Risk Factors for the Development of Multisite Pain in Children

Abstract

Objective

Chronic pain has economic costs on par with cardiovascular disease, diabetes, and cancer. Despite this impact on the health care system and an increasing awareness of the relationship between pain and mortality, efforts to identify simple symptom-based risk factors for the development of pain, particularly in children, have fallen short. This is critically important as pain that manifests during childhood often persists into adulthood. To date no longitudinal studies have examined symptoms in pain-free children that presage a new, multisite manifestation of pain in the future. We hypothesized that female sex, sleep problems and heightened somatic complaints at baseline would be associated with the risk of developing new multisite pain one year later.

Methods

Symptom assessments were completed by parents of youth (ages 9–10) enrolled in the Adolescent Brain Cognitive Development study. Multivariate logistic regression models focused on children who developed multisite pain one year later (n=331) and children who remained pain-free (n=3335).

Results

Female sex (OR=1.35; 95% CI=1.07, 1.71; p=0.01), elevated non-painful somatic complaints (OR=1.17; 95% CI=1.06, 1.29; p<0.01), total sleep problems (OR=1.20; 95% CI=1.07, 1.34; p< 0.01), and attentional issues (OR=1.22; 95% CI=1.10, 1.35; p<0.001) at baseline were associated with new multisite pain one year later. Baseline negative affect was not associated with new multisite pain.

Discussion

Identifying symptom-based risk factors for multisite pain in children is critical for early prevention. Somatic awareness, sleep and attention problems represent actionable targets for early detection, treatment, and possible prevention of multisite pain in youth.

INTRODUCTION

Chronic pain during childhood and adolescence is common, affecting nearly 20–25% of all youth.[1] Pain is also a significant personal and societal burden, impacting nearly every aspect of a child’s life, evidenced in poor school performance, activity limitations, isolation, and increased risk of future opioid misuse.[2, 3] Further, pain in childhood often persists into adulthood. A key prospective study found that >80% of youth with juvenile fibromyalgia, which is characterized by widespread musculoskeletal pain, continued to have fibromyalgia symptoms as adults.[4] Additionally, between 32 – 63% of youth with functional abdominal pain continued to have pain in adulthood, and often developed pain in other body regions (e.g., back pain and headache).[5] Retrospective research is consistent with these studies in that many adults with current chronic pain also experienced pain in childhood, indicating that the vulnerability to chronic pain may be established early in life.[6, 7]

Chronic pain, in children and adults, is often accompanied by a cluster of non-painful symptoms such as sleep disturbance, fatigue, negative mood, multisensory sensitivity, and cognitive problems.[8–10] However, the temporal relationship between pain and these related symptom domains is unclear. The purpose of this study is to determine if non-painful symptoms can serve as early warning signs that children are at risk of developing pain in the future. This is vital to reveal because, to the best of our knowledge, there are no published studies examining risk factors for pain in children who were entirely pain free at baseline. Most prior studies have relied upon samples of children who had localized pain complaints at baseline; unsurprisingly, current pain tends to be a very strong predictor of future pain manifestations.[11–14] Some studies have examined risk factors for new onset regional pain (e.g., low back pain or headache) in children, but fail to measure pain in other body regions, which is problematic given that pain conditions tend to co-occur within individuals (i.e., multisite pain).[15–17]

These limitations aside, female sex and poor sleep appear to be fairly consistent risk factors for pediatric pain across studies.[12, 14, 16, 18] Heightened somatic awareness has also been identified as a risk factor for pediatric[19] and adult pain.[20, 21] Few other predictors have been consistently found across studies—likely due to how pain is defined/measured, which predictor variables are included in the analyses, and large variation in age ranges of the study populations. This latter point is critical to consider as risk factors may change over time or depend on a particular developmental stage.[22] Some studies report that negative emotions (such as anxiety and depression), behavioral/conduct problems, and family factors such as parental pain and family income are risk factors for pediatric pain, though these findings do not replicate in other studies.[11–14, 16] Although these studies identify many potential risk factors for pain trajectories or persistence, the presence of pain at baseline, common to all of them, may well obscure other temporal relationships with non-painful symptoms.

The Adolescent Brain Cognitive Development (ABCD) study provides an excellent opportunity to study the longitudinal development of pain in childhood and adolescence. The ABCD study is a prospective cohort study of nearly 12,000 children exploring a multitude of factors affecting childhood development, including clinical and biological phenotyping data – all of which are released publicly.[23] At the baseline ABCD visit children were 9–10 years old, an age when most children are pre-pubertal or in the early stages of pubertal development. Importantly, ABCD includes parental assessments of pain and many related symptom domains, allowing us to examine symptoms associated with the development of pain in children who were pain-free at study entry. We focused on the development of new ‘multisite’ pain as the primary outcome. Multisite pain, compared to pain in a single area of the body, is more likely to reflect neurobiological vulnerability or central nervous system (CNS) sensitization (i.e., nociplastic pain) rather than a transient injury. Nociplastic pain can be present in many chronic pain conditions but is typically associated with juvenile and adult fibromyalgia, and importantly, once established it is exceedingly difficult to treat. We hypothesized that female sex, sleep problems and heightened somatic complaints at baseline would be associated with the risk of developing multisite pain one year later. However, given the conflicting findings of previous studies, we also included measures of other internalizing and externalizing behaviors to determine if they significantly predict new onset multisite pain.

METHODS

Sample

The ABCD study included 11,875 children (9–10 years old) from 21 sites across the United States.[24] The Institutional Review Board for each site approved the procedures. All participants provided written informed consent (parents) and informed assent (children). We downloaded all data used in this analysis from NDA Study 901 (Release 3.0), DOI 10.15154/1519007 accessed on 04/08/2021. Children whose parents reported that their children experienced no pain at baseline and had complete data (for measures relevant to this analysis) at the 1-year follow-up were included in the prediction analyses. In cases in which siblings or twins from a single family participated in ABCD, we limited analyses to the sibling or twin with the earliest enrollment date to exclude sources of dependence. Figure 1 depicts derivation of the sample included for this analysis.

Figure 1.

Consort diagram showing sample for new multisite pain analysis. Dashed boxes represent excluded participants.

Measures

Demographics.

Parental report provided child race/ethnicity, sex, and household income. For pubertal status, an average of the child self-report Pubertal Development Scale items was calculated,[25] and if a participant did not answer two or more of the five items for their respective sex, then a score was not calculated (and they were excluded from analyses; see Figure 1).

The Child Behavior Checklist (CBCL) and Sleep Disturbance Scale for Children (SDSC).

Parents completed the CBCL and SDSC at baseline and at the one-year follow-up visit.[23] In brief, we used all eight empirically derived behavioral/symptom subscales from the CBCL, which include the anxious/depressed (internal consistency, Cronbach α = 0.75), withdrawn/depressed (α = 0.67), somatic complaints (α = 0.22), social problems (α = 0.68), thought problems (α = 0.63), attention problems (α = 0.85), rule-breaking behavior (α = 0.68), and aggressive behavior subscales (α = 0.87).[26] We used the SDSC, which contains 26 items that can be divided into six subscales, to evaluate sleep disorders and as an overall measure of sleep disturbances.[27] We included the total sleep problems score from the SDSC (α = 0.51). Together, these measures cover a wide range of behavioral/symptom domains, including fundamental factors that have been tied to chronic pain in adult and pediatric populations represented by the “SPACE” acronym (i.e., sleep, pain, affect, cognition, energy)[8, 9], as well as additional behavioral domains (aggressive behavior, rule-breaking) that are associated with pain in children.[13, 15, 28] The primary outcome of interest was development of new multisite pain one year after reporting a pain-free baseline assessment. In the CBCL, parents reported if their child currently or within the past 6-months suffered from the following physical problems without known medical cause: 1) aches or pains (not stomach or headaches), 2) headaches, and 3) stomachaches. We categorized responses as 0 = “not true”, 1 = “somewhat or sometimes true”, 2 = “very true or often true”. We operationalized new multisite pain as parental endorsement of at least two of the three pain items at one year (either “sometimes,” or “often”), when no pain was reported at baseline. Children were characterized as pain-free if their parent reported ‘not true’ on all three pain items at baseline and one year.

Statistical Analyses

Prediction of New Multisite Pain

We used R (version 4.0.2) to perform all statistical analyses and determined associations between multisite pain and individual predictors including demographic, behavioral, and sleep factors using bivariate logistic regression. We used multivariate logistic regression to predict the primary outcome of new multisite pain. Independent predictor variables were the symptom domains described above. Since we used the pain items of the somatic complaints subscale to create the dependent variable at one year, and because children were pain-free by these criteria at baseline, the somatic complaints subscale for prediction of new multisite pain contained only those 8 questions relating to non-painful symptoms (e.g., “constipated”, “dizzy/lightheaded”, “nightmares”, “overtired without good reason”, “nausea/feels sick”, “problems with eyes”, “rashes or other skin problems”, “vomiting”). Demographic predictors included race/ethnicity, pubertal status, sex, and parental household income. We converted the values from the eight empirically derived behavioral/symptom subscales from the CBCL and the total sleep problems score from the SDSC to z-scores. To assist with interpretation of the results, pre-transformed means and standard deviations of the symptom subscales are presented in Supplemental Table 1. To address potential issues of perfect/quasi-perfect separation, we used bias-reduction methods for logistic regression. A stepwise backward elimination strategy reduced the full model to the final model. We also conducted variable selection by testing all possible combinations of fixed effect terms with Aikake’s Information Criteria (lower indicates better fit) as the selection criteria.

RESULTS

The analysis sample included 3666 children (42.0% female). Of these, 331 (9.0% developed new multisite pain at 1-year (46.8% female). Household income was the only demographic factor significantly associated with multisite pain (p=<0.01; Table 1). Bivariate analysis demonstrated the significance of the nine symptomatic variables (withdrawn/depressed, p=<0.01; others p<0.001; Table 2). The full multivariate model included all previously described covariates (Table 3). The final model contained four significant predictors: female sex, higher levels of total sleep problems, somatic complaints, and attentional issues (all p<.05; see Table 3). To understand the magnitude of these effects, being female was associated with 35% higher odds of developing multisite pain, a 1 SD increase in total sleep problems was associated with approximately 22% higher odds of developing new multisite pain, a 1 SD increase in somatic complaints was associated with 17% higher odds, and a 1 SD increase in attentional issues associated with 20% higher odds (See Table 3 for estimates from the final selected model). Of note, both measures of depressive and anxious symptoms were associated with multisite pain in bivariate analyses (anxious/depressed OR: 1.23, p<0.001; withdrawn/depressed OR: 1.16, p<0.01; see Table 2), but these associations were not statistically significant when adjusting for demographic and other symptomatic variables (anxious/depressed OR: 1.00, p=0.997; withdrawn/depressed OR: 0.97, p=0.68; see Table 3).

Table 1:

Sociodemographic Characteristics by Multisite Pain Status

Characteristic	Total (n = 3666)	New Multisite Pain (n = 331)	Pain Free (n = 3335)	p-value
Female Sex	1538 (42.0%)	155 (46.8%)	1383 (41.5%)	0.07
Pubertal Status: Mean (SD)	1.66 (0.49)	1.69 (0.49)	1.66 (0.49)	0.26
Race				0.43
White	1981 (54.0%)	186 (56.2%)	1795 (53.8%)
Black	511 (13.9%)	45 (13.6%)	466 (14.0%)
Hispanic	726 (19.8%)	60 (18.1%)	666 (20.0%)
Asian	92 (2.5%)	4 (1.2%)	88 (2.6%)
Other	356 (9.7%)	36 (10.9%)	320 (9.6%)
Household Income				<0.01
Less than $5,000	134 (3.7%)	7 (2.1%)	127 (3.8%)
$5,000 - $11,999	122 (3.3%)	7 (2.1%)	115 (3.4%)
$12,000 - $15,999	90 (2.5%)	10 (3.0%)	80 (2.4%)
$16,000 - $24,999	167 (4.6%)	27 (8.2%)	140 (4.2%)
$25,000 - $34,999	203 (5.5%)	22 (6.6%)	181 (5.4%)
$35,000 - $49,999	292 (8.0%)	30 (9.1%)	262 (7.9%)
$50,000 - $74,999	488 (13.3%)	51 (15.4%)	437 (13.1%)
$75,000 - $99,999	515 (14.0%)	48 (14.5%)	467 (14.0%)
$100,000 - $199,999	1161 (31.7%)	98 (29.6%)	1063 (31.9%)
$200,000+	494 (13.5%)	31 (9.4%)	463 (13.9%)

Note: All values reported as n (%). Multisite pain status was derived from parent-reported physical problems (aches or pains, headache, and stomachaches) with no known medical cause. Differences between groups were assessed using Chi-square tests.

Table 2:

Associations of Sociodemographic and Behavioral/Symptom Subscale Measures with Multisite Pain Status

Predictor	Bivariate Associations OR (95% CI)	p-value
Female Sex	1.24 (0.99, 1.56)	0.06
Pubertal Status	1.14 (0.91, 1.43)	0.25
Race (ref White)
Black	0.93 (0.66,1.30)	0.69
Hispanic	0.87 (0.64, 1.17)	0.37
Asian	0.44 (0.13, 1.07)	0.11
Other	1.09 (0.74, 1.56)	0.67
Household Income	0.97 (0.93, 1.01)	0.17
Anxious/ Depressed	1.23 (1.12, 1.36)	<0.001
Withdrawn/ Depressed	1.16 (1.05, 1.28)	<0.01
Somatic Complaints	1.31 (1.19, 1.43)	<0.001
Social Problems	1.27 (1.16, 1.39)	<0.001
Thought Problems	1.28 (1.16, 1.39)	<0.001
Attention Problems	1.33 (1.21, 1.47)	<0.001
Rule-Breaking Behavior	1.22 (1.11, 1.34)	<0.001
Aggressive Behavior	1.26 (1.15, 1.38)	<0.001
Sleep Problems	1.37 (1.24, 1.50)	<0.001

Note: OR = odds ratio; CI = confidence interval. All results were produced using logistic regression modeling. Statistical significance was measured via Wald test.

Table 3:

Model Estimates for the Association of Sociodemographic and Behavioral/Symptom Subscale Measures with Multisite Pain Status

Predictor	Full Model OR (95% CI)	p-value	Final Model OR (95% CI)	p-value
Female Sex	1.36 (1.07, 1.72)	0.01	1.35 (1.07, 1.71)	0.01
Pubertal Status	1.10 (0.87, 1.39)	0.43	-	-
Race (ref White)
Black	0.71 (0.47,1.05)	0.10	-	-
Hispanic	0.76 (0.54, 1.05)	0.10	-	-
Asian	0.42 (0.13, 1.04)	0.10	-	-
Other	0.97 (0.65, 1.41)	0.87	-	-
Household Income	0.97 (0.92, 1.03)	0.30	-	-
Anxious/ Depressed	1.00 (0.87, 1.14)	0.997	-	-
Withdrawn/ Depressed	0.97 (0.86, 1.10)	0.68	-	-
Somatic Complaints	1.17 (1.06, 1.30)	<0.01	1.17 (1.06, 1.29)	<0.01
Social Problems	1.03 (0.89, 1.20)	0.66	-	-
Thought Problems	0.98 (0.85,1.13)	0.79	-	-
Attention Problems	1.17 (1.01, 1.35)	0.03	1.20 (1.07, 1.34)	< 0.01
Rule-Breaking Behavior	1.04 (0.89, 1.21)	0.60	-	-
Aggressive Behavior	1.00 (0.84, 1.19)	0.999	-	-
Sleep Problems	1.22 (1.09, 1.37)	<0.001	1.22 (1.10, 1.35)	<0.001

Note: OR = odds ratio; CI = confidence interval. All results were produced using logistic regression modeling. Statistical significance was measured via Wald test.

DISCUSSION

To the best of our knowledge, this is the first study to determine risk factors for new-onset multisite pain in youth who were entirely pain-free at baseline. The development of multisite pain was more common among females than males. Further, three symptom domains were associated with new multisite pain: sleep problems, non-painful somatic complaints, and attention issues. In contrast we did not find that negative affect was associated with new multisite pain. These findings provide insight into the temporal relationship between symptoms commonly associated with pain, but that have been difficult to place in predictive/pathophysiologic pathways. These symptoms may therefore deserve renewed attention as early indicators of pain vulnerability and as potential therapeutic targets in children.

The concept of central nervous system (CNS) sensitization, now referred to as ‘nociplastic’ pain, is thought to play a critical role in the development of the most treatment refractory forms of chronic pain, which are often characterized by multisite or widespread pain.[29] ‘Pain vulnerable’ brain networks may set the stage for the development of pain later in life, and we recently addressed this question directly using a subset of the ABCD sample analyzed here. Using functional neuroimaging data we found increased functional connectivity between brain regions involved in pain processing, including the insula, primary motor/somatosensory cortices and mid-cingulate, at baseline in children who develop multisite pain one year later relative to controls who remain pain-free.[30] Many of these findings also have been noted in adults with chronic nociplastic pain.[29] Previously, it was thought that these types of functional brain outcomes changed as individuals developed chronic pain (i.e., a “state” effect). Instead, the neuroimaging findings combined with the non-painful symptom predictors we present here suggest a more trait-like effect, highlighting a potential underlying vulnerability for the development of multisite pain. Of note, we found a significant effect of sex on the probability of new multisite pain, in which girls were more likely than boys to develop new multisite pain after one year. Previous studies in adolescents demonstrate that sex differences in pain begin to emerge during pubertal development,[31] and eventually chronic multisite/widespread pain conditions such as fibromyalgia disproportionately affect adult women at a ratio of approximately 2:1.[32] The underlying mechanism(s) driving the sex difference in pain remains unknown. Future studies will examine the trajectories of pain and brain development as the youth in the ABCD Study progress through puberty.

Household income was significantly different between multisite pain groups; however, it was not significant in the bivariate or full multivariate models. Previous cross-sectional studies have found that low socioeconomic status is associated with higher pain prevalence in children.[33] In the ABCD Study this association may strengthen with time as a meta-analysis found low socioeconomic status was a risk factor for musculoskeletal pain in children in studies with long-term (but not short-term) follow-up.[34] The mechanism underlying this association is unknown, but one potential explanation is that children in low-income households are more likely to experience adverse events which are linked to a range of poor health outcomes, including chronic pain.[35, 36] Future studies should examine the impact of socioeconomic factors, and the interaction with adverse events, on pain development.

Three baseline symptom domains predicted new multisite pain: total sleep problems, non-painful somatic complaints, and attentional issues. Total and selective sleep deprivation in healthy individuals rapidly produces a cluster of symptoms that includes multisite sensitivity to pain.[37] Epidemiological studies in adults have largely confirmed that sleep difficulties are associated with new-onset widespread pain and/or worsening of painful symptoms,[38, 39] as have studies with new-onset migraine/tension-type headache and chronic musculoskeletal pain in adolescents.[18, 28] These findings have been supported by experimental evidence demonstrating that induced sleep impairment enhances acute pain sensitivity.[37] In fact, a recent preclinical study even suggests that reduced alertness plays an intermediary role for pain development in sleep-deprived mice, a notable finding given the relationship between attention issues and pain discussed below.[40] Of the identified symptom domains associated with new onset pain, sleep may be the most ‘actionable’ target for early interventions in youth because there are well-established screening tools and psychological and behavioral treatments (e.g., cognitive-behavioral therapy; CBT) aimed at improving sleep in youth.[41, 42] In a recent review, Law and colleagues identified future directions for research and clinical practice, including the need for more clinical trials of CBT in youth with chronic pain and the need to implement routine screening of sleep disturbances in youth.[42]

Bothersome somatic symptoms like dry eye, perceived issues with the skin, nausea, and dizziness, which has been termed somatic awareness, are also well-established forerunners of new-onset pain in adults. For instance, in the large, multisite Orofacial Pain Prospective Evaluation and Risk Assessment Study (OPPERA), heightened somatic awareness was the best overall predictor of new-onset orofacial pain out of a large family of psychosocial constructs,[21] and the same relationship has been shown for orofacial pain in adolescents.[19] Increased somatic awareness has also been found to predict new-onset chronic widespread pain in adults.[20] The current results are consistent with the hypothesis that bothersome somatic symptoms (many of which seem to be related to the sensory hyperresponsiveness seen in nociplastic pain) may serve as an early indicator of neurobiological sensitization that will ultimately result in multisite pain.[21] It is important to note, though, that the internal consistency of the CBCL somatic complaints subscale was poor in this sample, likely due to the removal of the three pain items, so this result should be interpreted with caution.

Perhaps the most surprising finding in these analyses was the significant role of attentional issues in predicting new multisite pain. “Cognitive fog,” referring to perceived memory and attentional issues coupled with a state of confusion, is a well-known symptom related to chronic widespread pain.[43] Attentional issues are often conceived of as consequences of chronic pain or attendant issues like sleep problems, but here we find an independent association that does not appear to be the byproduct of other early bothersome symptoms. This is consistent with a previous study in older adults which found that cognitive complaints at baseline were associated with new-onset widespread pain.[39] As attentional issues clearly overlap strongly with those seen in ADHD, previous studies indicating sensory deficits in children with ADHD and their unaffected siblings may indicate shared neurobiological vulnerabilities to ADHD and abnormal sensory processing.[44]

In contrast, we did not find evidence that negative emotions, such as anxiety or depressive symptoms, placed children at an increased risk of developing multisite pain. This is somewhat surprising, as relationships between affective disorders and pain are well-established in studies of chronic pain in adolescents and adults.[45, 46] This also contradicts earlier studies that have identified negative emotionality as a risk factor for new-onset widespread pain in children and adolescents.[11, 12] Another recent study using the ABCD cohort demonstrated that children with higher psychologic and somatic symptoms at baseline had higher odds of new or persistent pain.[47] However, most studies examining the relationship between pain and negative affect have been cross-sectional, and prospective studies have never included pain-free children when predicting outcomes. This is not to suggest that affect plays no role in chronic pain. Affect is a critical dimension of the pain experience, and negative affect increases the impact of chronic pain and may even increase the risk of pain becoming persistent or recurrent.[16, 47] It is possible too that diagnostic levels of affective symptoms, which were not assessed here, may show different relationships with new multisite pain.

There are several strengths in the current analyses, most important being that the children were pain-free at baseline, and the ABCD study is designed to reflect as much as possible the underlying characteristics of the U.S. population. We also used several symptom domains that have been commonly associated with chronic pain in adults and children allowing for a more comprehensive assessment of risk. However, there are also limitations. At this phase of ABCD study, only parents completed the CBCL and the SDSC about their child, which might lead to different results than those obtained by self-report. In future waves of ABCD data collection, children will complete a pain questionnaire, allowing for a direct evaluation of the effect of parental report. We were limited to two time points, which means that while the symptoms observed at 1-year are newly reported, we cannot know if children experienced pain earlier in life that resolved before the baseline assessment, or if after the second time point, pain persisted and became a chronic issue. Furthermore, while the CBCL is a well-validated pediatric survey, it was not developed specifically to measure pain in children. This means our operationalization of multisite pain is limited to available questions and departs from other criteria that may have a clear temporal (chronicity) requirement. The CBCL also does not measure pain severity or interference, and based on our definition of multisite pain, it is possible that children with multisite musculoskeletal pain were excluded from our analyses. Since the CBCL specifies pain “without known medical cause”, it is also possible that children with pain due to medical causes were included in the pain-free group. Future studies with more ABCD study time points and additional pain-specific outcomes measures will be better able to address the natural history of chronic multisite pain. For example, it is unknown if the risk factors identified here are stable across late childhood and adolescence, or if risk factors change across development and are only associated with pain at specific “critical periods”.[22]

Chronic pain has economic costs on par with cardiovascular disease, diabetes, and cancer.[48] Despite this impact on the health care system and an increasing awareness of the relationship between pain in adolescence and opioid misuse in adulthood,[3] efforts to identify simple symptom-based risk factors for the development of pain, particularly in children, have fallen short. This is critically important as multisite pain, once established, is very difficult to treat. For many individuals this results in pain becoming a lifelong burden. The current study represents an important step forward in identifying the prodromal symptoms of new multisite pain in children and may help us to identify vulnerable children before pain develops. In addition to female sex, we have identified three symptom domains that precede the development of new multisite pain in children – sleep problems, somatic complaints, and attentional issues. These domains represent actionable targets for mechanistic investigation, early detection and treatment, and possible prevention of multisite pain.