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


Protocol Registration

This systematic review and meta-analysis adheres to Cochrane methodology[21] and PRISMA guidelines,[31] and this protocol was registered on PROSPERO (CRD42020172451).

Types of Studies

We searched the literature for randomised controlled trials (RCTs) and nonrandomised comparative observational trials of pharmacological and physical therapies as we anticipated there would be few published RCTs. For psychological trials, we only included RCTs because of the larger number of trials published in this area. We included only superiority studies in our primary analyses but included equivalence design studies for pharmacological and physical therapies in secondary, subgroup, and sensitivity analysis.

Across all intervention types, we included peer-reviewed publications and records from trial registries where results were available, with more than 10 participants per arm posttreatment. We also identified ongoing trials in the area. We excluded conference abstracts and dissertations that were not peer reviewed.

Types of Participants

We included studies of children (ages 0–19 years) with chronic pain (pain lasting for longer than 3 months), including disease-related pain (eg, juvenile idiopathic arthritis, cancer, and palliative conditions) and idiopathic chronic pain conditions (eg, musculoskeletal pain and neuropathic pain). We included studies of children and adults if children were reported separately, made up >50% of the sample, or the mean age of the sample was <19 years.

Types of Interventions

For pharmacological interventions, we included any analgesic drug delivered by any route and dose, including adjuvant therapies, such as antidepressants, acetaminophen (paracetamol), opioids, nonsteroidal anti-inflammatory drugs, anticonvulsants, antiepileptic drugs, and ketamine. Eligible physical interventions had to involve bodily movement including, but not exclusively, physiotherapy, muscle strengthening, sports, and conditioning. Passive physical interventions such as massage or manipulation were excluded. Eligible psychological interventions included any that delivered recognisable psychological content (eg, cognitive behavioural therapy [CBT], behaviour therapy, hypnosis, or acceptance and commitment therapy). These interventions could be delivered remotely but must include interaction with the programme or include homework (ie, must not be completed passively such as reading about CBT). We also included interventions directed at parents of children with pain. We planned to include studies that consisted of a combination of pharmacological, physical, or psychological interventions but did not find any trials in this area. We excluded complementary and alternative therapies, which were beyond the scope of this review. Studies investigating these interventions are typically low quality.[37]

Trials comparing interventions to active or placebo comparisons, treatment as usual, waitlist control, or other pharmacological or physical therapy interventions were included.

Types of Outcomes

We followed the PedIMMPACT consensus statement on core outcome domains in paediatric chronic and recurrent pain clinical trials.[30]

Critical outcomes included:

  1. continuous pain intensity and 30% and 50% reduction in pain intensity,

  2. health-related quality of life (HRQOL),

  3. functional disability,

  4. role functioning,

  5. emotional functioning,

  6. sleep, and

  7. adverse events (AEs): treatment-related serious AEs, treatment-related AEs, and other AEs (including misuse of prescription medicines and nonmedical use of related psychoactive substances, and opioid use disorder if reported).

Important outcomes included

  1. activity participation,

  2. global judgement of satisfaction with treatment,

  3. patient global impression of change, and

  4. fatigue.

We prioritised child-reported to observer-reported outcomes where both were reported.

Search Strategy and Selection of Studies

We searched Cochrane CENTRAL, MEDLINE and MEDLINE in Process (OVID), and EMBASE (OVID) from inception to April 2020 for pharmacological and physical therapy trials. For psychological therapy trials, we updated searches using the 3 aforementioned databases in addition to PsycINFO (EBSCO) from previous Cochrane reviews,[13,14,27] and we searched for trials in children with cancer from inception of each database to March 2020. No date or language restrictions were applied. Search strategies can be found in Appendix A (available at http://links.lww.com/PAIN/B369).

In an initial accelerated screening process, 2 independent reviewers were required to exclude a title or abstract, and only one reviewer was required to include. Two reviewers then independently screened full reports of each potential study. We resolved any disagreements in the screening or extraction process by consensus with a third reviewer. We screened relevant systematic reviews identified by the search for further trials.

Using DistillerSR,[12] one reviewer extracted data from studies and assessed risk of bias and a second reviewer cross-checked extraction. Disagreements were resolved through discussion. We extracted study characteristics (eg, design of study, number of participants, intervention and comparison, dose or duration, route or mode of intervention, funding, and conflicts of interests), participant characteristics (eg, age, sex, medical or pain condition, and length of condition), and outcome data. When outcome data were not reported within the peer-reviewed article, we contacted study authors for the data through email.

Risk of Bias Assessment for Randomised and Nonrandomised Controlled Trials

Randomised Controlled Trials. We used the Cochrane Risk of Bias tool[21] to assess risk of bias in RCTs.

Random Sequence Generation (Checking for Possible Selection Bias): We assessed the method used to generate the allocation sequence as low risk of bias (any truly random process, eg, random number table or computer random number generator) or unclear risk of bias (method used to generate sequence not clearly stated). Studies using a nonrandom process (eg, odd or even date of birth and hospital or clinic record number) were rated as high risk of bias.

Allocation Concealment (Checking for Possible Selection Bias): The method used to conceal allocation to interventions before assignment determines whether intervention allocation could have been foreseen in advance of or during recruitment, or changed after assignment. We assessed the methods as low risk of bias (eg, telephone or central randomisation or consecutively numbered sealed opaque envelopes) or unclear risk of bias (method not clearly stated). Studies that do not conceal allocation (eg, open list) were rated as high risk of bias.

Blinding of Participants and Personnel (Checking for Possible Performance Bias): For pharmacological interventions, we assessed the methods used to blind study participants and personnel from knowledge of which intervention a participant received. We assessed methods as low risk of bias (study states that it was blinded and describes the method used to achieve blinding, such as identical tablets matched in appearance or smell or a double-dummy technique) or unclear risk of bias (study states that it was blinded but does not provide an adequate description of how it was achieved). Studies that are not double blinded were considered to have a high risk of bias for pharmacological trials.

Blinding of participants and personnel is less relevant for physical and psychological interventions because it is almost impossible to blind either party. Therefore, we did not assess this domain for those interventions.

Blinding of Outcome Assessment (Checking for Possible Detection Bias): We assessed the methods used to blind study participants and outcome assessors from knowledge of which intervention a participant received. We assessed the methods as low risk of bias (study has a clear statement that outcome assessors were unaware of treatment allocation and ideally describes how this was achieved or whether outcome assessments were completed electronically by participants) or unclear risk of bias (study states that outcome assessors were blind to treatment allocation but lacks a clear statement on how it was achieved). Studies where outcome assessment is not blinded were rated as high risk of bias.

Incomplete Outcome Data (Checking for Possible Attrition Bias due to the Amount, Nature, and Handling of Incomplete Outcome Data): We assessed the methods used to deal with incomplete data as low risk of bias (<10% of participants did not complete the study or used "baseline observation carried forward" analysis), unclear risk of bias (used "last observation carried forward" analysis), or high risk of bias (used "completer" analysis).

Selective Reporting (Checking for Reporting Bias): We assessed whether primary and secondary outcome measures were prespecified and whether these were consistent with those reported: low risk (protocol or trial registration available and adhered to), unclear risk (insufficient information to judge low or high risk of bias), or high risk (one or more outcomes listed in the protocol or trial registration are missing or additional primary outcomes included that were not prespecified).

Nonrandomised Trials. The list of the most important confounders (age, sex, and baseline pain intensity) and cointerventions was determined in consultation with appropriate clinical experts.

The ROBINS-I[35] tool considers 7 domains of bias: 2 domains of bias preintervention (bias due to confounding and bias in selection of participants into the study), one domain of bias at intervention (bias in the measurement of interventions), and 4 domains of bias postintervention (bias due to departures from intended interventions, bias due to missing data, bias in measurement of outcomes, and bias in selection of the reported result).

Each domain of bias contains signalling questions to facilitate judgements of risk of bias. The response options for the signalling questions included yes, probably yes, probably no, no, and no information.

The "risk of bias" judgement options for each domain were as follows:

  1. Low risk of bias: The study is comparable with a well-performed randomised trial regarding this domain.

  2. Moderate risk of bias: The study is sound for a nonrandomised study regarding this domain but cannot be considered comparable with a well-performed randomised trial.

  3. Serious risk of bias: The study has some important problems in this domain.

  4. Critical risk of bias: The study is too problematic in this domain to provide any useful evidence on the effects of intervention.

No information on which to base a judgement about risk of bias for this domain.

Summarizing and Interpreting Results: Evidence Profiles

We used the GRADE approach to interpret findings and created a "GRADE Profile" table following the GRADE Handbook.[20] The tables provide the effect estimate and the associated certainty of evidence for each outcome of interest. It also allowed us to judge the overall certainty of evidence for the PICO question.

The certainty of evidence for RCTs started at high certainty, but we downgraded it to moderate, low, or very low for the following reasons:

  1. Limitations in study design or execution (risk of bias): We downgraded by one level if more than 50% of risk of bias assessments were unclear or high risk of bias and by 2 levels if more than 75% of risk of bias assessments were unclear or high risk of bias.

  2. Inconsistency of results: We downgraded by one level if heterogeneity exceeded 50% in the analysis and by 2 levels if heterogeneity exceeded 75% and if we were not able to explain it, for example, through subgroup analysis.

  3. Indirectness of evidence: We downgraded by one level if there were few pain conditions included in the analyses.

  4. Imprecision: We downgraded by one level if there were fewer than 400 participants or 2 studies included in the analysis, if there were wide confidence intervals, or if most studies contributing to the analysis included fewer than 30 participants per arm. We downgraded by 2 levels if there were fewer than 200 participants or if 2 or fewer studies contributed to the analysis.

  5. Publication bias: To assess publication bias, we tested for asymmetry in a funnel plot if there were at least 10 studies in a meta-analysis and if the studies were of different sizes. If we suspected publication bias, we downgraded by one level. For nonrandomised studies, certainty may also be upgraded if the pooled estimates represent a large magnitude of effect or if there is a strong dose–response gradient.

Factors that may increase the certainty of evidence are as follows:

  1. A large magnitude of effect.

  2. All plausible confounding variables would reduce a demonstrated effect or suggest a spurious effect when results show no effect.

  3. A dose–response gradient.

Data Analysis

We conducted comparisons of each class of intervention (pharmacological, physical, and psychological) separately, compared with placebo, waitlist, or active control. We analysed each comparison at immediate posttreatment and up to 12 months follow-up. Where trials included more than 2 arms, we combined the 2 most similar arms if appropriate.

We took a 2-step approach to data analysis, preferentially analysing RCTs and analysing RCTs separately to nonrandomised studies, which we described narratively. We described results from crossover trials. We used random-effects models for all meta-analyses because of heterogeneity of studies and data. For dichotomous outcomes, we used risk ratios or risk differences where there are a low number of events reported, with 95% confidence intervals. For continuous outcomes, we used standardised mean difference. We interpreted standardised mean differences as small effects (0.2), medium effects (0.5), and large effects (0.8).[2] Heterogeneity was assessed using the I2 statistic and interpreted following the Cochrane Handbook.[21]

For pharmacological therapies, we performed subgroup analyses by type of drugs in the primary analyses. For the psychological and physical studies, we included all intervention types and conducted subgroup analyses by therapy type where enough data (ie, >10 studies) were available. Subgroup analyses were conducted to assess differences in control type (active or placebo), age (infants <1 year, children 1–9 years inclusive, and adolescents 10–19 years inclusive), chronic pain condition following the ICD-11 classifications, dose or duration of treatment, and route of delivery (pharmacological treatments alone).

We planned sensitivity analysis by restricting the analysis to studies with a low risk of bias or those studies that included >200 participants per treatment arm if sufficient studies (ie, > 10) could be combined.