Psychological therapies for the management of chronic and recurrent pain in children and adolescents

Emma Fisher; Emily Law; Joanne Dudeney; Tonya M Palermo; Gavin Stewart; Christopher Eccleston

doi:10.1002/14651858.CD003968.pub5

. 2018 Oct 1;2018(9):CD003968. doi: 10.1002/14651858.CD003968.pub5

Psychological therapies for the management of chronic and recurrent pain in children and adolescents

Emma Fisher ^1,^✉, Emily Law ², Joanne Dudeney ³, Tonya M Palermo ², Gavin Stewart ⁴, Christopher Eccleston ⁵

Editor: Cochrane Pain, Palliative and Supportive Care Group

PMCID: PMC6257251 NIHMSID: NIHMS991188 PMID: 30270423

Abstract

Background

This is an update of the original Cochrane review first published in Issue 1, 2003, and previously updated in 2009, 2012 and 2014. Chronic pain, defined as pain that recurs or persists for more than three months, is common in childhood. Chronic pain can affect nearly every aspect of daily life and is associated with disability, anxiety, and depressive symptoms.

Objectives

The aim of this review was to update the published evidence on the efficacy of psychological treatments for chronic and recurrent pain in children and adolescents.

The primary objective of this updated review was to determine any effect of psychological therapy on the clinical outcomes of pain intensity and disability for chronic and recurrent pain in children and adolescents compared with active treatment, waiting‐list, or treatment‐as‐usual care.

The secondary objective was to examine the impact of psychological therapies on children's depressive symptoms and anxiety symptoms, and determine adverse events.

Search methods

Searches were undertaken of CENTRAL, MEDLINE, MEDLINE in Process, Embase, and PsycINFO databases. We searched for further RCTs in the references of all identified studies, meta‐analyses, and reviews, and trial registry databases. The most recent search was conducted in May 2018.

Selection criteria

RCTs with at least 10 participants in each arm post‐treatment comparing psychological therapies with active treatment, treatment‐as‐usual, or waiting‐list control for children or adolescents with recurrent or chronic pain were eligible for inclusion. We excluded trials conducted remotely via the Internet.

Data collection and analysis

We analysed included studies and we assessed quality of outcomes. We combined all treatments into one class named 'psychological treatments'. We separated the trials by the number of participants that were included in each arm; trials with > 20 participants per arm versus trials with < 20 participants per arm. We split pain conditions into headache and mixed chronic pain conditions. We assessed the impact of both conditions on four outcomes: pain, disability, depression, and anxiety. We extracted data at two time points; post‐treatment (immediately or the earliest data available following end of treatment) and at follow‐up (between three and 12 months post‐treatment).

Main results

We identified 10 new studies (an additional 869 participants) in the updated search. The review thus included a total of 47 studies, with 2884 children and adolescents completing treatment (mean age 12.65 years, SD 2.21 years). Twenty‐three studies addressed treatments for headache (including migraine); 10 for abdominal pain; two studies treated participants with either a primary diagnosis of abdominal pain or irritable bowel syndrome, two studies treated adolescents with fibromyalgia, two studies included adolescents with temporomandibular disorders, three were for the treatment of pain associated with sickle cell disease, and two studies treated adolescents with inflammatory bowel disease. Finally, three studies included adolescents with mixed pain conditions. Overall, we judged the included studies to be at unclear or high risk of bias.

Children with headache pain

We found that psychological therapies reduced pain frequency post‐treatment for children and adolescents with headaches (risk ratio (RR) 2.35, 95% confidence interval (CI) 1.67 to 3.30, P < 0.01, number needed to treat for an additional beneficial outcome (NNTB) = 2.86), but these effects were not maintained at follow‐up. We did not find a beneficial effect of psychological therapies on reducing disability in young people post‐treatment (SMD ‐0.26, 95% CI ‐0.56 to 0.03), but we did find a beneficial effect in a small number of studies at follow‐up (SMD ‐0.34, 95% CI ‐0.54 to ‐0.15). We found no beneficial effect of psychological interventions on depression or anxiety symptoms.

Children with mixed pain conditions

We found that psychological therapies reduced pain intensity post‐treatment for children and adolescents with mixed pain conditions (SMD ‐0.43, 95% CI ‐0.67 to ‐0.19, P < 0.01), but these effects were not maintained at follow‐up. We did find beneficial effects of psychological therapies on reducing disability for young people with mixed pain conditions post‐treatment (SMD ‐0.34, 95% CI ‐0.54 to ‐0.15) and at follow‐up (SMD ‐0.27, 95% CI ‐0.49 to ‐0.06). We found no beneficial effect of psychological interventions on depression symptoms. In contrast, we found a beneficial effect on anxiety at post‐treatment in children with mixed pain conditions (SMD ‐0.16, 95% CI ‐0.29 to ‐0.03), but this was not maintained at follow‐up.

Across all pain conditions, we found that adverse events were reported in seven trials, of which two studies reported adverse events that were study‐related.

Quality of evidence

We found the quality of evidence for all outcomes to be low or very low, mostly downgraded for unexplained heterogeneity, limitations in study design, imprecise and sparse data, or suspicion of publication bias. This means our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect, or we have very little confidence in the effect estimate; or the true effect is likely to be substantially different from the estimate of effect.

Authors' conclusions

Psychological treatments delivered predominantly face‐to‐face might be effective for reducing pain outcomes for children and adolescents with headache or other chronic pain conditions post‐treatment. However, there were no effects at follow‐up. Psychological therapies were also beneficial for reducing disability in children with mixed chronic pain conditions at post‐treatment and follow‐up, and for children with headache at follow‐up. We found no beneficial effect of therapies for improving depression or anxiety. The conclusions of this update replicate and add to those of a previous version of the review which found that psychological therapies were effective in reducing pain frequency/intensity for children with headache and mixed chronic pain conditions post‐treatment.

Plain language summary

Psychological therapies for the management of chronic and recurrent pain in children and adolescents

Bottom line

Psychological therapies reduce pain frequency immediately following treatment for children and adolescents with chronic headache and reduce pain intensity for children and adolescents with mixed chronic pain conditions. Psychological therapies also reduce disability for children and adolescents with mixed chronic pain conditions immediately following treatment and up to 12 months later, and for children with headache conditions up to 12 months later.

Background

Chronic pain or pain that lasts for longer than three months is common in young people. Psychological therapies (e.g. relaxation, hypnosis, coping skills training, biofeedback, and cognitive behavioural therapy) may help people manage pain and its disabling consequences. Therapies can be delivered face‐to‐face by a therapist, via the Internet, by telephone call, or by computer programme. This review focused on treatments that are delivered face‐to‐face by a therapist, which includes therapies delivered by telephone or via a book with exercise instructions. For children and adolescents, there is evidence that relaxation by itself and cognitive behavioural therapy (treatment that helps people test and revise their thoughts and actions) are effective in reducing the intensity of pain in chronic headache, recurrent abdominal pain, fibromyalgia, and sickle cell disease immediately after treatment.

Study Characteristics

This review included 47 studies with 2884 participants. The average age of the children and adolescents was 12.6 years. Most studies included young people with headache (23 studies) or stomach pain (10 studies), The remaining studies investigated children with irritable bowel syndrome, fibromyalgia, temporomandibular disorders, sickle cell disease, inflammatory bowel disease, or included samples with various chronic pain conditions.

Key results

Psychological therapies reduced pain frequency immediately following treatment for children and adolescents with chronic headache, and pain intensity and anxiety for children and adolescents with other chronic pain conditions. Psychological therapies also reduced disability for children and adolescents with non‐headache chronic pain conditions immediately following treatment and for children with headache and mixed chronic pain conditions up to 12 months later. We did not find any benefit of psychological treatments on reducing anxiety for children with headache or for depression in children with headache or mixed chronic pain conditions.

Quality of evidence

We judged all outcomes to be low or very low‐quality, meaning our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect or we have very little confidence in the effect estimate; or the true effect is likely to be substantially different from the estimate of effect.

Summary of findings

Summary of findings 1. Summary of findings.

Psychological therapies compared with any control for children and adolescents with frequent headaches
Patient or population: Children and adolescents with chronic pain Settings: Community and hospitals Intervention: Psychological therapies (cognitive behavioural therapy or behavioural therapy) Comparison: Any control (active, treatment‐as‐usual, wait‐list)
Outcomes	Probable outcome with control	Probable outcome with intervention	NNTB and/or relative effect (95% CI)	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
Pain: 50% reduction in headache frequency Post‐treatment Lower scores = fewer headaches	10 per 1000	24 per 1000	NNTB = 2.86; RR 2.35 (1.67 to 3.30)	644 participants (15 studies)	⊕⊝⊝⊝ very low^b,c,h
Pain: 50% reduction in headache frequency Follow‐up (up to 12 months) Lower scores = fewer headaches	10 per 1000	27 per 1000	NNTB = 3.16; RR 2.73 (0.98 to 7.63)	223 participants (5 studies)	⊕⊝⊝⊝ very low^b,c,e,f,g,h
Disability Post‐treatment Lower scores = lower reported disability		The mean disability in the intervention groups was 0.26 lower (95% CI ‐0.56 to 0.03)		446 participants (6 studies)	⊕⊝⊝⊝ very low^d,f,g
Disability Follow‐up Lower scores = lower reported disability		The mean disability in the intervention groups was 0.37 lower (95% CI ‐0.65 to ‐0.10)		209 participants (3 studies)	⊕⊝⊝⊝ very low^f,g
Anxiety Post‐treatment Lower scores = lower reported anxiety		The mean anxiety in the intervention groups was 0.11 lower (95% CI ‐0.39 to 0.17)		439 participants (7 studies)	⊕⊝⊝⊝ very low ^a,d,f,h,i
Anxiety Follow‐up Lower scores = lower reported anxiety		The mean anxiety in the intervention groups was 0.12 lower (95% CI ‐0.46 to 0.21)		271 participants (4 studies)	⊕⊝⊝⊝ very low^a,f,g
Depression Post‐treatment Lower scores = lower reported depression		The mean depression in the intervention groups was 0.08 lower (95% CI ‐0.28 to 0.11)		400 participants (6 studies)	⊕⊝⊝⊝ very low^a,f,g
CI: Confidence interval; RR: Risk Ratio; NNTB; Number needed to treat to benefit.
GRADE Working Group grades of evidence High‐quality: we are very confident that the true effect lies close to that of the estimate of the effect; Moderate‐quality: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of effect, but there is a possibility that it is substantially different; Low‐quality: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect; Very low‐quality: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

Open in a new tab

^a50 to 75% risk of bias ratings were unclear/high.

^b> 75% of risk of bias ratings were unclear or high.

^cConfidence intervals were wide.

^dHeterogeneity (I²) was 46 to 65%.

^eHeterogeneity (I²) was 66 to 100%.

^f75 to 100% of studies eligible to be included in the analysis were not included in the analysis.

^gSmall number of participants contributing to the outcome.

^hAsymmetrical funnel plots suggesting publication bias.

ⁱThere was mostly unclear/high risk of bias in the selective reporting category.

Summary of findings 2. Summary of findings.

Psychological therapies compared with any control for children and adolescents with chronic pain conditions (mixed)
Patient or population: Children and adolescents with chronic pain Settings: Community and hospitals Intervention: Psychological therapies (cognitive behavioural therapy or behavioural therapy) Comparison: Any control (active, treatment‐as‐usual, wait‐list)
Outcomes	Probable outcome with control	Probable outcome with intervention	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
Pain Post‐treatment Lower scores = lower reported pain intensity		The mean pain intensity in the intervention group was 0.43 lower (95% CI ‐0.67 to ‐0.19)	1210 participants (16 studies)	⊕⊝⊝⊝ very low^d,f
Pain Follow‐up Lower scores = lower reported pain intensity		The mean pain intensity in the intervention group was 0.08 lower (95% CI ‐0.30 to 0.13)	763 participants (9 studies)	⊕⊝⊝⊝ very low^b,c,e,f
Disability Post‐treatment Lower scores = lower reported disability		The mean disability in the intervention group was 0.34 lower (95% CI ‐0.54 to ‐0.15)	1226 participants (14 studies)	⊕⊕⊝⊝ low^c,f
Disability Follow‐up Lower scores = lower reported disability		The mean disability in the intervention group was 0.27 lower (95% CI ‐0.49 to ‐0.06)	866 participants (9 studies)	⊕⊕⊝⊝ low^c,e
Anxiety Post‐treatment Lower scores = lower reported anxiety		The mean anxiety in the intervention group was 0.16 lower (95% CI ‐0.29 to ‐0.03)	883 participants (8 studies)	⊕⊕⊝⊝ low^f
Anxiety Follow‐up Lower scores = lower reported anxiety		The mean anxiety in the intervention group was 0.01 lower (95% CI ‐0.20 to 0.18)	805 participants (8 studies)	⊕⊕⊝⊝ low^b,f
Depression Post‐treatment Lower scores = lower reported depression		The mean disability in the intervention group was 0.05 lower (95% CI ‐0.23 to 0.12)	757 participants (8 studies)	⊕⊝⊝⊝ verylow^b,e,f
CI: Confidence interval.
GRADE Working Group grades of evidence High‐quality: we are very confident that the true effect lies close to that of the estimate of the effect; Moderate‐quality: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of effect, but there is a possibility that it is substantially different; Low‐quality: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect; Very low‐quality: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

Open in a new tab

^a50 to 75% risk of bias ratings were unclear/high.

^bConfidence intervals were wide.

^cHeterogeneity (I²) was 46 to 65%.

^dHeterogeneity (I²) was 66 to 100%.

^e50 to 75% of studies eligible to be included in the analysis were not included in the analysis.

^fAsymmetrical funnel plots suggesting publication bias.

Background

Description of the condition

This review is an update of a previously published review in the Cochrane Library on 'Psychological therapies for the management of chronic and recurrent pain in children and adolescents' (Eccleston 2003; Eccleston 2009; Eccleston 2012; Eccleston 2014). Chronic (pain lasting more than three months) and recurrent pain is a common problem in young people. Recent epidemiology suggests a prevalence of 15% to 30%, with 8% of children described as having severe and frequent pain (King 2011; Perquin 2000; Perquin 2001; Stanford 2008). The most common pain locations are the head, abdomen, and limbs (King 2011). Girls more commonly report all types of chronic and recurrent pain, and there is a peak in incidence at ages 14 to 15 years (Stanford 2008). Young people report pain to be distressing and interfering, and in some cases this can be severely debilitating, affecting all aspects of a child's life, and the lives of the parents and family members (Palermo 2005; Palermo 2014). The deleterious effects of chronic pain in childhood can also extend to adulthood (Horst 2014; Walker 2012).

Description of the intervention

Psychological treatments for children and adolescents with chronic pain conditions are specifically designed to alter psychological processes thought to underlie, or significantly contribute to, pain, distress, and disability. The design of psychological treatments is informed by specific theories of the causes of human behaviour, or has developed pragmatically through observation and study of response to intervention. Behavioural and cognitive treatments designed to ameliorate pain, distress, and disability were first introduced in adults over 40 years ago (Fordyce 1968; Keefe 2004), and were used to inform the development of psychological treatments for children and adolescents with chronic pain. In paediatric practice, the treatments have different therapeutic aims and components than those for adults. In general, psychological treatments for children and adolescents aim to control pain and modify situational, emotional, familial, and behavioural factors that play a role in the onset and maintenance of pain (Palermo 2012).

Treatments were originally delivered in a face‐to‐face format in which the patients and therapists worked together in person to implement therapeutic strategies. Methods of remote delivery of psychological treatments for children with chronic pain conditions have also been developed; these are the subject of a separate Cochrane review 'Psychological therapies (remotely delivered) for the management of chronic and recurrent pain in children and adolescents' (Fisher 2015). A companion review of psychological treatments for the management of chronic pain in adults is also published (Williams 2012). Pharmacological interventions are also delivered to children and adolescents with chronic pain and a recent suite of Cochrane reviews have investigated the efficacy of opioids, paracetamol, antidepressants, anti‐epiletic drugs, and NSAIDs, but there is little evidence in young people (Cooper 2017a; Cooper 2017b; Cooper 2017c; Cooper 2017d; Eccleston 2017).

How the intervention might work

A variety of intervention strategies have been designed to reduce pain, increase comfort, and reduce associated disability and dysfunction in children with pain conditions. Behavioural strategies include relaxation training, biofeedback, and behavioural management programmes (e.g. teaching parents strategies to reinforce adaptive behaviours such as school attendance). Cognitive strategies include hypnosis, stress management, guided imagery, and cognitive coping skills (Palermo 2012). Cognitive‐behavioural therapy programmes incorporate elements of both behavioural and cognitive strategies. Parent interventions may include operant strategies, communication strategies, or problem‐solving skills (PST). PST is aimed at decreasing distress in parents of children with chronic pain by teaching problem‐solving skills, including steps to define a problem, generate possible solutions, implement a solution, and then evaluate (D'Zurilla 1999; D'Zurilla 2007). The efficacy of parent therapies are investigated in a sister review (Eccleston 2015) which is currently being updated.

Given that headache and abdominal pain are the most common types of recurrent pain in children, most of the treatment literature has focused on these two populations. By far the most commonly described treatment is relaxation training or biofeedback, or both, for headache (Law 2017), and recommendations have been made to offer psychological treatment as a matter of routine care for children with headaches (Ernst 2015; Palermo 2014). Older trials investigated efficacy of treatment predominantly for children with headache, comparing different elements of relaxation training and biofeedback and with different treatment doses, and treatment setting (clinic, school, and home; Griffiths 1996; Larsson 1987a). More recently, the quality of trials has improved to include larger sample sizes, multiple recruitment sites, and active control comparator conditions (e.g. Kashikar‐Zuck 2012; Levy 2016; Levy 2017; Palermo 2016; Powers 2013). Indeed, psychological therapies have now been developed and evaluated for children with a variety of chronic pain conditions including abdominal pain (Levy 2017), musculoskeletal pain (Kashikar‐Zuck 2012) and disease‐related pain (Barakat 2010; Levy 2016).

In clinical practice, psychological therapies for children with chronic pain are often delivered as one component of a multidisciplinary treatment programme (Palermo 2012). Such programmes aim to restore function and ameliorate pain through physical rehabilitation, psychological pain management strategies, and medical strategies (Palermo 2012). This treatment is typically offered in specialised outpatient clinics or more intensive day treatment or inpatient rehabilitation programs housed within tertiary medical centres (Hechler 2015).

Why it is important to do this review

Several reviews have documented the effectiveness of psychological therapies for children with headache, abdominal, and disease‐related pain (Fisher 2015; Huertas‐Ceballos 2008; Kibby 1998; Ng 2017; Walco 1999). Reviews have used data pooling techniques for studies of children with headache (Eccleston 2014; Fisher 2015; Ng 2017). In the previously published Cochrane review (Eccleston 2014), we found that psychological treatments were effective in reducing pain intensity in youth with headache and mixed chronic pain conditions. Fisher 2015 and Ng 2017 reported similar findings for children and adolescents with headache. Since the protocol of this review, there has been growing awareness of the limitations of smaller trials. Small studies are now a significant problem in pain research (Moore 2013) and are typically pilot studies or older trials. Therefore, in this update we have presented subgroup analyses to investigate smaller versus larger trials.

Objectives

The aim of this review was to update the published evidence on the efficacy of psychological treatments for chronic and recurrent pain in children and adolescents.

The primary objective of this updated review was to determine any effect of psychological therapy on the clinical outcomes of pain intensity and disability for chronic and recurrent pain in children and adolescents compared with active treatment, waiting‐list, or treatment‐as‐usual care.
The secondary objective was to examine the impact of psychological therapies on children's depressive symptoms and anxiety symptoms, and determine adverse events.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) comparing a credible psychological treatment, or a compound treatment with credible primary psychological content, to an active treatment, treatment‐as‐usual, or waiting‐list control. We judged content as credible if it was based on an extant psychological theory or framework. We excluded studies if the pain was associated with life‐limiting conditions (e.g. cancer) or if the therapy was delivered remotely using methods such as telephone or Internet.

We included studies if they:

were available as a full report of a RCT;
had a design that placed a psychological treatment as an active treatment of primary interest;
had a psychological treatment with definable psychotherapeutic content (although not necessarily delivered by someone with psychological qualifications);
were published (or electronically pre‐published) in a peer‐reviewed scientific journal;
had participants who reported chronic pain (i.e. pain that recurs or persists for at least three months duration);
had 10 or more participants in each treatment arm at the end‐of‐treatment assessment; and
included a psychological intervention that was delivered in person (face‐to‐face treatment), via telephone and at home via a written instruction booklet. We excluded studies delivered via technology (e.g. Internet, smartphone).

Types of participants

Children and adolescents (< 18 years) reporting chronic or recurrent pain in any body site, not associated with cancer. We separated conditions into headache conditions or mixed pain conditions. Mixed pain conditions (previously referred to as 'non‐headache conditions') refer to other types of chronic pain (e.g. recurrent abdominal pain, musculoskeletal pain, disease‐related pain).

Types of interventions

We included studies if at least one trial arm consisted of a psychological intervention (not delivered via technology), and a comparator arm consisted of active treatment, treatment‐as‐usual, or waiting‐list control. We excluded primary interventions that were delivered remotely via other methods (e.g. Internet, telephone).

Types of outcome measures

We assessed and recorded data from all measurement instruments reported in each study. We extracted data from the most appropriate measurement instruments for the outcomes below. We decided on the most appropriate measurement instruments as measures that were psychometrically established and frequently used across the studies.

Primary outcomes

Pain intensity
Pain‐related disability

Secondary outcomes

Depression
Anxiety
Adverse events

Search methods for identification of studies

Electronic searches

For this update, searching the following databases identified RCTs of any psychological therapy for paediatric chronic or recurrent pain:

CENTRAL (CRSO): searched 22/1/14 to 1/5/18
MEDLINE and MEDLINE in Process (OVID): searched January 2017 to 1/5/18
Embase (OVID): searched January 2014 to 2018 week 18
PsycINFO (OVID): searched January 2014 to May week 1 2018

Searching other resources

We searched clinicaltrials.gov for possible ongoing or completed trials in this area on 1 May 2018. We also examined reference lists and citation searches of included studies and relevant systematic reviews for other potential RCTs.

Data collection and analysis

Selection of studies

For this update, two authors sifted abstracts (EF, JD), and a third author (TP) arbitrated any disagreements. We selected studies for inclusion based on the following criteria: the study had to be a RCT in design and published in a peer‐reviewed journal, included children (< 18 years of age) who had chronic or recurrent pain (non‐cancer pain), included a psychological intervention as an active treatment, and had >10 participants in each arm at each extraction time point. Studies that had not been peer‐reviewed were excluded in order to keep the quality of included studies high. Consistent with our last update (Eccleston 2014), remotely‐delivered interventions were excluded from this review. We considered psychological interventions for inclusion if they had credible, recognisable psychological/psychotherapeutic content and were specifically designed to change the child's behaviour, cognition, or social‐environmental contingencies. All trials included in our previous systematic review and meta‐analysis were considered automatically eligible for inclusion (Eccleston 2014).

Data extraction and management

Data extracted included: details relating to the design of the study, the participants, primary diagnosis, characteristics of treatment (e.g. treatment setting, treatment delivered, and length of treatment), adverse events, outcome measurement tools used, and outcome data for computation of effect sizes. We contacted trial authors via email to obtain data necessary for effect size calculations if data were missing for primary outcomes of interest. We entered data suitable for pooling into RevMan 5.3 (RevMan 2014).

Assessment of risk of bias in included studies

We measured risk of bias using the recommended Cochrane 'Risk of bias' tool (Higgins 2011). We assessed five categories from this tool: random sequence generation (selection bias), allocation concealment (selection bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), and selective reporting (reporting bias). We excluded 'Blinding of participant or personnel' (performance bias) for the purposes of this review as we deemed it redundant because of the nature of delivering or receiving a psychological intervention. We made judgements on the 'risk of bias' categories using the following rules.

Selection bias (random sequence generation and allocation concealment)

We based judgements of random sequence generation on whether authors used a convincing method of randomisation. We judged studies that provided an adequate method of randomisation as at low risk of bias. We judged studies that did not provide a convincing method of randomisation as at unclear risk of bias. Studies that did not randomise participants were excluded from this review.

We based judgements of allocation concealment on whether there were convincing methods used for random allocation to take place. We did not deem studies as biased if participants were stratified by age or gender. We judged studies that provided a convincing method of allocation concealment (e.g. opaque envelopes) as at low risk of bias. We judged studies that did not report allocation concealment as at unclear risk of bias.

Detection bias (blinding of outcome assessment)

We based judgements of blinding of outcome assessment on whether the measures were taken by a third party who was blind to the treatment condition. We judged studies that reported an outcome assessor blinded to treatment as having a low risk of bias. We judged studies where a description was not provided, as at unclear risk of bias.

Attrition bias (incomplete outcome data)

We based judgements of incomplete outcome data on whether attrition was fully reported. Authors had to report attrition at each measurement time point (post‐treatment and follow‐up), and state whether there were any significant differences between completers and non‐completers. We judged studies as having low risk of bias if studies reported attrition and no differences between completers and non‐completers. We judged studies to be at unclear risk of bias if they reported attrition but did not report if there were differences between completers and non‐completers. Finally, we judged studies to be at high risk of bias if they did not report attrition.

Reporting bias (selecting reporting bias)

Finally, we based judgements of selective reporting bias on whether data could be fully extracted for analyses in this review. We judged studies that reported all outcomes in the manuscript as having low risk of bias. We marked this category as 'unclear bias' if authors provided data when requested. We allocated high risk of bias to studies that did not report all outcomes.

Measures of treatment effect

We combined all treatments labelled as psychological in the following meta‐analyses, and designated these as 'Treatment'. Similarly, we combined all control conditions and designated these as 'Control'. We combined the intervention or control arms if more than one intervention or control group was reported to create a single pairwise comparison in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). The studies were divided into two groups based on pain condition. We labelled the first group 'headache' and the second group 'mixed chronic pain conditions'. We also selected two assessment points: post‐treatment and follow‐up. Post‐treatment was the assessment point occurring soonest following treatment (often after a delay of several weeks to allow for recording of episodic pain), and follow‐up was the assessment point at least three months after the post‐treatment assessment point, but not more than 12 months. We selected the longer time point if there were two or more follow‐up assessments within this time frame. Therefore, we designed four separate comparisons comprising two forms of comparator (Treatment, Control) and two assessment time points (post‐treatment and follow‐up). They were labelled as follows.

Treatment versus control (headache) post‐treatment.
Treatment versus control (headache) follow‐up.
Treatment versus control (mixed chronic pain) post‐treatment.
Treatment versus control (mixed chronic pain) follow‐up.

For each comparison, we identified five outcomes labelled 'Pain', 'Disability', 'Depression', 'Anxiety', and 'Adverse events'. From each trial we selected the measure considered most appropriate for each outcome. We applied two rules to guide the choice of outcome measure. First, if an outcome measure was established and occurred frequently among studies it was selected over more novel instruments. Second, given a choice between single item and multi‐item self‐report tools, multi‐item tools were chosen on the basis of inferred increased reliability. Studies did not necessarily report data on all five outcomes. For headache treatments, we preferentially extracted the proportion of participants achieving a clinically significant (50%) reduction in headache frequency as the outcome for pain. We extracted the proportion of participants achieving a 50% reduction in pain intensity when headache frequency was not reported. Thus, for pain outcomes for headache treatments, we used relative ratios or risk ratios (RR) and we calculated numbers needed to treat for an additional beneficial outcome (NNTBs). All other extracted outcomes were continuous. Due to different populations and measures used, we analysed data using random effect models and standardised mean differences. We calculated effect sizes for the continuous outcomes which could be interpreted as follows; small = 0.2, medium = 0.5, large = 0.8 (Cohen 1992).

Due to the number of analyses in this review, we checked for model overfitting using Aikaike Information Criteria (AIC) comparing meta‐analysis and the additive combination of risk of bias and sample size in random‐effects meta‐regressions without an interaction term.

Unit of analysis issues

Randomisation occurred at the individual level. When studies included both children with headache and children with mixed pain conditions, we included the data in both analyses.

Dealing with missing data

We contacted study authors if there were missing data in the peer reviewed manuscript needed for data analysis.

Assessment of heterogeneity

We followed the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) to interpret heterogeneity in our meta‐analyses. We interpreted heterogeneity in analyses as follows: 0 to 40% as not important, 30 to 60% as moderate, 50 to 90% as substantial, and 75 to 100% as considerable heterogeneity.

Assessment of reporting biases

We checked for publication bias by using contour‐enhanced funnel plots, tests of funnel plot asymmetry, and meta‐regressions including total sample size as a moderator.

Quality of Evidence

Two review authors (EF, JD) rated the quality of the outcomes. We used the GRADE (Grades of Recommendation, Assessment, Development and Evaluation) system to rank the quality of the evidence using the RevMan 'Summary of Findings' table, and the guidelines provided in Chapter 12.2 of the CochraneHandbook for Systematic Reviews of Interventions (Higgins 2011).

The GRADE approach uses five considerations (study limitations, consistency of effect, imprecision, indirectness, and publication bias) to assess the quality of the body of evidence for each outcome. The GRADE system uses the following criteria for assigning grades of evidence:

High: we are very confident that the true effect lies close to that of the estimate of the effect;
Moderate: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of effect, but there is a possibility that it is substantially different;
Low: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect;
Very low: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

The GRADE system uses the following criteria for assigning a quality level to a body of evidence (Chapter 12, Higgins 2011).

High: randomised trials; or double‐upgraded observational studies;
Moderate: downgraded randomised trials; or upgraded observational studies;
Low: double‐downgraded randomised trials; or observational studies;
Very low: triple‐downgraded randomised trials; or downgraded observational studies; or case series/case reports.

Factors that may decrease the quality level of a body of evidence are:

limitations in the design and implementation of available studies suggesting high likelihood of bias;
indirectness of evidence (indirect population, intervention, control, outcomes);
unexplained heterogeneity or inconsistency of results (including problems with subgroup analyses);
imprecision of results (wide confidence intervals);
high probability of publication bias.

Factors that may increase the quality level of a body of evidence are:

large magnitude of effect;
all plausible confounding would reduce a demonstrated effect or suggest a spurious effect when results show no effect;
dose‐response gradient.

We decreased the grade rating by one (‐ 1) or two (‐ 2) if we identified:

Serious (‐ 1) or very serious (‐ 2) limitations to study quality;
Important inconsistency (‐ 1);
Some (‐ 1) or major (‐ 2) uncertainty about directness;
Imprecise or sparse data (‐ 1);
High probability of reporting bias (‐ 1).

There are sometimes reasons to downgrade an outcome directly to 'very low‐quality', as recommended by GRADE guidelines (Guyatt 2013). Where relevant, we describe all our reasons for making this judgement.

'Summary of findings' table

We included two 'Summary of findings' tables to present the main findings in a transparent and simple tabular format. In particular, we included key information concerning the quality of evidence, the magnitude of effect of the interventions examined, and the sum of available data on the outcomes, pain intensity, and disability post‐treatment and at follow‐up. We also included anxiety and depression post‐treatment, and anxiety at follow‐up.

Subgroup analysis and investigation of heterogeneity

In this update, we included two sets of subgroup analyses for each analysis, where data were available. We chose 20 participants as a cutoff to be consistent with other reviews (e.g. Eccleston 2015 (currently being updated); Williams 2012) which have raised their minimum n from 10 to 20 participants per arm. As a group of authors, we still feel it is premature to raise the minimum n to 20 participants per arm and therefore, in this update, we conducted the following subgroup analyses:

trials including fewer than 20 participants in each arm (n < 20);
trials including more than 20 participants in each arm (n > 20).

Sensitivity analysis

For this update, we conducted sensitivity analyses on analyses that included more than 10 studies. We removed studies where we judged high or unclear risk of reporting biases to determine whether higher quality studies reported similar effects.

Results

Description of studies

See: Characteristics of included studies; Characteristics of excluded studies.

Results of the search

Five separate searches were undertaken using databases from inception to May 2018. Details of the previous four searches can be found in Appendix 1. In the most recent search, we searched databases from January 2014 to May 2018 (see Figure 1). The current search yielded 3021 abstracts (2110 abstracts after duplication) and we included nine trials (Chen 2014; Daniel 2015; Greenley 2015; Hickman 2015; Levy 2016; Levy 2017; Palermo 2016; Wahlund 2003; Wahlund 2015). Due to the revised inclusion criteria to include trials delivered via the telephone, we also included one trial (Cottrell 2007) from Fisher 2015. Therefore, we included 10 new studies (n = 869 at post‐treatment), a total of 47 RCTs (49 papers) (Abram 2007; Alfven 2007; Barakat 2010; Barry 1997; Bussone 1998; Chen 2014; Cottrell 2007; Daniel 2015; Duarte 2006; Fichtel 2001; Gil 1997; Greenley 2015; Griffiths 1996; Grob 2013; Gulewitsch 2013; Hechler 2014; Hickman 2015; Humphreys 2000; Kashikar‐Zuck 2005; Kashikar‐Zuck 2012; Kroener‐Herwig 2002; Labbe 1984; Labbe 1995; Larsson 1987a; Larsson 1987b; Larsson 1990; Larsson 1996; Levy 2010; Levy 2016; Levy 2017; McGrath 1988; McGrath 1992; Osterhaus 1997; Palermo 2016; Passchier 1990; Powers 2013; Richter 1986; Robins 2005; Sanders 1994; Sartory 1998; Scharff 2002; Van der Veek 2013; Van Tilburg 2009; Vlieger 2007; Wahlund 2003; Wahlund 2015; Wicksell 2009).

Included studies

The total number of participants completing treatments from the 47 studies was 2884. Of the 47 studies, one had four treatment arms, 12 had three arms, and 30 had two arms. Twenty‐five studies included fewer than 20 participants per arm. We included these in separate subgroup analyses in this update (Barakat 2010; Barry 1997; Bussone 1998; Cottrell 2007; Duarte 2006; Fichtel 2001; Griffiths 1996; Grob 2013; Gulewitsch 2013; Hickman 2015; Humphreys 2000; Kashikar‐Zuck 2005; Kroener‐Herwig 2002; Labbe 1984; Labbe 1995; Larsson 1987a; Larsson 1987b; Larsson 1990; Larsson 1996; Osterhaus 1997; Richter 1986; Sartory 1998; Scharff 2002; Van Tilburg 2009; Wicksell 2009). Two studies delivered treatment to parents only (Levy 2017; Palermo 2016). Incidentally, both studies included more than 20 participants per arm. Across all studies, the mean number of participants per study at the end of treatment was 68 (standard deviation (SD) = 45.04). For studies including fewer and more than 20 participants per arm, the mean number of participants per study at the end of treatment was 36 (SD = 8.17) for studies including less than 20 participants per arm, and 91 (SD = 48.61) for studies including more than 20 participants per arm. Girls outnumbered boys in most studies (mean 66% girls). Child age was reported in 44 studies (Mean 12.65 years, SD 2.21 years).

Young people were recruited from a range of healthcare settings and other sources. Thirty trials recruited from hospital or clinic settings and 12 recruited participants from a mixture of of advertisements, schools, community, or hospital settings. Five trials did not report their recruitment source.

There were 23 trials of treatments for children with headache (including migraine). Of the remainder, 10 were for abdominal pain (Alfven 2007; Duarte 2006; Grob 2013; Humphreys 2000; Levy 2010; Levy 2017; Robins 2005; Sanders 1994; Van der Veek 2013; Van Tilburg 2009) and two studies treated participants with either a primary diagnosis of abdominal pain or irritable bowel syndrome (Gulewitsch 2013; Vlieger 2007). Two studies treated children with fibromyalgia (Kashikar‐Zuck 2005; Kashikar‐Zuck 2012), two studies included children with temporomandibular disorders (Wahlund 2003; Wahlund 2015), three were for the treatment of pain associated with sickle cell disease (Barakat 2010; Daniel 2015; Gil 1997), and two studies treated children and adolescents with inflammatory bowel disease (Greenley 2015; Levy 2016). Finally, three studies included mixed pain conditions including headache and mixed chronic pain conditions (Hechler 2014; Palermo 2016; Wicksell 2009) For these three studies, we included data in both analyses, as appropriate.

We classified treatment arms on the basis of their content and of the label given by the study authors. We classified interventions into four broad groups. The first is best described as behavioural, typically relaxation‐based, with or without biofeedback, and including autogenic or hypnotherapeutic content (Bussone 1998; Chen 2014; Fichtel 2001; Labbe 1984; Labbe 1995; Larsson 1987a; Larsson 1987b; Larsson 1990; Larsson 1996; McGrath 1988; McGrath 1992; Passchier 1990; Wahlund 2003; Wahlund 2015; Van Tilburg 2009; Vlieger 2007). The second is best described as cognitive behavioural therapy, including coping skills training (Abram 2007; Alfven 2007; Barakat 2010; Barry 1997; Duarte 2006; Gil 1997; Griffiths 1996; Grob 2013; Gulewitsch 2013; Hickman 2015; Humphreys 2000; Kashikar‐Zuck 2005; Kashikar‐Zuck 2012; Kroener‐Herwig 2002; Levy 2010; Levy 2017; McGrath 1992; Osterhaus 1997; Powers 2013; Richter 1986; Robins 2005; Sanders 1994; Sartory 1998; Scharff 2002; Van der Veek 2013; Wicksell 2009). Problem‐solving therapy was the third category, which involved problem‐solving strategies delivered to the family (Daniel 2015). We also included interventions delivered exclusively to parents, which is new to this update. Parent interventions included cognitive‐behavioural therapy (Levy 2016) and problem‐solving therapy (Palermo 2016), with a primary aim to improve parenting behaviour or parent mental health, or both, and a secondary aim to improve children's pain, disability, and/or emotional functioning. One trial (Hechler 2014) evaluated the efficacy of a three‐week intensive inpatient pain rehabilitation programme, which included psychological therapy.

We categorised different control conditions into either treatment‐as‐usual (n = 8), active (e.g. education, sham therapy; n = 27) or wait‐list (n = 12). We were able to extract post‐treatment data for outcomes included in this review from 31 studies post‐treatment, and for 16 studies at follow‐up. Thirty‐nine studies reported the treatment length; this was typically short (mean = 6 hours 36 minutes for headache studies, mean = 5 hours for mixed chronic pain studies (Table 3, note ‐ these averages excluded Hechler 2014 as they delivered an in‐patient programme which would have skewed the findings). Six studies did not report the duration of psychological treatment (Alfven 2007; Chen 2014; Humphreys 2000; Larsson 1990; Sartory 1998; Wahlund 2003).

1. Duration of treatment and setting by condition.

Headache Studies
Author	Illness	Treatment duration (hours)	Setting
Abram 2007	Headache	1.5	Clinic
Barry 1997	Headache	3	Unknown
Bussone 1998	Headache	7	Clinic
Chen 2014	Headache	Unknown	Unknown
Cottrell 2007	Headache	4 hours plus tasks	Home
Fichtel 2001	Headache	6.75	Clinic
Griffiths 1996	Headache	12	Home/clinic
Hickman 2015	Headache	2.8	Home/clinic
Hechler 2014*	Mixed	136.5 (3‐week intensive therapy)	Clinic
Kroener‐Herwig 2002	Headache	12	Clinic
Labbe 1984	Headache	6.7	Clinic
Labbe 1995	Headache	7.5	Clinic
Larsson 1987a	Headache	6.75	School
Larsson 1987b	Headache	5	School
Larsson 1990	Headache	Unknown	Home
Larsson 1996	Headache	3.3	Clinic
McGrath 1988	Headache	6	Unknown
McGrath 1992	Headache	8	Home/clinic
Osterhaus 1997	Headache	9.3	Clinic
Passchier 1990	Headache	2.5	School
Palermo 2016*	Mixed	5	Home/clinic
Powers 2013	Headache	13	Clinic
Richter 1986	Headache	9	Unknown
Sartory 1998	Headache	Unknown	Clinic
Scharff 2002	Headache	4	Clinic
Wicksell 2009*	Mixed	10	Clinic
Mixed Chronic Pain Studies
Author	Illness	Treatment duration hours)	Setting
Alfven 2007	RAP	Unknown	Clinic
Barakat 2010	SCD	6	Home
Daniel 2015	SCD	9.5	Home/clinic
Duarte 2006	RAP	3.3	Unknown
Gil 1997	SCD	0.75	Clinic
Greenley 2015	IBD	4	Home
Grob 2013	RAP	9	Clinic
Gulewitsch 2013	RAP/IBS	2	Clinic
Hechler 2014*	Mixed	136.5 (3‐week intensive therapy)	Clinic
Humphreys 2000	RAP	Unknown	Clinic
Kashikar‐Zuck 2005	Fibromyalgia	6	Clinic
Kashikar‐Zuck 2012	Fibromyalgia	7.5	Unknown
Levy 2010	RAP	4	Home/clinic
Levy 2016	IBD	3.5	Home/clinic
Levy 2017	RAP	3	Home/clinic
Palermo 2016*	Mixed	5	Home/clinic
Robins 2005	RAP	3.5	Clinic
Sanders 1994	RAP	6	Clinic
Van der Veek 2013	RAP	4.5	Clinic
Van Tilburg 2009	RAP	1.8	Home
Vlieger 2007	RAP/IBS	5	Clinic
Wahlund 2003	TMD	Unknown	Unknown
Wahlund 2015	TMD	6	Clinic
Wicksell 2009*	Mixed	10	Clinic

Open in a new tab

*Headache and mixed chronic pain studies were entered twice. ^{IBD: inflammatory bowel disease IBS: irritable bowel syndrome JIA: Juvenile idiopathic arthritis RAP: Recurrent abdominal pain SCD: Sickle cell disease TMD: temporomandibular disorders}

The setting of treatment delivery varied between studies (Table 3). Twenty‐three studies delivered treatment in a clinic, eight were based either in a clinic or at home, and five delivered treatment to families at home, so exposure to treatment was uncontrolled. One study delivered treatment in an inpatient hospital setting. A further three were based in schools and seven studies were unclear as to the location of treatment delivery. Home maintenance or practice of treatment was a common and important feature of many studies, but overall treatment exposure including home practice was not reported.

Excluded studies

Nineteen studies were excluded, of which four were new to this update (Jastrowski Mano 2013; Korterink 2016; Rapoff 2014; Rutten 2017). We excluded Connelly 2006, Hicks 2006, Palermo 2009, Rapoff 2014, Stinson 2010, Trautmann 2010 as they were delivered remotely, so did not meet the new inclusion criteria. Eight studies were excluded as they had fewer than 10 participants in a treatment arm at the end of treatment (Fentress 1986; Jastrowski Mano 2013;Kroener‐Herwig 1998; Larsson 1986; Sanders 1989; Trautmann 2008; Weydert 2006; Youssef 2009), three studies were judged to have insufficient psychological content in the treatment (Koenig 2013; Korterink 2016; Olness 1987), one study reported only follow‐up data of more than one year (Vlieger 2012), and one trial was had a non‐inferiority design (Rutten 2017).

Risk of bias in included studies

We rated all included studies for risk of bias in five categories: random sequence generation (selection bias), allocation concealment (selection bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), and selective reporting (reporting bias) (Figure 2; Figure 3).

Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.

Methodological quality summary: review authors' judgements about each methodological quality item for each included study.

Allocation

Random sequence generation

We found 16 trials used a convincing method of randomisation which we judged as having low risk of bias and we found a further 31 trials were judged at unclear risk of bias on random sequence generation, as they did not provide an adequate description of the method of randomisation. We scored none as having high risk of bias as the inclusion criteria required that studies randomised participants.

Allocation concealment

We found 12 trials adequately concealed allocation of participants, 31 had unclear allocation concealment, and we judged four studies as having high risk of bias.

Blinding

We found 11 trials that described adequate blinding of outcome assessors and, therefore, we allocated these trials as having low risk of bias. We found the remaining trials did not report on blinding and so were judged as having unclear risk of bias.

Incomplete outcome data

We found 17 studies reported attrition fully, reporting that there was no significant difference between completers and non‐completers. We found 24 studies only partially reported attrition and so we judged them to be at unclear risk of bias and we judged six studies to have a high risk of bias as they did not report attrition.

Selective reporting

We found 25 studies reported data fully, which could be extracted and used in analyses; seven studies did not fully report data in the published trial, but provided data when contacted via email; we judged 15 studies to have high risk of bias for selective reporting as they did not provide full extractable data.

Effects of interventions

See: Table 1; Table 2

For this update, we conducted subgroup analyses of trials that included fewer than 20 participants per arm or more than 20 participants per arm. A scorecard outlining the treatment effects is included in Table 4.

2. Scorecard of findings.

Psychological treatments for children and adolescents with headache pain
	Pain	Disability	Depression	Anxiety
Post‐treatment	Effect (15)	No effect (6)	No effect (6)	No effect (7)
< 20/arm	Effect (13)	No effect (2)	No effect (3)	No effect (4)
> 20/arm	Effect (2)	Effect (4)	No effect (3)	No effect (3)
Follow‐up	No effect (5)	Effect (3)	No effect (3)	No effect (4)
< 20/arm	Effect (4)	Unknown*	Unknown*	No effect (2)
> 20/arm	Unknown*	Effect (2)	No effect (2)	No effect (2)
Psychological treatments for children and adolescents with mixed pain conditions
	Pain	Disability	Depression	Anxiety
Post‐treatment	Effect (16)	Effect (14)	No effect (8)	Effect (8)
< 20/arm	Effect (7)	Effect (6)	No effect (2)	Unknown*
> 20/arm	No effect (9)	Effect (8)	No effect (6)	Effect (7)
Follow‐up	No effect (9)	Effect (9)	No effect (7)	No effect (8)
< 20/arm	Effect (2)	No effect (2)	Unknown*	Unknown*
> 20/arm	No effect (7)	Effect (7)	No effect (6)	No effect (7)

Open in a new tab

Unknown (no data); Unknown* (only one study); Number in brackets denotes number of studies in analysis.

Treatment versus control (headache) post‐treatment

Headache frequency/intensity, post‐treatment

We included 15 studies with 644 participants in an analysis of the effects of treatment on headache frequency/intensity post‐treatment (Barry 1997; Fichtel 2001; Griffiths 1996; Kroener‐Herwig 2002; Labbe 1984; Labbe 1995; Larsson 1987a; Larsson 1987b; Larsson 1990; Larsson 1996; McGrath 1992; Osterhaus 1997; Powers 2013; Sartory 1998; Scharff 2002). Overall, the analysis suggested a risk ratio (RR) of 2.35 (95% confidence interval (CI) 1.67 to 3.30; P < 0.01) for a beneficial reduction in headache frequency (number needed to treat for an additional beneficial outcome (NNTB) = 2.86) (Analysis 1.1; Figure 4). However, the GRADE quality rating for this outcome was very low, meaning we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect. We downgraded this outcome due to limitations in design, imprecision of results, and asymmetrical funnel plot.

1.1 — Comparison 1: Treatment versus control (headache) post‐treatment, Outcome 1: Pain

Forest plot of comparison: 1 Treatment versus control (headache) post‐treatment, outcome: 1.1 Pain.

Subgroup analysis:

We found 13 studies (n = 437) that included fewer than 20 participants per arm for a beneficial reduction in headache frequency post‐treatment (RR 2.86; 95% CI 1.73 to 4.72, P < 0.01; NNTB = 2.59). Two studies (McGrath 1992; Powers 2013; n = 207), including more than 20 participants per arm, also had a beneficial effect on reducing headache frequency (RR 1.88, 95% CI 1.36 to 2.58, P < 0.01; NNTB = 3.58).

Disability, post‐treatment

We included six studies with 446 participants in the analysis of the effects of treatment on disability (Chen 2014; Hechler 2014; Hickman 2015; Palermo 2016; Powers 2013; Wicksell 2009). Overall, we did not find a beneficial effect of psychological therapies on reducing disability in children with headache (Standardised mean difference (SMD) ‐0.26, 95% CI ‐0.56 to 0.03, P = 0.08; Analysis 1.2). The GRADE quality rating for this outcome was very low, downgraded due to limitations in the design, unexplained heterogeneity, and for sparse data.

1.2 — Comparison 1: Treatment versus control (headache) post‐treatment, Outcome 2: Disability

Subgroup analysis:

First, in two studies (Hickman 2015; Wicksell 2009, n = 61) that included fewer than 20 participants per arm, we did not find a beneficial effect of psychological treatment (SMD 0.04, 95% CI ‐0.47 to 0.54, P = 0.88). Second, for studies that included more than 20 participants per arm (Chen 2014; Hechler 2014; Palermo 2016; Powers 2013, n = 385), we found a medium beneficial effect of treatment on reducing disability (SMD ‐0.35, 95% CI ‐0.69 to 0.00, P = 0.05).

Depression, post‐treatment

We entered six studies with 400 participants in an analysis of the effects of treatment on depression (Griffiths 1996; Hechler 2014; Hickman 2015; Palermo 2016; Powers 2013; Wicksell 2009). We found psychological therapies did not show a beneficial effect for reducing depression for children with headache (SMD ‐0.08, 95% CI ‐0.28 to 0.11, P = 0.41; Analysis 1.3). We judged this outcome as having a very low‐quality rating, meaning we have very little confidence in the effect estimate. We downgraded the outcome due to limitations of study design, high probability of publication bias, and sparse data.