Cognitive Therapy, Mindfulness-based Stress Reduction, and Behavior Therapy for the Treatment of Chronic Pain

Randomized Controlled Trial

John W. Burns; Mark P. Jensen; Beverly Thorn; Teresa A. Lillis; James Carmody; Andrea K. Newman; Francis Keefe


Pain. 2022;163(2):376-389. 

In This Article


Most studies of psychosocial treatments for chronic pain have compared their efficacy with largely inert control groups. The result has been a plethora of evidence indicating that people with chronic pain benefit from such treatment approaches. One implicit goal of developing new approaches to treatment is to have the beneficial effects of the new treatments exceed the benefits of the extant treatments. To this end, fewer studies have compared the effects of 2 or more active treatments; those that have compared active treatments have found, by and large, that different psychosocial treatments for chronic pain produce similar effects on primary outcomes. This observation is based on the results of both underpowered and adequately powered comparison studies. Consistent with this prior research, and in an adequately powered study, we found that CT, MBSR, and BT produced similar pretreatment to posttreatment effects on primary outcomes and revealed similar levels of maintenance of treatment gains at a 6-month follow-up. Proposing a new metric for determining relative treatment superiority, we also examined whether one or more of the active treatments would produce treatment benefits more rapidly than the other treatment(s). Although we did find that the 3 treatments differed significantly from TAU on average by session 6, this effect emerged across all 3 active treatments. In summary, not only did CT, MBSR, and BT produce similar levels of improvements and maintenance of gains on 5 outcome domains, they did so at approximately the same rates.

Treatment Condition × Time interactions for the 8 session assessments revealed that CT, MBSR, and BT showed greater improvements during treatment than TAU on all 5 outcome domains, and the active treatment conditions did not differ significantly from each other on any outcomes at posttreatment. At the same time, pretreatment to posttreatment within treatment effect sizes ranged from Cohen's d = 0.15 to 0.37, and posttreatment effects comparing TAU with the other groups ranged from Cohen's d = 0.21 to d = 0.61. These results suggest that all 3 treatments produced on average small to medium effect size improvements on the 5 outcome variables. Regarding clinically meaningful changes, we found that 25% to 33% of participants receiving CT, BT, and MBSR reported pain intensity reductions of 30% or greater, whereas, on average, only 7% of them reported meaningful increases in pain intensity. Moreover, Treatment Condition × Time interactions for the posttreatment to 6-month follow-up epoch were nonsignificant, as were all but one of the within-treatment effects of time. These results indicate that the pretreatment to posttreatment improvements for all 3 conditions were maintained, with the single exception of pain interference for MBSR. These results are consistent with other well-powered comparison studies,[7,25,37] underscoring the pervasive findings that (1) different psychosocial pain treatments produce similar levels of clinically meaningful outcomes on average and (2) psychosocial pain treatments maintain these improvements, for the most part, after treatment. It should be noted, however, that a recent meta-analysis[44] found very small benefits in pain reduction for CBT compared with active control groups, such as pain education, among a collective N of 3235 research participants.

Condition × Time interaction analyses for the 8 sessions also allowed us to analyze regions of significance. These regions identified the approximate treatment session at which the CT, MBSR, and BT groups began to differ significantly from TAU on each of the 5 outcome factors. Put otherwise, we were able to determine the relative rates at which improvement occurred. There was notable variability in these regions between the conditions and among the outcome factors. Indeed, participants in the BT and MBSR conditions differed significantly from the TAU group on sleep disturbance by session 2; this without any explicit therapeutic attention devoted to easing such problems. Across outcome factors, we found that the 3 active treatments began to differ significantly from TAU on average by sessions 5 or 6. On one level, this finding again highlights the similarities of ostensibly different treatments. Here, distinct treatments showed similar rates of improvements. On another level, this finding regarding the sixth session hints at the possibility that at least some psychosocial treatments for chronic pain may be reduced in length (eg, from 8 to 6 sessions) with only small reductions in overall efficacy. Recent work also supports the possibility of reducing treatment length.[3,4] Still, it should be considered that all 3 active treatments had additional benefits past session 6.

On a third level, the finding regarding the sixth session may invite us to entertain questions about what it is that we are actually doing in these treatments. It is possible that 6 sessions worth of any psychosocial treatment may be necessary but that the particular therapeutic content of these 6 sessions may not be a critical determinant of outcomes. The specific techniques included in CT, MBSR, and BT may be less important for outcomes than people participating in any techniques rooted in these empirically supported psychosocial treatments for chronic pain.

We can advance at least 2 reasons to explain the bulk of our and past findings. First, it is possible that many psychosocial treatments for chronic pain are all viable treatment options by virtue of sharing positive effects on outcomes. Thus, people with chronic pain can choose from an array of efficacious approaches. However, the quandary of equivalent outcomes leaves people to choose from a field of winners with no clear criteria by which to select a treatment approach. Second, despite different theories and different techniques, the equivalence of outcomes suggests the possibility that extant approaches may be just so potent and no more. We could therefore choose to abandon the search for new and better treatments that developing new treatments that are unlikely to be better treatments is not a valuable way to spend time, energy, and scarce resources. We could rest on our laurels by relying on CBT or other established psychosocial treatments as the best that we can do. However, we believe this conclusion is premature.

One approach is to identify potential treatment moderators.[10] Investigators have noted that focusing only on mean levels of treatment response misses the potential information reflected in variability in treatment response. Put simply, people respond to treatments with different levels of improvement varying around the sample mean from low to high. It may be the case that people characterized by certain factors may respond more or less favorably to certain treatments. Thus, treatment response to, for example, CBT could be magnified by matching people with CBT-relevant characteristics to CBT and not to another treatment. A second approach to help understand the implications of similar outcomes across different treatments is to determine if they operate through different mechanisms, even when they are shown to have the same or similar effects on outcomes.[11,19] If different treatments are found to operate by different underlying mechanisms, then this information could be used to inform algorithms for better patient-treatment matching. We plan to examine questions of mechanisms and moderators using the data from this trial in future reports.

Some limitations should be delineated. First, we effectively dismantled CBT into CT and BT treatments, thus potentially weakening the component effects. CBT was divided in this way because we wanted to separate, as best we could, cognitive and behavioral therapeutic techniques to examine the potentially specific effects of cognitive and behavioral mechanisms. Information about and analyses of mechanisms will be conveyed in future reports. Second, the mean pain intensity score at baseline was 5.55/10 (SD = 2.11), the mean pain interference score at baseline was 3.68/6 (1.27), and the mean CESD score was 9.48 (5.81). Based on 6 recent RCTs of psychosocial chronic treatments,[7,21,25,26,36,37] pain intensity ratings at baseline ranged from 5.4/10 to 6.4/10 with a mean of 5.9. Pain interference ratings ranged from 3.5/6 to 4.8/6 with a mean of 4.22. Finally, CESD scores ranged from 10.83 to 12.98 with a mean of 11.9. Thus, our sample at baseline reported, on average, lower levels of pain intensity, interference, and depressive symptoms than participants in these other studies, and results should be interpreted in this light. Third, this and other recent comparative outcome trials have focused on CBT and MBSR; treatments which may have a great deal of theoretical and procedural overlap. Future comparative trials may need to focus on treatment approaches that make different assumptions about principles and processes of change, such as emotional awareness and expression therapy.[26] Fourth, we did not assess the degree to which participants adhered to the different skills they had learned during treatment. Thus, we do not know whether or how much the treatment sessions actually altered pain coping activities, broadly speaking, as practiced in the daily lives of the participants.

In sum, results of this study are consistent with recent comparative outcome studies. Cognitive therapy, MBSR, and BT all produced moderate improvements to approximately the same degree, and we add in this report, at approximately the same rate. Despite differences in theories, proposed mechanisms and therapeutic techniques, the persistent finding of equivalent outcomes suggests the possibility that these differences may be merely apparent. Although we are left with a number of viable psychosocial treatments, at bottom, there is clearly a problem in not having produced at least one superior treatment. As we suggest, to move beyond equivalence, we may need to adopt new research strategies and analytic methods that expand our understanding of possible differences between treatments beyond outcomes alone.