bPsychiatry, University of Calgary, Calgary, AB, Canada
cAlberta Children's Hospital Research Institute
dHotchkiss Brain Institute
eMathison Centre for Mental Health Research and Education
fDepartment of Neuroscience, Monash University, Melbourne, Australia
gDepartment of Anesthesiology and Pain Medicine, University of Washington, Seattle, WA, United States
hDepartment of Anesthesia, University of Calgary, Calgary, AB, Canada
iDepartment of Psychology, York University and Hospital for Sick Children, Toronto, ON, Canada
jOwerko Centre, University of Calgary, Calgary, AB, Canada
*Corresponding author. Address: Department of Psychology, The University of Calgary, 2500 University Dr NW, Calgary, AB, T2N 1N4, Canada. Tel.: +1 (403) 220-4969. E-mail address: firstname.lastname@example.org (M. Noel).
Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Web site ( www.painjournalonline.com ).
Epidemiological and cross-sectional studies have shown that post-traumatic stress disorder symptoms (PTSS) are common and impairing in youth with chronic pain. Yet, the co-occurrence of PTSS and pediatric chronic pain has not been examined longitudinally, which has limited understanding of theoretically proposed mechanisms (eg, sleep disturbance) underlying the PTSS-pain relationship over time. This longitudinal study aimed to fill this gap. Participants included 138 youth (Mage = 14.29, 75% girls) referred to a tertiary-level outpatient chronic pain program and one of their parents. At baseline, youth reported their pain intensity and interference, PTSS, and subjective sleep disturbances (ie, sleep quality and insomnia). Youth and parents completed semistructured diagnostic interviews to determine the child's post-traumatic stress disorder diagnostic status, and youth completed an objective assessment of sleep patterns for 7 days using actigraphy. At 3-month follow-up, youth once again completed the diagnostic interview and reported their pain intensity, pain interference, and PTSS. Partially latent cross-lagged structural equation panel models revealed that, controlling for pain intensity, pain interference and PTSS co-occurred at baseline, but not at follow-up (while controlling for baseline levels). Higher levels of baseline PTSS were predictive of increases in pain interference at follow-up. Furthermore, subjective sleep disturbances mediated the relationship between baseline PTSS and follow-up pain interference. These findings lend support to conceptual models of PTSS–pain co-occurrence and highlight a critical need to assess and address trauma and sleep disturbances in youth with chronic pain.
In youth with chronic pain, higher baseline post-traumatic stress disorder symptoms predicted worsening of pain interference over time, with subjectively reported sleep mediating the relationship.
Pediatric chronic pain is prevalent, 29 disabling, 54 costly, 63 may persist into adulthood, 67 and often co-occurs with internalizing mental health issues. 66 Post-traumatic stress disorder symptoms (PTSS) are prevalent in youth with chronic pain. 46 In 1 study, 32% of children with chronic pain assessed at an interdisciplinary pain clinic reported clinically significant levels of PTSS as compared to 8% of pain-free youth. 48 The co-occurrence of pediatric chronic pain and PTSS has been conceptualized by shared symptoms (eg, avoidance and hypervigilance) that mutually maintain each other over time. 20 Although a recent systematic review in adults found mixed support for the mutual maintenance of PTSS and chronic pain over time, 57 studies have not examined the longitudinal relationships between PTSS and chronic pain in youth or their underlying mechanisms.
Sleep disturbance has been posited as a key factor underlying the co-occurrence between PTSS and chronic pain. 20 Cross-sectional evidence from youth with chronic pain revealed that subjective sleep disturbances mediated the relationships between PTSS and pain intensity and interference. 47 Moreover, insomnia mediated the relationship between pain intensity and functional disability. 25 Adolescents exposed to traumatic events in childhood are more likely to report insomnia. 68 However, longitudinal examinations of subjective sleep disturbances in pediatric pain and PTSS mutual maintenance are lacking. Furthermore, self-report of sleep (eg, sleep quality) provides data on behavioral aspects of sleep, but actigraphic assessment of sleep patterns (eg, sleep duration) can complement this information. In adults with chronic pain and PTSS, objectively assessed sleep disturbances through polysomnography constitute a distinct profile characterized by reduced slow wave sleep and repeated awakenings 32 that have been linked to worse outcomes (eg, intensified PTSS 33 and pain 60 ). In pediatric chronic pain samples, actigraphic measured sleep duration was associated with higher levels of next-day pain intensity. 36 Youth diagnosed with post-traumatic stress disorder (PTSD), vs depression and controls, had more fragmented sleep and longer sleep-onset latency. 16 Thus, youth with PTSS and/or chronic pain may have disturbed sleep patterns. Yet, the role of sleep patterns in the maintenance of co-occurring pediatric chronic pain and PTSS has not been examined.
Using a cohort of youth with chronic pain, the present longitudinal study aimed to examine (1) the co-occurrence of PTSS and pain intensity and interference at baseline and 3-month follow-up and (2) the mediating role sleep quality and patterns (ie, sleep quality, insomnia, and sleep duration) in the relationship between PTSS and pain intensity and interference over time. Consistent with conceptual models of PTSS and chronic pain co-occurrence in youth, 20 we hypothesized significant relationships between PTSS and pain intensity and interference at each time point. Furthermore, given that untreated PTSS tend to worsen over time, 65 we hypothesized that higher levels of baseline PTSS and pain intensity and interference would predict worsening PTSS and pain intensity and interference at follow-up, with sleep mediating these relationships.
One hundred thirty-eight youth with chronic pain and one of their parents were recruited from the chronic pain rehabilitation programs at a tertiary-level pediatric hospital in Western Canada. The study was approved by the institution's health research ethic board. Youth were eligible for the study if they were between the ages of 10 and 18 years, identified as having chronic pain (ie, pain lasting ≥3 months), and reported ongoing pain during the recruitment screening process. Youth were excluded if they had cognitive impairment, developmental disorders, schizophrenia or other psychotic disorders, attention-deficit hyperactivity disorder, serious chronic health or life-threatening conditions (eg, cancer), and if they were unable to read or speak English. This study is part of a broader program of research that examined children's pain memory biases as another mutually maintaining mechanism underlying the PTSS and chronic pain relationship. The study exclusion criteria were in place to account for the influence of neurodevelopmental and psychotic disorders on youth memory. Treatment status (ie, whether youth had or had not received treatment at one of the pediatric pain programs at the tertiary-level pediatric hospital) was not an exclusion criterion.
Youth who had received or had been referred for treatment at a pediatric chronic pain program at a tertiary-level children's hospital within the past 2 years, and one of their parents, were contacted over the phone and screened for eligibility. The baseline assessment included the following elements: (1) Youth sleep patterns were objectively assessed using wrist actigraphy (ie, Actiwatch-2 mailed to participants with a set of instructions) for 7 consecutive days to obtain measures of sleep duration. Youth received links to complete daily sleep diaries to assist in scoring actigraphy data through email or text messages. Research staff monitored timely (ie, same day) completion of the sleep diaries; (2) youth past and present PTSD diagnostic status was assessed using a clinician-administered semistructured interview administered through phone with youth and parents separately; and (3) youth and parents completed baseline questionnaires using the Research Electronic Data Capture (REDCap), a secure online data collection tool. 18 Parents completed measures pertaining to their own and their children's sociodemographics. Youth completed measures assessing pain characteristics (intensity, interference, location, frequency, and duration), PTSS, sleep quality, and insomnia.
At 3-month follow-up, youth PTSD diagnostic status was assessed again using the semistructured diagnostic interview. Youth also reported their PTSS and pain intensity and interference using online REDCap questionnaires. Three months is a typical follow-up timepoint in pediatric chronic pain treatment programs. 9,19 Throughout the study, parents and youth received total of $65 gift cards as reimbursement for their time.
Parents reported on their own and their child's age, sex, and ethnicity, as well as their annual household income.
2.3.2. Pain characteristics
Youth completed the valid and reliable Pain Questionnaire 52 to report their pain location, frequency, and duration (in months). Specifically, youth used a validated body map to report their pain location. 56 Pain frequency was reported using a 5-point Likert scale ranging from “not at all” to “daily.” The Pain Questionnaire has been previously used in studies of youth with chronic pain. 47,53 Participants reported their average pain intensity using an 11-point Likert scale (0 = “no pain” and 10 = “worst pain possible”). The 4-item Patient-Reported Outcomes Measurement Information System (PROMIS-25) pain interference subscale assessed the degree to which pain interferes with everyday life. Youth reported their pain interference over the past 7 days on a 5-point Likert scale that ranged from 0 = “never” to 4 = “almost always.” The summed scores were standardized into t-scores (M = 50, SD = 10), with higher T-scores representing greater pain interference. This subscale has been shown to have good construct validity 26 and has been previously used in pediatric chronic pain samples. 44,47,48 Internal consistency was good (α = 0.81) in our sample.
2.3.3. Post-traumatic stress disorder diagnosis
Youth completed the post-traumatic stress disorder module of the Kiddie-SADS-Present and Lifetime Version (K-SADS-PL), 27 a semistructured diagnostic interview that assesses current and past episodes of psychopathology in children aged 6 to 18 years of age based on DSM-5 diagnostic criteria. Parents also completed the module answering questions about their children. In cases of disagreement, child report was prioritized given the internalizing nature of PTSD. The interview was administered through phone by individuals with at least an undergraduate degree in psychology who had undergone intensive training on the measure. All interviewees were supervised by a registered PhD-level clinical psychologist. In the PTSD module, youth were asked whether they had experienced any of the listed traumatic events (eg, witnessing a natural disaster or being abused) as well as the presence of associated PTSS (eg, hypervigilance and recurrent thoughts). If youth endorsed ≥1 symptom(s), the PTSD supplement was administered to determine whether a PTSD diagnosis was warranted. The K-SADS-PL is psychometrically sound and rigorous diagnostic assessment tool of mental health disorders in youth. 28
2.3.4. Post-traumatic stress disorder symptoms
To assess PTSS, the 27-item Child PTSD Symptom scale (CPSS-5) 13 was used. Youth were instructed to first think of a scary or upsetting event (eg, a car accident or abuse). Similar to previous research, 48 the recorded answers were coded into categories of traumatic life events ( Table 1 ). Participants then rated 20 symptoms on a 5-point Likert scale that ranges from 0 = “not at all” to 4 = “6 or more times a week/almost always.” The 20 symptoms map onto the DSM-5 PTSD symptom clusters (ie, re-experiencing/intrusion, avoidance, negative alterations in mood/cognition, and altered arousal/reactivity). 1 The total symptom score indicates the PTSS severity level, with a clinical cutoff score of 31. In the current sample, 19% of participants reported a score of 31 or higher at baseline (and 14% at follow-up), which is similar to previous research with youth from the same geographical region/country. 44,48 Subclinical levels of PTSS, as opposed to a diagnosis of PTSD, have been shown to be associated with decreased functioning. 14 As outlined in the conceptual model of PTSS-pain co-occurrence, 20 symptoms of PTSD, as opposed to a clinical diagnosis of PTSD (ie, whether or not they meet full diagnostic criteria for the disorder), are hypothesized to contribute to and exacerbate pain over time. Therefore, PTSD symptoms vs youth diagnostic status of PTSD were included in the subsequent analyses. The CPSS-5 is psychometrically sound, 13 has been previously used in pediatric chronic pain research, 44 and demonstrated excellent internal consistency (α = 0.95) in the current sample. Given that participants could endorse abuse (ie, they were asked to identify the nature of their “most distressing event” in their own words), research staff and graduate students regularly monitored the surveys for any indications of imminent harm (eg, child abuse, neglect, and suicidal intent). If and when this was indicated (n = 1 event), a member of the research staff or a graduate student followed an established protocol that was approved by the institution's research ethics board. Specifically, they immediately contacted the PI (senior author of this article), a PhD-level registered clinical psychologist, who then took action to report this to the appropriate services (eg, local Child and Family Services). No data on ongoing treatment for mental health (including PTSD) symptoms were collected for the purposes of the study. However, patients do not typically receive treatment for mental health in tertiary-level outpatient chronic pain programs.
Types of traumatic events reported by children at baseline.
2.3.5. Subjective sleep quality
The 10-item revised Adolescent Sleep–Wake Scale 10 was used to assess youth sleep quality. This scale has been used in youth with chronic pain 10 including in a previous cross-sectional study examining the role of sleep disturbance in the PTSS-pain relationship in youth. 47 Youth reported on the frequency of problems with sleep initiation, maintenance, and overall sleep quality within the past month on a 6-point Likert scale ranging from 1 = “always” to 6 = “never.” The revised Adolescent Sleep–Wake Scale consists of 3 subscales: (1) going to bed (eg, “In general, I try to put off or delay going to bed”), (2) falling asleep and reinitiating sleep (eg, “After waking up during the night, I have trouble getting comfortable”), and (3) returning to wakefulness (eg, “In the morning I wake up feeling rested and alert”). The subscale scores are averaged to obtain the total score, with higher scores representing better sleep quality. In our sample, internal consistency of this measure was good (α = 0.82).
The Insomnia Severity Index (ISI) 41 was used to assess insomnia symptoms (ie, falling and staying asleep). Youth were asked to rate the severity of their insomnia symptoms over the past 2 weeks on a 5-point Likert scale from 0 = “none” to 4 = “very severe.” In the current sample, 54% of participants reported a score of 10 or greater, which is the clinical cutoff indicating the presence of mild insomnia. 41 The ISI has been shown to be a reliable and valid self-report measure of insomnia 3 and has been used in previous pediatric chronic pain research. 25,50 Internal consistency of this measure was very good (α = 0.86) in our sample.
2.3.7. Objective sleep patterns (actigraphy)
Similar to previous research, 51,62 participants' sleep patterns were assessed for 7 consecutive days using the Actiwatch-2 (Philips Respironics, Inc., Murrysville, PA), a watch-like device worn on the nondominant wrist that uses proprietary software to calculate sleep–wake patterns based on movement. Movements are registered through an omnidirectional mercury switch every 60 seconds (ie, epoch) and scored based on their amplitude and frequency. Actigraphy data were extracted using the Philips Actiware software (version 6), automatically created sleep intervals were removed, and data were manually scored. Sleep onset and sleep offset were defined by 10 consecutive sleep epoch counts. Specifically, 10 consecutive sleep epoch counts (with maximum of 1 active epoch count) indicated sleep onset and offset 37 ; the consecutive counts were identified using visual inspection of the data. Event markers (ie, youth pushed a button on the Actiwatch when they were in bed intending to fall asleep) and sleep diaries were used to confirm sleep onset and offset. Daily diaries and event markers were used to confirm approximate times of sleep onset and offset. In case of discrepancies between actigraphy and daily diaries (on average < 30 minutes), individual actograms were discussed within the study team and consensus on scoring was reached. Sleep duration, the total number of sleep epochs from sleep onset to sleep offset, wake after sleep onset, and sleep efficiency were calculated using the Philips Actiware software. Consistent with the previous research, 51 all measures were averaged over the total number of nights that youth wore the Actiwatch (minimum 5 nights; range: 5-9). Actigraphy provides comparable estimates to polysomnography for sleep duration in youth. 23
2.4. Statistical analyses
Descriptive statistics and correlational analyses were conducted using the Statistical Package for the Social Sciences (SPSS), version 24. 22 Categorical variables were characterized using frequency statistics; mean values and SDs were calculated for continuous variables. Pearson bivariate two-tailed correlational analyses were used to analyze the associations between key variables. T- and χ2 tests as well as bivariate correlations were conducted to examine differences in key variables as a function of sociodemographic characteristics (ie, age and sex).
Subsequent analyses consisted of partially latent cross-lagged structural equation panel models. Structural equation modeling was performed using Analysis Of a Moment Structures (AMOS) 25.0. 2 Analyses adjusted for sex. In the first model ( Fig. 1 , top panel) examining bidirectional influences of pain interference, pain intensity, and PTSS, 3-month pain interference, intensity, and PTSS are regressed on their respective baseline levels as well as each other at baseline. By adjusting for baseline levels of our outcomes, this model examines whether baseline PTSS predict change over time in pain interference and intensity and whether baseline pain interference and intensity predict change over time in each other as well as PTSS. In our second model ( Fig. 1 , bottom panel), latent subjective sleep variable at baseline was indicated by participant reports on the ISI and the Adolescent Sleep–Wake Scale (ie, 2 self-report measures of sleep sharing a significant amount of variance, r = −0.67, P < 0.001) and was predicted by baseline pain interference, intensity, and PTSS. Given weak correlations between sleep time and the other actigraphy parameters (ie, wake after sleep onset and sleep efficiency), an objective sleep latent variable was not created. Instead, objective sleep was indicated by sleep duration, one of the most commonly used objective sleep parameter in research on PTSD and pain. 12,34,36,37 Objective and subjective (ie, sleep quality and insomnia) sleep then predicted change over time in PTSS, pain interference, and pain intensity adjusting for their respective baseline levels.
Nontrimmed models of bidirectional relations between pain and PTSD symptoms at baseline and 3-month follow-up (top panel) and the role of subjective and objective sleep variables in the relations between PTSD symptoms and pain at baseline and 3-month follow-up (bottom panel).
As measures of goodness of fit, we presented χ2, ratio of χ2 to degrees of freedom, comparative fit index (CFI), and root-mean-square error of approximation (RMSEA). Generally, a χ2/df less than 2, 8 CFI values greater than 0.90, 21 and an RMSEA of less than 0.08 with a nonsignificant P of close fit (pClose) 31 indicate acceptable fit. AMOS cannot estimate indirect effects in the presence of missing data. Accordingly, missing data were singly imputed through regression imputation with Full Information Maximum Likelihood (FIML) estimation. Estabrook and Neale 11 note that FIML most accurately estimates individual scores with missing data relative to other methods (eg, mean imputation). This approach is generally acknowledged to be preferable to other methods for dealing with missing data, such as listwise deletion or mean imputation, as these latter approaches are more likely to yield biased estimates. 38,42,57 A bias-corrected bootstrapping procedure was used to estimate confidence intervals of indirect effects. Bootstrapping is a nonparametric method based on resampling (5000 in the current study) with replacement. 4,59 We present 95% confidence intervals as well as parameter estimates associated with the indirect effect. Specifically, we examined indirect effects of baseline pain interference, intensity, and PTSD on 3-month outcomes through objective and subjective sleep. For clarity, the paths included in the mediation analyses are bolded ( Fig. 2 , bottom panel). On the basis of Wald tests, and in the interest of parsimony, nonsignificant paths were removed in order beginning with the least significant paths. After the deletion of nonsignificant paths, χ2 difference tests were conducted to ensure the trimmed model did not fit the data significantly worse as compared to the full model.
Trimmed models of bidirectional relations between pain and PTSD symptoms (PTSS) at baseline and 3-month follow-up (top panel) and the role of sleep in the relationship between pain and PTSS at baseline and 3-month follow-up (bottom panel). *P < 0.05, †P < 0.01, ‡P < 0.001. Indirect effects of subjective sleep: (1) baseline to follow-up pain interference, b = 0.08, 95% CI [0.03-0.18], P = 0.012; (2) baseline PTSS to follow-up pain interference, b = 0.08, 95% CI [0.02-0.16], P = 0.013. ASWS, Adolescent Sleep–Wake Scale; ISI, Insomnia Severity Index
3.1. Missing data
Of the 138 participants, 95% (n = 131) completed all items of baseline measures and 5% (n = 7) had 1 or more items missing; 89% (n = 123) of participants completed 5 to 9 nights of actigraphy [M = 6.95, SD = 0.41] ( Fig. 3 ). Seventy-nine percent of participants (n = 109) completed or partially completed follow-up measures, and 82% (n = 113) completed a follow-up clinical interview. At 3 months, 1 or more items of follow-up measures were missing for 18% (PTSS; n = 20), 11% (pain interference; n = 12), and 9% (pain intensity; n = 10) of participants. There were no significant differences on any of the baseline key variables among participants who completed the study vs those who were lost at follow-up (ps > 0.05). Missing data were estimated using FIML that is only valid when data are missing at random (ie, there are no systematic biases in which participants have missing data). Little's Missing Completely at Random (MCAR) test 38 confirmed that missingness was unrelated to any variable in our study: χ2 (122) = 116.76, P = 0.62. Thus, data were viewed as missing at random for analyses.
3.2. Descriptive statistics
Sociodemographic data are summarized in Table 2 . The majority of participants were female (75.4%); the average age was 14.29 years (SD = 2.30, range 9-18). At baseline, youth reported an average pain intensity level of 5.65/10 (SD = 1.80) and an average pain interference level of 55.06 (T-score; SD = 9.13). At follow-up, average pain intensity was 5.40/10 (SD = 1.98), and pain interference was 53.61 (SD = 9.04). Overall, 50% of participants experienced pain “daily,” and 47.2% reported multiple pain locations. The average duration of their pain was 39.17 months (SD = 38.66).
Sociodemographic characteristics of the sample.
Descriptive statistics of key variables are summarized in Table 3 . Youth age was associated with key variables. Older children reported significantly lower sleep quality (r = −0.26, P < 0.01), more severe insomnia (r = 0.22, P < 0.05), and shorter sleep duration (r = −0.34, P < 0.001). Furthermore, older children reported higher levels of pain interference (baseline: r = 0.18, P < 0.05; follow-up r = 0.23, P < 0.05) and baseline PTSS (r = 0.21, P < 0.05). Girls, as compared to boys, reported higher levels of pain interference (baseline: t  = 2.01, P < 0.05; follow-up: t  = 2.89, P < 0.01). Based on the objective assessment of sleep, girls, vs boys, slept for a longer duration (t [39.17] = 2.28, P < 0.05). Girls did not significantly differ from boys with regard to PTSS or subjective sleep variables (ps > 0.05). Bivariate correlations between the key variables and outcomes are presented in Table 4 . Supplementary analyses are available at http://links.lww.com/PAIN/A927 .
Correlations among key variables.
3.3. Structural equation modelling
Our initial model of PTSS and pain co-occurrence that did not include sleep ( Fig. 1 , top panel) was fully saturated, and we therefore do not report fit indices given that they will by default be perfect. Sex was not significantly related to baseline and follow-up levels of pain intensity or PTSS. Baseline levels of pain intensity were not significantly associated with follow-up PTSS or pain interference. Pain interference at baseline was not significantly related to follow-up pain intensity and PTSS. Therefore, these paths were removed from the model. The model ( Fig. 2 , top panel) demonstrated excellent fit to the data χ2 (10) = 5.33, P = 0.87, χ2/df = 0.53, CFI > 0.99, RMSEA < 0.001, pClose = 0.953. Pain interference, intensity, and PTSS all showed high levels of rank-order stability, although PTSS appeared to be more highly stable than pain interference. Over and above sex and baseline pain interference levels, elevated PTSS at baseline predicted increases in pain interference at 3-month follow-up. There were no effects of baseline pain interference or PTSS on follow-up pain intensity. This suggests that PTSS may contribute to increases in pain interference, but not intensity, over time, and that neither pain interference nor pain intensity influence PTSS over time.
The second model ( Fig. 1 , bottom panel) then included subjective and objective sleep variables. This model showed excellent fit to the data: χ2 (8) = 16.15, P = 0.04, χ2/df = 2.02, CFI = 0.977, RMSEA = 0.086, pClose = 0.145. Given the strong correlation between sleep quality and insomnia severity (r = −0.67, P < 0.001), the 2 variables were combined to create a latent subjective sleep quality variable. Observed indicators of subjective sleep (ie, sleep quality and insomnia symptoms) showed high standardized loadings on their respective latent variable (βs ≥ |0.73|). Baseline levels of pain intensity and pain interference did not significantly predict sleep duration, and baseline pain intensity was not significantly associated with subjective sleep quality. Baseline pain intensity and children's sex were not significantly associated with follow-up levels of PTSS, and baseline pain interference did not significantly predict follow-up pain intensity or PTSS. Baseline PTSS did not significantly predict follow-up levels of pain intensity or pain interference. Subjective sleep quality latent variable was not significantly predictive of follow-up pain intensity or PTSS. Furthermore, sex was not significantly associated with subjective sleep quality latent variable, baseline PTSS, and baseline/follow-up pain intensity. Sleep duration did not predict any of the key outcomes (ie, follow-up levels of PTSS, pain intensity, and pain interference). These paths were therefore removed from the model.
The trimmed model continued to show excellent fit to the data ( Fig. 2 , bottom panel): χ2 (29) = 44.68, P = 0.03, χ2/df = 1.54, CFI = 0.971, RMSEA = 0.063, pClose = 0.26 and did not fit the data significantly worse than the full model, Δχ2 (21) = 28.53, P = 0.13. Baseline levels of PTSS and sex were associated with sleep duration, such that higher levels of PTSS were linked to shorter sleep duration, and boys slept less as compared to girls. Elevated pain interference and PTSS at baseline additively predicted worse subjective sleep quality controlling for the effects of each other. Worse subjective sleep quality in turn predicted increased pain interference at follow-up. There was a significant indirect effect from greater baseline pain interference to greater follow-up pain interference through worse subjective sleep quality (b = 0.08, 95% CI [0.03-0.18], P = 0.012). That is, elevated pain interference may contribute to worse subjective sleep quality, which, in turn, increases pain interference over time. There was additionally a significant indirect effect from greater baseline PTSS to greater follow-up pain interference through worse subjective sleep quality (b = 0.08, 95% CI [0.02-0.16], P = 0.013). Thus, PTSS may worsen pain interference over time due to worsened subjectively reported sleep quality.
The current study is the first to apply cross-lagged structural equation models to longitudinally examine the co-occurrence and mutual maintenance of PTSS and pain in a sample of youth with primary pain disorders. The findings lend support to the conceptual model of PTSS and pediatric chronic pain co-occurrence. 20 The model underscores symptoms (eg, hyperarousal through sleep disturbance) that are common between PTSD and chronic pain and that contribute to their mutual maintenance. Consistent with the model and existing empirical evidence, 48 the current findings demonstrated the co-occurrence of PTSS and chronic pain in youth at baseline but not at follow-up (while controlling for baseline levels). Moreover, over and above the effect of each other, PTSS and baseline pain interference predicted worsened pain interference at 3-month follow-up. This is consistent with prospective research with youth after traumatic brain injury, showing that PTSS more strongly predicted pain at 3- and 6-month follow-ups than vice versa. 6 These findings suggest that PTSS is a stronger driver of pain interference than vice versa and should be assessed and targeted from the outset of treatment to comprehensively manage pain-related interference over time.
Consistent with our hypotheses, findings revealed the unique contribution of subjective sleep quality in the worsening of pain interference and the co-occurrence of PTSS and pain interference over time. Sleep disturbance, as indicated by worse self-reported sleep quality and insomnia symptoms, was found to underlie the relationships between baseline and 3-month pain interference. Specifically, the rate of change in baseline to follow-up pain interference levels was mediated by subjective sleep quality with higher levels of baseline pain interference contributing to worse sleep quality, which, in turn, was associated with increased pain interference at follow-up. Furthermore, sleep quality mediated the relationship between baseline PTSS and pain interference at follow-up, such that higher levels of baseline PTSS were associated with reduced sleep quality that, consequently, predicted higher levels of pain interference at 3 months. These findings support a 64 conceptual model of the sleep-pain relationship in pediatric pain that posits a bidirectional relationship between sleep quality and pain levels that is influenced by psychological (eg, trauma) and physiological/biological (eg, sex) factors.
Counter to our hypotheses, significant associations were not found between actigraphic measured sleep patterns (ie, sleep duration) and follow-up levels of PTSS or pain interference. These findings are contrary to research demonstrating the predictive role of objectively measured sleep patterns (eg, shorter sleep duration) in next-day pain intensity levels among youth with chronic pain. 36 It remains unclear whether sleep quantity and sleep patterns play a role in PTSS or pain interference. The lack of significant relationships in the current study may be due to averaging actigraphy variables, thus losing night-to-night impact of sleep disturbances.
Lack of significant associations between objective sleep patterns and subjective sleep quality variables is expected and consistent with existing research. 51 Sleep deficiency is a broad domain encompassing both the quantity and quality of sleep. Self-reports of sleep quality and insomnia severity may be indicative of and dependent on negative cognitive and/or anxiety-related biases. For example, elevations in general anxiety levels may translate into sleep-specific worry that prevent youth from falling asleep. Sleep quantity, on the other hand, may be affected by a different set of behavioral (eg, routines) and environmental (eg, school start times) factors. Future work is needed to more comprehensively understand the contribution of various sleep domains to pain intensity and pain interference.
Although the current study was limited to behavioural measures, future research should also examine the neurobiological underpinnings of these relationships. The hypothalamic–pituitary–adrenal (HPA) axis acts as the primary circuit underlying the brain's response to stress, but also plays a significant role in the homeostatic balance between alertness and sleep promotion. Sleep disruption increases HPA axis activation and stress reactivity. 7,40 Moreover, maladaptive stress responses to pain perpetuate the release of stress hormones and prolong painful experiences. 17 PTSD and sleep deprivation inflict a serious impact on the anterior cingulate cortex through alterations in both its structure and function. 43 These changes to the anterior cingulate cortex reflect impaired modulatory control over the limbic system, leading to PTSS. 39,43,49 Hippocampal volumetric reductions have been consistently reported in PTSD research, 15,30,58,69 while sleep disturbances and chronic stress have also been found to reduce hippocampal neurogenesis and promote neuronal loss. 24 The hippocampus has been implicated in the consolidation of traumatic memories, which fosters sleep deprivation and, in turn, maintains or exacerbates PTSS. 43,45 Finally, the insular cortex plays a pivotal role in sleep, pain, and PTSS. Reduced insular volume in response to both PTSD and sleep deprivation impairs interoceptive and emotional processing, while increased insular activity enhances pain perception, contributing to the perpetuation of sleep disturbances. 43
The current findings are of clinical importance. Given high rates of PTSS among youth with chronic pain, 46,48 and the impairing influence of PTSS on pain interference over time, findings underscore the importance of assessing PTSS among youth with chronic pain. For youth with elevated PTSS, incorporation of trauma-focused modules in treatment protocols may contribute to improved functioning and quality of life. A larger body of treatment research has examined co-occurring chronic pain and PTSD in adults and treatment protocols targeting both conditions have been developed. 5 Interdisciplinary stepped-care treatment programs addressing both conditions have been successfully incorporated into the Veteran Affairs System of Care and have resulted in increased physical and everyday functioning and improved mental health symptoms. 5 Developing integrative treatments for pediatric populations is warranted.
The unique contributions of sleep quality and insomnia symptoms to the maintenance of pain interference indicate another clinical intervention target. A pilot trial of brief cognitive-behaviour therapy for insomnia in youth with insomnia comorbid with physical and/or psychological comorbidities has demonstrated improvements in sleep quality, as well as health-related quality of life and psychological symptoms (ie, decreased depressive and/or anxiety symptoms). 50 Our findings suggest that effectively targeting subjective sleep quality and insomnia could potentially buffer against worsening pain interference over time.
The current findings should be considered in light of limitations. First, youth PTSS, pain interference, and sleep quality were assessed at 2 timepoints, which partially limits the explanatory/causal role of sleep disturbances in the PTSS-pain interrelationship. That is, mediation analyses presume a causal order of effects between variables, with predictors necessarily temporally preceding outcomes. Although sleep disturbances measured at baseline predicted change over time in pain interference, an additional follow-up assessment would permit examination of whether pain interference and PTSS precede and predict the development of sleep disturbances over time. Second, objective assessment of sleep disturbances was limited to actigraphy. The indices available through actigraphy do not fully capture all of the processes that are believed to underlie the PTSS-pain association (ie, sleep architecture, such as disrupted slow wave sleep, decreased rapid eye movement sleep). Furthermore, actigraphy merely records activity, or rather a lack of activity, as a proxy for sleep, rather than directly measuring the length and depth of the sleep stages. 35 Polysomnography, the current gold standard of objective sleep assessment, assesses sleep architecture and may better elucidate distinct facets of sleep quality/disturbances that maintain PTSD and primary pain disorder symptoms. Third, the models did not control for other internalizing (eg, anxiety) or externalizing (eg, ADHD) symptoms that are common in youth with chronic pain. 46,61 Future research should examine and disentangle the relationships between mental health (eg, anxiety, depressive, PTSD, and ADHD) symptoms and their role in the onset and/or maintenance of pediatric chronic pain. In addition, future longitudinal research is needed to examine the mutual maintenance of pain and PTSS in samples of youth who have not yet developed chronic pain conditions (eg, youth with new-onset pain due to injuries). Finally, 21% of participants did not complete follow-up assessment, which may have resulted in a positively skewed follow-up assessment (ie, participants may have dropped out due to decreases in functioning).
In conclusion, this study prospectively examined the co-occurrence and mutual maintenance of PTSS and pain in youth with primary pain disorders. Using cross-lagged models, we demonstrated that PTSS and pain interference co-occur with higher baseline PTSS predicting worsening of pain interference over time. Sleep quality and insomnia mediated the association between baseline and follow-up pain interference, as well as the relationship between baseline PTSS and follow-up pain interference. Our findings support and complement existing conceptual models and highlight the importance of assessing and addressing PTSS and sleep in pediatric samples with chronic pain. Furthermore, the findings confirm that sleep quality is one of the mechanisms that contribute to the PTSS-pain interference interrelationship. Sleep quality is modifiable, making it a critically important treatment target in youth with primary pain disorders.
Conflict of interest statement
The authors have no conflicts of interest to declare.
Appendix A. Supplemental digital content
Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/A927 .
This work was supported by funding from the Chronic Pain Network and the Alberta Children's Hospital Research Institute awarded to Dr. Noel. Ms. Pavlova was supported by the Alberta Strategy for Patient-Oriented Research Graduate Studentship. J. Katz is supported by a Canadian Institutes of Health Research Canada Research Chair in Health Psychology.
. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders (DSM-5). Arlington: American Psychiatric Association, 2013.