CONTEXT:

Migraine is a common neurologic disorder in children and adolescents. However, a comparison of multiple nonpharmacological treatments is lacking.

OBJECTIVE:

To examine whether nonpharmacological treatments are more effective than waiting list and whether there are differences between interventions regarding efficacy.

DATA SOURCES:

Systematic review and network meta-analysis of studies in Medline, Cochrane, Embase, and PsycINFO published through August 5, 2019.

STUDY SELECTION:

Randomized controlled trials of nonpharmacological treatments in children and adolescents diagnosed with episodic migraine.

DATA EXTRACTION:

Effect sizes, calculated as standardized mean differences (SMDs) for the primary outcome efficacy, were assessed in a random-effects model.

RESULTS:

Twelve studies (N = 576) were included. When interventions were classified into groups on the basis of similarity of treatment components, self-administered treatments, biofeedback, relaxation, psychological treatments, and psychological placebos were significantly more effective than waiting list with effect sizes ranging between SMD = 1.14 (95% confidence interval, 0.09 to 2.19) for long-term psychological placebos to SMD = 1.44 (95% confidence interval, 0.26 to 2.62) for short-term self-administered treatments. However, when all interventions were examined individually (ie, 1 node per intervention), none were significantly more effective compared with waiting list, mainly because of lack of statistical power.

LIMITATIONS:

Because of our focus on pediatric migraine, only a small number of studies could be included.

CONCLUSIONS:

Our findings reveal that components of nonpharmacological interventions are effective in treating pediatric migraine. Some effects have to be interpreted carefully because they are based on small studies. Future researchers should identify factors associated with individual responses in large, multicentered studies.

Migraine affects all aspects of a child’s functioning and can lead to negative affective states (eg, anger, depression, anxiety) and increased psychosocial problems (eg, school absences, decreased home and family interactions, and decreased socialization with peers).1,2  The quality of life ratings of children suffering from migraine are considerably lower compared with healthy peers and similar to that of children with arthritis and cancer.2  Treatment of migraine in children and adolescents is usually threefold: acute therapy, preventive treatment, and biobehavioral therapy.3  It is important to note that generally, pharmacologic and nonpharmacologic interventions share similar treatment goals, namely to reduce headache days and pain-related disability. Acute treatment includes nonspecific medications (including nonsteroidal anti-inflammatory drugs and general pain relievers), whereas prophylactic medication is considered preventive treatment to reduce the frequency of migraine attacks and improve disability (see our network meta-analysis [NMA] on prophylactic medication).4  Finally, biobehavioral treatment aims to improve migraine management and is divided into the components of treatment adherence, lifestyle management, and psychological interventions.1  Psychological interventions include cognitive behavioral therapy (CBT), relaxation (eg, progressive muscle relaxation), biofeedback, mindfulness meditations, acceptance-based treatments, parents and family education, consultation with school, and hypnosis and training in self-hypnosis.5  This biopsychosocial and multidisciplinary approach is considered essential for effective migraine management. Interestingly, researchers in a recent study looking at alterations in brain function after CBT for migraine not only found a reduction in headache frequency, but also changes in brain areas directly involved in pain processing.6  This points to the importance of psychological treatments such as CBT that go beyond improved management of the pain problem. Researchers in previous studies on the efficacy of psychological interventions have demonstrated positive results, but of moderate effect sizes; a meta-analysis on psychological treatments of recurrent headaches (including migraine or tension-type headache or combined) found a significant, but only moderate, treatment effect compared with waiting list or no treatment control groups.7  The beneficial effects of treatment seem to be maintained or even slightly increased posttreatment and at follow-up.7,8  Yet researchers of a Cochrane review who looked at psychological interventions for children with headache pain and mixed pain conditions did not find a beneficial effect on the reduction of disability, depression, or anxiety symptoms for children and adolescents suffering from headaches, including migraine.9  For chronic migraine (ie, 15 or more headache days per month with a majority having migraine features10 ), researchers in 1 trial reported a combination of CBT and amitriptyline to result in a greater reduction in days with headache compared with headache education and amitriptyline, lending support to the efficacy of CBT for the treatment of migraine in children and adolescents.11  Notably, most meta-analyses combined different forms of headache and migraine in their inclusion criteria.

With regard to other nonpharmacological interventions, biofeedback, especially thermal biofeedback (ie, volitional hand-warming), has been studied extensively in pediatric migraine and results suggest a considerable robustness of biofeedback efficacy.12  For relaxation, the results appear to be mixed: although several studies revealed that progressive muscle relaxation seems to be effective,13  autogenic relaxation revealed limited effects compared with waiting list controls and no effect when compared with an attention control condition in a review of the literature.14  This highlights a common issue, namely that effect sizes of an intervention largely depend on the comparison that is made: usually, treatments show larger effect sizes when compared with a passive control group (such as waiting list, treatment as usual) and smaller effect sizes when compared with an active control group (such as education, support groups).15  Also, a considerable placebo effect in children and adolescents has been reported not only for psychological disorders16  but also for headaches, including migraine.5,17 

We set out to conduct a systematic review and NMA. First, we aimed to examine whether nonpharmacological treatments are more effective than waiting list. Second, we aimed to systematically compare the different nonpharmacological treatment options for pediatric migraine relative to each other with regard to their efficacy and safety. The use of NMA uniquely allows the integration of multiple direct and indirect comparisons of nonpharmacological interventions and hence generate strong conclusions about their relative efficacy and safety.

This systematic review and NMA is registered with PROSPERO (CRD42019146666).

We conducted a systematic review and NMA in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) statement.18,19  We searched Medline, Cochrane, Embase, and PsycINFO from inception until August 5, 2019. Furthermore, additional trials were identified from an existing systematic review of migraine treatments.20  For search terms, see Supplemental Information 1. In total, the search led to 9012 articles (Supplemental Fig 3). Notably, the search also identified pharmacological studies; these results have been published elsewhere.4  The screening and selection processes were conducted independently by 5 coauthors (C.L., H.K., J.K., J.B., and T.L.L. in pairs of 2).

Randomized controlled trials (RCTs) of nonpharmacological interventions for children and adolescents <18 years were included in this study. Participants had to have a diagnosis of episodic migraine (with or without aura) according to the International Headache Society (IHS) criteria, or criteria for migraine diagnosis had to be in close agreement with the IHS classification. Trials had to specifically state that they were focusing on pediatric migraine but were not required to specify the exact diagnosis of migraine to be included. Eligible trial designs included RCTs that make head-to-head comparisons of at least 2 nonpharmacological interventions as well as RCTs that compare at least 1 nonpharmacological intervention with a control group. To be included, trials had to report at least 1 migraine-related outcome. Crossover studies were only included if we were able to extract the results of the first period separately. Studies in which migraine was associated with other neurologic disorders as well as studies on menstrual migraine were excluded.

All eligible data were extracted in duplicate (C.L., H.K., J.B., J.K., and T.L.L. in pairs of 2) on a standardized form. We resolved disagreements through consensus and, if indicated, consultation with a third reviewer. For continuous outcomes, means and SDs were extracted. If SDs were not reported, we calculated them from SEs, confidence intervals (CIs), or other measures.21,22  In cases where we were not able to calculate SDs, we imputed them with the help of SDs reported for the same outcome measure.23  If the sample size was missing in the table of analysis, we used the sample size of the baseline data. In cases where the continuous outcome data were not reported, we extracted the number of patients who fulfilled the response criterion as defined by the authors of the included study.

We defined the primary efficacy outcomes in line with the recent guidelines of the IHS for children and adolescents with migraine.24  We created a hierarchy of measures and chose the highest available priority in each individual study. In descending order, the primary efficacy outcome measures were defined as (1) number of headache days per month, (2) the number of migraine days per month, (3) frequency of headache attacks (means and SDs), (4) frequency of migraine attacks (means and SDs), or (5) headache index and activity. If no information on these primary efficacy outcomes were available, the proportion of responders was used instead. Preferably, we defined responders as patients with at least 50% reduction in the number of headache days. If these data were not reported, we used (in descending order of preference) patients (1) with at least a 50% reduction in the number of migraine days, (2) with at least a 50% reduction in frequency of headache attacks, (3) with at least 50% reduction in frequency of migraine attacks, (4) with at least a 50% reduction in headache index, or the global assessment of improvement by patients and physicians.

We defined different time periods to reduce between-study heterogeneity and to increase the comparability between studies: 8 weeks or 2 months after randomization (post), 3 to 4 months after randomization (follow-up [FU] 1), 5 to 6 months after randomization (FU 2), and >6 months after randomization (FU 3). For the main analysis, we considered outcomes reported between 3 and 4 months after randomization. If no data were available for that time period, outcomes at 8 weeks or 2 months after randomization were applied. For the long-term analysis, we considered all reported outcomes from FU 3, FU 2, and FU 1 with a priority for longer follow-up.

We assessed the quality of the included studies using the Cochrane risk of bias tool for RCTs.25  Two reviewers rated each study, with conflicts resolved through consensus or, if necessary, consultation with a third reviewer.

For our primary efficacy outcomes (ie, continuous outcome data), we calculated the effect sizes of the interventions applying the standardized mean difference (SMD). The magnitude of the SMDs was interpreted as small, moderate or large, with 0.20, 0.50, and 0.80 SD units, respectively.26  For our secondary efficacy outcomes (ie, categorical outcome data), we calculated odds ratios as effect sizes between groups22  and transformed them into SMDs in line with the recommendations in the Cochrane Handbook of Systematic Reviews.27  We decided to primarily calculate with continuous outcome data, because dichotomizing continuous scores into categorical scores is associated with a loss of information, reduced power, and leading to an artificial boundary.28  We decided to apply random-effects models rather than fixed-effects models because the included studies were assumed to be heterogeneous.29 

We conducted NMAs using R package “netmeta,”30  which applies a frequentist method based on a graph-theoretical approach.31  With regard to the nodes in the network, we applied 2 different approaches (ie, lumping versus splitting approach). Applying the lumping approach, we formed broad classifications of interventions with similar components allocated into groups. For this broad classification, we modeled interventions as “class-effects,” assuming interventions to be similar within 1 class. For the splitting approach, we differentiated between all the interventions and plotted networks accordingly. We decided to rank the included interventions, using P scores that are the frequentist analog of the surface under the cumulative ranking curve32  and can be similarly interpreted. P scores (range: 0–1) represent the extent of certainty that a treatment is better than another, averaged over all competing treatments. For all treatment comparisons in a NMA, we assumed a common between-study heterogeneity. Different statistics were used to quantify heterogeneity: the (within design) Q statistic,33  the between-study variance τ2, and the heterogeneity statistic I2.34  The I2 value can be interpreted as follows: 0% to 40%: might not be important; 30% to 60%: may represent moderate heterogeneity; 50% to 90%: may represent substantial heterogeneity; 75% to 100%: represents considerable heterogeneity.35  We calculated the agreement (ie, consistency) between direct and indirect treatment effects, with local (ie, separating direct from indirect evidence)36  and global (ie, design-by-treatment interaction test)37  statistical approaches.

The certainty of evidence for the network estimates of the efficacy outcomes was evaluated by using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) ratings,38  which were conducted in CINeMA (Confidence in Network Meta-Analysis).39  In GRADE, the quality of a body of evidence is defined as the study limitations, imprecision, inconsistency, indirectness and publication bias.38  Prediction intervals were calculated that consider between-study heterogeneity and estimate what true treatment effects can be expected in future studies under similar settings.40  To date, no standardized methodology to assess across-studies bias (publication bias) in NMAs exists. Therefore, a comparison-adjusted funnel plot and the Egger test for funnel plot asymmetry were examined.41 

In total, 14 RCTs met our inclusion criteria, but we had to exclude 2 studies because they did not connect with the other studies in the network42,43  (see Supplemental Table 2). Overall, 12 RCTs (comprising 576 patients) conducted between 1986 and 2019 and comparing various interventions (ie, relaxation stress management, relaxation education, relaxation, biofeedback, biofeedback stress management, biofeedback relaxation education, transcendental meditation, autogenic feedback, autogenic training, psychological treatments, progressive muscle relaxation, self-administered psychological treatments, education, and hypnotherapy) with different control groups (ie, psychological placebo, sham biofeedback, and waiting list) could be included.4455  The characteristics of the 12 studies included in the NMA are presented in Table 1. For one of the included studies, authors provided us the summary data for the subsample of pediatric migraine: results were only reported across different forms of primary headaches in the published article.53  Researchers in 1 study reported only categorical data52 ; those in another study reported continuous data insufficiently and categorical data had to be used.45  In both cases, odds ratios were transformed into SMDs. A list of outcome measure that were analyzed for the individual trials can be found in the Supplemental Table 3.

TABLE 1

Demographics and Study Characteristics

Study IDInterventionaIntervention 2aIntervention 3aDurationControlaDiagnosisDiagnostic CriteriaMean Age per Group, y (SD)% Female per GroupCountryNo. Study SitesRisk of Biasb
Allen and Shriver,44  1998 Biofeedback Biofeedback plus stress management — 6 wk — Migraine IHS 1988 I1: 12.3 I2: 12.1 (SD: nr.) I1: 69.2% I2: 100% United States Unclear Moderate 
McGrath et al,45  1992 Psychological treatment Self-administered treatment — 8 wk Psychological placebo Migraine Intermittent, paroxysmal headache All groups: range = 11–18 (SD: nr.) All groups: 72.4% Canada Moderate 
Oelkers-Ax et al,42  2008c Music therapy Butterbur root extract — 12 wk Placebo Migraine with and without aura IHS 1988 I1: 9.9 (SD: 1.4), I2: 10.6 (SD: 1.2), control: 10.6 (SD: 1.5) I1: 15.0%, I2: 90.0%, control: 31.6% Germany Moderate 
Pintov et al,43  1997c Acupuncture — — 10 wk Sham acupuncture Migraine Prensky criteria I1: 9.8 (SD: 1.2), control: 10.4 (SD: 1.6) I1: 58.3%, control: 60.0% Israel Low 
Rapoff et al,46  2014 Psychological treatment — — 4 wk (then: 2 wk after) Education Migraine with and without aura IHS 2004 I1:10.2 (SD: 2.0), control: 10.2 (SD: 1.5) I1: 55.6%, control: 88.2% United States Moderate 
Sartory et al,47  1998 Relaxation plus stress management Biofeedback plus stress management Metoprololc 6 wk — Migraine with and without aura IHS 1988 I1:11.4 (SD: 2.4), I2: 11.8 (SD: 2.0), I3: 10.5 (1.9) I1: 40.0%, I2: 46.7%, I3: 26.6% Germany Unclear Moderate 
Fentress et al,48  1986 Biofeedback plus relaxation plus education Relaxation plus education — 11 wk Waiting list Migraine Recurrent intermittent headache All groups: 10.1 (SD: nr.) All groups: 61.11% United States Unclear Moderate 
Fichtel and Larsson,49  2001 Progressive muscle relaxation — — 8–10 session (weeks unclear) Waiting list Migraine and TTH or only migraine IHS 1988 All groups: 15.4 (SD: 1.55) I1: 75%, I2: 62.5% Sweden Moderate 
Labbé and Williamson,55  1984 Autogenic feedback — — 7 wk Waiting list Migraine Secondary diagnosis of vascular or migraine headache I1: 10.6, control: 10.8 (SD: nr.) IF1: 50%, control: 50% United States Unclear Moderate 
Labbé,50  1995 Autogenic feedback Autogenic training — 7 wk Waiting list Vascular or migraine headache Reported at least 2 migraine headaches per month All groups: 12.0 (SD: nr.) All groups: 43.33% United States Unclear Moderate 
McGrath et al,51  1988 Progressive muscle relaxation Self-administered treatment — 6 wk Psychological placebo Migraine Intermittent, paroxysmal headache All groups: 13.1 (SD: nr.) All groups: 69.7% Canada Moderate 
Scharff et al,52  2002 Hand-warming biofeedback plus stress management Sham (hand-cooling) biofeedback — 6 wk Waiting list Migraine with and without aura IHS 1988 I1: 13.3 (SD: 2.5), I2: 13.2 (SD: 2.0), control: 12.0 (SD: 2.7) I1: 69.2%, I2: 45.5%, control: 83.3% United States Moderate 
Jong et al,53  2019 Hypnotherapy Transcendental meditation Progressive muscle relaxation 12 wk — Migraine ICHD-2 criteria I1: 11.2, I2: 12.2, I3: 11.5 I1: 43.0%, I2: 33%, I3: 50% Nether-lands Low 
Richter et al,54  1985 Relaxation Psychological treatment — 6 wk Psychological placebo Migraine Intermittent, paroxysmal headache All groups: 12.9 All groups: 67% Canada Moderate 
Study IDInterventionaIntervention 2aIntervention 3aDurationControlaDiagnosisDiagnostic CriteriaMean Age per Group, y (SD)% Female per GroupCountryNo. Study SitesRisk of Biasb
Allen and Shriver,44  1998 Biofeedback Biofeedback plus stress management — 6 wk — Migraine IHS 1988 I1: 12.3 I2: 12.1 (SD: nr.) I1: 69.2% I2: 100% United States Unclear Moderate 
McGrath et al,45  1992 Psychological treatment Self-administered treatment — 8 wk Psychological placebo Migraine Intermittent, paroxysmal headache All groups: range = 11–18 (SD: nr.) All groups: 72.4% Canada Moderate 
Oelkers-Ax et al,42  2008c Music therapy Butterbur root extract — 12 wk Placebo Migraine with and without aura IHS 1988 I1: 9.9 (SD: 1.4), I2: 10.6 (SD: 1.2), control: 10.6 (SD: 1.5) I1: 15.0%, I2: 90.0%, control: 31.6% Germany Moderate 
Pintov et al,43  1997c Acupuncture — — 10 wk Sham acupuncture Migraine Prensky criteria I1: 9.8 (SD: 1.2), control: 10.4 (SD: 1.6) I1: 58.3%, control: 60.0% Israel Low 
Rapoff et al,46  2014 Psychological treatment — — 4 wk (then: 2 wk after) Education Migraine with and without aura IHS 2004 I1:10.2 (SD: 2.0), control: 10.2 (SD: 1.5) I1: 55.6%, control: 88.2% United States Moderate 
Sartory et al,47  1998 Relaxation plus stress management Biofeedback plus stress management Metoprololc 6 wk — Migraine with and without aura IHS 1988 I1:11.4 (SD: 2.4), I2: 11.8 (SD: 2.0), I3: 10.5 (1.9) I1: 40.0%, I2: 46.7%, I3: 26.6% Germany Unclear Moderate 
Fentress et al,48  1986 Biofeedback plus relaxation plus education Relaxation plus education — 11 wk Waiting list Migraine Recurrent intermittent headache All groups: 10.1 (SD: nr.) All groups: 61.11% United States Unclear Moderate 
Fichtel and Larsson,49  2001 Progressive muscle relaxation — — 8–10 session (weeks unclear) Waiting list Migraine and TTH or only migraine IHS 1988 All groups: 15.4 (SD: 1.55) I1: 75%, I2: 62.5% Sweden Moderate 
Labbé and Williamson,55  1984 Autogenic feedback — — 7 wk Waiting list Migraine Secondary diagnosis of vascular or migraine headache I1: 10.6, control: 10.8 (SD: nr.) IF1: 50%, control: 50% United States Unclear Moderate 
Labbé,50  1995 Autogenic feedback Autogenic training — 7 wk Waiting list Vascular or migraine headache Reported at least 2 migraine headaches per month All groups: 12.0 (SD: nr.) All groups: 43.33% United States Unclear Moderate 
McGrath et al,51  1988 Progressive muscle relaxation Self-administered treatment — 6 wk Psychological placebo Migraine Intermittent, paroxysmal headache All groups: 13.1 (SD: nr.) All groups: 69.7% Canada Moderate 
Scharff et al,52  2002 Hand-warming biofeedback plus stress management Sham (hand-cooling) biofeedback — 6 wk Waiting list Migraine with and without aura IHS 1988 I1: 13.3 (SD: 2.5), I2: 13.2 (SD: 2.0), control: 12.0 (SD: 2.7) I1: 69.2%, I2: 45.5%, control: 83.3% United States Moderate 
Jong et al,53  2019 Hypnotherapy Transcendental meditation Progressive muscle relaxation 12 wk — Migraine ICHD-2 criteria I1: 11.2, I2: 12.2, I3: 11.5 I1: 43.0%, I2: 33%, I3: 50% Nether-lands Low 
Richter et al,54  1985 Relaxation Psychological treatment — 6 wk Psychological placebo Migraine Intermittent, paroxysmal headache All groups: 12.9 All groups: 67% Canada Moderate 

I1, intervention 1; I2, intervention 2; I3, intervention 3; ID, identifier; ICHD, International Classification of Headache Disorders; nr., not reported; TTH, tension-type headache; —, not applicable.

a

The label of the intervention used in this table equals the name of the respective node. A more detailed description of the interventions can be found in Supplemental Table 2.

b

Based on adequacy of random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, usage of observer-rated outcomes, completeness of outcome data, specification of main outcomes a priori, reporting on all outcomes, and conduction of ITT analyses. Individual items were summarized to produce a total score of 1 (low risk of bias), 2 (moderate or unclear risk of bias), or 3 (high risk of bias).

c

These studies and arms were not included because they did not connect with the other studies in the network.

In total, the mean age was 12.01 years (SD 2.1) and 353 (61%) of the sample population were female. In 6 studies, researchers used the IHS criteria (4 used the first edition,44,47,49,52  2 used the second edition46,53 ); authors of the remaining 6 studies used similar, albeit not completely analogous, diagnostic criteria (ie, they merged the ICHD-I and ICHD-II criterion C and D symptoms).45,48,50,51,54,55  The median duration of the nonpharmacological intervention was 6 weeks (range: 4–12 weeks). Furthermore, 2 studies (17%) were multicenter studies. Six (50%) of the included trials recruited children and adolescents from the United States, 3 (25%) from Europe, and 3 (25%) from Canada. Details on the interventions (Supplemental Table 2) and raw data of included studies (Supplemental Table 3) can be found in the Supplemental Information.

First, we clustered the nonpharmacological interventions into broad classes. For example, biofeedback, biofeedback stress management, biofeedback relaxation education, and autogenic feedback were taken together and modeled in the “biofeedback” node (see Supplemental Information 3 for details). This resulted in the following intervention nodes: biofeedback, relaxation, psychological treatment, education, and self-administered psychological treatments. The control group nodes comprised psychological placebo, sham biofeedback, and waiting list. Two studies had to be excluded because the interventions could only be assigned to the same intervention node: researchers in 1 of the studies44  examined 2 forms of biofeedback, the other study53  was focused on 3 different forms of relaxation.

Short-Term Analysis

Figure 1A shows the network of all comparisons for short-term efficacy for the broadly defined intervention nodes. Self-administered treatments (SMD: 1.44; 95% CI, 0.26 to 2.62), biofeedback (SMD: 1.41; 95% CI, 0.64 to 2.17), relaxation (SMD: 1.38; 95% CI, 0.61 to 2.14), psychological treatments (SMD: 1.36; 95% CI, 0.15 to 2.57), and psychological placebos (SMD: 1.17; 95% CI, 0.06 to 2.27) were significantly more effective than the waiting list (see Fig 2A for forest plot). Head-to-head comparisons of all included nonpharmacological treatments can be found in the league table (see Supplemental Table 4).

FIGURE 1

A, NMA of eligible comparisons for efficacy in the short-term: lumping approach. B, NMA of eligible comparisons for efficacy in the long-term: lumping approach. C, NMA of eligible comparisons for efficacy in the short-term: splitting approach. D, NMA of eligible comparisons for efficacy in the long-term: splitting approach. Note: width of the lines is proportional to the number of trials comparing every pair of interventions.

FIGURE 1

A, NMA of eligible comparisons for efficacy in the short-term: lumping approach. B, NMA of eligible comparisons for efficacy in the long-term: lumping approach. C, NMA of eligible comparisons for efficacy in the short-term: splitting approach. D, NMA of eligible comparisons for efficacy in the long-term: splitting approach. Note: width of the lines is proportional to the number of trials comparing every pair of interventions.

Close modal
FIGURE 2

A, Forest plot of NMA of all trials for efficacy in the short-term: lumping approach. B, Forest plot of NMA of all trials for efficacy in the long-term: lumping approach. C, Forest plot of NMA of all trials for efficacy in the short-term: splitting approach. D, Forest plot of NMA of all trials for efficacy in the long-term: splitting approach. Note: nonpharmacological interventions were compared with waiting list, which was the reference group. The brackets behind the intervention names indicate the following: number of studies over number of patients in which the intervention was examined. P score, Ranking of nonpharmacological intervention using P values.

FIGURE 2

A, Forest plot of NMA of all trials for efficacy in the short-term: lumping approach. B, Forest plot of NMA of all trials for efficacy in the long-term: lumping approach. C, Forest plot of NMA of all trials for efficacy in the short-term: splitting approach. D, Forest plot of NMA of all trials for efficacy in the long-term: splitting approach. Note: nonpharmacological interventions were compared with waiting list, which was the reference group. The brackets behind the intervention names indicate the following: number of studies over number of patients in which the intervention was examined. P score, Ranking of nonpharmacological intervention using P values.

Close modal

Long-Term Analysis

Figure 1B shows the network of all comparisons for long-term efficacy applying a broad classification rule for the nodes. Self-administered treatments (SMD: 1.40; 95% CI, 0.28 to 2.52), relaxation (SMD: 1.35; 95% CI, 0.60 to 2.09), psychological treatments (SMD: 1.33; 95% CI, 0.18 to 2.47), biofeedback (SMD: 1.21; 95% CI, 0.47 to 1.94), and psychological placebos (SMD: 1.14; 95% CI, 0.09 to 2.19) revealed significantly higher effects than the waiting list (see Fig 2B for forest plot). Head-to-head comparisons of all included nonpharmacological treatments are displayed in the league table (see Supplemental Table 5).

In all 12 studies, researchers provided sufficient data in the prespecified time periods to be included in the splitting approach for both the short-term and the long-term efficacy analyses.

Short-Term Analysis

Figure 1C shows the network of all comparisons for short-term efficacy. No nonpharmacological intervention was significantly more effective than the waiting list, which was the reference group. The nonsignificant SMDs ranged from −0.20 (95% CI, −3.20 to 2.79 for hypnotherapy) to 2.50 (95% CI, −0.75 to 5.75 for relaxation stress management) (see Fig 2C for details). There were no significant differences between the included nonpharmacological interventions (see Supplemental Table 6 for details). Although the relaxation stress management appeared to have the largest P score (P = .81) with respect to short-term efficacy outcomes (see Fig 2C), it was only examined in 1 trial with a sample size of 15 patients per arm.47  The prediction intervals for all comparisons showed nonsignificant effects and included both clinically beneficial and detrimental effects.

Long-Term Analysis

Three studies did not report long-term efficacy for the control groups.48,50,51  For these studies, we included the last complete data set (ie, in all cases FU 1). Additionally, 5 studies only reported FU 1 and not FU 2 or FU 3.43,46,49,52,54  For the remaining 4 studies, FU 2 or FU 3 served as long-term outcomes. Figure 1D shows the network of all comparisons for efficacy in the long-term.

All nonpharmacological interventions revealed nonsignificant SMDs, ranging from 0.03 (95% CI, −2.41 to 2.47 for hypnotherapy) to 2.50 (95% CI, −0.26 to 5.26 for relaxation stress management) (see Fig 2D). Additionally, none of the included nonpharmacological interventions did differ significantly from one another (see Supplemental Table 7 for details). Similarly, the 95% prediction intervals for all comparisons included the null value and contained both clinically beneficial and detrimental effects.

The certainty of evidence for the network estimates of the short- and long-term efficacy outcomes was examined by using GRADE.38  The results for study limitations (within study bias), publication bias (across-studies bias), indirectness, imprecision, heterogeneity, and incoherence can be found in the supplement (Supplemental Information 47, Supplemental Figs 4 and 5). All certainty of evidence analyses are based on the large networks with finer-grained nodes of the individual interventions (Supplemental Information 47, Supplemental Figs 4 and 85).

None of the included studies reported safety or acceptability outcomes. Hence, we were unable to conduct these analyses.

The detailed PRISMA for NMA can be found in the Supplemental Information 8.

We conducted the first NMA on the efficacy of nonpharmacological interventions for pediatric migraine and included 12 studies with a total of 576 patients. Interestingly, our NMA revealed different findings depending on the structure of the nodes in the network. In a first step, we lumped the interventions into broader classes. Significant short- and long-term effects were found for most nonpharmacological interventions (ie, relaxation, biofeedback, psychological treatments, and self-administered psychological treatments) as well as for 1 control group (ie, psychological placebo) when compared with the waiting list. In terms of our second aim (ie, to systematically compare the different nonpharmacological treatment options relative to each other), the included interventions did not significantly differ from one another. In a second step, we split the various published interventions into individual nodes. Here, in the short- and long-term analyses, none of the included interventions were significantly more effective than the waiting list, which was the control group that most interventions were compared with and that connected 2 otherwise unconnected networks. Also, none of the interventions differed significantly from each other in the head-to-head comparisons.

There are several reasons why we conducted 2 analyses that apply different strategies for the formation of the nodes. As stated in a recent literature review, there is currently no consensual method to support the node-making process in NMA.56  One reason for this is that the descriptions of nonpharmacological interventions in RCTs often lack essential information.57  Generally, the lumping versus splitting approach involves finding a compromise between clinical and statistical considerations. The lumping approach, ie, our broad classification, has the statistical advantage of increased power and allowing for a more accurate estimation of intervention effect sizes, especially when grouping interventions with similar components together.58  The splitting approach, ie, our individual nodes, has the advantage of enhancing clinical interpretability and utility. This is because the individual interventions that have been taken together in the lumping approach share similar treatment components (that justify the lumping process); however, they are not clinically interchangeable.59  For example, in our analyses, we combined several treatment arms into the node “relaxation” for the lumping approach. These individual interventions apply a variety of approaches linked to relaxation (eg, progressive muscle relaxation, autogenic training, hypnotherapy, and transcendental meditation). Hence, although the splitting approach compares all available individual interventions with each other, the lumping approach gives us an estimation of the effects of different types of interventions available for pediatric migraine. Both types of node-making were based on expert consensus in our group of authors that includes researchers, clinicians, and statisticians, which is in line with recommendations.56 

The difference in results between the 2 types of analyses (ie, lumping approach versus splitting approach) is noteworthy. Compared with traditional meta-analysis, the indirect comparisons that the NMA considers influence the respective effect sizes of the included interventions. The advantage of a NMA is that different interventions can be examined separately and relative to each other.60  Although the more detailed analysis represent the actual breadth of published studies and displays the actual types of studies that have been conducted, the smaller amount of studies in each node and the resulting lack of power led to a lack of significance, mostly because of large confidence and prediction intervals.

When compared with previous reviews and meta-analyses in the same area of research,8,9,61  our findings are comparable to some extent: researchers in other studies found considerably large, positive, and significant effect sizes for pain frequency at posttreatment of some nonpharmacological interventions, such as relaxation-based interventions, biofeedback, and CBT. Although the inclusion criteria differed substantially from previous reviews in that we exclusively focused on pediatric migraine and excluded other types of headaches, our analyses with the broader intervention nodes found similar effect sizes for relaxation, biofeedback, psychological treatments, and self-administered psychological treatments. Interestingly, these effects were also maintained in the long-term. This is in contrast to our previous analysis of prophylactic pharmacologic interventions in pediatric migraine, where we found effects in the short- but not the long-term.4  One reason for this might be an effect of training: biofeedback, for example, works through the control of otherwise involuntary physiologic responses, such as skin temperature, blood volume pulse, and brain activity.62  Previous studies in tension-type headache and migraine also reported a durable effect of biofeedback and relaxation trainings.63,64  In the short- and in the long-term analyses, self-administered treatments resulted in the largest P score with a value of 0.68 and 0.70, respectively. The P score is interpreted analogously to surface under the cumulative ranking curve in the Bayesian framework; a value of 0.68 means that 68% of competitor interventions are worse than the target intervention.65  Notably, researchers of both studies that examined self-administered treatments45,51  concluded that self-control of migraine may be a main factor in the effectiveness of interventions for pediatric migraine. Of note, in one of the studies, the self-administered treatment (called “own best efforts”) was designed as a control condition.51  Patients had 1 single session to discuss the use of the headache diary. In the other study, patients took part in an 8-week home-based treatment program focusing on cognitive and behavioral stress coping and relaxation strategies.45 

Interestingly, psychological placebos were significantly effective interventions when compared with waiting list controls in the short- and long-term. In the included studies, the psychological placebo for example consisted of 6 individual, 1-hour weekly sessions in which children were taught to recognize and label their emotions, relate them to their life situation, and discuss their feelings daily with a friend or parent.51  In the second study with a psychological placebo,45  participants were given a list of common triggers for migraine, such as different foods, met with a therapist once to become aware of and avoid triggers, and were taught brainstorming skills to deal with stressful situations. After that, participants were contacted weekly by the therapist by phone. The finding of significant effects for psychological placebos is not surprising, given that researchers have shown psychological placebos to be effective in other settings,66,67  which can in part be explained by various components of the placebo effect (such as a supportive patient–practitioner relationship and a therapeutic ritual68 ), which are present even in these control conditions.

The more detailed analyses in which the individual published interventions comprised 17 nodes could not confirm those significant effects found when combining the studies into 8 broader nodes. Various reasons why this might be the case come to mind. The individual nodes were composed of predominantly small studies, date back up to 3 decades, and are mainly single-center. Some of the comparisons in our network consist of only 1 indirect comparison (eg, relaxation stress management versus waiting list, n = 15). Therefore, some effects are potentially due to the so-called small study effect, a bias that describes how smaller studies reveal different treatment effects than large ones.41,69,70  Potential causes for the small study effect include selective outcome reporting,7173  clinical heterogeneity between patients in large and small trials (eg, smaller trials tend to include a more homogeneous sample of patients or include patients at higher risk), publication bias, and selective analysis reporting.74  In conducting 2 different analyses (ie, the lumping approach versus the splitting approach), we tried to eliminate the small study effect, because broader nodes increase the number of participants per node. In addition, the apparent small study effect is often actually due to differences in quality, patient inclusion criteria, baseline severity, and definition of outcome between small and large trials.41  With regard to the detection of the small study effect, an asymmetric funnel plot implies differences between the estimated derived from large and small trials.75  In the case of the presented NMAs, the funnel plots were nonsignificant for the short- and the long-term efficacy (see Supplemental Figs 4 and 5).

Overall, and in line with previous analyses, some of the effect sizes were considerably large in both types of analyses (ie, lumping and splitting approaches), with SMDs up to SMD = 2.50 for relaxation stress management and SMD = 1.81 for biofeedback stress management in the short- and long-term, pointing to a potentially clinically meaningful effect in some patients. Such large effect sizes are somewhat unexpected, especially when compared with prophylactic pharmacologic interventions with SMDs typically <1.0,4,76  but in line with a meta-analysis that focused on biofeedback for migraine only (and no other type of headache).53  This might point to the potential of these nonpharmacological interventions in children and adolescents. Notably, these included both beneficial and detrimental effects and suggest that interindividual differences play an important role: one subgroup of patients for whom nonpharmacological interventions lead to substantial pain relief versus a subgroup of patients who will not respond to these interventions or even experience a worsening of symptoms. Our results indicate that these differential reactions to the same interventions might exist, which is in line with the precision medicine approach;77  however, our findings do not inform about the characteristics of these subgroups.

Noteworthy, none of the studies included in our NMA reported safety or acceptability outcomes. This is concerning, given that nonpharmacological interventions can lead to negative consequences, such as side effects, nonimprovement of symptoms, or symptom deterioration.78,79  However, to date, negative consequences of nonpharmacological interventions, especially psychotherapy, are underreported and not adequately explored.80 

Our analysis has some limitations: first, we could only include a small amount of studies examining nonpharmacological interventions for pediatric migraine. This is in contrast to previous analyses,8  because we decided to specifically focus on migraine to increase the clinical usefulness of our results. However, an NMA uniquely allows for the formation of a network of intervention of any set of studies that links 3 or more interventions via direct comparisons, which then results in multiple ways to assess indirect comparisons between the interventions. All direct and indirect evidence can then be exploited; this simultaneous comparison is suggested to yield more precise estimates of the relative intervention effects than any single direct or indirect estimate.81  Conclusively, indirect evidence, when combined with direct evidence, increases the power and precision of treatment effects.82  Second, the included studies showed considerable statistical heterogeneity, indicating that patients and trials differed substantially. With the small and heterogeneous set of studies, the definition of nodes (ie, classes of interventions) was challenging. Merging different interventions (ie, lumping) has shown to increase heterogeneity within nodes, because similar interventions might not have the same treatment effect. Splitting treatments, in contrast, leads to decreased power.83  We tried to find a compromise between lumping and splitting by conducting 2 analyses with different sizes of nodes. A statistical alternative would have been to conduct several comparison-specific pairwise meta-analyses. However, in contrast to meta-analyses, NMAs can easily incorporate multiarm studies and evaluate disagreements between different sources of evidence. Third, none of the studies reported any safety or acceptability outcomes. Therefore, we were not able to evaluate the risk profile of these interventions and to make a statement about the risk-benefit balance. Fourth, the majority of our comparisons in the long-term analysis were based on FU 1 (ie, 3–4 months after randomization), because no later time points were reported. Moreover, 3 studies did report FU 2 or FU 3, but not for their respective control groups. It is therefore impossible to draw final conclusions on the maintenance of the effect. Fifth, because of limited reporting of follow-up assessment in the included studies, the applied time periods in the short- and long-term analyses did overlap in some cases. Sixth, researchers in 2 studies analyzed pediatric migraine in combination with tension-type headache49  and vascular headache.50  For these studies, the change of migraine headache cannot be clearly distinguished from the change of other types of pediatric headache. Seventh, half of the studies used similar, albeit not completely analogous, diagnostic criteria for migraine compared with the other half of the studies.45,48,50,51,54,55  We were not able to conduct an additional sensitivity analysis because the exclusion of these studies resulted in 2 separate subnetworks. Finally, 2 studies did not connect with the network and therefore had to be excluded. Hence, the efficacy of acupuncture and music therapy for this specific age group could not be evaluated.

Because none of the studies included in this NMA provided safety or acceptability outcomes, it was impossible to compare the interventions in this regard. We therefore highly recommend including measures of safety and acceptability in every trial, independent of the type of intervention (pharmacological and nonpharmacological). This is also in line with recommendations on outcome measures for pediatric pain trials such as PedIMMPACT (Pediatric Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials).84  Along with the idea that uncertainty in high-quality systematic reviews and NMAs indicates that research has yet to answer the primary question,85  our findings reveal that there are significant research gaps in the area of pediatric migraine. This is mainly due to the lack of data, a problem that is well-known across pediatric pain research.86 

Future researchers should control for natural course of pediatric migraine, and, preferably in studies with more participants and study sites, further evaluate the safety and efficacy of those interventions with large effect sizes in our NMA. Furthermore, to correspond with recent guidelines of the IHS for pediatric migraine,24  future RCTs should include the number of headache days as their primary outcome, which was not the case for the included studies. This is because the reporting of headache days allows the use of relatively simple trial diaries compared with the reporting of migraine days.24  Finally, and importantly, we urge future NMAs to report the node-making process in detail, especially for nonpharmacological treatments that are usually complex and heterogeneous.56 

Clinically, our results reveal that relaxation, biofeedback, psychological treatments, and self-administered psychological treatments are effective interventions for pediatric migraine. When analyzed in broad nodes and lumped on the basis of similarity of treatment components, these interventions revealed large and presumably clinically meaningful effect sizes for some patients. The more detailed analysis (ie, splitting approach), however, suggested that some of the effects have to be considered carefully because they are based on small studies. Moreover, because the CIs were considerably large for the individual treatment nodes, it remains important to find strategies on how to best identify those patients who profit most from individual nonpharmacologic interventions, in line with of the goals of precision medicine.77  Statistically, our results highlight that the decision about the formation of the nodes (ie, the lumping versus splitting approach) in an NMA has a strong impact on the significance and interpretability of the findings. We encourage the reader to weigh the clinical and statistical considerations that are associated with the different formations of the nodes.

Dr Locher conceived and designed the study, selected the articles and extracted the data, analyzed the data, wrote the first draft of the manuscript, interpreted the data, contributed to the final version of the manuscript, and agreed to be accountable for all aspects of the work; Dr Koechlin conceived and designed the study, selected the articles and extracted the data, wrote the first draft of the manuscript, interpreted the data, contributed to the final version of the manuscript, and agreed to be accountable for all aspects of the work; Dr Kossowsky conceived and designed the study, performed the literature search, selected the articles and extracted the data, interpreted the data, revised the article, contributed to the final version of the manuscript, and agreed to be accountable for all aspects of the work; Dr Meissner conceived and designed the study, performed the literature search, selected the articles and extracted the data, revised the article, contributed to the final version of the manuscript, and agreed to be accountable for all aspects of the work; Dr Lam selected the articles, extracted the data, revised the article, approved the final version of the manuscript, and agreed to be accountable for all aspects of the work; Mr Barthel selected the articles and extracted the data, revised the article, approved the final version of the manuscript, and agreed to be accountable for all aspects of the work; Dr Gaab supported to the interpretation of the data, revised the article, contributed to the final version of the manuscript, revised the article, and agreed to be accountable for all aspects of the work; Dr Berde supported to the interpretation of the data, revised the article, contributed to the final version of the manuscript, revised the article, and agreed to be accountable for all aspects of the work; Dr Schwarzer supported the analysis of the data, revised the article, contributed to the final version of the manuscript, and agreed to be accountable for all aspects of the work; Dr Linde conceived and designed the study, revised the article, contributed to the final version of the manuscript, and agreed to be accountable for all aspects of the work; and all authors read and met the International Committee of Medical Journal Editors criteria for authorship and agree with the results and conclusions of this article; and all authors approve the final manuscript as submitted.

This trial has been registered at with PROSPERO Registration (https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=146666) (identifier CRD42019146666; Nonpharmacological interventions for pediatric migraine: a systematic review and network meta-analysis).

FUNDING: Dr Koechlin is sponsored by the Swiss National Science Foundation fellowship P400PS_186658. Dr Locher received funding for this project from the Swiss National Science Foundation: P400PS_180730. Dr Meissner received support from the Schweizer-Arau-Foundation and the Theophrastus Foundation, Germany. This work was supported in part by the Sara Page Mayo Endowment for Pediatric Pain Research, Education, and Treatment.

CBT

cognitive behavioral therapy

CI

confidence interval

FU

follow-up

GRADE

Grading of Recommendations Assessment, Development, and Evaluation

ICHD

International Classification of Headache Disorders

IHS

International Headache Society

NMA

network meta-analysis

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-analysis

RCT

randomized controlled trial

SMD

standardized mean difference

1
Kabbouche
MA
,
Gilman
DK
.
Management of migraine in adolescents
.
Neuropsychiatr Dis Treat
.
2008
;
4
(
3
):
535
548
2
Powers
SW
,
Patton
SR
,
Hommel
KA
,
Hershey
AD
.
Quality of life in childhood migraines: clinical impact and comparison to other chronic illnesses
.
Pediatrics
.
2003
;
112
(
1 pt 1
):
e1
e5
3
Hershey
AD
.
Current approaches to the diagnosis and management of paediatric migraine
.
Lancet Neurol
.
2010
;
9
(
2
):
190
204
4
Locher
C
,
Kossowsky
J
,
Koechlin
H
, et al
.
Efficacy, safety, and acceptability of pharmacologic treatments for pediatric migraine prophylaxis: a systematic review and network meta-analysis
.
JAMA Pediatr
.
2020
;
174
(
4
):
341
349
5
Sieberg
CB
,
Huguet
A
,
von Baeyer
CL
,
Seshia
S
.
Psychological interventions for headache in children and adolescents
.
Can J Neurol Sci
.
2012
;
39
(
1
):
26
34
6
Nahman-Averbuch
H
,
Schneider
VJ
 II
,
Chamberlin
LA
, et al
.
Alterations in brain function after cognitive behavioral therapy for migraine in children and adolescents
.
Headache
.
2020
;
60
(
6
):
1165
1182
7
Trautmann
E
,
Lackschewitz
H
,
Kröner-Herwig
B
.
Psychological treatment of recurrent headache in children and adolescents–a meta-analysis
.
Cephalalgia
.
2006
;
26
(
12
):
1411
1426
8
Ng
QX
,
Venkatanarayanan
N
,
Kumar
L
.
A systematic review and meta-analysis of the efficacy of cognitive behavioral therapy for the management of pediatric migraine
.
Headache
.
2017
;
57
(
3
):
349
362
9
Fisher
E
,
Law
E
,
Dudeney
J
,
Palermo
TM
,
Stewart
G
,
Eccleston
C
.
Psychological therapies for the management of chronic and recurrent pain in children and adolescents
.
Cochrane Database Syst Rev
.
2018
;
9
(
9
):
CD003968
10
Headache Classification Subcommittee of the International Headache Society
. The International Classification of Headache Disorders: 2nd Edition. In:
Cephalalgia
, vol.
24
.
2004
:
9
160
11
Powers
SW
,
Kashikar-Zuck
SM
,
Allen
JR
, et al
.
Cognitive behavioral therapy plus amitriptyline for chronic migraine in children and adolescents: a randomized clinical trial
.
JAMA
.
2013
;
310
(
24
):
2622
2630
12
Hermann
C
,
Blanchard
EB
.
Biofeedback in the treatment of headache and other childhood pain
.
Appl Psychophysiol Biofeedback
.
2002
;
27
(
2
):
143
162
13
Evers
S
,
Pothmann
R
,
Überall
M
,
Naumann
E
,
Gerber
W-D
.
Therapie idiopathischer Kopfschmerzen im Kindesalter. Empfehlungen der Deutschen Migräne-und Kopfschmerzgesellschaft (DMKG)
.
Schmerz
.
2002
;
16
(
1
):
48
56
14
Damen
L
,
Bruijn
J
,
Koes
BW
,
Berger
MY
,
Passchier
J
,
Verhagen
AP
.
Prophylactic treatment of migraine in children. Part 1. A systematic review of non-pharmacological trials
.
Cephalalgia
.
2006
;
26
(
4
):
373
383
15
Mohr
DC
,
Spring
B
,
Freedland
KE
, et al
.
The selection and design of control conditions for randomized controlled trials of psychological interventions
.
Psychother Psychosom
.
2009
;
78
(
5
):
275
284
16
Locher
C
,
Koechlin
H
,
Zion
SR
, et al
.
Efficacy and safety of selective serotonin reuptake inhibitors, serotonin-norepinephrine reuptake inhibitors, and placebo for common psychiatric disorders among children and adolescents: a systematic review and meta-analysis
.
JAMA Psychiatry
.
2017
;
74
(
10
):
1011
1020
17
Diener
H-C
,
Schorn
CF
,
Bingel
U
,
Dodick
DW
.
The importance of placebo in headache research
.
Cephalalgia
.
2008
;
28
(
10
):
1003
1011
18
Liberati
A
,
Altman
DG
,
Tetzlaff
J
, et al
.
The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration
.
PLoS Med
.
2009
;
6
(
7
):
e1000100
19
Moher
D
,
Liberati
A
,
Tetzlaff
J
,
Altman
DG
;
PRISMA Group
.
Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement
.
J Clin Epidemiol
.
2009
;
62
(
10
):
1006
1012
20
Meissner
K
,
Fässler
M
,
Rücker
G
, et al
.
Differential effectiveness of placebo treatments: a systematic review of migraine prophylaxis
.
JAMA Intern Med
.
2013
;
173
(
21
):
1941
1951
21
Higgins
JPT
,
Deeks
JJ
.
Selecting studies and collecting data. In: Cochrane Handbook for Systematic Reviews of Interventions 5.1.0. 2011. Available at: https://handbook-5-1.cochrane.org/index.htm#chapter_7/7_selecting_studies_and_collecting_data.htm. Accessed September 18, 2019
22
Lipsey
MW
,
Wilson
DB
.
Practical Meta-Analysis
.
Thousand Oaks, CA
:
Sage Publications, Inc
;
2001
23
Higgins
JP
,
Deeks
JJ
,
Altman
DG
. Special Topics in Statistics. In:
Cochrane Handbook for Systematic Reviews of Interventions
.
Hoboken, NJ
:
John Wiley & Sons, Ltd
;
2008
:
481
529
24
Abu-Arafeh
I
,
Hershey
AD
,
Diener
H-C
,
Tassorelli
C
;
Clinical Trials Standing Committee and the Child and Adolescent Standing Committee of the International Headache Society
. Guidelines of the International Headache Society for Controlled Trials of Preventive Treatment of Migraine in Children and Adolescents, 1st Edition. In:
Cephalalgia
, vol.
39
.
2019
:
803
816
25
Higgins
JPT
,
Savovic
J
,
Page
MJ
,
Elber
RG
,
Sterne
JAC
.
Assessing risk of bias in a randomized trial
. In:
Cochrane Handbook for Systematic Reviews of Interventions Version 6.0
.
2019
.
Available at: www.training.cochrane.org/handbook. Accessed October 1, 2019
26
Cohen
J
.
Statistical Power Analysis for the Behavioral Sciences
, 2nd ed.
Hillsdale, NJ
:
Taylor & Francis Inc
;
1988
27
Higgins
JPT
,
Chandler
J
,
Cumpston
M
,
Li
T
,
Page
MJ
,
Welch
VA
, eds..
Cochrane Handbook for Systematic Reviews of Interventions
, 2nd ed.
Hoboken, NJ
:
John Wiley & Sons
;
2019
28
Moncrieff
J
,
Kirsch
I
.
Efficacy of antidepressants in adults
.
BMJ
.
2005
;
331
(
7509
):
155
157
29
Borenstein
M
, ed..
Introduction to Meta-Analysis
.
Hoboken, NJ
:
John Wiley & Sons
;
2009
30
Rücker
G
,
Krahn
U
,
König
J
,
Efthimiou
O
,
Schwarzer
G
.
Netmeta: network meta-analysis using frequentist methods. 2019. Available at: https://CRAN.R-project.org/package=netmeta. Accessed October 29, 2019
31
Rücker
G
.
Network meta-analysis, electrical networks and graph theory
.
Res Synth Methods
.
2012
;
3
(
4
):
312
324
32
Salanti
G
,
Ades
AE
,
Ioannidis
JPA
.
Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial
.
J Clin Epidemiol
.
2011
;
64
(
2
):
163
171
33
Cochran
WG
.
The comparison of percentages in matched samples
.
Biometrika
.
1950
;
37
(
3–4
):
256
266
34
Higgins
JPT
,
Thompson
SG
,
Deeks
JJ
,
Altman
DG
.
Measuring inconsistency in meta-analyses
.
BMJ
.
2003
;
327
(
7414
):
557
560
35
Deeks
JJ
,
Higgins
JPT
,
Altman
DG
. Analysing Data and Undertaking Meta-Analyses. In:
Higgins
JPT
,
Green
S
, eds.
Cochrane Handbook for Systematic Reviews of Interventions
.
Hoboken, NJ
:
John Wiley & Sons
;
2011
:
241
284
36
Dias
S
,
Welton
NJ
,
Caldwell
DM
,
Ades
AE
.
Checking consistency in mixed treatment comparison meta-analysis
.
Stat Med
.
2010
;
29
(
7–8
):
932
944
37
Higgins
JPT
,
Jackson
D
,
Barrett
JK
,
Lu
G
,
Ades
AE
,
White
IR
.
Consistency and inconsistency in network meta-analysis: concepts and models for multi-arm studies
.
Res Synth Methods
.
2012
;
3
(
2
):
98
110
38
Salanti
G
,
Del Giovane
C
,
Chaimani
A
,
Caldwell
DM
,
Higgins
JP
.
Evaluating the quality of evidence from a network meta-analysis
.
PLoS One
.
2014
;
9
(
7
):
e99682
39
CINeMA
.
Confidence in network meta-analysis. 2017. Available at: http://cinema.ispm.ch. Accessed September 29, 2019
40
IntHout
J
,
Ioannidis
JP
,
Rovers
MM
,
Goeman
JJ
.
Plea for routinely presenting prediction intervals in meta-analysis
.
BMJ Open
.
2016
;
6
(
7
):
e010247
41
Chaimani
A
,
Salanti
G
.
Using network meta-analysis to evaluate the existence of small-study effects in a network of interventions
.
Res Synth Methods
.
2012
;
3
(
2
):
161
176
42
Oelkers-Ax
R
,
Leins
A
,
Parzer
P
, et al
.
Butterbur root extract and music therapy in the prevention of childhood migraine: an explorative study
.
Eur J Pain
.
2008
;
12
(
3
):
301
313
43
Pintov
S
,
Lahat
E
,
Alstein
M
,
Vogel
Z
,
Barg
J
.
Acupuncture and the opioid system: implications in management of migraine
.
Pediatr Neurol
.
1997
;
17
(
2
):
129
133
44
Allen
KD
,
Shriver
MD
.
Role of parent-mediated pain behavior management strategies in biofeedback treatment of childhood migraines
.
Behav Ther
.
1998
;
29
(
3
):
477
490
45
McGrath
PJ
,
Humphreys
P
,
Keene
D
, et al
.
The efficacy and efficiency of a self-administered treatment for adolescent migraine
.
Pain
.
1992
;
49
(
3
):
321
324
46
Rapoff
MA
,
Connelly
M
,
Bickel
JL
, et al
.
Headstrong intervention for pediatric migraine headache: a randomized clinical trial
.
J Headache Pain
.
2014
;
15
(
1
):
12
47
Sartory
G
,
Müller
B
,
Metsch
J
,
Pothmann
R
.
A comparison of psychological and pharmacological treatment of pediatric migraine
.
Behav Res Ther
.
1998
;
36
(
12
):
1155
1170
48
Fentress
DW
,
Masek
BJ
,
Mehegan
JE
,
Benson
H
.
Biofeedback and relaxation-response training in the treatment of pediatric migraine
.
Dev Med Child Neurol
.
1986
;
28
(
2
):
139
146
49
Fichtel A
,
Larsson
B
.
Does relaxation treatment have differential effects on migraine and tension-type headache in adolescents?
Headache
.
2001
;
41
(
3
):
290
296
50
Labbé
EE
.
Treatment of childhood migraine with autogenic training and skin temperature biofeedback: a component analysis
.
Headache
.
1995
;
35
(
1
):
10
13
51
McGrath
PJ
,
Humphreys
P
,
Goodman
JT
, et al
.
Relaxation prophylaxis for childhood migraine: a randomized placebo-controlled trial
.
Dev Med Child Neurol
.
1988
;
30
(
5
):
626
631
52
Scharff
L
,
Marcus
DA
,
Masek
BJ
.
A controlled study of minimal-contact thermal biofeedback treatment in children with migraine
.
J Pediatr Psychol
.
2002
;
27
(
2
):
109
119
53
Jong
MC
,
Boers
I
,
van Wietmarschen
HA
, et al
.
Hypnotherapy or transcendental meditation versus progressive muscle relaxation exercises in the treatment of children with primary headaches: a multi-centre, pragmatic, randomised clinical study
.
Eur J Pediatr
.
2019
;
178
(
2
):
147
154
54
Richter
IL
,
McGrath
PJ
,
Humphreys
PJ
,
Goodman
JT
,
Firestone
P
,
Keene
D
.
Cognitive and relaxation treatment of paediatric migraine
.
Pain
.
1986
;
25
(
2
):
195
203
55
Labbé
EL
,
Williamson
DA
.
Treatment of childhood migraine using autogenic feedback training
.
J Consult Clin Psychol
.
1984
;
52
(
6
):
968
976
56
James
A
,
Yavchitz
A
,
Ravaud
P
,
Boutron
I
.
Node-making process in network meta-analysis of nonpharmacological treatment are poorly reported
.
J Clin Epidemiol
.
2018
;
97
:
95
102
57
Hoffmann
TC
,
Erueti
C
,
Glasziou
PP
.
Poor description of non-pharmacological interventions: analysis of consecutive sample of randomised trials
.
BMJ
.
2013
;
347
:
f3755
58
Shi
C
,
Westby
M
,
Norman
G
,
Dumville
JC
,
Cullum
N
.
Node-making processes in network meta-analysis of nonpharmacological interventions should be well planned and reported
.
J Clin Epidemiol
.
2018
;
101
:
124
125
59
Rouse
B
,
Chaimani
A
,
Li
T
.
Network meta-analysis: an introduction for clinicians
.
Intern Emerg Med
.
2017
;
12
(
1
):
103
111
60
Salanti
G
.
Indirect and mixed-treatment comparison, network, or multiple-treatments meta-analysis: many names, many benefits, many concerns for the next generation evidence synthesis tool
.
Res Synth Methods
.
2012
;
3
(
2
):
80
97
61
Stubberud
A
,
Varkey
E
,
McCrory
DC
,
Pedersen
SA
,
Linde
M
.
Biofeedback as prophylaxis for pediatric migraine: a meta-analysis
.
Pediatrics
.
2016
;
138
(
2
):
e20160675
62
McKee
MG
.
Biofeedback: an overview in the context of heart-brain medicine
.
Cleve Clin J Med
.
2008
;
75
(
suppl 2
):
S31
S34
63
Grazzi
L
,
Andrasik
F
,
D’Amico
D
,
Leone
M
,
Moschiano
F
,
Bussone
G
.
Electromyographic biofeedback-assisted relaxation training in juvenile episodic tension-type headache: clinical outcome at three-year follow-up
.
Cephalalgia
.
2001
;
21
(
8
):
798
803
64
Silver
BV
,
Blanchard
EB
,
Williamson
DA
,
Theobald
DE
,
Brown
DA
.
Temperature biofeedback and relaxation training in the treatment of migraine headaches. One-year follow-up
.
Biofeedback Self Regul
.
1979
;
4
(
4
):
359
366
65
Rücker
G
,
Schwarzer
G
.
Resolve conflicting rankings of outcomes in network meta-analysis: partial ordering of treatments
.
Res Synth Methods
.
2017
;
8
(
4
):
526
536
66
Gaab
J
,
Kossowsky
J
,
Ehlert
U
,
Locher
C
.
Effects and components of placebos with a psychological treatment rationale - three randomized-controlled studies
.
Sci Rep
.
2019
;
9
(
1
):
1421
67
Locher
C
,
Koechlin
H
,
Gaab
J
,
Gerger
H
.
The other side of the coin: nocebo effects and psychotherapy
.
Front Psychiatry
.
2019
;
10
:
555
68
Kaptchuk
TJ
,
Kelley
JM
,
Conboy
LA
, et al
.
Components of placebo effect: randomised controlled trial in patients with irritable bowel syndrome
.
BMJ
.
2008
;
336
(
7651
):
999
1003
69
Sterling
TD
,
Rosenbaum
WL
,
Weinkam
JJ
.
Publication decisions revisited: the effect of the outcome of statistical tests on the decision to publish and vice versa
.
Am Stat
.
1995
;
49
(
1
):
108
112
70
Sterne
JA
,
Gavaghan
D
,
Egger
M
.
Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature
.
J Clin Epidemiol
.
2000
;
53
(
11
):
1119
1129
71
Chan
A-W
,
Hróbjartsson
A
,
Haahr
MT
,
Gøtzsche
PC
,
Altman
DG
.
Empirical evidence for selective reporting of outcomes in randomized trials: comparison of protocols to published articles
.
JAMA
.
2004
;
291
(
20
):
2457
2465
72
Chan
A-W
,
Krleza-Jerić
K
,
Schmid
I
,
Altman
DG
.
Outcome reporting bias in randomized trials funded by the Canadian Institutes of Health Research
.
CMAJ
.
2004
;
171
(
7
):
735
740
73
Williamson
PR
,
Gamble
C
.
Identification and impact of outcome selection bias in meta-analysis
.
Stat Med
.
2005
;
24
(
10
):
1547
1561
74
Ioannidis
JPA
,
Munafò
MR
,
Fusar-Poli
P
,
Nosek
BA
,
David
SP
.
Publication and other reporting biases in cognitive sciences: detection, prevalence, and prevention
.
Trends Cogn Sci
.
2014
;
18
(
5
):
235
241
75
Chaimani
A
,
Higgins
JP
,
Mavridis
D
,
Spyridonos
P
,
Salanti
G
.
Graphical tools for network meta-analysis in STATA
.
PLoS One
.
2013
;
8
(
10
):
e76654
76
El-Chammas
K
,
Keyes
J
,
Thompson
N
,
Vijayakumar
J
,
Becher
D
,
Jackson
JL
.
Pharmacologic treatment of pediatric headaches: a meta-analysis
.
JAMA Pediatr
.
2013
;
167
(
3
):
250
258
77
Collins
FS
,
Varmus
H
.
A new initiative on precision medicine
.
N Engl J Med
.
2015
;
372
(
9
):
793
795
78
Barlow
DH
.
Negative effects from psychological treatments: a perspective
.
Am Psychol
.
2010
;
65
(
1
):
13
20
79
Lilienfeld
SO
.
Psychological treatments that cause harm
.
Perspect Psychol Sci
.
2007
;
2
(
1
):
53
70
80
Jonsson
U
,
Alaie
I
,
Parling
T
,
Arnberg
FK
.
Reporting of harms in randomized controlled trials of psychological interventions for mental and behavioral disorders: a review of current practice
.
Contemp Clin Trials
.
2014
;
38
(
1
):
1
8
81
Chaimani
A
,
Caldwell
DM
,
Li
T
,
Higgins
JP
,
Salanti
G
. Undertaking Network Meta-Analyses. In:
Cochrane Handbook for Systematic Reviews of Interventions, Version 6.1
.
Hoboken, NJ
:
John Wiley & Sons
;
2020
:
285
320
82
Thorlund
K
,
Mills
EJ
.
Sample size and power considerations in network meta-analysis
.
Syst Rev
.
2012
;
1
:
41
83
Caldwell
DM
,
Welton
NJ
.
Approaches for synthesising complex mental health interventions in meta-analysis
.
Evid Based Ment Health
.
2016
;
19
(
1
):
16
21
84
McGrath
PJ
,
Walco
GA
,
Turk
DC
, et al.;
PedIMMPACT
.
Core outcome domains and measures for pediatric acute and chronic/recurrent pain clinical trials: PedIMMPACT recommendations
.
J Pain
.
2008
;
9
(
9
):
771
783
85
Li
T
,
Vedula
SS
,
Scherer
R
,
Dickersin
K
.
What comparative effectiveness research is needed? A framework for using guidelines and systematic reviews to identify evidence gaps and research priorities
.
Ann Intern Med
.
2012
;
156
(
5
):
367
377
86
Eccleston
C
,
Fisher
E
,
Cooper
TE
, et al
.
Pharmacological interventions for chronic pain in children: an overview of systematic reviews
.
Pain
.
2019
;
160
(
8
):
1698
1707

Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

Supplementary data