Association Between Psychological Interventions and Chronic Pain Outcomes in Older Adults: A Systematic Review and Meta-analysis

Key Points

Question  Do older adults with chronic pain benefit from psychological therapies?

Findings  In this systematic review and meta-analysis including 22 studies with 2608 participants, psychological interventions that used cognitive behavioral therapy modalities were associated with statistically significant benefits in terms of reduced pain and catastrophizing beliefs as well as improved self-efficacy for managing pain. Benefits were small and documented at the time of treatment completion; with the exception of pain reduction, evidence is lacking for the persistence of observed benefits in other assessments conducted up to 6 months later.

Meaning  Among older adults with chronic pain, psychological therapies have a small, but statistically significant, benefit for reducing pain and catastrophizing beliefs and improving self-efficacy for managing pain.

Importance  Chronic noncancer pain (hereafter referred to as chronic pain) is common among older adults and managed frequently with pharmacotherapies that produce suboptimal outcomes. Psychological treatments are recommended, but little information is available regarding their efficacy in older adults.

Objective  To determine the efficacy of psychological interventions in older adults with chronic pain and whether treatment effects vary by participant, intervention, and study characteristics.

Data Sources  MEDLINE, Embase, PsycINFO, and the Cochrane Library were searched from inception to March 29, 2017.

Study Selection  Analysis included studies that (1) used a randomized trial design, (2) evaluated a psychological intervention that used cognitive behavioral modalities alone or in combination with another strategy, (3) enrolled individuals with chronic pain (pain ≥3 months) with a sample mean age of 60 years or older, and (4) reported preintervention and postintervention quantitative data.

Data Extraction and Synthesis  Two of the authors independently extracted data. A mixed-model meta-analysis tested the effects of treatment on outcomes. Analyses were performed to investigate the association between participant (eg, age), intervention (eg, treatment mode delivery), and study (eg, methodologic quality) characteristics with outcomes.

Main Outcomes and Measures  Pain intensity was the primary outcome; secondary outcomes included pain interference, depressive symptoms, anxiety, catastrophizing beliefs, self-efficacy for managing pain, physical function, and physical health.

Results  Twenty-two studies with 2608 participants (1799 [69.0%] women) were analyzed. Participants’ mean (SD) age was 71.9 (7.1) years. Differences of standardized mean differences (d_D) at posttreatment were pain intensity (d_D = −0.181, P = .006), pain interference (d_D = −0.133, P = .12), depressive symptoms (d_D = −0.128, P = .14), anxiety (d_D = −0.205, P = .09), catastrophizing beliefs (d_D = −0.184, P = .046), self-efficacy (d_D = 0.193, P = .02), physical function (d_D = 0.006, P = .96), and physical health (d_D = 0.160, P = .24). There was evidence of effects persisting beyond the posttreatment assessment only for pain (d_D = −0.251, P = .002). In moderator analyses, only mode of therapy (group vs individual) demonstrated a consistent effect in favor of group-based therapy.

Conclusions and Relevance  Psychological interventions for the treatment of chronic pain in older adults have small benefits, including reducing pain and catastrophizing beliefs and improving pain self-efficacy for managing pain. These results were strongest when delivered using group-based approaches. Research is needed to develop and test strategies that enhance the efficacy of psychological approaches and sustainability of treatment effects among older adults with chronic pain.

Chronic noncancer pain (hereafter referred to as chronic pain) is one of the most common conditions encountered by health care professionals.¹ Chronic pain is particularly common among individuals aged 60 years or older and is associated with substantial disability and health care costs.^1-5 Among older adults, management of chronic pain is complicated by age-related physiologic changes, competing comorbidities that limit treatment options, patient barriers (eg, fear of deleterious adverse effects from medications),⁶ health care professional barriers (eg, lack of knowledge),⁷ large adverse effect profiles of commonly administered pharmacologic therapies,⁸ and a limited evidence base to guide treatment.⁵

Given these limitations and concerns, as well as the ongoing opioid epidemic,⁹ nonpharmacologic therapies that use cognitive behavioral therapy (CBT) approaches, including cognitive and behavioral coping skills training, cognitive restructuring, and behavioral activation techniques, have received increased attention as treatments for individuals with chronic pain.^10-19 Cognitive behaviorally based approaches have been shown to have small but statistically significant associations with pain, mood, and disability in nonelderly adults with chronic pain.¹⁰ In recent years, the number of studies investigating nonpharmacologic approaches for the treatment of chronic pain in older populations is expanding. One quantitative review examined the effectiveness of psychological approaches for chronic pain in older adults in studies published between January 1975 and March 2008.¹¹ This review, which included 12 clinical trials and 5 uncontrolled pilot studies, found that psychological approaches were moderately effective in reducing pain but did not have a significant effect on depressive symptoms, physical functioning, or pain medication use.¹¹

Given the increasing interest in nonpharmacologic modalities and the expanding number of published studies in this area over the past decade, we conducted a comprehensive systematic review and meta-analysis to evaluate the efficacy of CBT-based approaches for chronic pain among older adults. As a secondary aim, we conducted preplanned analyses to determine whether specific participant (eg, pain type), intervention (eg, mode of therapy delivery), and study (eg, methodologic quality score) characteristics moderated treatment outcomes.

This systematic review and meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement.²⁰ Comprehensive searches of MEDLINE (Ovid), Embase (Ovid), PsycINFO (EBSCOhost), and the Cochrane Library (Wiley) were performed to identify clinical trials that evaluated 1 or more psychological intervention for the treatment of chronic pain in older adults published from database inception through June 28, 2016. An updated search was performed (from June 29, 2016, through March 29, 2017) to identify new publications. Subject headings and keywords included chronic pain, noncancer pain, musculoskeletal pain, osteoarthritis, arthritis, rheumatoid, back pain, mindfulness, cognitive therapy, behavioral therapy, cognitive-behavioral, CBT, older adults, and elderly. The BMJ Clinical Evidence Randomized Controlled Trial Strategy filter²¹ was applied to the MEDLINE and Embase searches. We used a combination of keywords and subject headings representing trials in the PsycINFO and Cochrane searches. Additional studies were identified by reference searching in Scopus using the “cited by” and “view references” features. The MEDLINE search strategy is provided in eTable 1 in the Supplement.

We included studies if they (1) used a randomized clinical trial design; (2) evaluated CBT techniques (see Table 1 for details) as a stand-alone treatment or in combination with another strategy (eg, exercise); (3) enrolled participants with chronic pain, defined as pain lasting 3 months or longer at the time of enrollment; (4) focused on older individuals as reflected by a sample mean age of 60 years or older; and (5) reported preintervention and postintervention quantitative data (ie, means and SDs or SEs) for each group for each assessment (or data for baseline and change scores). We excluded studies that targeted patients with pain due to cancer or chronic headache to be consistent with a Cochrane review on this topic,¹⁰ did not report full quantitative results on outcomes and were not willing or able to share their data upon request, were published in languages other than English, or were not published in a peer-reviewed journal.

After duplicates were removed, 2 of us (B.N., R.B.) screened titles and abstracts of the identified searches independently followed by a full-text inspection of potentially eligible articles to determine eligibility, with disagreements resolved by consensus. The study selection process appears in eFigure 1 in the Supplement.

Five studies used 3 arms.^22-26 To ensure comparability across all studies, we included 1 intervention and 1 control group from these 5 studies. Two of the studies used 2 control groups.^22,23 We included the active control group in both studies. One study evaluated 2 intervention modalities (ie, a physical therapist delivered pain-coping skills training vs physical therapist delivered pain-coping skills training along with an exercise training component) and 1 control group.²⁴ We pooled the data from the 2 intervention groups. One study reported 2 intervention groups, including a group-based intervention and an individual-based intervention; posttreatment outcomes were reported as pooled data from the 2 intervention groups.²⁵ Finally, 1 study used 2 intervention groups: 1 delivered CBT for pain and insomnia and the other delivered CBT for pain only. We extracted data from the CBT for pain-only group.²⁶

The research team extracted data on outcomes that prior research has shown to be positively affected by psychological therapies^{10,11,15,16,23} as well as outcomes that were assessed in a minimum of 4 studies in the sample. We abstracted data on 8 outcomes in 3 domains: pain (pain intensity, pain-related interference), psychological (depressive symptoms, anxiety, catastrophizing beliefs, and self-efficacy for managing pain), and functional (self-reported physical function and physical health). For studies that used more than 1 pain intensity measure (eg, current pain, average pain), we used the mean of the measures as the outcome. Finally, we extracted data on 2 additional outcomes given their importance—change in pain medication use and adverse events.²⁷Table 2 reports the outcomes assessed in each study.

For each outcome, mean and SD data were extracted for intervention and control groups at pretreatment and posttreatment, as well as for all other follow-up assessments. We categorized any assessment that took place less than 12 weeks after the completion of treatment as posttreatment. Due to study differences in the follow-up assessments, we operationalized any assessment occurring between 12 or more weeks to 24 or fewer weeks after treatment completion as a mid-term outcome, while those taking place longer than 24 weeks after treatment completion were considered long-term outcomes.

Data extraction included bibliographic information, demographic characteristics of the sample and clinical characteristics when present, and data on the intervention, as well as outcome data. Two of us (B.N., R.B.) performed double data entry independently, and the resulting databases were then compared with each other. Discrepancies were resolved through consensus.

To judge the methodologic quality of the retained articles, we used the quality rating scale developed by Yates et al,⁴⁶ a valid and reliable instrument designed to evaluate the quality of randomized clinical trials examining psychological interventions in individuals with chronic pain. Total scores range from 0 to 35, with higher scores indicating better methodologic quality.⁴⁶ Two of us (B.N., E.K.) performed this assessment independently; discrepancies were resolved through discussion. We did not exclude any study from the analysis based on quality score but examined in moderator analyses whether the quality score affected treatment outcomes. We also assessed the risk of methodologic bias (appropriateness of randomization, allocation bias, and measurement bias) using the Yates tool.

Meta-analyses were carried out in statistical mixed models. The dependent variables were the standardized mean differences over time for control and intervention (outcome mean differences divided by the SD of the difference). The primary model included treatment (control vs intervention), time of assessment (a repeated measure: baseline, first follow-up [treatment completion], mid-term follow-up, and long-term follow-up) as fixed classification factors, the interaction between these factors, and studies as levels of a random classification factor. An unstructured error was specified. Random effects take into account heterogeneity among studies.

The effect of the intervention on study outcomes was examined by the treatment × time interaction in this model and the treatment effect specific to each of 3 time contrasts: baseline to first follow-up (treatment completion), baseline to mid-term follow-up, and baseline to long-term follow-up. The baseline vs first follow-up contrast was the primary outcome of interest because all studies reported outcomes at this assessment. We carried out further examination of effects at the later time points by models that looked at the baseline vs first follow-up limited to the studies that provided data at later assessments.

Results are reported in terms of differences of standardized mean differences (d_D) because of the numerous measurement instruments and scale ranges used by the studies for each outcome. As a guide to the magnitude of the treatment effects, we show what d_D represents for several outcome scales.

Additional independent variables were examined, including study characteristics (study quality [high vs low], year of publication, and pilot study vs larger-scale randomized clinical trial), intervention characteristics (mode of treatment delivery [group vs individual]), level of therapist training (evidence that therapists had appropriate training in intervention components prior to the trial: adequate vs inadequate), treatment fidelity (adherence to the therapist manual: adequate or inadequate, mode of therapy [group vs individual], and duration of the intervention phase in weeks), and participant characteristics (pain type [musculoskeletal vs other], proportion of women in the sample, mean age of sample, and pain duration in years). Each of these variables was added to the primary model (as a fixed classification factor for categorical variables and as a covariate for quantitative variables; separate models for each variable) as well as its interaction with treatment and time. To examine whether, for example, specific study-level methodologic characteristics moderated treatment effects—whether effects were stronger for or limited to certain levels of these characteristics—the focus was on the interaction with treatment (overall and for specific time contrasts), including examination of homogeneity of regressions for the covariates.⁴⁷

In this type of meta-analysis, it is clear that an assumption of studies as fixed (a single true effect size for all studies) is inappropriate. True effect sizes will vary by studies not just owing to sampling error but also to differences in sample composition (eg, age, ethnicity, and educational level), methods of assessment and study protocol, variable definitions, overall study quality, and numerous other factors. We used mixed models in which studies are assumed to be random (sampled from a population of studies). Effect sizes are assumed to differ by studies.

For the sake of completeness, we computed the Cochran Q statistic and Higgins-Thompson H² and I² values to examine heterogeneity across studies.⁴⁸ We examined the question of publication bias by constructing a funnel plot with 1/(SE), a measure of sample size, plotted against effect size.⁴⁹

The database searches identified 2391 articles; 238 were selected based on title and abstract for full-text review to determine eligibility. We included 22 studies (23 publications) with a total of 2608 participants (mean [SD] age was 71.9 [7.1] years and 1799 [69.0%] were women) in the final sample (Table 2).^22-26,28-45

Table 2 shows studies that evaluated participants with various types of chronic pain, including back pain, pain due to osteoarthritis or rheumatoid arthritis, and mixed pain types. We categorized types of chronic pain into 2 groups: musculoskeletal (17 studies) and other (5 studies). The other category included patients with rheumatoid arthritis (n = 1) and mixed pain types (n = 4). The mean (SD) pain duration was 16.1 (13.9) years.

The mean (range) length of the intervention period was 9.4 (4-35) weeks. The mean (range) number of treatment sessions was 8.4 (6-14). Fifteen studies delivered an intervention using a group-based approach, and the most common mode of treatment delivery was in person, which was used by 19 studies (Table 2).

eTable 2 in the Supplement presents the Yates quality scores for all 22 studies. The mean (range) quality score was 24.5 (13-33). In the Yates et al⁴⁶ study, articles with a score or 22.7 or greater were deemed to have excellent methodologic quality. Twelve (55%) studies met the criterion for taking steps to minimize the possibility of measurement bias, while 10 (45%) were judged to be at low risk for allocation bias (eTable 2 in the Supplement). The funnel plot showed no clustering of studies in the lower right of the funnel that would indicate lack of publication of smaller or nonsignificant studies (eFigure 2A, B, and C in the Supplement).

The Cochran Q and I² scores for the key variable—pain intensity—were 25.9% and 27.6%, which did not indicate a high degree of heterogeneity. Other outcomes showed similarly modest heterogeneity.

Six studies assessed for change in pain medication use but used different scales that precluded generation of a summary effect size.^22,34-37,44 None of the studies reported any treatment-related reduction in pain medication use, including opioid use. The results of the meta-analyses appear in Table 3. Quiz Ref IDDifferences of standardized mean differences (d_D) and corresponding P values for the outcomes at posttreatment are pain intensity (d_D = −0.181, P = .006), pain interference (d_D = −0.133, P = .12), depressive symptoms (d_D = −0.128, P = .14), anxiety (d_D = −0.205, P = .09), catastrophizing beliefs (d_D = −0.184, P = .046), self-efficacy for managing pain (d_D = 0.193, P = .02), physical function (d_D = 0.006, P = .96), and physical health (d_D = 0.160, P = .24). The d_D terms reported above (for pain, catastrophizing, and self-efficacy) correspond to a baseline to posttreatment reduction in pain intensity on a 0 to 10 scale of 0.49, in catastrophizing on a 0 to 6 scale of 0.32, and an improvement in self-efficacy on a 0 to 60 scale of 4.11 points.

Quiz Ref IDThe treatment result for pain intensity persists up to 6 months after treatment completion (d_D = −0.251, P = .002). There is no evidence of treatment results persisting in assessments conducted greater than 24 weeks after treatment completion, but this evaluation is confounded by the small number of studies with data at that assessment and the mixed results of models using only those studies in baseline vs posttreatment comparisons.

The Figure displays a forest plot showing the effect size for each study and corresponding weight given to each for pain intensity at posttreatment, the key contrast of this study. The pooled analysis indicates a significant benefit in favor of treatment relative to controls.

Quiz Ref IDOnly 3 studies assessed for adverse events.^24,25,40 Two found transient increases in pain associated with an exercise and behavioral skills training protocol,^24,25 while the third reported no serious adverse events associated with a mindfulness meditation-based intervention.⁴⁰

To determine whether treatment effects differed by level of the potentially moderating variables, we examined 11 independent variables, 1 at a time, as additions to the primary model, including their interactions with treatment and time of assessment. These variables were conceptualized in 3 areas: participant characteristics, intervention characteristics, and study characteristics. Across all outcomes and possible moderators, only mode of therapy showed a coherent pattern of results. Other moderators were nonsignificant, and there were no indications of negative results for any subgroup. For the majority of outcomes, including pain intensity, treatment differences were stronger for or limited to group therapy for the baseline to postintervention comparison (for pain intensity, d_D = −0.202, P = .008 for group therapy; d_D = −0.120, P = .38 for individual therapy). However, this interaction was not significant for pain intensity at mid-term follow-up (P = .01 for group, P = .07 for individual). Too few studies conducted long-term follow-ups for a meaningful examination of moderation at this time point.

Psychological therapies for individuals with chronic pain have received increased attention in the wake of the ongoing opioid epidemic in the United States.^9,50 Various initiatives have been launched to address the opioid crisis, including the release of the Centers for Disease Control and Prevention opioid guidelines for patients with chronic pain.⁵¹ The Centers for Disease Control and Prevention guidelines encourage clinicians to prescribe nonpharmacologic therapies, such as CBT, for patients with chronic pain. Our results are relevant to the management of chronic pain in older adults by demonstrating that psychological interventions have salutary, albeit small, benefits for treatment of pain, catastrophizing beliefs, and self-efficacy. Mean treatment results demonstrated in the present study obscure variations at the individual patient level. Some older patients with chronic pain may receive substantial benefit through psychological therapy, while others may not benefit. There is no evidence that the beneficial results identified at the completion of treatment persisted up to 6 months for outcomes other than pain reduction. There were too few studies reporting long-term outcomes to determine completely whether this finding was due to decreased power or to a tapering of treatment benefits over time.

Quiz Ref IDThe observed benefits were strongest when delivered using group-based approaches. Potential mechanisms that could account for this finding include access to peer support, social facilitation of target behaviors, and public commitment to therapy goals.⁵² No other results of participant, intervention, or study characteristics were found. Treatment benefits were equally likely to occur in older men and women irrespective of age and duration of chronic pain.

Our results add to the existing literature by demonstrating that older adults—an understudied population with respect to the benefits of psychological therapies for chronic pain^15,16—can benefit from these treatment approaches. Our findings are similar to those reported in the Cochrane review,¹⁰ which demonstrated that, among nonelderly adults with chronic pain, CBT has a small effect on pain at posttreatment. Unlike the present study, the effect documented in the Cochrane review did not persist at 6 months. Our findings of a small benefit with respect to pain mitigation are also similar to those of a recent report evaluating psychological therapies for nonelderly adults with low back pain.⁵³

Our study has several limitations that warrant consideration. Our search was limited to English-language studies, which may have eliminated otherwise eligible trials. In addition, trials with negative results may fail to report full outcome data but report only that there was no significant difference in outcomes analyzed. However, only 1 study was excluded from our sample for this reason. Quiz Ref IDOther factors limiting the generalizability of our findings include a lack of diversity in study populations (eg, focus on white individuals and young-old populations). Furthermore, the intensity of the interventions did not vary greatly, making it difficult to discern whether differences in treatment dose affect outcomes. Finally, few studies evaluated outcomes more than 6 months after treatment completion, so the long-term effects of these approaches remain poorly understood.

Our findings support guideline recommendations^54,55 that encourage clinicians to consider psychological treatments in the care of older patients with chronic pain, particularly those delivered using a group-based approach. Clinicians should learn and share with patients basic information about psychological approaches to managing pain. Inquiring about patients’ treatment expectations, including perceived benefits and potential harms, is also important.¹ Leveraging social supports to encourage patients’ continued use of psychological techniques (eg, distraction, relaxation techniques) over time is also warranted.¹ Finally, management of chronic pain in older adults should be multimodal, including use of both pharmacologic and nonpharmacologic approaches.^54-57 Physical treatments in the form of exercise and other movement-based approaches have demonstrated benefits in the form of reduced pain and improved functioning, are safe to use in older adults, and should also be considered.^54-60

Research is needed to better understand the mechanisms responsible for the effects of psychological therapies on chronic pain and ways to augment these effects. Although our results indicate that group-based vs individually delivered approaches produce superior outcomes, we still do not know which components of psychological therapies are most efficacious and in which subgroups of older adults. Research is also needed to determine what influences these approaches have on older adults’ use of pain medications, particularly opioids, and to ascertain any harms associated with their use. In addition, research is necessary to ascertain whether factors such as degree of cognitive impairment, race/ethnicity status, and level of support to adopt and use psychological techniques moderate treatment outcomes and, if so, to what degree. Further research should also explore whether treatment effects can be enhanced and sustained by leveraging research findings in the areas of temporal horizons⁶¹ and age-related changes in emotional and cognitive processing.^62,63 Prior findings indicate that older adults prioritize well-being in the present moment,⁶⁴ which may make them reluctant to engage in long-term treatment programs, and they are more likely to process information if it is presented in a positive frame.⁶⁵ These insights could be leveraged by emphasizing the immediate benefits of an intervention (eg, social engagement in a group setting) and crafting positively framed feedback messages to promote long-term adherence.⁶⁶ Finally, more research is needed regarding the role of mobile health technologies as tools to help deliver treatments and whether these devices can enhance adherence to the psychological techniques over time.⁶⁷

Psychological interventions for the treatment of chronic pain in older adults have beneficial, albeit small, associations with pain and catastrophizing as well as self-efficacy for managing pain. These benefits, documented at the completion of treatment, were found to persist up to 6 months later only for pain intensity reduction. Efforts are therefore needed to develop and test psychological interventions that generate more robust treatment effects that are sustainable in this growing population of patients.

Accepted for Publication: January 31, 2018.

Corresponding Author: M. Carrington Reid, MD, PhD, Division of Geriatrics and Palliative Medicine, Department of Medicine, Weill Cornell Medicine, 525 E 68th St, New York, NY 10065 (mcr2004@med.cornell.edu).

Published Online: May 7, 2018. doi:10.1001/jamainternmed.2018.0756

Author Contributions: Dr Reid had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Niknejad, Bolier, Henderson, Delgado, Kozlov, Reid.

Acquisition, analysis, or interpretation of data: Bolier, Henderson, Delgado, Kozlov, Löckenhoff, Reid.

Drafting of the manuscript: Niknejad, Bolier, Henderson, Delgado, Reid.

Critical revision of the manuscript for important intellectual content: Bolier, Henderson, Kozlov, Löckenhoff, Reid.

Statistical analysis: Henderson.

Administrative, technical, or material support: Niknejad, Bolier, Henderson, Delgado.

Study supervision: Bolier, Henderson, Reid.

Conflict of Interest Disclosures: None reported.

Funding/Support: Drs Löckenhoff and Reid are supported by Edward R. Roybal Translational Research on Aging award P30AG022845 from the National Institute on Aging. Dr Reid is also supported by National Institute on Aging award K24AGO53462, an investigator-initiated award from Pfizer Pharmaceuticals, and the Howard and Phyllis Schwartz Philanthropic Fund. Dr Kozlov is supported by National Institute on Aging grant T32AG049666.

Role of the Funder/Sponsor: The funders and sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data, preparation, review or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or other funders.