BACKGROUND: Slipped capital femoral epiphysis (SCFE) is believed to be associated with childhood obesity, although the strength of the association is unknown.
METHODS: We performed a cohort study using routine data from health screening examinations at primary school entry (5–6 years old) in Scotland, linked to a nationwide hospital admissions database. A subgroup had a further screening examination at primary school exit (11–12 years old).
RESULTS: BMI was available for 597 017 children at 5 to 6 years old in school and 39 468 at 11 to 12 years old. There were 4.26 million child-years at risk for SCFE. Among children with obesity at 5 to 6 years old, 75% remained obese at 11 to 12 years old. There was a strong biological gradient between childhood BMI at 5 to 6 years old and SCFE, with the risk of disease increasing by a factor of 1.7 (95% confidence interval [CI] 1.5–1.9) for each integer increase in BMI z score. The risk of SCFE was almost negligible among children with the lowest BMI. Those with severe obesity at 5 to 6 years old had 5.9 times greater risk of SCFE (95% CI 3.9–9.0) compared with those with a normal BMI; those with severe obesity at 11 to 12 years had 17.0 times the risk of SCFE (95% CI 5.9–49.0).
CONCLUSIONS: High childhood BMI is strongly associated with SCFE. The magnitude of the association, temporal relationship, and dose response added to the plausible mechanism offer the strongest evidence available to support a causal association.
- CDC —
- Centers for Disease Control and Prevention
- CHI —
- Community Health Index
- CHSP-P1 —
- Child Health Systems Programme-Primary Year 1
- CHSP-P7 —
- Child Health Systems Programme-Primary Year 7
- CI —
- confidence interval
- ICD-10 —
- International Classification of Diseases, 10th Revision
- ISD —
- Information Services Division
- NHS —
- National Health Service
- SCFE —
- slipped capital femoral epiphysis
- SMR01 —
- Scottish Morbidity Record
- UK90 —
- United Kingdom 1990 reference population
What’s Known on This Subject:
An association between slipped capital femoral epiphysis and childhood obesity has long been suggested, although there have been no robust attempts to explore this association. The current evidence is almost exclusively based on small, low-quality case series from specialist centers.
What This Study Adds:
Using routinely collected BMI from ∼600 000 children and 4.25 million child-years of follow-up, we provide robust evidence to support a causal association between obesity and slipped capital femoral epiphysis: a strong association, temporal relationship, and marked dose response.
Childhood obesity is a global problem and a major cause of lifelong morbidity.1–3 In a report on ending childhood obesity, the World Health Organization highlighted a lack of awareness of the consequences of childhood obesity.4 Long-term outcomes of childhood obesity are well described5–8; however, there is poor understanding of short-term outcomes that may cause early childhood disability. Slipped capital femoral epiphysis (SCFE) is a disease of the growth plate (physis) that causes profound lifelong disability and is believed to be caused by obesity.9–14 Although an association between SCFE and childhood obesity has been suggested,14 it has not yet been definitively demonstrated. Clarifying the relationship between SCFE and childhood obesity has been identified as a priority for the American Academy of Orthopedic Surgeons.15
SCFE alters the shape of the hip, resulting in bone impingement, and it is 1 of the most common reasons for hip replacement surgery in adolescence and early adulthood.16 It affects 1 in 1300 individuals during childhood,17 typically requires urgent surgery, and often results in deformity. Early detection and surgery can minimize the severity of deformity, although the disease frequently goes undetected for many months, often because the pain poorly localizes to the thigh or knee, creating confusion for children, parents, and clinicians alike.17,18 Diagnostic delays worsen clinical outcomes and can have significant medicolegal consequences for a range of clinicians.18
The current evidence for an association between SCFE and childhood obesity arises from observations of increasing SCFE incidence rates coupled with rising childhood obesity10,11 and retrospective case series from specialist centers.12–14,19 We sought to define the strength of this association using a nationwide population cohort study to determine whether there is evidence for a causal relationship between obesity and SCFE.
Study Design, Setting, and Population
This was a historic cohort study in which we linked health care data sets within Scotland. The cohort was formed from 2 sources of routine universal childhood height and weight measurements at primary school entry (5–6 years old).
Cohort 1 comprises the Study of Trends in Obesity in North East Scotland (STONES), which was collected from the Grampian region of Scotland and represents ∼10% of the Scottish population. Information was collected from 1970 onward. The Scottish Community Health Index (CHI) number, which is a unique identity number among all Scottish residents, was collected for children born after 1992. Before 1992, children were matched to other data sets on the basis of initials, sex, and date of birth. This population has been described previously.20
Cohort 2 comprises the Child Health Systems Programme-Primary 1 (CHSP-P1), which is a nationwide child health surveillance program. CHSP-P1 began in 1995 and encompassed all of Scotland by 2003. CHI numbers were collected throughout.
Both cohorts were linked to the Scottish Morbidity Record (SMR01) up to December 2016. The SMR01 is an episode-based record relating to inpatients and day case patients discharged from all Scottish hospitals. The SMR01 was computerized in 1968 and has been used for the financial management of hospitals since 1989. Ongoing data quality assessment through periodic random sampling by National Health Service (NHS) Scotland reveals high data quality with an accuracy of 89.0% (95% confidence interval [CI] 87.9%–90.1%) for the main conditions coded within the SMR01.21
Children’s height and weight were routinely recorded at school entry (5–6 years old) to monitor obesity trends across Scotland. Measurements were collected by school nurses, although no information was available on measurement equipment. Other variables were sex, exact age of entry into the cohort, year of cohort entry, and (after 2001) an area-based quintile measure of socioeconomic status (2001 Carstairs score).
For a subgroup of children, their height and weight were additionally recorded at exit from primary school (aged 11–12 years) in a linked child health surveillance program (Child Health Systems Programme-Primary 7 [CHSP-P7]).
Within the SMR01, we sought an electronic diagnostic record representing SCFE (International Classification of Diseases, 10th Revision [ICD-10] M93.0 [slipped upper-femoral epiphysis] or International Classification of Diseases, Ninth Revision 732.2 [unspecified slipped upper-femoral epiphysis]). Cases were restricted to codes recorded at age >5 years and <18 years (ie, the period “at risk for SCFE”). Previous work in England in which researchers used the SCFE ICD-10 code in linked databases has revealed it to be specific for the identification of SCFE.17 The first date of diagnostic code entry within the medical record was considered to be the index date. An individual could only contribute 1 SCFE diagnosis because laterality was not coded, so it would have been unclear whether additional SCFE diagnosis codes truly represented a contralateral event or a secondary admission related to the initial diagnosis (eg, removal of metalwork).
Children contributed to follow-up until they (1) reached 18 years old, (2) received a diagnosis of SCFE, or (3) were censored in December 2016, when data were extracted.
BMI was calculated and expressed as a z score of the United Kingdom 1990 reference population (UK90) adjusted for age and sex.22 The transformation of BMI data to z scores was performed by using the LMS method and the zanthro package within Stata 14.1 (Stata Corp, College Station, TX).23 z scores are a measure of how many SDs a score varies from the mean. There is debate around the clinical cutoff definitions for obesity in children; however, to aid in interpretation, BMI was categorized according to cutoffs recognized by both UK90 and the Centers for Disease Control and Prevention (CDC)24–26 (underweight, below fifth percentile; normal weight, 5th–85th percentile; overweight, ≥85th percentile; and obese, ≥95th percentile). We further stratified obesity as mild or moderate obesity (≥95th–99th percentile) and severe (morbid) obesity (≥99th percentile). z scores were converted to the clinical cutoff BMIs to improve clinical relevance and interpretation (ie, ≥95th percentile equates to a z score ≥1.645).
During data cleaning, any height or weight recorded as “0” was replaced with “missing.” All data were explored graphically, during which we initially identified a decimal error in 1 year of source data for height and/or weight, which was addressed. A height or weight outside ±5 SDs was excluded because these were likely to be spurious (height, n = 311; weight, n = 1159), which has been the approach previously used in the interpretation of these data sets.20
The analysis was conducted by using Stata 14.1. The incidence of SCFE was calculated and stratified according to BMI at primary school entry and exit (when available). Poisson CIs were calculated for rate estimations.
A Cox proportional hazards regression model was fitted to estimate the SCFE hazard by using the covariates of age at cohort entry, sex, quintile of socioeconomic deprivation (eg, from most affluent [first quintile] to least affluent [fifth quintile]), and BMI z score. The relationship between Schoenfeld residuals and event time was examined to formally test the proportional hazards assumption. Deprivation was considered within the analysis because of the known association between deprivation and obesity.20 The measure of area deprivation used was the 2001 Carstairs score expressed as quintiles27 and fitted as a categorical variable. Carstairs is an area-based measure of material deprivation routinely used by the Scottish Government that includes measures of unemployment, car ownership, overcrowding, and social class. Scores are assigned to postal code sectors, with the mean population in each postal code sector being 5012 individuals. Quintile 1 represents the most affluent, and quintile 5 represents the least affluent.
The cumulative age to SCFE diagnosis was examined by separating data into the ≥85th percentile (ie, children with overweight and obesity) and <85th percentile (ie, underweight or normal). The categories were compared by using log-rank tests for equality of survivor functions.
The study protocol, data request application, and Stata code are available in the Supplemental Information. Reporting is in line with the Reporting of Studies Conducted Using Observational Routinely Collected Data Statement. Raw data are not available to be shared owing to the terms of the data-sharing agreement. The threshold for statistical significance was P < .05.
The cohort included routine health records of 615 950 5- to 6-year-old children at school entry. BMI could be calculated for 597 017 (97%) children, of whom 11.9% were overweight and 9.2% were obese. The mean age at cohort entry was 66.2 months (5 years and 6 months; interquartile range 63.0–72.0 months). Total follow-up among children for whom BMI was known in the SCFE risk period (6–18 years old) was 4.26 million child-years. The mean follow-up time was 7 years and 1 month.
A screening examination at exit from primary school (11–12 years old) was available for 39 468 of the children from the initial cohort. BMI was available for 38 458 children at both school entry and exit. BMI was broadly consistent at both time points (Table 1). Of the 3973 children who were obese (≥95th percentile) at 5 to 6 years old, 2963 (75%) remained obese at 11 to 12 years old. Among those who were overweight (n = 5086) at 5 to 6 years old, 39% were obese at 11 to 12 years old. This was in contrast to those who were underweight (below the fifth percentile) at 5 to 6 years old (n = 973), of whom 2% were obese at 11 to 12 years old.
During the follow-up period, 209 children received a diagnosis of SCFE. BMI was available for 195 of these children at 5 to 6 years old and was also available for 32 children at 11 to 12 years old. One case was excluded because weight was >5 SDs above the population mean and was likely spurious. SCFE diagnoses were recorded in 117 boys and 92 girls (5:4 male/female ratio).
The crude incidence of SCFE was 4.7 per 100 000 6 to 18 child-years of risk (95% CI 4.1–5.4), although the crude incidence rate is an underestimate because the cohort had an uneven distribution of follow-up. The cohort expanded in recent years, which resulted in disproportionately greater numbers of children with shorter follow-up. The age of SCFE onset is known to be nonuniform across childhood, with the peak age at diagnosis being among 11-year-olds (incidence of 13.4 per 100 000 child-years; 95% CI 10.0–17.7; Supplemental Table 4). The incidence when adjusted to the age structure of the European Standard Population was 5.45 per 100 000 6 to 18 child-years of risk (95% CI 4.8–6.1).28
There was a strong association between BMI at 5 to 6 years old and SCFE. The incidence of SCFE increased by a factor of 1.7 (95% CI 1.5–1.9; P < .001) for each integer increase in BMI z score (Fig 1). Of the children who developed SCFE, 59 (30%) were obese at 5 to 6 years old and 25 (13%) were overweight. The incidence rate ratio for developing SCFE compared with children of normal weight at 5 to 6 years old was 5.9 (95% CI 3.9–9.0) among those with severe obesity, 3.8 (95% CI 2.6–5.8) among those with mild or moderate obesity, and 1.5 (95% CI 0.9–2.3) among those with overweight (Table 2). Assuming that this association between SCFE and obesity was causal, the proportion of SCFE cases that would be eliminated if the BMI of the entire population were to fit within the BMI range of the fifth to 85th percentile, defined as normal (ie, the attributable risk or excess risk, which is the difference in disease rates between an exposed population and an unexposed population), is 78% among children with obesity and 31% among overweight children.
Among the smaller subpopulation of children for whom BMI was available at exit from primary school (aged 11–12 years), there were 32 children with SCFE. The magnitude of the association at 11 to 12 years old was even stronger (Table 2). The incidence of SCFE in the 11- to 12-year-old children with obesity was 23.8 per 100 000 6 to 18 child-years of risk (95% CI 14.8–36.4) compared with 1.9 per 100 000 6 to 18 child-years (95% CI 0.6–4.5) among those of normal weight (risk ratio 12.3; 95% CI 4.6–32.6). The risk in those with severe obesity was greatest, with the risk of SCFE being 17.0 times greater (95% CI 5.9–49.0) in this group compared with those with normal BMI. Of these children with SCFE, 26 (∼80%) were overweight or obese compared with 35.8% of whole population of 11- to 12-year-olds.
The age at SCFE diagnosis was significantly lower among those with overweight or obesity. Children with overweight and obesity were diagnosed 1 year earlier than children with normal weight or underweight (P < .002; Fig 2). The age at disease onset decreased by 3.3 months (95% CI 0.5–6.0; P = .02) with each integer increase of BMI z score.
A Carstairs score was available for 495 954 children (of whom 130 were affected by SCFE), and the incidence of SCFE was lowest in the most affluent quintile (Supplemental Table 5). Those in the 3 most deprived quintiles had a similar risk of SCFE.
In the Cox proportional hazards model, we used the covariates of age at cohort entry, sex, quintile of socioeconomic deprivation, and BMI z score. Only the BMI z score and deprivation score contributed significantly (Table 3).
We identify a strong association between childhood obesity and SCFE, with increasing childhood BMI both increasing the risk and reducing the age at disease onset. Obesity was recorded before any child was affected, which reveals that the association was temporal. Even children of normal weight are at risk for SCFE, although notably less so than those who are overweight or obese. Children with the lowest BMI at 5 to 6 years old had an almost negligible lifetime risk of SCFE, those with a normal BMI had an approximate risk of 1:2500, those with overweight had an approximate risk of 1:1750, those with mild or moderate obesity had a risk of 1:650, and those with severe obesity had a lifetime risk of 1:450. Although there were less data available for children at 11 to 12 years old, obesity at this age had the strongest association with SCFE, with the lifetime risk among those with severe obesity being 17.0 times greater (95% CI 5.9–49.0) than in those with a normal BMI, equating to a lifetime risk of ∼1:250. With this study, we also support longitudinal studies in which researchers have suggested that obesity at primary school entry (kindergarten) is intimately associated with obesity later in childhood.29
Mechanical studies have revealed that childhood obesity may generate forces sufficient to overcome the yield point of the physis.30 The peak age of SCFE is around puberty, and rapid growth of the bone is believed to lower the mechanical yield point for physeal injury. It appears that obesity around puberty, rather than earlier in childhood, is the most important time point in the development of the disease. SCFE histologically occurs through the zone of hypertrophy, which is the location at which the supporting matrix of the physis is particularly redundant.31 Therefore, there is biological plausibility through a mechanical disease mechanism for obesity causing SCFE.
Previous case series from specialist centers have revealed an association between SCFE and obesity,12,13,19 although these studies suffered from referral bias and poor generalizability to the wider population. Furthermore, the temporal relationship between the disease and obesity has been difficult to establish (ie, did children become obese because of hip disease, or did hip disease develop because of obesity). The only previous cohort study of SCFE included a health care cohort from family medicine,17 and revealed that predisease BMI was 1.43 SDs above the mean (95% CI 1.20–1.68). However, the researchers in this study did not standardize the timing of BMI measurement, were unable to determine a dose response, had no controls, used a health care population, and were prone to bias because BMI was more likely to be recovered among individuals who were not healthy.
The age and sex distribution of SCFE in this study was consistent with that of previous studies,19,32 and incidence rates were comparable to those identified in England and Wales (incidence 4.8; 95% CI 4.4–5.2 cases per 100 000 0–16-year-olds)17 and in Scotland.10
The relationship with socioeconomic deprivation has previously been proposed as a risk factor for SCFE,17 although worsening deprivation and increasing childhood obesity are known to be intrinsically linked in the United Kingdom.33 Even after adjusting for obesity, socioeconomic deprivation remained an independent risk factor. However, the relationship between the 2 is so intertwined that they may be difficult to adequately separate, particularly by using area-based measures of deprivation, which may introduce an ecological fallacy.
This study has many strengths compared with previous attempts to understand the association between obesity and SCFE, although there are still limitations. We were unable to quantify the effects of ethnicity because this was poorly recorded within the data set. However, the population of Scotland in the 2011 census was 96.0% white,34 so unless ethnicity exerted an overwhelming effect, its availability would be unlikely to help discern a small difference in disease vulnerability. No adjustment was made for comorbid disease associations; however, in a previous cohort identified from health records, researchers failed to find any strong evidence for an association with other childhood diseases.17 We did not account for children dying or leaving Scotland, although this is unlikely to introduce bias because this is a nondirectional effect related to obesity. A small number of children for whom measurements were >5 SDs from the mean were excluded to remove spurious data, but some genuinely extreme BMI values may have been falsely excluded. We cannot be certain regarding the exact sensitivity and specificity of the diagnostic codes used, although previous work has revealed that they are reliable.17
The reduced age at onset in children with obesity is a novel finding. It is conceivable that diagnoses may be more readily made, and therefore made sooner, in children with obesity owing to clinicians’ heightened awareness of disease in this group. However, SCFE generally causes marked pain or a limp and is diagnosed on the basis of clear radiographic findings. Therefore, it is unlikely that children with obesity are overdiagnosed or that children without obesity are underdiagnosed. Two biological explanations are that obesity may lower the age of puberty and advance skeletal maturation, which may therefore account for the earlier SCFE age at onset among children with obesity,35 and a greater mechanical load may trigger earlier physeal failure in children with obesity.
A confounding relationship is an alternative explanation for the observed effect between obesity and SCFE (ie, a factor that is independently associated with both obesity and physeal failure). Abnormalities in serum leptin have been suggested as a possible independent risk factor for SCFE, with the suggestion being that this may be a confounder.36 However, the positive association between leptin and obesity more likely suggests that leptin is a disease mediator or simply a proxy measure of obesity exposure.37
We demonstrate that childhood obesity is a major risk factor for the development of SCFE. The temporal relationship, dose response, and magnitude of the association build on the existing biological plausibility and findings in previous lower-quality studies to offer the strongest possible support for a causal relationship between childhood obesity and SCFE.
We thank Emma Morely of STEPS Charity Worldwide (www.steps-charity.org.uk), the patient charity that helped direct the research agenda and will assist in the dissemination of results. We also thank the Information Services Division (ISD) of NHS Scotland for the provision of data from ISD Scotland, particularly Andrew Duffy, the research coordinator within National Services Scotland.
- Accepted August 6, 2018.
- Address correspondence to Daniel C. Perry, PhD, Institute of Translational Medicine, University of Liverpool, Alder Hey Children's Hospital, Liverpool L12 2AP, UK. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Dr Perry is funded by a UK National Institute for Health Research Clinician Scientist Award (grant NIHR/CS/2014/14/012). This article presents independent research funded by the National Institute for Health Research (NIHR). The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Ng M,
- Fleming T,
- Robinson M, et al
- World Health Organization
- Law C
- American Academy of Orthopedic Surgeons
- Porter M,
- Borroff M,
- Gregg P,
- Howard P,
- MacGregor A,
- Tucker K
- Perry DC,
- Metcalfe D,
- Costa ML,
- Van Staa T
- Smith SM,
- Craig LCA,
- Raja AE,
- McNeill G,
- Turner SW
- Information Services Division–National Services Scotland
- Freeman JV,
- Cole TJ,
- Chinn S,
- Jones PR,
- White EM,
- Preece MA
- Vidmar SI,
- Cole TJ,
- Pan H
- Barlow SE; Expert Committee
- Cole TJ,
- Freeman JV,
- Preece MA
- GPD Team
- Pace M,
- Cayotte E,
- Lanzieri G, et al
- Ippolito E,
- Mickelson MR,
- Ponseti IV
- Baker C
- National Records of Scotland
- Copyright © 2018 by the American Academy of Pediatrics