Exploring the relationship between pain catastrophizing and migraine in youth: A longitudinal clinical cohort study

Abstract

Objective

This study explored the relationship between pain catastrophizing and migraine‐related outcomes (i.e., migraine‐related disability and headache frequency) between visits with a neurologist in a clinical population of children and adolescents with migraine.

Background

Evidence from adult populations suggests that pain catastrophizing, the tendency to magnify the threat value of, and ruminate and feel helpless about, pain may be associated with migraine‐related outcomes, but the association in children and adolescents is less clear.

Methods

In this prospective longitudinal clinical cohort study, children and adolescents aged 8–18 years with migraine completed headache questionnaires and a validated measure of pain catastrophizing (Pain Catastrophizing Scale for Children) at baseline and initial follow‐up visits with a neurologist. Recruitment spanned from May 2019 to July 2023. Headache frequency and migraine‐related disability (Pediatric Migraine Disability Assessment) were assessed at both visits. Migraine outcomes at follow‐up were examined in relation to baseline pain catastrophizing scores in models that controlled for sex, age, preventive treatment use, baseline headache frequency, and baseline disability.

Results

For this study, 121 consenting participants were included. In models adjusted for age, sex, baseline headache frequency, baseline disability, and preventive treatment use, baseline pain catastrophizing scores were significantly associated with disability scores at follow‐up (β = 0.81, 95% confidence interval [CI] = 0.13–1.48, p = 0.020), but not with headache frequency at follow‐up (β = 0.04, 95% CI = −0.10 to 0.19, p = 0.575). When examining the specific subscales of pain catastrophizing in an adjusted model, only baseline pain magnification (β = 6.73, 95% CI = 2.95–10.51, p = 0.001) had a significant association with disability at follow‐up, while feelings of helplessness (β = 0.08, 95% CI = −2.11 to 2.27, p = 0.944) and rumination did not (β = −1.83, 95% CI = −4.22 to 0.56, p = 0.133). In a subset of participants with pain catastrophizing measured at both visits (n = 65), pain catastrophizing total and subscale scores did not significantly differ between visits.

Conclusion

Baseline pain catastrophizing scores were associated with migraine‐related disability, but not headache frequency, at follow‐up in a clinical population of children and adolescents with migraine. Pain magnification specifically appeared to drive this association. Future studies should aim to replicate our results and to investigate if interventions aimed specifically at reducing pain magnification may help to mitigate migraine‐related disability in children and adolescents.

Plain Language Summary

Pain catastrophizing is a term that describes a person's tendency to magnify (i.e., to feel threatened) by pain, to ruminate (i.e., to think a lot) about pain, and to feel helpless about pain. We examined whether pain catastrophizing was linked to migraine‐related disability and headache frequency in 121 children and adolescents visiting a neurologist. Higher pain catastrophizing at a first visit, specifically the tendency to magnify pain, was linked to higher levels of migraine‐related disability at first follow‐up visit, but was not linked with headache frequency at follow‐up.

Abbreviations

CHREB

Conjoint Health Research Ethics Board

HrQoL

Health‐related quality of life

ICHD‐3

International Classification of Headache Disorders‐3rd edition

ICHD‐II

International Classification of Headache Disorders‐2nd edition

PCS‐C

Pain Catastrophizing Scale for Children

PedMIDAS

Pediatric‐Migraine Disability Assessment Scale

REDCap

Research Electronic Data Capture

INTRODUCTION

Migraine is a neurological disease and a primary headache disorder characterized by recurrent moderate to severe attacks. Migraine is the most common cause of neurological disability in youth aged 5–19 years old. ¹ , ² The prevalence of migraine among children and adolescents is approximately 11%, ³ with ~50% of cases persisting into adulthood. ⁴ Migraine in children and adolescents can cause significant disability, impacting health‐related quality of life (HrQoL), school attendance and performance, attendance in extracurricular activities, social relationships, and functioning in tasks at home. ⁵ , ⁶ , ⁷ Given the significant disability associated with migraine in children and adolescents, it is imperative that more research be done to understand treatment targets that may help to mitigate symptoms as well as migraine‐related disability. ⁸ , ⁹ Given that chronic pain disorders, inclusive of migraine, arise in a complex biopsychosocial context, minimizing disability requires both pharmacological and non‐pharmacological strategies, the latter of which address modifiable behaviors and cognitions that can maintain disability.

Pain catastrophizing is a thinking style characterized by the tendency to magnify the threat value of, ruminate about, and feel helpless about their ability to cope with pain. ¹⁰ In the broader chronic pain literature, pain catastrophizing has been linked to impaired treatment outcomes, high rates of medical consultation, diminished functioning, and lower HrQoL. ¹¹ Many studies have also aimed to assess the relationship between pain catastrophizing and migraine outcomes, ⁸ , ⁹ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ but few of these studies were carried out among children and adolescents, ⁹ , ²⁰ , ²¹ and none of the pediatric studies examined outcomes in a longitudinal manner. There is some evidence from adult studies that targeting pain catastrophizing through cognitive behavioral therapy or acceptance and commitment therapy may help to improve migraine outcomes, ²² , ²³ but the association between pain catastrophizing and migraine outcomes in children and adolescents, and their potential as a treatment targets, are less clear.

To address this gap, we aimed to determine: (1) the relationship between baseline pain catastrophizing levels and migraine‐related outcomes (i.e., migraine‐related disability and headache frequency) at clinical follow‐up, and (2) if pain catastrophizing levels change between visits with a neurologist in a clinical population of children and adolescents with migraine. For our first aim, we hypothesized that higher pain catastrophizing scores would be associated with lesser improvement in migraine‐related disability and headache frequency longitudinally. For our second aim, though we are unaware of any published studies with this same aim, we hypothesized that pain catastrophizing scores would be lower at follow‐up as compared to baseline visits. We hypothesized this based on theorizing that, as migraine‐related outcomes improve between visits, this may lead to improvements in pain catastrophizing.

METHODS

Design and setting

This prospective longitudinal clinical cohort study includes data from 121 children and adolescents with migraine. This is a secondary analysis of data from a study aiming to determine the longitudinal relationship between mental health and migraine outcomes in children and adolescents. ²⁴ Participants were seen by a pediatric neurologist with specialty training in headache medicine at the Alberta Children's Hospital located in Calgary, Alberta, Canada, and data from standardized questionnaires administered at initial consultation and first follow‐up visits were included. Participants were treated according to current guidelines on the treatment of migraine in children and adolescents. ²⁵ , ²⁶ All participants were prescribed evidence‐based acute migraine interventions, while participants with higher attack frequencies and/or migraine‐related disability were also offered preventive interventions using a shared‐decision making model; these interventions could comprise the following modalities or a combination thereof: pill‐based interventions (pharmaceuticals or nutraceuticals), injection‐based interventions (monoclonal antibodies, onabotulinumtoxinA injections, and/or nerve blocks), neuromodulation‐based interventions, and/or psychological interventions. All participants provided written informed consent and/or assent, as appropriate, for their data to be used for research and publication purposes. The study protocol was approved by the Conjoint Health Research Ethics Board at the University of Calgary (CHREB; REB19‐1319).

Participants

Eligible participants were consenting children and adolescents between the ages of 8 to 18 years old with a diagnosis of migraine, as determined by the treating neurologist using the International Classification of Headache Disorders – Third Edition (ICHD‐3) diagnostic criteria. ²⁷ The recruitment period spanned from May 2019 to July 2023. Participants were excluded if they did not have migraine and pain catastrophizing data from an initial consultation appointment, if they did not have migraine outcome data from a subsequent first follow‐up visit with the neurologist, and if they were not fluent in written and spoken English.

A total of 287 participants were approached for participation between May 2019 and July 2023. Of these, 215 patients consented to participate, 175 had a subsequent follow‐up visit between May 2019 and July 2023, and 121 had baseline pain catastrophizing scores and at least one migraine outcome at baseline and follow‐up visits available, thereby meeting criteria for the present analysis (Figure 1). Most of these participants were also included in a previously published research study. ²⁴ Therefore, a total of 121 participants were included in the present study, with 75 of these participants having total PCS‐C data available for both visits.

FIGURE 1.

Flowchart showing number of participants included and excluded.

Data collection

Participants were sent electronic questionnaires via emailed links prior to baseline and follow‐up appointments or were given paper copies of the questionnaires during the appointment. While the participants were instructed to complete the surveys themselves, parents, legal guardians, and primary caregivers were permitted to help them as needed. The treating neurologist reviewed and confirmed the headache questionnaire data in a semi‐structured interview. Pain catastrophizing data were not reviewed by the neurologist to avoid reporting bias. Data were housed in the institutional iteration of the Research Electronic Data Capture software (REDCap, managed by the University of Calgary Clinical Research Unit). Manual data cleaning was performed to detect missing data, outlying data and, where necessary and possible, chart reviews in the electronic health record were done to impute missing data and rectify outlying data.

Measurements/scales

Demographics, migraine history, migraine outcomes, and treatment details were ascertained using standardized headache questionnaires administered at both baseline and follow‐up. Headache frequency (range 0–31 days/month) was assessed using an item from a published standardized questionnaire that was validated for use in children and adolescents against the International Classification of Headache Disorders ‐ Third Edition (ICHD‐3) pediatric migraine criteria. ²⁸ The headache questionnaires also included the Pediatric‐Migraine Disability Assessment Scale (PedMIDAS), which is a validated questionnaire frequently used in the assessment of migraine in children and adolescents. ²⁹ It comprises a six‐item scale assessing the impact of migraine and its effects on school, home, and extracurricular activities in the past 3 months (score range 0–240). ⁶ , ⁷ , ³⁰

The Pain Catastrophizing Scale for Children (PCS‐C) ¹⁰ was used to quantify catastrophic thinking about pain. The PCS‐C contains 13 statements, rated on a five‐point Likert scale, that describe a thinking style that reflects the tendency to magnify the threat value of, and ruminate and feel helpless about, pain. ¹⁰ The total PCS‐C score for all 13 items ranges from 0 to 52, with three subscale scores: magnification (range 0–12; e.g., “I wonder whether something serious may happen”), rumination (range 0–16; e.g., “I can't keep it [the pain] out of my mind”), and feelings of helplessness (range 0–24; e.g., “there is nothing I can do to stop the pain”). Scores for these subscales are calculated based on responses to specific questions in the PCS‐C. Children and adolescents with chronic pain, including headache disorders, as well as healthy children and adolescents have all undergone testing with the PCS‐C, which has proven to be valid, reliable, and internally consistent in these populations. ¹⁰ Furthermore, the PCS‐C has an established clinical cutoff score: in a large sample of youth with chronic pain, scores of 26 or higher were considered to be elevated based on associations with various clinical outcomes. ³¹

Statistical methods

No a priori sample size calculations were carried out for this study specifically, as it was a secondary analysis of available data from our prior study on longitudinal migraine outcomes and mental health in youth. ²⁴

The baseline variables were summarized using frequencies, proportions, means (standard deviations), or medians [quartiles 1, 3], as appropriate. Shapiro–Wilk tests and QQ plots were used to determine the distribution of continuous variables. Along with a total score, the PCS‐C scores were subdivided into three categories according to established subscale scores (magnification, rumination, and feelings of helplessness). Participants with and without complete PCS‐C data were compared in their clinical characteristics using t‐tests or Mann–Whitney U tests (for non‐normal distribution) for continuous data, as appropriate, and chi‐squared or Fisher's exact tests (for small cell sizes) for categorical data, as appropriate.

The relationships between the migraine outcomes at follow‐up (i.e., PedMIDAS and headache frequency) and baseline PCS‐C scores were initially examined using simple linear regression models. Multivariable models, adjusting for age, sex, baseline headache frequency, baseline PedMIDAS score, and whether the participant was taking preventive interventions (yes/no), were used to estimate the association between the outcomes and initial PCS‐C total and subscale scores. Variance inflation factors (VIF) were used to assess multicollinearity in the models, with values >2.5 being considered significant for possible multicollinearity. The assumption of normality within the models was evaluated using histograms and QQ plots of model residuals. Model fits for the multiple linear regression models were assessed using adjusted R squared values and the reliability of the reported R squared values were assessed using F‐tests.

For participants who had PCS‐C scores at both the baseline and follow‐up visits, analyses were performed to determine if PCS‐C scores changed between visits using paired t‐tests. For all analyses, two‐tailed hypothesis tests were used, with p values <0.05 considered statistically significant. Stata software (version 17.0, College Station, TX, StataCorp LLC) was used for all analyses.

RESULTS

Participant characteristics

There were 121 participants in this study, 75 of whom had pain catastrophizing data available for both visits. Participant demographic characteristics (Table 1) were as follows: 66.9% (n = 81) were female, median age = 14 years [quartile 1 = 11, quartile 3 = 16] ranging from 8 to 18 years old, 48.0% (n = 58) were diagnosed with chronic migraine at baseline, and the baseline median PCS‐C symptom score was 23 [13, 33] with a range of 0 to 48. For the average attack severity at baseline, 3.3% of the participants (n = 4) reported mild severity, 51.2% (n = 62) reported moderate severity, 35.5% (n = 43) reported severe severity, and 9.9% (n = 12) did not provide these data. The median baseline headache frequency was 15 days/month [6, 30], and the median follow‐up headache frequency was 10 days/month [4, 25], while the median baseline PedMIDAS score was 31 [12, 64.5] and the median follow‐up PedMIDAS score was 24 [6, 57]. At baseline, only 29 participants (23.9%) were taking preventive treatment, and at follow‐up this number increased to 75 (61.9%) participants (details on preventive treatment types are available in Table S1). While only six participants (4.9%) were receiving a psychological intervention at baseline, 21 (17.4%) were receiving a psychological intervention at follow‐up (e.g., cognitive‐behavioral therapy for pain, biofeedback, etc.). Participants with and without complete PCS‐C data (missing either PCS‐C total scores or PCS‐C sub‐scale scores for baseline and/or follow‐up) were similar overall, but did differ in some baseline characteristics: those without PCS‐C follow‐up scores had significantly longer visit intervals, lower headache frequencies at follow up, were less likely to be taking a preventive at follow‐up, were less likely to have allodynia, and had lower PedMIDAS scores at follow‐up (Table S2).

TABLE 1.

Demographic and descriptive characteristics of included research population (n = 121).

Characteristic	Median [Q1, Q3]	Range	Number (%)	Number missing (%)
Sex				0 (0)
Male	N/A	N/A	39 (32)
Female	N/A	N/A	81 (67)
Gender diverse	N/A	N/A	1 (1)
Age (years)	14 [11, 16]	8 to 18	N/A	0 (0)
Race
White	N/A	N/A	55 (45)	50 (41) a
Black	N/A	N/A	2 (2)
Asian	N/A	N/A	4 (3)
Indigenous	N/A	N/A	4 (3)
Unknown	N/A	N/A	0 (0)
Other	N/A	N/A	6 (5)
Do not want to answer	N/A	N/A	1 (1)
Headache duration (months)	36.5 [12.2, 60.8]	0.1 to 158.2	N/A	1 (1)
Interval between visits (days)	204 [130.7, 273]	18 to 679	N/A	0 (0)
Migraine with aura
Baseline	N/A	N/A	39 (32)	0 (0)
Follow‐up	N/A	N/A	39 (32)	0 (0)
Chronic migraine diagnosis
Baseline	N/A	N/A	58 (48)	0 (0)
Follow‐up	N/A	N/A	48 (40)	0 (0)
Headache frequency (days/month)
Baseline	15 [6, 30]	0 to 31	N/A	0 (0)
Follow‐up	10 [4, 25]	0 to 31	N/A	0 (0)
Average attack duration (minutes)
Baseline	240 [120, 720]	10 to ‘constant’	N/A	12 (10)
Follow‐up	180 [90, 600]	0 to ‘constant’	N/A	7 (6)
Average severity
Baseline				12 (10)
Mild	N/A	N/A	4 (3)
Moderate	N/A	N/A	62 (51)
Severe	N/A	N/A	43 (36)
Follow‐up				14 (12)
Mild	N/A	N/A	16 (13)
Moderate	N/A	N/A	63 (52)
Severe	N/A	N/A	28 (23)
Average severity b
Baseline	6 [5, 7]	2 to 10	N/A	12 (10)
Follow‐up	6 [4.8, 7]	0 to 10	N/A	5 (4)
Allodynia
Baseline	N/A	N/A	75 (62)	0 (0)
Follow‐up	N/A	N/A	69 (57)	0 (0)
PedMIDAS score
Baseline	31 [12, 64.5]	0 to 240	N/A	2 (2)
Follow‐up	24 [6, 57]	0 to 198	N/A	0 (0)
PedMIDAS change Score	−8 [−26.5, 10.5]	−240 to 105	N/A	2 (2)
PedMIDAS grade
Baseline				2 (2)
1	N/A	N/A	27 (22)
2	N/A	N/A	30 (25)
3	N/A	N/A	23 (19)
4	N/A	N/A	39 (32)
Follow‐up				0 (0)
1	N/A	N/A	43 (36)
2	N/A	N/A	23 (19)
3	N/A	N/A	23 (19)
4	N/A	N/A	32 (26)
PCS‐C total symptom score
Baseline	23 [13, 33]	0 to 48	N/A	0 (0)
Follow‐up	21 (11.5, 31.5)	1 to 50	N/A	46 (38)
PCS‐C magnification score
Baseline	3 [1, 5.5]	0 to 12	N/A	14 (12)
Follow‐up	3 [1, 6]	0 to 12	N/A	52 (43)
PCS‐C rumination score
Baseline	10 [6, 13]	0 to 16	N/A	14 (12)
Follow‐up	10 [6, 13]	0 to 16	N/A	52 (43)
PCS‐C feelings of helplessness score
Baseline	10 [5, 13.5]	0 to 23	N/A	14 (12)
Follow‐up	8 [3, 15]	0 to 24	N/A	52 (43)
PCS‐C total above clinical cutoff score (>26)
Baseline	N/A	N/A	50 (41.3)	0 (0)
Follow‐up			29 (38.6)	46 (38.0)
Baseline PCS‐C total score by diagnosis at baseline
Episodic migraine	17 [12, 27]	N/A	N/A	N/A
Chronic migraine	26.5 [17, 35]	N/A	N/A	N/A
Follow‐up PCS‐C total score by diagnosis at follow‐up
Episodic migraine	17 [11, 29]	N/A	N/A	N/A
Chronic migraine	26.5 [16, 34.5]	N/A	N/A	N/A
PROMIS depression short form T score
Baseline	50.6 [40.4, 60]	N/A	N/A	0 (0)
Follow‐up	53.3 [35.2, 62.1]	N/A	N/A	43 (35.5)
PROMIS anxiety short form T score
Baseline	49.8 [41.8, 58.7]	N/A	N/A	1 (0.1)
Follow‐up	52.5 [43, 63.4]	N/A	N/A	41 (33.8)

Abbreviation: PCS‐C, Pain Catastrophizing Scale for Children.

^a The large proportion of missing values for race reflects the addition of a race question to the questionnaire after enrolment began, such that initial participants did not have this data collected.

^b Scores are based on a scale from 0 to 10, where 0 = no pain and 10 = worst pain ever.

Univariable linear regression models

Both PedMIDAS scores and headache frequency at follow‐up were significantly associated with baseline PCS‐C total symptom scores in univariable linear regression models (Table 2). All PCS‐C subscales were also significantly associated with PedMIDAS scores and headache frequency at follow‐up in the univariable models, except for headache frequency at follow‐up and baseline rumination scores, which were not significantly associated (Table 2).

TABLE 2.

Univariate linear regression models assessing the association between baseline PCS‐C scores and headache frequency and PedMIDAS at follow‐up.

Variable	Beta	95% Confidence interval (lower, upper)	Standard error	t‐Value	p‐Value
PedMIDAS at follow‐up as outcome variable
PCS‐C total (n = 121)	1.40	0.73–2.07	0.34	4.12	<0.001*
Magnification (n = 107)	7.13	4.60–9.67	1.28	5.58	<0.001*
Rumination (n = 107)	1.93	0.01–3.85	0.97	1.99	0.049*
Helplessness (n = 107)	2.73	1.32–4.14	0.71	3.84	<0.001*
Headache frequency at follow‐up as outcome variable
PCS‐C total (n = 121)	0.26	0.09–0.42	0.08	3.13	0.002*
Magnification (n = 107)	1.21	0.56–1.87	0.33	3.66	<0.001*
Rumination (n = 107)	0.41	−0.06 to 0.87	0.24	1.72	0.088
Helplessness (n = 107)	0.50	0.15–0.85	0.18	2.83	0.006*

^*Indicates statistical significance at the p < 0.05 level.

Multiple linear regression models

In the multiple linear regression models, baseline PCS‐C total scores were significantly associated with PedMIDAS at follow‐up (model 1; β = 0.81, 95% CI = 0.13–1.48, p = 0.020), but not with headache frequency at follow‐up (model 2; β = 0.04, 95% CI = ‐0.10–0.19, p = 0.575), after adjusting for age, sex, baseline PedMIDAS, baseline headache frequency, and preventive treatment use (Table 3). In an adjusted model examining PCS‐C subscale scores as exposures, baseline pain magnification scores (model 3; β = 6.73, 95% CI = 2.95–10.51, p = 0.001) were significantly associated with PedMIDAS scores at follow‐up, while baseline pain rumination scores (β = −1.83, 95% CI = −4.22 to 0.56, p = 0.133) and baseline feelings of helplessness scores (β = 0.08, 95% CI = −2.11 to 2.27, p = 0.944) were not (Table 4).

TABLE 3.

Multiple linear regression models assessing the association between baseline PCS‐C total scores and headache frequency and PedMIDAS scores at follow‐up.

Variable	Beta	95% Confidence interval (lower, upper)	Standard error	t‐Value	p_Value
Model 1: PedMIDAS at follow‐up as outcome variable (n = 119)
PCS‐C baseline total	0.81	0.13–1.48	0.33	2.36	0.020*
Sex (referent = male)
Female	−1.26	−17.88 to 15.34	8.38	−0.15	0.880
Gender diverse	−31.52	−113.95 to 50.91	41.6	−0.76	0.450
Age	0.81	−2.14 to 3.75	1.48	0.54	0.588
Baseline headache frequency	0.35	−0.53 to 1.24	0.44	0.80	0.427
Baseline PedMIDAS	0.29	0.14–0.44	0.08	3.84	<0.001*
Taking preventative treatment (yes/no)	−6.45	−24.63 to 11.74	9.18	−0.70	0.484
Model 2: Headache frequency at follow‐up as outcome variable (n = 119)
PCS‐C baseline total	0.04	−0.10 to 0.19	0.074	0.56	0.575
Sex (referent = male)
Female	1.41	−2.17 to 4.99	1.81	0.78	0.437
Gender diverse	−4.95	−22.73 to 12.83	8.97	−0.55	0.582
Age	0.23	−0.41 to 0.86	0.32	0.71	0.479
Baseline headache frequency	0.57	0.38–0.76	0.10	5.95	<0.001*
Baseline PedMIDAS score	0.009	−0.02 to 0.04	0.016	0.55	0.580
Taking preventative treatment (yes/no)	1.83	−2.90 to 5.75	1.98	0.92	0.357

^*Indicates statistical significance at the p < 0.05 level.

TABLE 4.

Multiple linear regression models assessing the association between baseline PCS‐C subscale scores and PedMIDAS scores at follow‐up (Model 3; n = 106).

Variable	Beta	95% confidence interval (lower, upper)	Standard error	t‐Value	p‐Value
PCS‐C magnification subscale score	6.73	2.95–10.51	1.90	3.53	0.001*
PCS‐C rumination subscale score	−1.83	−4.22 to 0.56	1.20	−1.52	0.133
PCS‐C feeling helpless subscale score	0.08	−2.11 to 2.27	1.10	0.07	0.944
Sex (referent = male)
Female	−5.90	−22.17 to 10.37	8.20	−0.72	0.473
Gender diverse	−51.40	−128.99 to 26.17	39.08	−1.32	0.192
Age	−0.050	−2.95 to 2.86	1.46	−0.03	0.973
Baseline headache frequency	0.45	−0.45 to 1.35	0.45	1.00	0.321
Baseline PedMIDAS	0.19	0.038–0.35	0.08	2.47	0.015*
Taking preventative treatment (yes/no)	1.88	−16.45 to 20.22	9.24	0.20	0.321

^*Indicates statistical significance at the p < 0.05 level.

Changes in pain catastrophizing between visits

The analyses examining changes in pain catastrophizing total scores between visits included the 75 participants who had PCS‐C data available for both visits. Only 65 participants had data available to compare the PCS‐C subscales between visits, further reducing the sample size for those analyses. Neither the total PCS‐C score nor any of its subscales significantly differed between visits (Table 5).

TABLE 5.

Change in pain catastrophizing total and sub‐scale scores between visits.

Variable	Mean difference between visits	95% Confidence interval for mean difference	p‐Value
PCS‐C total (n = 75)	−1.37	−4.31 to 1.57	0.355
Magnification (n = 65)	−0.75	−1.51 to 0.04	0.051
Rumination (n = 65)	−0.11	−1.29 to 1.08	0.856
Helplessness (n = 65)	−0.62	−2.24 to 1.01	0.453

Model characteristics and performance

There were no multicollinearity problems identified between the variables in the multivariable linear regression models, with the exception of the feelings of helplessness subscale score in the adjusted model examining the associations between PedMIDAS scores at follow‐up and PCS‐C subscale scores (i.e., model 3). The VIF for baseline feelings of helplessness scores was 3.03, slightly above our pre‐specified conservative VIF cutoff value of 2.5, and indicative of a moderate correlation between feelings of helplessness and the other variables in the model. As this variable was not significantly associated with the outcome in the model, we were not concerned about threats to overall model interpretation in this context. There were no violations of the assumption of model normality according to examination of the histograms and QQ plots representing model residuals. Adjusted R squared (adj. R ²) values were as follows for the multiple linear regression models: (1) model 1 adj. R ² = 0.26, (2) model 2 adj. R ² = 0.37, and (3) model 3 adj. R ² = 0.31. Importantly, the adjusted R squared value of model 3 was larger than model 1, suggesting that modeling PCS‐C scores with subscales as opposed to total scores resulted in improved model fit. All F‐tests were statistically significant, suggesting that adjusted R squared estimates were reliable and that we can reject the null hypothesis that all regression coefficients in the models were equal to zero: (1) model 1 F(7,111) = 7.00, p < 0.0001; (2) model 2 F(7,111) = 10.79, p = 0.0001; and (3) model 3 F(9,96) = 6.23, p < 0.0001.

DISCUSSION

This study examined the longitudinal association between pain catastrophizing and migraine outcomes in a clinical population of youth with migraine. We hypothesized that increased pain catastrophizing at baseline would be associated with higher migraine‐related disability and higher headache frequency at follow‐up. In support of our hypothesis, we did find a significant association between pain catastrophizing, specifically pain magnification, and migraine‐related disability. Contrary to our hypothesis, we did not observe a relationship between baseline pain catastrophizing scores and headache frequency. We had also hypothesized that pain catastrophizing scores would decrease between visits with a neurologist, but did not observe this in our data.

In terms of key findings, specifically, we found that baseline pain magnification was associated with migraine‐related disability at follow‐up and we observed a large magnitude association: after adjusting for baseline PedMIDAS scores and other key confounders (age, sex, etc.), we found that, for each 1 point increase in pain magnification, PedMIDAS scores at follow‐up were 6.73 points higher, which we feel is a clinically meaningful difference as it represents ~6–7 days more with disability from migraine in important functional spheres (i.e., home, school, extracurriculars) over the past 3 months. Importantly, our model that assessed PCS‐C subscale scores in relation to follow‐up PedMIDAS scores (i.e., model 3), had a better fit than our model using PCS‐C total scores (i.e., model 1), suggesting that the subscales may be more valuable than total scores, and that pain magnification may be the most important clinical target in relation to targeting improved migraine‐related disability in children and adolescents with migraine. Pain magnification is a person's tendency to evaluate their pain as threatening, ³² and here we see that this may be a particularly powerful cognitive predictor of future disability in children and adolescents with migraine. Perhaps pain magnification is more strongly associated with behaviors that contribute to reduced functioning, compared to other aspects of pain catastrophizing, because it may perpetuate fear avoidance ³³ (i.e., if I perceive my migraine attacks as threatening, I need to protect myself from the threat of attacks by avoiding activities that could increase exposure to the threat). This finding largely aligns with published studies examining pain catastrophizing among adults with migraine, which have shown that higher pain catastrophizing scores are associated with poorer migraine outcomes, ⁸ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ though this is not uniformly true. ¹⁹ Pediatric studies assessing how pain catastrophizing may be related to migraine outcomes in youth are far fewer in number. ⁹ , ²⁰ , ²¹ Notably, none of these published pediatric studies have found a significant association between pain catastrophizing and migraine‐related disability to date, though all prior pediatric studies have examined cross‐sectional associations, while our interest was in understanding if baseline pain catastrophizing is associated with migraine outcomes at follow‐up. ⁹ , ²⁰ , ²¹ Our finding, of course, requires replication. However, if future studies do replicate our finding that pain magnification is specifically and strongly associated with future migraine‐related disability, this could pave the way to more parsimonious and more accessible psychological interventions that target pain magnification in children and adolescents with migraine. In the interim, given the observed strong association between baseline pain magnification and migraine‐related disability at follow‐up, it may be warranted to screen youth with migraine for pain catastrophizing, and to prioritize youth with elevated scores (particularly in pain magnification), for referrals to behavioral interventions (e.g., cognitive behavioral therapy for pain).

Contrary to our hypothesis, we did not find an association between baseline pain catastrophizing and headache frequency. To our knowledge, there is only one other pediatric study that also examined this relationship, albeit indirectly; in that study, pain catastrophizing levels did not differ between youth with episodic vs. chronic migraine, suggesting that headache frequency and pain catastrophizing may not be related. ²¹ Intuitively, this null association makes sense, as headache frequency is likely a more direct reflection of disease activity, while migraine‐related disability is a more complex, multifactorial outcome that reflects not only disease activity, but also behaviors and cognitions that are more subject to the influence of psychosocial factors. It is also possible that there is a relationship between pain catastrophizing and headache frequency, but that it is modest and that each of the pediatric studies, inclusive of the present study, have been underpowered for this outcome. Either way, larger studies that have been specifically powered to detect associations between headache frequency and pain catastrophizing in youth with migraine are warranted based on the gaps in the current pediatric literature.

As observed in our data among the subsample of participants who had pain catastrophizing scores available for both visits, there were no statistically significant changes between baseline and first follow‐up visit pain catastrophizing scores. This could have been due to a lack of statistical power, as our sample size for this particular question was small (n = 75 for total scores, n = 65 for sub‐scale scores). However, the magnitudes of the mean changes in pain catastrophizing total and subscale scores were small, and unlikely to be clinically significant (change scores ranging from 0.11–1.37 points). This finding suggests that seeing a neurologist for migraine has no influence on pain catastrophizing in children and adolescents, and that targeted interventions to reduce pain catastrophizing may be warranted in this population, in particular for pain magnification that may be related to future migraine‐related disability.

Our findings must be considered with respect to the limitations of our study. First, our study utilized data from a real‐world clinical practice where visit intervals vary depending on patient needs and were more prolonged than typical for approximately half of the recruitment period given that the study encompassed the coronavirus disease 2019 (COVID‐19) pandemic. These factors lead to discrepancies in the intervals between baseline and follow‐up visits between patients. By not having a standardized time frame for baseline and follow‐up visits, we may have missed important time points in between, or may have not measured the correct intervals required to identify significant changes in pain catastrophizing scores. However, we did verify whether there was an association between visit interval and the follow‐up migraine‐related disability score in the context of our significant findings for this outcome; in a univariate model, there was no significant association between visit interval and follow‐up PedMIDAS. This is why we chose to omit visit interval from the modeling, to keep the models parsimonious and only include key covariates and our planned exposures. Second, this study was a secondary analysis of data from a prior published study, ²⁴ and we did not specifically power the current analysis for our secondary study objectives, which increases our chances of both false positive and negative findings. However, the significant association found between baseline pain magnification and follow‐up migraine‐related disability was large in magnitude with the lower end of the confidence interval in the adjusted model being 2.95; therefore, we think that it is unlikely to be a false positive finding, as the lower end of the confidence interval is far from zero and would still reflect a clinically meaningful difference in follow‐up PedMIDAS scores (i.e., for every 1 point increase in pain magnification at baseline, there would be a PedMIDAS increase of 2.95 at follow‐up). Also, there were certainly variables that we did not include in our models that are unmeasured confounders in the relationships assessed (e.g., socioeconomic status, adherence to treatments). Investigating pain catastrophizing as predictor of longitudinal migraine outcomes is inherently limited in that it focuses on cognitions related to pain and does not capture cognitions related to migraine‐associated symptoms. We were only able to examine changes in pain catastrophizing between visits in a subsample of 65 participants. A comparison of baseline characteristics between participants who provided follow‐up PCS‐C data and those without such data (either total PCS‐C scores or one of the PCS‐C sub‐scale scores) shows the groups differed in certain characteristics (as shown in Supplementary Table S2). Participants with missing PCS‐C data had a longer interval between baseline and first follow‐up visits, lower monthly headache frequency at follow‐up, had a lower proportion who were taking preventive treatment at follow‐up, a lower proportion with allodynia, and lower PedMIDAS scores. These differences suggest that the participants who were missing some follow‐up PCS‐C data may have had less disabling migraine attacks, which could explain why such participants had longer intervals between clinical visits, and that participants included in subsequent analyses had more severe or disabling migraine attacks by comparison. These differences between participants with vs. without follow‐up pain catastrophizing scores may have influenced our results in the subgroup analysis. Lastly, the single‐center design of our study limits external validity, and our observed effects (or lack thereof) may not be generalizable to the entire pediatric migraine population. That being said, we did compare our sample's PCS‐C scores, qualitatively, with PCS‐C scores in other samples of children and adolescents with migraine in the literature, and the distribution of our scores overlapped with those of other published samples (median and mean within range of other studies). ⁹ , ²⁰ , ²¹ Recruitment during the COVID‐19 pandemic for part of our study may render our results less generalizable to non‐pandemic times, though participant data on headache outcomes and PCS‐C scores (as described above) did not differ substantially from other studies of clinical populations of children and adolescents with migraine who were exclusively recruited outside of the COVID‐19 pandemic. Despite these limitations, this study extends our knowledge of pain catastrophizing among children and adolescents with migraine as there is sparse prior literature on this topic and only cross‐sectional data published prior to our study. ⁹ , ²⁰ , ²¹ It also highlights some areas for future research, like the pressing need to see if the observed association between baseline pain magnification and future migraine‐related disability can be replicated, as it may have important clinical implications for how to target more accessible, cheaper, and less intensive psychological interventions in this population.

CONCLUSION

We found a large, statistically and clinically significant association between baseline pain magnification scores and follow‐up migraine‐related disability in children and adolescents. Headache frequency at follow‐up was not associated with baseline pain catastrophizing. Our work requires replication in future studies specifically powered to examine how pain catastrophizing is associated with migraine outcomes in children and adolescents. If pain magnification is consistently found to specifically predict migraine‐related disability, targeted and brief psychological interventions that are more easily accessible than existing more intensive psychological interventions (e.g., cognitive behavioral therapy for pain) may be warranted to improve treatment access and outcomes for children and adolescents with migraine.

AUTHOR CONTRIBUTIONS

Alexis Espanioli: Data curation; writing – original draft. Nynke J. van den Hoogen: Writing – review and editing. Jonathan Kuziek: Data curation; resources; writing – review and editing. Kirsten Sjonnesen: Writing – review and editing. Melanie Noel: Conceptualization; writing – review and editing. Serena L. Orr: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; supervision; writing – review and editing.

FUNDING INFORMATION

This research was supported by a summer studentship award from the O'Brien Institute for Public Health, the Alberta Children's Hospital Research Institute and the Department of Pediatrics, Cumming School of Medicine, University of Calgary. The O'Brien Institute for Public Health, the Alberta Children's Hospital Research Institute, and the Department of Pediatrics provided funding for this project but had no role in design or execution of the project.

CONFLICT OF INTEREST STATEMENT

Alexis Espanioli, Nynke Van Den Hoogen, Jonathan Kuziek, Kirsten Sjonnesen, and Melanie Noel have no conflicts to declare. Serena L. Orr receives royalties from Cambridge University Press. She serves on the editorial boards of Headache, Neurology, and the American Migraine Foundation. She also has research funding from the Canadian Institutes of Health Research, the Alberta Children's Hospital Research Institute, and the American Headache Society. Dr. Orr has received compensation from the Université de Sherbrooke for delivering a rounds presentation.

Supporting information

Table S1.

Table S2.

Espanioli A, van den Hoogen NJ, Kuziek J, Sjonnesen K, Noel M, Orr SL. Exploring the relationship between pain catastrophizing and migraine in youth: A longitudinal clinical cohort study. Headache. 2025;65:1369‐1380. doi: 10.1111/head.15039