Efficacy and Safety of Pharmacological, Physical, and Psychological Interventions for the Management of Chronic Pain in Children

A WHO Systematic Review and Meta-analysis

Emma Fisher; Gemma Villanueva; Nicholas Henschke; Sarah J. Nevitt; William Zempsky; Katrin Probyn; Brian Buckley; Tess E. Cooper; Navil Sethna; Christopher Eccleston


Pain. 2022;163(1):e1-e19. 

In This Article


Summary of Findings

Overall, we included 34 pharmacological therapy trials (29 RCTs, 5 nonrandomised studies, and 37 reports) and 5 ongoing RCTs (no results), 24 physical therapy RCTs (33 reports) and 12 further ongoing trials (11 RCTs and 1 nonrandomised trial [which reported results]), and 63 psychological therapy RCTs and 16 ongoing RCT studies (Figures 1–3). We did not identify any studies with data of children or adolescents with cancer diagnoses, needing palliative care, or with an intellectual disability. Included and excluded study references with reasons are included in Appendix B (available at http://links.lww.com/PAIN/B370) by intervention modality. Sample and study characteristics are shown in Table 1.

Figure 1.

PRISMA flow diagram: pharmacological therapies.

Figure 2.

PRISMA flow diagram: physical therapies.

Figure 3.

PRISMA flow diagram: psychological therapies.

Assessments of risk of bias judgements for each study categorised by treatment modality are shown in Figures 4–9, described in Appendix C (available at http://links.lww.com/PAIN/B369), and individual judgements are included in Appendix D (available at http://links.lww.com/PAIN/B371). Overall, we identified unclear or high risk of bias for most studies and across most risk of bias domains. High risk of bias was judged most often in incomplete outcome data and selective reporting bias. Low risk of bias was judged most often for random sequence generation and allocation.

Figure 4.

Risk of bias: pharmacological interventions summary.

Figure 5.

Risk of bias: physical interventions summary.

Figure 6.

Risk of bias: psychological interventions summary.

Figure 7.

Risk of bias: pharmacological interventions individual study summary.

Figure 8.

Risk of bias: physical interventions individual study summary.

Figure 9.

Risk of bias: psychological interventions individual study summary.

Despite the number of pharmacological and physical therapy trials included, relatively few analyses could be conducted because of poor reporting of outcomes. For the primary analyses of superiority trials, there were 16 pharmacological trials, 12 physical therapy trials, and 63 psychological trials.

Pharmacological Trials. Of the 34 trials, most were conducted in North America (n = 12) or Europe (n = 11) and included 4091 participants at baseline. Thirty studies reported sex; 6 studies included only female participants (878 girls). Overall, there were 2738 girls (984 boys; 74% female) included. The average age of participants was 12.27 years.

Most studies (n = 28) reported a diagnostic criterion. Across the trials, only 3 required a minimum pain intensity as part of their inclusion criteria. Eleven trials in total reported baseline pain, which ranged from 3/10 to 8/10 across studies.

Of the 34 studies, 16 studies included an arm that was not another pharmacological intervention. When considering the funding, 18 studies did not report the source of funding for the trial. Four studies were supported by government or university funding. Eight studies were supported by pharmaceutical companies. Two studies were funded in partnership between a research foundation and pharmaceutical company. Two studies did not report funding, but we suspect they were funded by pharmaceutical companies because they thanked them in the acknowledgments or authors were employees of the company. There were 23 studies that did not report the presence of conflicts of interests in the article. Five studies explicitly reported that the authors had no conflicts of interest. Two studies did not report conflicts of interest, but one study reported that an author received consulting fees of less than $10,000 from Bristol Meyers Squibb, and in the other study, most authors were employees of Pfizer Inc. The remaining 4 studies reported conflicts of interest.

Physical Trials. Of the 25 trials that had results, 4 studies each were conducted in the United Kingdom and the United States, 3 in Turkey, 2 each in Canada and Sweden, and one each in Australia, Brazil, Chile, Denmark, India, Iran, Israel, the Netherlands, Portugal, and Saudi Arabia. The 25 included studies with results randomised 1470 participants. There were more girls (73%; n = 984) than boys (n = 373) enrolled into treatment, as reported by 24 studies. Two dysmenorrhea studies only enrolled female participants (n = 223). The mean age of participants was 13 years.

We found 12 studies used a diagnostic criterion before enrolling children into the trial. Two trials only enrolled participants with a minimum pain of 2 of 10 or more, and one trial used 4 of 10 or more. The remaining trials did not have a minimum pain intensity inclusion criterion.

We found 25 physical intervention studies, of which 11 studies compared a physical intervention to a nonphysical intervention control, standard care, or waitlist control. One additional study included 4 arms and compared 2 PT interventions (stretching vs stretching and massage) and 2 control arms (massage vs no treatment). We therefore combined the 2 stretching interventions and compared with control.

Two studies reported conflicts of interests; Stephens et al.[34] reported one author had received consultancy fees, speaking fees, or honoraria (less than $10,000) from Bristol Myers Squibb, and Ahlqwist et al.[1] reported one or more of the author(s) has or have received or will receive benefits for personal or professional use from a commercial party related directly or indirectly to the subject of this manuscript and benefits will be directed solely to a research fund, foundation, educational institution, or other nonprofit organization that the author(s) has or have been associated with. Nineteen studies reported no conflicts of interest, and 4 studies did not provide a conflicts of interest statement. Seven studies did not report a funding statement, one study reported that no funding was received for the study, one study was sponsored by industry, and the remaining 16 studies were supported by government, hospital, or research charity grants.

Psychological Trials. We found 34 studies conducted in North America, 11 in Sweden, 6 in the Netherlands, 6 in Germany, 2 in Australia, and one in Brazil, Spain, Italy, and China each. Five thousand twenty-five participants were randomised to a treatment or control condition. Children were on average aged 12.87 years primarily recruited through hospital settings. Girls outnumbered boys (67%; 3172 girls to 1542 boys).

Twenty-one studies did not have a diagnostic criterion for including participants, but the remaining 42 studies did report that participants had to fulfil a diagnostic criterion. No studies reported a defined minimum pain for inclusion, and there were a range of baseline pain intensity scores (ranging from 1.3/10 to 6.8/10).

Of the 63 trials, 13 included multiple arms. We found 43 arms of CBT, 15 arms of relaxation, 7 arms of behavioural therapy, 3 arms of hypnosis, 2 arms of problem-solving therapy, and one arm of acceptance commitment therapy. We found 36 active control arms, 16 standard or usual care arms, and 17 waitlist control arms (note, because some studies included multiple arms, these numbers will not add up to 63). When considering route of delivery, we found 45 arms that delivered treatment face-to-face and 26 arms of treatment delivered remotely. Most (n = 13) were delivered through the Internet or smartphone, with the remaining arms being delivered through the telephone, audiotapes, or CD ROM. Funding for trials were primarily from government grants or research foundations (n = 48). Two studies reported specifically no funding, and the remaining 13 studies did not report funding. Twenty-seven studies explicitly reported no conflicts of interest related to the study, 34 studies did not include a conflict of interest statement, and 2 studies reported potential conflicts of interest; one study reported an author was an employee of OrthoNeuro Inc, and a second study reported an author had received consulting fees, speaking fees, or honoraria from Pfizer.


All meta-analysis results are summarised in Table 2, forest plots are included in Appendix E, and GRADE tables in Appendix F (available at http://links.lww.com/PAIN/B369). We could only conduct subgroup analyses for psychological trials because there were too few studies to combine for pharmacological and physical studies. Analyses of equivalence studies and subgroup analyses of psychological trials are described in Appendix C, G, and H (available at http://links.lww.com/PAIN/B369 and http://links.lww.com/PAIN/B372).

Critical Outcomes. Pain: Overall, we found that pharmacological interventions reduced pain intensity posttreatment (moderate certainty of evidence). Within drug class, anticonvulsants (pregabalin) produced a reduction in pain intensity, but no benefit was found for antidepressants (amitriptyline, citalopram, or duloxetine) vs placebo or NSAIDs (ibuprofen) vs active control (acupressure or sham acupressure). At follow-up, 2 studies comparing antidepressants (amitriptyline or citalopram) to placebo showed no benefit of antidepressants (very low certainty). For 30% and 50% pain reduction, there were no benefits of any drug vs control (very low certainty). Single studies of anticonvulsants (pregabalin) vs placebo showed no effect (very low certainty), but antidepressants (duloxetine) vs placebo was beneficial (very low certainty) in both analyses. No meta-analyses could be conducted at follow-up.

For physical therapy studies, there may be a moderate benefit of reducing pain intensity posttreatment, but we are very uncertain about this effect (very low certainty) and this was not maintained at follow-up (very low certainty). No data could be analysed for pain reduction for physical therapy trials.

For psychological therapies, there may be small benefits of reducing pain intensity posttreatment (low certainty) and at follow-up (low certainty). Only one study provided data for 30% pain reduction. Fourteen studies reported on 50% headache reduction in children and adolescents, and we found that psychological interventions may be beneficial at reducing pain intensity posttreatment (low certainty) and at follow-up (very low certainty).

Health-related Quality of Life: No meta-analyses could be conducted for pharmacological therapies at either time point or physical therapies at follow-up. At posttreatment, physical therapies showed no benefit at improving HRQOL (very low certainty). Similarly, no benefit was found for psychological therapies at improving HRQOL posttreatment or at follow-up (both low certainty of evidence).

Functional Disability: No meta-analyses could be conducted for pharmacological therapies at either time point, or physical therapies at follow-up. Physical therapies showed a moderate beneficial effect of reducing functional disability posttreatment, but we are very uncertain about this effect (very low certainty). There may be small beneficial effects of psychological therapies at reducing functional disability at both posttreatment (low certainty) and follow-up (moderate certainty).

Role Functioning: No meta-analyses could be conducted for pharmacological therapies at either time point, or physical therapies at follow-up. Psychological therapies did not show a benefit at reducing school absence in children with chronic pain conditions at posttreatment (very low certainty) or follow-up (very low certainty).

Emotional Functioning: For depression and anxiety, 3 pharmacological studies compared antidepressants (citalopram, amitriptyline, or duloxetine) with placebo posttreatment and one study compared citalopram with placebo at follow-up. The analyses did not show any benefit for reducing depression or anxiety (posttreatment: low certainty and follow-up: very low certainty). Similarly, no benefit was found for physical therapies posttreatment or at follow-up (both very low certainty) or psychological therapies post treatment (depression: high certainty and anxiety: moderate certainty) or at follow-up (high certainty for both outcomes) for reducing depression or anxiety.

Sleep: No meta-analyses could be conducted for pharmacological therapies at either time point, or physical therapies at follow-up. Psychological interventions were not beneficial at improving sleep quality posttreatment (low certainty) or at follow-up (very low certainty).

Adverse Events: Pharmacological studies did not report whether serious adverse events (SAEs) were treatment-related or not; therefore, we included all reports of SAEs in the analysis. Overall, there were very few reports of SAEs in pharmacological trials (3/767 participants in the treatment group and 0/361 participants in the control group; 4 studies; very low certainty). SAEs included appendicitis, suicidal ideation, cholelithiasis, and major depression. When analysing the studies according to their pharmacological classification, we found no differences between individual pharmacological drug classifications (pregabalin, duloxetine, sumatriptan, and naproxen) and placebo.

Similar to SAEs, pharmacological trials did not distinguish between treatment-related AEs and non–treatment-related AEs, and therefore, we included all reports of AEs in the same analysis. We found that 153 of 490 participants reported AEs in the treatment groups and 104 of 291 in the control group (3 studies; very low certainty). When investigating the individual pharmacological classifications, we found no differences between pregabalin or sumatriptan and naproxen vs placebo. However, participants receiving duloxetine reported higher numbers of AEs compared with placebo. There were very few reported AEs for physical therapies (very low certainty), and no difference was identified between treatment and control groups.

We could not meta-analyse AEs or SAEs for psychological trials or SAEs for physical therapies. From the evidence that was reported, very few AEs or SAEs occurred in either group. However, reporting was poor across both treatment modalities, resulting in a very low-certainty evidence.

Important Outcomes. We could only perform a meta-analysis of psychological therapies for global judgement of satisfaction with treatment, posttreatment. Participants in the treatment group were more satisfied posttreatment, compared with control (moderate certainty), and at follow-up (very low certainty). No other meta-analysis could be conducted for important outcomes across treatment modalities because of lack of studies.