Predicting persistent low back pain in schoolchildren: A prospective cohort study

Authors

  • Gareth T. Jones,

    Corresponding author
    1. Aberdeen Pain Research Collaboration (Epidemiology Group), University of Aberdeen, Aberdeen, UK
    • Epidemiology Group, Institute of Applied Health Sciences, University of Aberdeen, School of Medicine and Dentistry, Polwarth Building, Foresterhill, Aberdeen, AB25 2ZD, UK
    Search for more papers by this author
  • Gary J. Macfarlane

    1. Aberdeen Pain Research Collaboration (Epidemiology Group), University of Aberdeen, Aberdeen, UK
    Search for more papers by this author

Abstract

Objective

Low back pain (LBP) is common in children but the prognostic indicators are poorly understood. We report the results of a prospective study to determine the risk factors for chronic LBP in children.

Methods

A total of 330 children with LBP were identified from a cross-sectional survey in schools in Northwest England. Data were collected by self-completion questionnaire on a number of potential risk factors for LBP persistence, including lifestyle factors, the occurrence of other symptoms, behavioral and emotional characteristics, and symptom severity and duration. Participants were then followed over the following 4 years to determine persistent symptoms.

Results

Complete followup data were available from 178 children, of whom 46 (26%) reported persistent LBP. Forward stepwise Poisson regression identified 5 independent predictors of pain persistence: peer relationship problems, being of smaller stature, the prior report of widespread body pain, long duration of LBP episodes, and radiating leg pain. Of children with none of these factors at baseline, <5% went on to report persistent LBP. In contrast, of those with 4 or 5 factors, nearly 80% experienced persistent symptoms.

Conclusion

Although childhood consultations for LBP are infrequent, we have shown that ∼25% of children 11–14 years of age with LBP still report symptoms 4 years later. These children can be identified early by a combination of clinical markers and etiologic factors. This provides a basis for considering interventions for secondary prevention; the challenge will be to determine whether we can intervene to alter symptom trajectory at an early stage.

INTRODUCTION

Low back pain (LBP) in children is common: the lifetime prevalence of the condition has been reported to be 40–85% at age 18 years (1, 2), and 20–25% of children will report symptoms in any 1-year period (3–6). Symptoms are more common in girls than boys, and occurrence increases with age (2, 7–10). Despite a number of cross-sectional associations, there is little evidence to suggest that a high level of physical and/or sedentary activity is associated with an increase in the risk of LBP (11, 12). Similarly, although a number of authors have demonstrated cross-sectional associations between LBP and high mechanical load (schoolbag weight) (13, 14), recent prospective studies have shown that, for the majority, this does not confer an increase in risk (12). Rather, evidence is accruing to suggest that emotional, social, and behavioral factors (3, 12, 15) and the occurrence of other somatic symptoms, such as headaches, abdominal pain, and sleep problems (12, 16), have a strong influence on symptom onset.

We have previously reported that among children with LBP, 94% reported some form of disability as assessed using a modified version of the Hannover Functional Ability Questionnaire (7). However, evidence in this area is not consistent; others have reported far lower levels of disability (5), and few children report symptoms sufficiently severe such that they are prevented from attending school or playing sports (4, 9, 13). Childhood consultations for LBP are infrequent (4, 7, 9) and, when a consultation does occur, an organic cause is seldom found (17). At some point, virtually every structure in the lumbar spine has been implicated as a possible source of LBP (18), although objective evidence of “abnormality” does not correlate well with pain reporting (19) and only poorly predicts pain onset (20).

A number of studies have reported that among children with LBP, at least one-third will have recurrent episodes (5, 6, 13), and there is some evidence to suggest that children with LBP are those most likely to experience LBP (19) or other pain (21) in later life. Further, there is evidence from large birth cohort studies to suggest that children with various pain syndromes, in particular abdominal pain and headache, experience an increased risk of abdominal pain, headache, musculoskeletal pain, and a variety of adverse psychological outcomes in later life (22–25). In adults, recent studies have highlighted the importance of psychological factors, physical/mechanical factors, and a number of episode-specific characteristics such as the duration of symptoms, radiating sciatic pain, and physical limitation (26–28). However, there is a paucity of data from studies in childhood, and little is known about the epidemiology of persistent LBP in children.

Although LBP is common in children, the majority of the symptoms are self-limiting and have few clinical consequences. However, for some small proportion, symptoms may be persistent or recurrent, and it is these individuals that may be at highest risk of developing long-term musculoskeletal symptoms in adulthood and other sequelae. Before it is possible to consider an early intervention for such cases with the aim of preventing or limiting adverse long-term health outcomes, it is important to determine whether it is possible at an early stage to identify the children who are most likely to have persistent symptoms. Thus, the aims of the current study were to determine, in a cohort of children reporting LBP, what proportion go on to report chronic symptoms, and to identify modifiable risk factors for LBP persistence.

PARTICIPANTS AND METHODS

A 4-year prospective cohort study of schoolchildren in Northwest England was conducted. Details of the baseline survey have been reported previously (7, 29). In brief, 1,496 children ages 11–14 years from 39 schools (both fee-paying and state-funded) were invited to participate. Participants were given a self-completion questionnaire in the classroom with the investigators present, and were identified as having LBP if they answered positively to both of the following questions: in the past month have you had LBP that lasted for 1 day or longer, and in the past month have you had any pain that has lasted for 1 day or longer in the preshaded area? (Figure 1).

Figure 1.

Preshaded manikin used as part of a self-completed questionnaire to identify lower back pain in children.

Information was also collected on potential risk factors for LBP persistence: lifestyle factors (physical activity, sports participation, sedentary activity) and the occurrence of other symptoms (headache, abdominal pain, widespread body pain, daytime tiredness) (29). Behavioral and emotional characteristics were assessed using the Strengths and Difficulties Questionnaire (30), an instrument designed to assess behavior, emotion, and relationships in 5 domains: prosocial behavior (a strength), hyperactivity, emotional problems, peer problems, and conduct problems (difficulties). Height, weight, body mass index (BMI), and flexibility were measured. For flexibility, the participants stood with feet together and legs straight, bent forward as far as was comfortable, and the fingertip-to-floor distance was then measured.

Among children with LBP, information was collected on symptom duration, severity, and disability using a modified Hannover Functional Ability Questionnaire (7, 31). This instrument assessed limitations in 9 daily activities such as reaching to get a book from a high shelf, sitting up in bed, and standing for 10 minutes. These items were summed and categorized as low (0–1 limitations), moderate (2–3 limitations), or high (4–9 limitations).

Ethical approval for the study was received from the University of Manchester Senate Committee on the Ethics of Research on Human Beings.

Followup.

All children with LBP at baseline were eligible for followup. Participants were given a self-completion questionnaire 1 year and 4 years after baseline. This was done in school where possible, or by postal questionnaire with up to 3 reminders for initial nonresponders. At each time point, LBP was assessed in the same manner as at baseline.

Statistical analysis.

Participants were classified as having persistent LBP if they answered positively to both back pain questions at baseline, the 1 year, and the 4-year time point. Potential risk factors for persistent LBP were assessed using Poisson regression. Thus, results are expressed as risk ratios (RRs) with 95% confidence intervals (95% CIs), the latter being derived using robust estimators of SE (32). All analysis was conducted using Stata, version 10.0 (StataCorp, College Station, TX).

In the first instance, all potential etiologic factors were examined for their univariable predictive ability with respect to persistent LBP, after being adjusted for age and sex. Thereafter, a forward stepwise multivariable Poisson model was constructed to identify independent predictors of outcome. For this, variables were selected for potential inclusion based on biologic and/or statistical criteria. Thus, age and sex were assumed on biologic grounds to be important, and were included in the model. All other variables were selected for potential inclusion if, on age- and sex-adjusted univariable regression, they predicted persistent LBP with a significance of P ≤ 0.2 as assessed using a Wald's test. Variables were entered into the final model at P ≤ 0.10 and removed at P ≥ 0.15. This resulted in an etiologic multivariable model (model 1). This approach in selecting variables for multivariable regression (i.e., selecting candidate variables with a significance of greater than the conventional P < 0.05) has been recommended to ensure the potential inclusion in the final model of all variables of possible importance (33).

Second, factors relating to initial LBP severity/disability were examined. Using the same eligibility, inclusion, and exclusion criteria as above, these clinical variables were offered to model 1. With the exception of age and sex, etiologic variables already in the model were allowed to leave the model if they met the appropriate criterion (P ≥ 0.15). This procedure resulted in a complete multivariable model (model 2).

Third, for all potentially modifiable risk factors identified, the population attributable risk associated with these factors was calculated to estimate the potential beneficial effect of intervening on these factors in the future.

Finally, older children may have been more likely to have left school between baseline and the 4-year followup and, thus, might have been difficult to trace and/or been reluctant to participate further in the study, potentially leading to a low participation rate and/or a differential followup between younger and older children. Therefore, to assess external validity we examined any differences between followup participants and nonparticipants with respect to baseline characteristics and constructed a third multivariable model (model 3), weighting the analysis based on these differences.

RESULTS

At baseline, 1,376 children (92%) participated and answered both questions on LBP. LBP was reported by 330 (24%); the prevalence increased with age (χmath image = 20.9; P < 0.001) and was higher in girls than in boys (χ2 = 14.1; P < 0.001). Data at both followup points were available from 178 children (54%), 46 (26%) of whom reported persistent LBP. There was no difference with age and, although girls were 50% more likely to report persistent LBP, this was not statistically significant (RR 1.5, 95% CI 0.8–2.6).

Etiologic predictors.

Children shorter than the median height (158 cm) experienced a doubling in the risk of persistent LBP (RR 2.1, 95% CI 1.2–3.8). However, there was no association between persistent symptoms and body weight (RR 1.1, 95% CI 0.6–1.9), BMI (RR 0.8, 95% CI 0.4–1.3), or flexibility (RR 1.2, 95% CI 0.7–2.0). Similarly, there was no evidence of an increase in risk with the majority of the lifestyle factors. However, there was some evidence to suggest that compared with those who walked to school, children who traveled by other means (predominantly by car) experienced a 70% increase in the risk of persistent LBP (RR 1.7, 95% CI 0.8–3.3). However, this was not statistically significant.

Children who reported high levels of peer problems experienced more than a doubling in risk of persistent LBP (RR 2.3, 95% CI 1.3–4.2). However, those with a high level of psychosocial difficulties overall were no more likely to report persistent symptoms than other children (Table 1). Neither was there a significant association with prosocial behavior.

Table 1. Risk of persistent LBP associated with behavioral and emotional factors*
ExposurePersistent LBPRR (95% CI)
No. (%) yesNo. no
  • *

    LBP = low back pain; RR = risk ratio; 95% CI = 95% confidence interval.

  • Split at the median value.

  • Adjusted for age and sex.

  • §

    Combined score of hyperactivity, emotional problems, peer problems, and conduct problems.

Peer relationship problems   
 Low11 (15.1)621.0
 High34 (33.7)672.3 (1.3–4.2)
Emotional problems   
 Low16 (19.8)651.0
 High30 (31.3)661.5 (0.9–2.6)
Conduct problems   
 Low22 (25.3)651.0
 High24 (26.4)671.1 (0.7–1.8)
Hyperactivity   
 Low26 (25.7)751.0
 High20 (26.0)571.0 (0.6–1.6)
Total difficulties§   
 Low16 (23.9)511.0
 High29 (27.4)771.1 (0.7–1.9)
Prosocial behavior   
 Low18 (20.7)691.0
 High28 (30.8)631.5 (0.9–2.4)

Children who reported widespread pain experienced a 50% increase in the risk of persistent LBP, of borderline significance (RR 1.5, 95% CI 0.9–2.5). There was no significant association with headaches, abdominal pain, or daytime tiredness (Table 2).

Table 2. Risk of persistent LBP associated with other somatic symptoms*
ExposurePersistent LBPRR (95% CI)
No. (%) yesNo. no
  • *

    See Table 1 for definitions.

  • Adjusted for age and sex.

  • 0 = not at all tired and 10 = extreme tiredness, split at the median value.

Widespread pain in the past month   
 No26 (21.8)931.0
 Yes20 (33.9)391.5 (0.9–2.5)
Headaches in the past month   
 None6 (16.2)311.0
 1–7 days32 (30.8)721.8 (0.8–4.1)
 >7 days8 (22.2)281.3 (0.5–3.3)
Abdominal pain in the past month   
 None10 (19.6)411.0
 1–7 days26 (26.5)721.3 (0.6–2.6)
 >7 days10 (34.5)191.5 (0.7–3.6)
Daytime tiredness (0–10)   
 0–416 (25.0)481.0
 5–1028 (25.7)811.1 (0.7–1.9)

Multivariable model: etiologic model.

Three factors met the criteria for inclusion in the multivariable model: peer problems, smaller stature, and widespread pain. As these factors were added to the model sequentially, all variables already in the model were assessed and none met the exclusion criteria. Thus, these 3 factors emerged as statistically independent predictors of persistent LBP, each of which was associated with a 50–150% increase in the risk of persistent symptoms (Table 3, model 1).

Table 3. Multivariable models*
ExposureAge- and sex-adjusted RR (95% CI)
Model 1Model 2Model 3§
  • *

    See Table 1 for definitions.

  • Original etiologic model.

  • Complete model: model 1 plus aspects of initial LBP severity/disability.

  • §

    Weighted model: model 2 weighted by distribution of age and peer problems in target population.

Peer relationship problems   
 Low1.01.01.0
 High2.4 (1.3–4.2)2.2 (1.2–4.0)2.3 (1.2–4.2)
Baseline height (above/below median)   
 120–158 cm1.01.01.0
 159–190 cm2.2 (1.2–4.0)2.2 (1.2–3.8)2.1 (1.1–3.7)
Widespread pain in the past month   
 No1.01.01.0
 Yes1.5 (0.9–2.4)1.5 (0.9–2.4)1.4 (0.9–2.3)
When did your LBP start?   
 <1 month ago1.01.0
 1–3 months ago0.7 (0.3–1.4)0.6 (0.3–1.3)
 3–12 months ago1.0 (0.5–2.0)0.9 (0.5–1.9)
 >12 months ago2.0 (1.1–3.5)1.9 (1.05–3.5)
Radiating leg pain   
 No1.01.0
 Yes1.6 (0.99–2.5)1.6 (1.02–2.7)

LBP severity/disability.

Children who reported that their LBP first occurred >12 months previously and those who reported that it usually lasted for >7 days experienced more than a doubling in risk of persistent symptoms (RR 2.4, 95% CI 1.3–4.4 and RR 2.6, 95% CI 1.4–4.9, respectively). Similarly, those with LBP on the day of the baseline survey, those with severe symptoms (top tertile of a 100-mm visual analog scale), and those with radiating leg pain experienced an increase in the risk of persistent symptoms (Table 4).

Table 4. Risk of persistent LBP associated with initial LBP severity and disability*
ExposurePersistent LBPRR (95% CI)
No. (%) yesNo. no
  • *

    See Table 1 for definitions.

  • Adjusted for age and sex.

  • At the worst during the past month, where 0 = no pain at all and 10 = worst pain I can imagine, split into tertiles.

When did your LBP start?   
 <1 month ago13 (22.4)451.0
 1–3 months ago9 (17.0)440.7 (0.3–1.5)
 3–12 months ago9 (23.7)291.1 (0.5–2.3)
 >12 months ago14 (51.9)132.4 (1.3–4.4)
How long does your LBP usually last?   
 <24 hours26 (20.6)1001.0
 1–7 days14 (34.1)271.6 (0.9–2.8)
 >7 days6 (54.5)52.6 (1.4–4.9)
Do you have LBP today?   
 No24 (22.9)811.0
 Yes21 (30.9)471.4 (0.9–2.3)
LBP severity (0–10)   
 0–312 (21.4)441.0
 4–616 (21.1)601.0 (0.5–2.0)
 7–1017 (37.8)281.8 (0.96–3.3)
Radiating leg pain   
 No24 (19.0)1021.0
 Yes20 (42.6)272.2 (1.4–3.6)
Modified Hannover Functional Ability Questionnaire   
 0–1 limitations2 (8.7)211.0
 2–3 limitations13 (28.9)323.5 (0.9–14.0)
 4–9 limitations25 (35.7)454.1 (1.05–16.2)

At baseline, 91% of children reported some limitation to daily activities. Compared with children reporting a low level (0–1 limitations), those reporting moderate (2–3 limitations) or high (4–9 limitations) levels experienced a >3-fold increase in the risk of persistent LBP (RR 3.5, 95% CI 0.9–14.0 and RR 4.1, 95% CI 1.05–16.2, respectively). On further examination, it was revealed that only some limitations were associated with an increased risk. Specifically, children who reported difficulty carrying their schoolbag and those who reported difficulty with standing in a line for 10 minutes experienced an approximately 2–3-fold increase in risk (RR 2.1, 95% CI 1.1–4.0 and RR 2.7, 95% CI 1.5–4.9, respectively). Other limitations (reaching to a high shelf, sitting up in bed, and bending down to put on socks) did not predict persistent LBP.

Multivariable model: complete model.

Two factors relating to LBP severity/disability emerged as independent predictors of outcome: long duration of LBP episodes and radiating leg pain. All etiologic variables remained in the model. Thus, 5 factors were identified as independent predictors of persistent LBP (Table 3, model 2).

To examine the performance of the complete model (model 2), a count was made for each child of how many of the 5 predictors they had at baseline (for this, LBP duration was dichotomized to <1 year versus >1 year), and the frequency of persistent LBP was examined with respect to this new summary variable. The proportions of children reporting 0, 1, 2, 3, 4, and 5 of these factors were approximately 16%, 27%, 32%, 16%, 8%, and 0.5%, respectively. Of children with none of these factors, <5% went on to report persistent LBP. In contrast, nearly 80% of those with 4–5 factors experienced persistent LBP (Figure 2).

Figure 2.

Performance of the multivariable model. LBP = low back pain.

Of the variables in the model, only peer problems were potentially modifiable. It is possible to calculate the risk difference between exposed/nonexposed groups as a proportion of the risk in the exposed (34) and, from this, to calculate the population attributable risk. We calculated that the population attributable risk associated with peer problems was ∼41%.

External validity.

There was no difference in followup between boys and girls (RRgirls 1.0, 95% CI 0.8–1.2). However, older children were less likely to participate (RR13–14 years/11–12 years 0.8, 95% CI 0.6–0.9). After adjusting for age, no important predictors of outcome were associated with followup participation except peer relationship problems; children who reported peer problems were 20% more likely to participate at followup than children who did not (RR 1.2, 95% CI 0.99–1.5). However, a subsequent analysis weighting the results back to the distribution of age and peer problems in the original sample produced risk estimates similar to those of the main analysis (Table 3, model 3).

DISCUSSION

It is widely believed that LBP in schoolchildren is common, and largely self-limiting. We have demonstrated, however, that 25% of children ages 11–14 years with LBP still report symptoms 4 years subsequently. Further, we have identified 5 factors that are highly discriminatory in terms of the risk of persistent symptoms. Long duration of LBP episodes, radiating leg pain, and widespread pain independently predict symptom persistence. In addition, peer problems and being of smaller stature are also important independent predictors.

There are several methodologic issues to consider. First, although baseline participation was high, only 54% of children were captured at both followup time points. High loss to followup would decrease the sample size and increase the likelihood of a null result, although this would not necessarily alter the actual risk estimates. More important is the issue of differential followup, in which participants/nonparticipants differ with respect to key variables. However, there were few differences between those who were successfully followed and those who were not, with the exception of age and peer problems: nonparticipants were older and reported fewer problems. We believe, however, that this has not introduced a bias; age was not a predictor of persistent LBP, and an analysis weighting the results back to the baseline distribution of both of these variables produced risk estimates only marginally different from the main analysis and did not alter the conclusions.

Second, information on LBP was collected from 3 surveys. Therefore, it is not possible to distinguish between persistent and recurrent LBP. It may be that cases we have identified with reportedly persistent symptoms actually had repeated episodes of pain, each of different etiology, e.g., multiple injuries. Others have observed in longitudinal studies that when asked repeatedly about LBP, children are actually very bad at recalling their answers from the previous year (9). This might suggest that pain reported on multiple occasions are multiple, discrete episodes of pain rather than one chronic condition. However, if this were the case in the current study, one might expect to see a relationship between LBP and physical activity or sports participation. No such relationships were observed, although from the current data, repeated discrete episodes of LBP cannot be ruled out.

To our knowledge, this is the first large-scale epidemiologic study to determine among children in the population with LBP those who are likely to have persistent symptoms. We have previously reported the occurrence of LBP in a wider sample, from which the current cohort was derived (7). Combining this data, we estimate that ∼6% of all children report persistent LBP over a 4-year period. In contrast to the longitudinal approach of the current study, others who have attempted to capture chronic LBP have conducted cross-sectional or case–control studies and have used retrospective measurements of exposures. This can lead to spurious relationships caused by differential recall between children with and without chronic LBP. This not-with-standing, there are a number of parallels between our results and the findings of other authors. Salminen et al (5) reported “recurrent or continuous” LBP in 7.8% of 14-year-old Finnish schoolchildren (5), and 35% still reported symptoms at age 23 years (35); whereas Mirovsky et al (36) found that 62% of children ages <16 years with nonspecific LBP in a hospital spine unit still reported pain 8 years later (36).

A number of authors have shown that LBP in childhood is mild, self-limiting, rarely sufficient to warrant nonattendance at school or sports (4, 6, 13), and that consultations for LBP are uncommon (4, 7, 9). However, our current results suggest that some of the strongest predictors of persistence are measures of symptom severity and duration. These results are similar to those from recent adult studies that found that episode-specific factors, such as duration of symptoms, were associated with poor prognosis (26, 27, 37). It is intuitive that symptoms of longer initial duration are more likely to be persistent. However, we have demonstrated that pain severity, in particular radiating leg pain, predicts persistence independently of pain duration. The 95% CIs associated with moderate/high levels of disability are wide (95% CI 0.9–14.0 and 95% CI 1.05–16.0, respectively), which hinders quick interpretation for the clinician. However, this is largely a function of the small number of cases in the reference category; only 2 of the 23 children with 0–1 limitations at baseline went on to develop persistent symptoms. In contrast, children with moderate (2–3 limitations) or severe (4–9 limitations) symptoms at baseline experienced an approximately 3.5–4-fold increase in the risk of persistent LBP. Further, on a practical note, the difference in risk associated with different elements of the modified Hannover Functional Ability Questionnaire would suggest that it may be possible to capture the predictive effect of the instrument in the clinic using fewer individual items.

Children with widespread pain in the month prior to baseline also experienced an increase in the risk of persistent LBP. Others have reported similar findings: Pellisé et al demonstrated that only 42.6% of adolescents with LBP had this symptom in isolation, and 8.5% also reported whole-body pain (38). This suggests that, in some children at least, LBP may be one of a cluster of general musculoskeletal symptoms. This is of particular note, in light of recent evidence that children who report common symptoms experience a range of adverse physical and mental health outcomes in later life (22, 23, 25).

In girls, persistent LBP was unrelated to menstruation (data not shown), and no pubertal data were available for boys. It would have been interesting to have more extensive data on pubertal development in girls and boys, particularly in light of the finding relating to height: shorter children were significantly more likely to develop persistent symptoms. Others have reported similar findings; Mirovsky et al (36) found that of children with LBP, those pain-free 8 years later were taller at their first assessment (36), and Fairbank et al (1) reported an increased trunk length in children with back pain compared with those without (1). Possible mechanisms for this association are unclear, but may be associated with posture, mechanical disadvantage (torque) during bending and twisting, or disc pathology. Alternatively, it may be that shorter children report more psychosocial problems than their peers. Although this is possible, the independent contributions of (smaller) stature and psychosocial problems to the multivariable model would suggest that this is not the explanation (at least, not the full explanation) for the observed association with height. Also, an additional analysis (data not shown) revealed no significant interactions between height and any of the other variables in the final model.

We found few associations between lifestyle and persistent LBP. Despite common beliefs to the contrary, there is increasing evidence that factors such as physical exercise play a little role in LBP onset (11, 12, 39, 40). We have now demonstrated, prospectively, these factors are also unimportant in terms of symptom persistence. However, consistent with studies in adult populations (27), adverse psychosocial exposures were associated with a significant increased risk of persistent LBP. We dichotomized the Strengths and Difficulties Questionnaire peer problems scale at the median value. Therefore, some children categorized as having high levels of problems may actually have been experiencing relatively minor levels of psychosocial difficulties. Should this be the case, such misclassification will have biased our findings toward the null and therefore the strong association in the current study is even more remarkable. We have identified among children with LBP 5 independent factors that are able to discriminate between those with a <5% risk of persistent symptoms (among those with none of these factors) and those with ∼80% risk (among those with 4 or 5 factors). In practical terms, it is difficult to identify comprehensive advice to the health care worker confronted with a child with LBP, being that arguably only peer problems are amenable to intervention. However, the potential impact of being able to intervene on peer problems is considerable: 41% of cases of persistent LBP in the (child) population are due to this exposure. However, such a modification in daily life may be very difficult to achieve.

In summary, LBP is common in childhood, and a sizeable minority have persistent symptoms. These children can be identified early by a combination of clinical markers and etiologic factors. Recent evidence suggests that children with pain are more likely to become adults with chronic pain and/or other symptoms. Thus, the ability to identify children who are at high risk of persistent symptoms provides a basis for considering interventions for secondary prevention; the challenge will be to determine whether we can intervene to alter symptom trajectory at an early stage.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Jones had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Jones, Macfarlane.

Acquisition of data. Jones.

Analysis and interpretation of data. Jones, Macfarlane.

Acknowledgements

The authors thank the 39 schools and the children who participated in the study. In addition, we are grateful to Professors Alan Silman, Ann Charlton, and Deborah Symmons for advice on study conduct, and to Kath Watson, Stewart Taylor, Ann Papageorgiou, Priscilla Appelbe, Elizabeth Nahit, Isabelle Hunt, and Stephen Conn for assistance with data collection.

Ancillary