberrant Drug-Related Behavior: Outcomes from a Randomized Controlled Trial

Abstract

Objective

There is high unmet need for effective behavioral treatments for chronic pain patients at risk for or with demonstrated histories of opioid misuse. Despite growing evidence supporting technology-based delivery of self-management interventions for chronic pain, very few such programs target co-occurring chronic pain and aberrant drug-related behavior. This randomized controlled trial evaluated the effectiveness of a novel, web-based self-management intervention, grounded in cognitive behavior therapy, for chronic pain patients with aberrant drug-related behavior.

Methods

Opioid-treated chronic pain patients at a specialty pain practice who screened positive for aberrant drug-related behavior (N = 110) were randomized to receive treatment as usual plus the web-based program or treatment as usual alone. The primary outcomes of pain severity, pain interference, and aberrant drug-related behavior, and the secondary outcomes of pain catastrophizing and pain-related emergency department visits, were assessed during the 12-week intervention and at one and three months postintervention.

Results

Patients assigned to use the web-based program reported significantly greater reductions in aberrant drug-related behavior, pain catastrophizing, and pain-related emergency department visits—but not pain severity or pain interference—relative to those assigned to treatment as usual. The positive outcomes were observed during the 12-week intervention and for three months postintervention.

Conclusions

A web-based self-management program, when delivered in conjunction with standard specialty pain treatment, was effective in reducing chronic pain patients’ aberrant drug-related behavior, pain catastrophizing, and emergency department visits for pain. Technology-based self-management tools may be a promising therapeutic approach for the vulnerable group of chronic pain patients who have problems managing their opioid medication.

Introduction

Increasing recognition of the risks associated with long-term opioid therapy for chronic pain has led to changing indications and reconsideration of the standards for therapeutic monitoring and oversight [1–6]. There is widespread acceptance of the view that treatment adherence is essential for continued therapy [7,3]. Clinicians must assess patients for nonadherence behaviors—phenomena also termed aberrant drug-related behaviors (ADRB)—and should they occur, re-evaluate and intervene. The inability to establish adherence often justifies discontinuation of treatment [1,4,5,6].

ADRB encompasses a broad array of behaviors within the clinical context that are suggestive of opioid medication misuse. Examples range from occasional unsanctioned dose escalation, drug hoarding during periods of reduced pain intensity, and aggressive complaining about the need for more pain relief to taking opioid medications to relieve symptoms other than physical pain, use of alternative routes of administration of oral formulations (e.g., sniffing, injection), and repeated resistance to medication changes despite evidence of adverse effects [8–10].

Studies of opioid-treated chronic pain patients indicate that the prevalence of ADRB is often high but varies widely (e.g., 20–75%) due to heterogeneous subpopulations, definitions of ADRB, and assessment measures used [11,9,12]. Even at the lowest rates, ADRB is a serious concern, and clinical interventions that increase the likelihood of adherence would be highly valued, particularly if they targeted co-occurring chronic pain and ADRB.

Behavioral therapies have long been a part of a multimodal strategy recommended for the treatment of chronic pain [2]. Cognitive behavior therapy (CBT) has been most extensively evaluated, with meta-analyses and systematic reviews demonstrating its effectiveness in reducing pain-related impairments, pain catastrophizing, and, in some cases, pain intensity [13–16]. Emerging research has also begun to support the effectiveness of interventions that integrate CBT principles and techniques in the management of ADRB. One randomized controlled trial found that chronic pain patients with ADRB reported less medication misuse when assigned to CBT-informed motivational counseling plus multimodal medication monitoring as compared with treatment as usual (TAU) [7]. Another randomized controlled trial of group-based Mindfulness-Oriented Recovery Enhancement (MORE) for prescription opioid-misusing chronic pain patients found a significantly greater reduction in the number of MORE participants who met criteria for opioid use disorder immediately postintervention relative to participants in a generic support group, although these gains were not sustained at three-month follow-up [17].

Yet most patients with chronic pain do not receive CBT or other behavioral treatments due to barriers that include physicians’ lack of familiarity with behavioral modalities, a scarcity of qualified providers, and the lack of sufficient insurance coverage for these modalities [2,18,19]. Using the internet and/or mobile phones to deliver behavioral interventions for chronic pain may help surmount these barriers due to the wide reach, assured fidelity, flexibility, scalability, and relatively low implementation costs of such tools [20,19]. Existing programs of internet-delivered CBT for chronic pain have been found to be more effective than waitlist or usual-care controls and comparable in effectiveness to other active treatments [21–23,20].

A technology-delivered behavioral intervention for co-occurring chronic pain and ADRB would provide an important therapeutic option for opioid-treated pain patients. Take Charge of Pain was developed as a novel, web-based self-management intervention grounded in CBT for chronic pain patients with ADRB. This paper reports the results of a randomized controlled trial to evaluate the effectiveness of this program when delivered as an adjunct to TAU, relative to TAU alone, in reducing pain severity, pain interference, and ADRB.

Methods

The trial was registered at ClinicalTrials.gov (Protocol Identifier: NCT01498510) and approved by the Institutional Review Boards of National Development and Research Institutes and Beth Israel/Mount Sinai Medical Center.

Study Design

In a two-arm randomized controlled trial, chronic pain patients with ADRB were assigned in a 1:1 ratio to either TAU (N = 55) or TAU plus the web-based CBT intervention, Take Charge of Pain (N = 55), in an intent-to-treat design. This paper reports the relative effectiveness of these treatments on the study’s three primary outcomes of pain severity, pain interference, and ADRB, as well as the secondary outcomes of pain catastrophizing and emergency department (ED) visits for pain. (Several additional secondary outcomes were assessed but are not reported on herein). We hypothesized that patients assigned to the CBT arm would attain significantly better outcomes across all the variables of interest. A sample size of 110 (55 per group) was estimated to produce sufficient power (>0.80) to detect medium-large effects on the primary outcome variables, based on meta-analyses of CBT for chronic pain patients [13,14].

Participants

Participants were recruited between April, 2012, and July, 2014, from a large, New York City tertiary pain treatment practice. Patients were referred to the study by their pain provider, by self-referral after seeing study fliers posted in the treatment rooms of the pain practice, or by research staff outreach in the practice’s waiting room.

To be eligible for the trial, patients were required to be ≥18 years of age; report moderate to severe pain (defined as rating one’s worst pain in the past week as ≥5 on the 0–10-point Brief Pain Inventory [BPI]) [24] for at least three months; be receiving long-term opioid therapy for pain; and endorse at least four items (with any response >0) on the Current Opioid Misuse Measure (COMM) [8] in relation to the past 30 days. Exclusion criteria included mental illness severe enough to impair the ability to provide informed consent; insufficient English to participate; scheduled to have major surgery in the next six months; described by pain physician as likely to die within the next year; planning to move out of the area in the next three months; and primary headache- or cancer-related pain diagnoses.

Procedure

Following a preliminary telephone screening, prospective participants completed an in-person screening at the research study site to confirm their trial eligibility. Eligible patients then participated in an informed consent session and completed the baseline assessment. Select eligibility criteria (i.e., pain diagnoses, opioid prescription status, and whether patients were scheduled for surgery or approaching end of life) were verified via post hoc review of patients’ medical records and consultation with pain providers.

Immediately following completion of the baseline assessment, participants were randomly assigned to TAU or TAU plus the web-based CBT intervention. Permuted-block randomization was conducted, stratified by the patient’s pain provider and whether or not the patient met lifetime DSM-IV criteria for abuse or dependence on any substance (as assessed with the MINI International Neuropsychiatric Interview [MINI]) [25]. Each participant’s allocation was determined by means of an electronic spreadsheet prepared by the study’s statistician, which the interviewer consulted at the conclusion of the baseline assessment. (As is common in trials of behavioral interventions, research staff were not blinded to participants’ treatment assignments.) The CONSORT Diagram depicting participants’ flow through the trial protocol is presented in Figure 1.

Figure 1.

CONSORT flow diagram. *Patients determined to be ineligible at prescreening (conducted by phone) did not visit the research office to complete in-person screening. ^†This calculation does not include the program’s first introductory/training module, which participants were asked to complete at the end of the in-person baseline visit. CBT = cognitive behavior therapy; F/U = follow-up; TAU = treatment as usual

Participants were assessed at baseline and weeks 4, 8, and 12 during the 12-week active intervention phase, and at one- and three-month postintervention follow-ups in order to assess the durability of intervention effects and the possible delayed emergence of such effects. Follow-up assessment concluded in December 2014. Participants were compensated $50 for completing the baseline assessment and $40 for each subsequent assessment, but were not compensated for completing modules within the web-based CBT intervention. (Thus, total possible compensation was $250, regardless of treatment assignment.) All assessments were administered by research staff using a computer-based interface.

Measures

Response to Web-CBT Intervention

To assess the acceptability of the Take Charge of Pain program, participants assigned to the web-CBT condition completed a structured feedback survey at the 12-week assessment time point. This assessment contained seven visual analog scale (VAS) items, anchored by 0 on the low end and 10 on the high end, asking participants to rate different aspects of the web-based intervention including its usefulness and interest level, the amount of new information it contained, the extent to which it clarified misconceptions about chronic pain and about opioid medications, ease of understanding, and their overall satisfaction with the program.

Primary Outcomes

Pain severity and pain interference were measured with their respective subscales from the Multidimensional Pain Inventory (MPI) [26], a validated self-report instrument [27] that is widely used in studies with chronic pain patients. The three-item Pain Severity subscale assesses pain severity and suffering within the past week. The nine-item Pain Interference subscale assesses the interference of pain with daily and social activities, work, and family relationships.

ADRB was measured with the COMM [8], a 17-item self-report assessment intended to monitor current medication misuse among chronic pain patients on long-term opioid therapy. The COMM assesses the relative frequency of a thought or behavior over the past 30 days on a five-point scale anchored by 0 = never and 4 = very often. To reduce the likelihood of patients deliberately underreporting their medication misuse behaviors, several items do not explicitly refer to medication misuse, but rather, cognitive or emotional dimensions of ADRB.

Secondary Outcomes

Pain catastrophizing, defined as exaggerated negative thoughts or feelings about pain, was assessed with the Pain Catastrophizing Scale (PCS) [28]. This 13-item self-report measure has high internal consistency (alpha = 0.87) and assesses three domains of catastrophizing—helplessness, rumination, and magnification—on a five-point scale ranging from 0 = not at all to 4 = always.

At baseline, participants were asked how often they had visited an ED in the past six months and how many of these visits were specifically for pain. These questions about general and pain-related ED visits were repeated, in reference to the past 30 days, at each of the five subsequent assessment time points.

Use of and Perceived Benefit from Cognitive Behavior Therapy Skills

To examine patients’ engagement in the CBT skills and techniques included in the web-based program (e.g., controlled breathing, muscle relaxation, and pacing physical activities), participants in both conditions were asked, at one- and three-month postintervention follow-ups, whether they had practiced these skills in the past 30 days in an effort to manage pain (see Table 4 for a list of activities assessed). The perceived benefit derived from these activities was then assessed with eight VAS items asking, for each activity endorsed, the extent to which participants found the activity helpful in relieving pain, with responses ranging from 0 = not at all helpful to 10 = very helpful.

Table 4.

Participants’ engagement in and perceived benefit from CBT activities at postintervention follow-ups, by treatment group

Cognitive Behavior Therapy Activity	Did in Past Month to Manage Pain				Extent Activity Was Helpful in Easing Pain (1 = Not Helpful to 10 = Very Helpful)
	1 mo Post-tx		3 mo Post-tx		1 mo Post-tx		3 mo Post-tx
	TAU	Web-CBT	TAU	Web-CBT	TAU	Web-CBT	TAU	Web-CBT
Paced activities	67%	76%	69%	64%	6.3*	7.2*	5.7*	6.8*
Muscle relaxation	38%*	63%*	35%†	65%†	6.0	7.0	4.8†	6.8†
Controlled breathing	50%*	72%*	46%	58%	5.4	6.4	5.2	6.4
Focused attention	77%	80%	54%*	77%*	5.9	6.8	5.6*	6.9*
Physical activities	67%	74%	69%	71%	5.7*	6.8*	5.6*	6.8*
Recognized automatic thoughts	57%	56%	57%	54%	4.4†	6.1†	5.1*	6.5*
Challenged inaccurate thoughts	61%	71%	53%	63%	5.3*	6.4*	6.1	6.4
Assertive communication	65%	82%	69%	77%	6.4	6.4	6.2	6.4

Differences between conditions were assessed with t tests. Separate analyses were conducted for each of the two postintervention time points.

CBT = cognitive behavior therapy; COMM = Current Opioid Misuse Measure; MPI = Multidimensional Pain Inventory; TAU = treatment as usual.

^*Difference between conditions is significant at the P < 0.05 level.

^†Difference between conditions is significant at the P < 0.01 level.

Intervention

Treatment as Usual

Treatment as usual (TAU) consisted of the usual care provided to patients at the pain practice study site, which typically included opioid pharmacotherapy, along with other medications and medical interventions, such as nerve blocks and injections, as indicated. The practice’s multidisciplinary medical team included neurologists, anesthesiologists, nurse practitioners, and fellows in Pain Medicine. Psychological and behavioral treatment modalities were not provided at this practice. All patients were under the care of a physician pain specialist, who was responsible for prescribing and monitoring opioid therapy. The plan of care for each patient was developed by the physician, without input from study personnel.

TAU Plus Web-Based CBT

The development of the web-based Take Charge of Pain program followed an iterative process that integrated feedback from pain experts, pain medicine clinicians, and chronic pain patients. To ensure that the perspectives of chronic pain patients were well represented in this process, a series of focus groups were conducted to probe patients’ interest in and relevance of the planned content and solicit their input on the program’s structure and design. After a beta version had been developed, patients reviewed the draft program modules, completed knowledge pre- and post-tests, and provided structured feedback on domains such as acceptability, ease of use, and usefulness. Patient feedback confirmed and/or highlighted the salience of the planned content such as ambivalence about opioid treatment. A detailed description of the development of the web-based intervention, along with findings from this mixed-methods formative research, are reported in Moore et al. [29].

Based on CBT principles [30,31], the Take Charge of Pain program teaches patients strategies for restructuring dysfunctional thinking about pain and skills for coping with pain and reducing its impact on one’s life. The program’s 27 self-paced modules are housed within a home page designed to be accessed from a computer via the internet and take approximately 20–30 minutes to complete. Modules include didactic content and interactive exercises that illustrate common physical and psychological effects of chronic pain, as well as effective coping strategies. The modules utilize text, images, and animation to educate patients about pain and teach a variety of cognitive-behavioral skills, such as pacing activity, identifying and challenging automatic negative thoughts, controlled breathing, and muscle relaxation. The program also contains educational content about opioids, medication misuse, and strategies for improving medication management, including two dedicated modules each on “Myths and Facts about Opioids and Addiction” and “Identifying and Managing Triggers for Misusing Medication.” Modules are accessible in a fixed sequence so that core CBT skills for a chronic pain population (e.g., activity pacing, attention-diversion coping) are presented early in the sequence. To accommodate patients with lower literacy, all module text is accompanied by optional voiceover narration. Interactive features include an activity calendar with progress graphs and pain interference tracking. Participants were given a unique login name and password to access the program. A back-end system recorded usage data on a secure server, allowing research staff to track participants’ module completion. Participants’ access to the program was terminated at the end of the 12-week active intervention period.

At the end of the baseline visit, participants assigned to the TAU-plus-web-CBT condition were asked to complete the web-based program’s first introductory/training module at the study’s research office (so that research staff could be available to facilitate program access, answer questions, and provide technical assistance, as necessary). Thereafter, participants were advised to complete approximately two modules per week for the duration of the 12-week intervention (although this was merely a suggested schedule, and patients could complete the program at their own pace). Participants typically completed modules independently at home, but could opt to complete them in a private location at the study’s research office (located adjacent to the pain clinic) or in public libraries or other locations where they could access the internet with a desktop or laptop computer. Research staff provided regular phone and/or email prompts to remind participants to complete intervention modules (particularly participants who fell behind weekly benchmarks) and were available by phone, e-mail, and in-person for basic technical assistance as needed.

Data Analysis

Analyses included all randomized participants, consistent with an intent-to-treat approach, and, depending on the type of analysis, were performed using SPSS version 22.0 or SAS version 9.3 statistical software.

Each participant’s total fixed daily opioid dose at baseline was calculated by converting the milligram dosage of each fixed-dose opioid medication (excluding PRN medications, where present) into its morphine milligram equivalent dose (MED) and summing across all fixed-dose opioids prescribed. The conversion factors published by Von Korff et al. [32] were used to estimate MED for all opioid medications except tapendatol and buprenorphine. Other sources were used for MED conversions for tapendatol [33] and buprenorphine [34].

Descriptive statistics were prepared for the Feedback Surveys completed by web-CBT participants, including the mean score for each of the seven VAS items. Potential differences across conditions in the number of CBT activities performed postintervention and the perceived benefit derived from these activities were assessed with separate t tests for the one-month and three-month follow-ups.

Each primary and secondary outcome was evaluated in separate analyses using generalized linear mixed-effects piecewise regression models for repeated measures to examine treatment effects, time effects, and treatment-by-time interactions. Analyses for primary outcomes and pain catastrophizing examined effects observed during the 12-week intervention period (assessed at baseline and at weeks 4, 8, and 12); a separate set of analyses examined the durability or emergence of effects during the three-month postintervention follow-up period (assessed at one month and three months after completion of the intervention period). A mixed-effects model was also used to examine pain-related ED visits (and, for purposes of comparison, general ED visits); in this model, dispersion of visits was treated as a Poisson distribution because of the right-skewed (zero-inflated) nature of the data. An effect size index (Cohen’s d) was also calculated for each outcome measure at each postbaseline assessment point; this index was then adjusted to control for baseline levels of the variable [35].

Another set of analyses was used to assess potential differences across conditions in the number of pain-related ED visits reported by participants for the six months preceding baseline vs the six-month study period (i.e., the three-month intervention period plus the three-month follow-up period). First, to allow for a comparison between pre- and postintervention time periods of equivalent length, a composite postintervention variable was constructed by summing the number of pain-related ED visits reported by participants at each postbaseline assessment. Participants were then categorized according to whether they reported more, fewer, or the same number of pain-related ED visits postintervention than pre-intervention, and the proportions in each category across the two conditions were evaluated with a Mantel-Haenszel chi-square test.

Results

Participant Characteristics

Participants’ baseline demographic and selected pain-related characteristics are presented in Table 1. Overall, participants were racially diverse, mostly female, unemployed, and economically disadvantaged, with about half receiving public assistance and nearly 80% receiving Medicaid. Patients generally reported long-standing and severe chronic pain. More than two-thirds (69%) had experienced chronic pain for six or more years. On average, participants rated the severity of their worst pain in the past week as 8.45 on a 0–10 scale. Ninety percent were prescribed a fixed daily dose of opioid analgesics, reporting, on average, a high fixed daily dose (mean MED = 297 mg, SD = 545 mg, median = 124 mg). Additionally, 16% were prescribed PRN opioids, and 74% were prescribed nonopioid medications for pain, including mood stabilizers/anti-epileptics (37%), muscle relaxants (36%), and antidepressants (21%). Only a small minority (4.5%) reported any prior exposure to CBT for pain management.

Table 1.

Participants’ baseline demographic and pain characteristics

Characteristic*	Overall (N = 110), % (N)	TAU (N = 55), % (N)	Web-CBT (N = 55), % (N)
Age, M (SD), y	51.3 (10.9)	51.8 (12.2)	51.0 (9.4)
Gender
Male	36 (40)	38 (21)	35 (19)
Race†
White	45 (49)	49 (27)	40 (22)
Black/African American	35 (38)	29 (16)	40 (22)
Other	20 (22)	22 (12)	18 (10)
Ethnicity
Hispanic	20 (22)	22 (54)	19 (46)
Marital status
Married	18 (20)	16 (9)	20 (11)
Income sources (all that apply)‡
Public assistance	49 (54)	47 (26)	51 (28)
Disability	41 (45)	45 (25)	36 (20)
Retirement (e.g., social security, pension)	35 (38)	35 (19)	35 (19)
Employment	22 (24)	25 (14)	18 (10)
Friends/family	17 (19)	18 (10)	16 (9)
Health insurance (all that apply)‡
Medicaid	78 (86)	76 (42)	80 (44)
Medicare	29 (32)	33 (18)	25 (14)
Private insurance	15 (16)	16 (9)	13 (7)
Pain duration†
3 mo–5 y	31 (34)	44 (24)	18 (10)
6–10 y	28 (30)	22 (12)	33 (18)
>10 y	41 (45)	33 (18)	49 (27)
Pain surgery (≥1, lifetime)	51 (56)	52 (28)	51 (28)
Daily opioid dose, M (SD), fixed dose (MED)	297 (545); median = 124	251 (387); median = 113	343 (668); median = 154
Worst pain in past week (0–10) M (SD)	8.45 (1.43)	8.29 (1.55)	8.60 (1.30)
Visited emergency dept. for pain in past 6 mo	39 (43)	31 (17)	47 (26)

CBT = cognitive behavior therapy; MED = morphine milligram equivalent dose;TAU = treatment as usual.

^*There were no significant differences between groups for any variables.

^†One participant did not respond to this item.

^‡Percentages sum to >100 because participants could select multiple responses.

Participants’ mean baseline score on the MPI Pain Severity subscale was 4.67 (SD = 0.98, range = 2–6), and their mean MPI Pain Interference score was 4.74 (SD = 1.01, range = 2.2–6). Both of these scores are consistent with clinical norms established for the MPI in many chronic pain populations (e.g., [36,37]). On average, participants scored 15.7 on the COMM (SD = 7.57, range = 4–39) at baseline—above both the scale’s standard cut-point of 9 for detecting clinically significant ADRB established among chronic pain patients at a pain specialty clinic [8] and the higher cut-point of 13 suggested by Meltzer et al. [38] for detecting prescription drug use disorder based on a study of primary care patients with chronic pain at an urban safety net hospital. Participants’ average baseline pain catastrophizing score was 27.3 (SD = 12.2, range = 1–51), which again exceeds the clinical norm of 20.9 for a chronic pain patient population [39]. Thirty-nine percent of participants reported having visited an ED one or more times for a pain-related complaint in the six months prior to baseline. For all assessed baseline measures, there were no significant differences between the two treatment conditions.

Use of and Response to the Web-Based Program

The mean number of total modules completed was 19 (70%, SD = 10). Two participants opted to complete no additional program modules beyond the first introductory/training module (which all web-CBT-assigned participants were asked to complete at the end of the in-person basline visit). Usage metrics revealed a generally bimodal engagement pattern wherein a majority of participants successfully engaged with the program and tended to complete most or all modules; in contrast, a relatively small minority failed to engage with the program and completed few modules beyond the training module.

Participants’ mean VAS scores on feedback items indicated high program acceptability; on average, they rated the intervention 8.0 for “useful,” 8.0 for “interesting,” 8.0 for “new information learned,” 7.5 for “clarified misconceptions about chronic pain,” 6.8 for “clarified misconceptions about opioid medications,” 6.2 for “easy to undestand,” and 8.0 for overall satisfaction with the program.

Primary Outcomes

Participants’ mean scores at each assessment time point, by treatment condition, for MPI Pain Severity, MPI Pain Interference, COMM, PCS, and number of pain-related ED visits are presented in Table 2, along with effect size estimates. The statistical results of our mixed-effects outcome analyses are presented in Table 3.

Table 2.

Primary and secondary outcome measures, by treatment group, at each assessment time point

	Pain Severity (MPI) (0–6)			Pain Interference (MPI) (0–6)			Aberrant Drug-Related Behavior (COMM) (0–68)			Pain Catastrophizing Scale (0–52)			Number of Emergency Dept. Visits for Pain*
	M (SD)		Effect Size†	M (SD)		Effect Size	M (SD)		Effect Size	M (SD)		Effect Size	M (SD)		Effect Size
	TAU	Web-CBT		TAU	Web-CBT		TAU	Web-CBT		TAU	Web-CBT		TAU	Web-CBT
Baseline	4.73 (0.93)	4.61 (1.03)	—–	4.71 (0.94)	4.77 (1.07)	—–	15.22 (8.47)	16.11 (6.60)	—–	27.55 (13.23)	26.95 (11.15)	—–	0.73 (1.43)	1.16 (1.86)	—–
Week 4	4.61 (1.04)	4.69 (0.89)	0.143	4.49 (1.11)	4.55 (1.07)	−0.038	13.28 (7.26)	10.60 (5.85)	−0.505	25.58 (12.40)	23.37 (13.05)	−0.126	0.26 (0.71)	0.08 (0.27)	−0.413
Week 8	4.68 (0.98)	4.50 (1.15)	−0.069	4.44 (1.22)	4.34 (1.11)	−0.114	12.61 (7.71)	8.85 (5.79)	−0.590	25.35 (11.80)	18.92 (14.43)	−0.417	0.16 (0.54)	0.06 (0.24)	−0.325
Week 12	4.41 (1.01)	4.19 (1.01)	−0.125	4.32 (1.13)	4.28 (1.11)	−0.020	12.67 (7.88)	9.15 (5.99)	−0.539	24.12 (12.59)	18.87 (14.55)	−0.283	0.35 (1.73)	0.09 (0.35)	−0.354
1 mo post-tx	4.49 (1.30)	4.30 (1.14)	−0.025	4.43 (1.17)	4.29 (1.03)	−0.112	12.04 (7.35)	9.57 (6.63)	−0.393	24.21 (12.38)	20.87 (13.91)	−0.167	0.23 (0.85)	0.09 (0.46)	−0.360
3 mo post-tx	4.43 (1.22)	4.41 (1.09)	0.046	4.34 (1.29)	4.38 (1.01)	−0.003	11.38 (6.07)	9.10 (6.22)	−0.384	22.98 (12.94)	18.69 (13.27)	−0.275	0.25 (0.67)	0.10 (0.37)	−0.352

CBT = cognitive behavior therapy; COMM = Current Opioid Misuse Measure; MPI = Multidimensional Pain Inventory; TAU = treatment as usual.

^*At baseline, the number of emergency department visits for pain represents the number of visits in the past six months; at all other assessment points, it represents the number of visits in the past 30 days.

^†Effect size is the Cohen’s d for the difference in means between groups for each outcome measure at each assessment point, controlling for the corresponding baseline values. Negative values indicate a benefit for the treatment group (Web-CBT); positive values indicate a benefit for the control group (TAU). Effect sizes were computed using an online calculator available at https://www.campbellcollaboration.org/escalc/html/EffectSizeCalculator-SMD21.php and based on the methodology in Lipsey and Wilson [35].

Table 3.

Results of the outcome analysis using generalized mixed-effects piecewise regression

	In-Treatment Period		Contrast Between In-Treatment and Post-Treatment Periods
	Treatment Effect* (Additional Change in Web-CBT Group)	Time Effect† (Change in Both Groups)	Difference in Treatment Effect	Difference in Time Effect
Primary outcomes	Estimate (SE),‡P	Estimate (SE), P	Estimate (SE), P	Estimate (SE), P
Pain severity (MPI)	−0.03 (0.06), 0.547	−0.11 (0.04), <0.001	0.07 (0.09), 0.447	0.12 (0.05), 0.007
Pain interference (MPI)	−0.03 (0.05), 0.560	−0.13 (0.03), <0.001	0.04 (0.09), 0.681	0.15 (0.05), 0.001
Aberrant drug-related  behavior (COMM)	−1.26 (0.38), 0.001	−1.51 (0.19), <0.001	2.01 (0.66), 0.002	1.62 (0.33), <0.001
Secondary outcomes
Pain catastrophizing (PCS)	−1.08 (0.53), 0.040	−1.60 (0.26), <0.001	1.64 (0.90), 0.070	1.55 (0.45), 0.001
Pain-related emergency  dept. visits§	−0.70 (0.26), 0.007	−1.15 (0.19), <0.001	1.17 (0.44), 0.008	0.60 (0.25), 0.016

CBT = cognitive behavior therapy; COMM = Current Opioid Misuse Measure; MPI = Multidimensional Pain Inventory; TAU = treatment as usual.

^*Treatment effect is indicated by treatment type by time interaction—that is, the difference between treatment (Web-CBT) and control (TAU) groups in rate of improvement.

^†Time effect represents the aggregate for both treatment (Web-CBT) and control (TAU) groups.

^‡Negative values indicate a benefit for the treatment group (Web-CBT); positive values indicate a benefit for the control group (TAU).

^§Pain-related emergency department visits were assessed for the past six months at baseline and for the past 30 days at each subsequent assessment point.

A significant time effect was found for pain severity and pain interference, with patients in both conditions reporting significant reductions in baseline levels of these variables during the active intervention that were generally maintained in the postintervention period. However, no significant differences between treatment conditions were observed for either outcome. These results are visually depicted in Figure 2.

Figure 2.

Outcome means by treatment group over time. In all five graphs, the error bars represent two standard errors above and below the mean, equivalent to a 95% confidence interval. For emergency department visits for pain, baseline represents the number of visits in the past six months, and all other assessment points represent the past 30 days. CBT = cognitive behavior therapy; COMM = Current Opioid Misuse Measure; MPI = Multidimensional Pain Inventory; TAU = treatment as usual.

For ADRB, in contrast, a significant treatment-by-time effect was found, with patients assigned to use the web-based program reporting greater reductions in ADRB during the 12-week intervention than patients receiving treatment as usual (i.e., a 6.96-point reduction in mean COMM score vs a 2.55-point reduction, d = 0.539, P = 0.001) (Table 2). These reductions in ADRB among web-CBT participants were sustained during the three-month postintervention period, although TAU participants also reported reductions in the post-treatment period. As evident in Figure 2, the greatest reductions in web-CBT participants’ COMM scores occurred relatively early in the treatment period—that is, by the four- and eight-week time points.

Secondary Outcomes

A significant treatment-by-time effect was also found for pain catastrophizing; on average, participants in the web-CBT condition reported an 8.08-point reduction in baseline PCS score across the intervention period, as compared with a 3.43-point reduction reported by TAU participants (d = 0.283, P = 0.040) (Table 2). Again, the reductions in pain catastrophizing reported by web-CBT participants during the intervention were maintained at the three-month postintervention follow-up (Table 3 and Figure 2).

Additionally, the web-based intervention was associated with significant reductions in visits to the ED specifically for pain. Significantly more patients in the web-CBT condition (38%) than in the TAU condition (18%, P = 0.004) reported a reduction in the number of pain-related ED visits they made during the six-month study period relative to the six months before baseline. Similarly, results of the mixed-effects analysis indicate a significant treatment-by-time interaction for past-30-day pain-related ED visits, with web-CBT participants reporting a significantly greater reduction in such visits during the intervention period than TAU participants (d = 0.354, P = 0.007) (Table 3 and Figure 2)—a reduction that was maintained for three months post-treatment. In contrast, no significant treatment effect was found for the number of general ED visits.

Use of and Perceived Benefit from CBT Skills

At both postintervention follow-ups, patients assigned to the web-CBT condition were more likely than those assigned to the TAU condition to report engaging in—and benefiting from—most of the eight core CBT skills and activities described in the web-based program, such as controlled breathing, muscle relaxation, and focused attention (see Table 4). These between-group differences generally endured, or in some cases increased in magnitude, from the one-month to the three-month postintervention follow-up.

Discussion

Although no significant intervention effects were found for two of the study’s three primary outcomes—pain severity and pain interference—participants who used the Take Charge of Pain program had less ADRB, experienced less frequent catastrophic thinking about pain, and had fewer pain-related visits to the ED. Effect sizes for the web-based intervention were generally small to medium, with the strongest effects observed for ADRB. These positive outcomes were sustained after completion of the web-based program, and web-CBT participants were more likely to report practicing and benefiting from CBT activities postintervention than participants assigned to standard treatment alone.

To our knowledge, this is the first randomized controlled trial of an evidence-based, technology-delivered intervention for co-occurring chronic pain and ADRB in the research literature. Although the number of technology-based self-management interventions targeting chronic pain is growing, and several have recently demonstrated efficacy in improving outcomes such as pain intensity, pain interference, and pain catastrophizing [22,40–43], these tools do not target co-occurring chronic pain and ADRB. Given increasing recognition of the importance of risk management during opioid therapy for pain, chronic pain patients with ADRB represent an underserved and vulnerable group who are at high risk of termination from pain treatment [7,3]. A widely accessible self-management intervention has the potential to assist clinicians who must address ADRB in a clinically appropriate manner and support patients as they work to avoid negative outcomes, better manage their use of opioid medications, and improve their experience of living with chronic pain.

The study’s focus on a clinical population of opioid-treated pain patients sets it apart from much of the extant literature on technology-based interventions for chronic pain, most of which reports studies of community-based samples recruited online or from primary care settings [20]. Because community samples tend to include individuals with lower problem severity than clinical samples and, in the case of internet-recruited samples, may overrepresent individuals who are especially motivated for behavior change [21], it is important that these tools also be evaluated in clinical populations of pain patients. Moreover, participants in the present study represent a notably high-risk subpopulation of chronic pain patients who were receiving treatment at a pain specialty practice and met criteria for ADRB. Most were economically disadvantaged, and, consistent with other studies of pain specialty patients [44], had long-standing, severe chronic pain, as well as high rates of physical and psychological comorbidities and medical service utilization (e.g., surgeries, ED visits).

Although the positive outcomes produced by the Take Charge of Pain program may not be generalizable to the broader population of chronic pain patients, particularly those with less severe chronic pain or pain of more recent onset, or patients in primary care settings, the ability to effect change among patients so deeply affected by chronic pain and its associated biopsychosocial impairments is promising. One study of an online CBT program for chronic back pain [45], which found that participants recruited online but not those recruited from pain clinics reported reduced pain, speculated that self-management interventions of this type may be generally less effective when the pain syndromes are relatively severe or complex. Our finding of positive outcomes for the Take Charge of Pain program in a sample of patients with severe and complex pain problems suggests that the intervention might have even greater effectiveness if delivered to populations with less severe or longstanding pain. In addition, intervention effectiveness could potentially be strengthened by allowing patients longer-term access to the web-based program; three months may be a short window in which to effect improvement in a complex condition that, for nearly 70% of participants, had endured for six or more years. As the effectiveness of CBT is predicated on an individual’s mastery of new skills and internalization of new patterns of thinking and behavior, pain patients may also benefit from having more time to practice these skills and learn how to deploy them in a broad range of real-life situations.

The Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) recommends that clinical trials of chronic pain treatments consider not just the statistical significance of findings but also the clinical relevance of any improvements in outcome measures to evaluate whether the observed changes are likely to meaningfully affect patients’ experience of their condition [46]. Although IMMPACT’s Consensus Statement does not suggest a specific benchmark for changes in ADRB, it is important to note that web-CBT participants on average reported a 43% reduction in COMM score from baseline to the end of the 12-week intervention (from 16.11 to 9.15), a medium-size effect that was sustained for three months postintervention (9.10 at three-month follow-up). This decrease dropped them below the threshold of 13 suggested by Meltzer and colleagues [38] as indicative of prescription drug use disorder and is on the cusp of the scale’s official cut-point of 9 denoting “clinically significant” ADRB. In contrast, the COMM scores of TAU participants reduced by only 25% over the intervention and follow-up period and remained well above the threshold for clinically significant ADRB at all assessment time points.

The finding of no significant treatment effect for the web-based intervention in reducing pain severity is not uncommon for self-management/CBT interventions for chronic pain, both technology-delivered and traditional [45]. In fact, Ballantyne and Sullivan [47] have questioned whether pain intensity is the best metric by which to judge the success of behavioral treatments for chronic pain. As they argue, the degree of suffering or distress experienced as a result of chronic pain is not synonymous with pain intensity and may be more a relevant metric from both a phenomenological and a clinical perspective. Moreover, distress or suffering associated with chronic pain can be improved even in the absence of a reduction in pain intensity. In general, pain reduction is not a primary focus for CBT studies. Ballantyne and Sullivan explain that behavioral treatments, such as CBT interventions that explicitly target dysfunctional thoughts and behaviors associated with pain and aim to improve an individual’s ability to effectively cope with pain, are better suited to reducing pain-related distress than reducing pain intensity. In this light, it is promising that patients assigned to use the Take Charge of Pain program in conjunction with standard pain treatment reported significantly greater reductions in pain catastrophizing than patients assigned to standard treatment alone. Pain catastrophizing may have functioned as a mediator, predicting reductions in ADRB, but exploration of this possibility was beyond the scope of the current paper; results of the study’s mediational analyses will be reported separately. As a final note, the finding that the web-based program exerted a selective effect on ED visits for pain (and presumably, the pain complaints that prompted these visits), but did not appear to affect ED visits for other health problems, further supports the program’s effectiveness and suggests that the positive outcomes observed in this trial were a function of the specific content of the intervention.

Limitations

The present findings should be interpreted in light of several limitations. First, it should be noted that, in addition to catastrophizing and pain-related ED visits, several other secondary outcomes were assessed, but are not reported here. With regard to outcome measurement, all variables were assessed by patient self-report only; therefore, the accuracy of data on ADRB and ED visits, in particular, may be limited by poor recall, social desirability, and other sources of potential response bias. (This is less of a concern for the inherently subjective constructs of pain severity, pain interference, and catastrophizing.) Future trials to evaluate Take Charge of Pain and other self-management interventions for chronic pain patients could enhance rigor by cross-validating self-report data, where appropriate and feasible, with objective measures (e.g., urine drug screening) or other data sources (e.g., medical records, physician report). Several additional constructs that may have improved as a result of the skills training provided in the web-based program, such as pain-related distress, self-efficacy to manage pain, pain coping and global assessment of change, were not assessed. As noted above, the present findings may not generalize to other chronic pain populations. Participants were opioid-treated chronic pain patients receiving treatment at a tertiary pain program who screened positive for ADRB. Because these patients are likely to have more severe and complex pain syndromes than patients seen in less specialized settings such as primary care, study results cannot be generalized to the broader population of opioid-treated chronic pain patients. A further limitation relates to the duration of the observed intervention effects. Although the reductions in ADRB, pain catastrophizing, and pain-related ED visits reported by Web-CBT participants during the 12-week intervention were maintained for three months postintervention, the duration of these gains beyond the study period is unknown. At present, very few evaluations of technology-based behavioral interventions for chronic pain monitor patients for longer than three months postintervention, but longer-term studies should be conducted to further investigate the likely duration of treatment effects and evaluate various means of sustaining and enhancing gains (e.g., periodic booster sessions, regular updates of online content, incorporation of peer support elements or personalized feedback from clinicians).

Conclusion

The findings of this randomized controlled trial demonstrate that an internet-delivered self-management intervention, when delivered as an enhancement to traditional specialty pain teatment, can effectively improve some pain-related outcomes and the occurrence of opioid misuse in a population of opioid-treated chronic pain patients who were engaging in ADRB. Given the extremely high prevalence of chronic pain in the United States and worldwide [48,2], along with increased recognition of the limitations and risks of long-term opioid therapy, there is an urgent need for effective, evidence-supported behavioral therapies addressing chronic pain and misuse of opioid medication. The present results suggest that technology-based self-management interventions may provide a promising and practical means to increase widespread access to effective behavioral treatment for chronic pain and help reduce problems associated with medication misuse among opioid-treated pain patients.

Funding sources: This research was supported by a grant from the US National Institute on Drug Abuse (NIDA) of the National Institutes of Health (R01DA026887) to Andrew Rosenblum and Lisa A. Marsch. Honoria Guarino received a pilot grant from the Center for Technology and Behavioral Health at Dartmouth College, a NIDA-funded P30 Center, to conduct a qualitative process research component of this parent study.

Disclosure and conflicts of interest: During the past 3 years, Dennis C. Turk has consulted for Pfizer, Nektar, Develco, Ironwood, GlaxoSmithKline, Mallincrodt, Orexo, and Xydnia, and Russell K. Portenoy’s organization has received research grants from AstraZeneca and Pfizer. Lisa A. Marsch is affiliated with HealthSim, LLC, the business that developed the web-based intervention platform used in this study. This relationship is extensively managed by Dr. Marsch and her academic institution. There are no other potential conflicts of interest to report.