Association Between Eczema and Stature in 9 US Population-Based Studies

Importance  Atopic dermatitis (eczema) is a chronic inflammatory disorder that is associated with other chronic diseases (eg, asthma), major quality-of-life impairment, sleep disturbance, and the use of potent topical and sometimes systemic corticosteroids, all of which might affect growth in childhood and adolescence. However, previous smaller-scale studies found conflicting results.

Objective  To determine whether eczema is associated with short stature.

Design, Setting, and Participants  We used data from 9 US population-based studies, including the National Survey of Children’s Health (2003-2004 and 2007-2008), National Health Interview Survey (children’s health, 2008-2012; adult health, 2010 and 2012), and National Health and Nutrition Examination Survey (2003-2004 and 2005-2006). Participants included 264 326 children and adolescents and 83 511 adults.

Exposure  History of eczema.

Main Outcomes and Measures  Percentiles of height for age and sex in children and height in adults. We constructed multivariate survey linear or logistic regression models for individual studies with Box-Cox transformed or dichotomized height, respectively. Pooled analyses used generalized linear mixed models.

Results  Overall, eczema was not associated with significant differences of height (continuous or <5th or <25th percentiles) in any of the studies or in the pooled analyses. We found a significant interaction by age, such that eczema was associated with shorter stature at 12 to 13 but not 14 to 15 or 16 to 17 years of age or in adulthood. Moderate to severe eczema was associated with shorter stature (continuous and <25th percentile). In particular, short stature (<5th percentile) was associated with eczema only when accompanied by an indicator of insufficient sleep (ie, 0 to 3 nights of sufficient sleep per week) (1.3% of children with eczema) but was not associated with asthma, hay fever, or use of prescription medication. The interaction between eczema and insufficient sleep remained significant at 10 to 11 years of age (P = .003) but not at other ages (P > .08 for all).

Conclusions and Relevance  Eczema is not associated with short stature overall. However, a small subset of children and adolescents with severe eczema accompanied by prominent insufficient sleep may have potentially reversible vertical growth impairment.

Atopic dermatitis (AD) or eczema is a chronic inflammatory skin disorder that affects approximately 10% of children¹ and adults² in the United States. Eczema is associated with intense itch, fatigue, sleep disturbance,² and a number of comorbid medical disorders, including asthma, hay fever, food allergy,³ recurrent ear infections, visual and dental health problems,³ cutaneous and extracutaneous infections,⁴ mental health comorbidity,⁵ injuries requiring medical attention,⁶ and epilepsy.⁷

Eczema has a number of sequelae that may have a negative effect on vertical growth during childhood and adolescence, such as sleep impairment, chronic inflammation, treatment with systemic corticosteroids, and multiple comorbid chronic medical conditions. Some studies^8-10 previously demonstrated that children with eczema have significantly shorter stature than those without eczema. However, another study¹¹ found no association between eczema and stature. Some of these studies had smaller sample sizes that limited the ability to perform subset analyses and thereby identify the major disease factors that might drive short stature. Sleep disturbances secondary to eczema are of particular interest because growth hormone secretion is highly regulated by normal sleep architecture.^12-14 Patients with eczema have high rates of fatigue and sleep disturbance, which are major predictors of overall quality of life.¹⁵ We hypothesized that childhood eczema, particularly with greater severity and more sleep impairment, is associated with shorter stature. The present study analyzed data from 9 different large-scale studies to determine whether eczema is associated with short stature.

After approval by Northwestern University’s institutional review board, deidentified data were assessed from the following studies: the National Health Interview Survey (NHIS) (annually in 2008-2012 for children’s health and 2010 and 2012 for adult health), the 2003-2004 and 2007-2008 National Survey of Children’s Health (NSCH), and the 2003-2004 and 2005-2006 National Health and Nutrition Examination Survey (NHANES). Informed consent was obtained for each study as per protocol. The characteristics of these studies are summarized in eTable 1 in the Supplement. Using data from the US Census Bureau, sample weights were created for each study by their sponsors, which factored age, sex, race, ethnicity, household size, and educational attainment of the most educated household member using a multistage area probability sampling design. These sample weights are needed to provide nationally representative prevalence estimates for each state’s population of noninstitutionalized adults. All frequency data are presented as raw values, whereas prevalence estimates reflect this complex weighting.

All data processing and statistical analysis were performed using commercially available software (SAS, version 9.4; SAS Institute, Inc). The frequency of missing values is presented in eTable 2 in the Supplement. We performed complete data analysis for all variables; that is, participants with missing data were excluded, except for the following 4 exceptions for which the missing values included more than 5% of respondents: household income in the 2003-2004 and 2007-2008 NSCH and height in the 2003-2004 and 2005-2006 NHANES. Therefore, we used the multiple imputation of missing household income levels for the 2003-2004 NSCH¹⁶ and the 2007-2008 NSCH¹⁷ that were generated by the National Center for Health Statistics. The National Center for Health Statistics previously evaluated the nonresponse rates for the NSCH and found minimal or no response bias for multiple questions.¹⁸ Analysis of the pattern of missing values for height in the NHANES revealed that missing values were more likely in children aged 2 to 9 years and in respondents from households with 1 person, but we found no major differences across sex, race/ethnicity, household income, or history of eczema, asthma, or hay fever. Therefore, sensitivity analyses that excluded children aged 2 to 9 years and respondents from households with 1 person were performed.

We performed outlier detection for height and excluded the outliers. All heights were standardized to inches, and pediatric heights were converted into percentiles of height for age and sex using the growth charts from the Centers for Disease Control and Prevention,¹⁹ termed height percentiles throughout this report. Height percentiles were transformed using the Box-Cox family of power transformations⁴⁰ to approach normality of distribution and constancy of error variance. For models of pediatric height percentiles and adult height, multivariate linear regression models were constructed for individual studies using commercially available software (SURVEY Procedures; SAS Institute, Inc) that accounted for the surveys’ complex weighting factors. The dependent variables consisted of transformed height percentiles in children and adolescents and height in adults, and the independent variables included history of eczema and eczema severity. Coefficients from regression models should be interpreted with caution because the transformed height variables are not the same scale as the original raw variables. Furthermore, Box-Cox transformations are nonlinear; thus, reverse transformation of coefficients is not possible. Therefore, coefficients should be interpreted as measures of overall association by their direction (positive or negative) and significance. Because similar results were found in models of untransformed variables, only data for transformed variables are shown.

For models of short stature, pediatric height percentiles were dichotomized (<5th or <25th percentile). In addition, adult heights were divided empirically into less than the 5th or less than the 25th percentile by combining 1 or 5 vigintiles (20 groups). Again, multivariate logistic regression models were constructed for dichotomized percentiles in individual studies using software (SURVEY Procedures) that accounted for the surveys’ complex weighting factors. In addition, generalized linear mixed models with identity (height percentiles) or logit (<5th or <25th percentile) link function were constructed for pooled analysis of all studies that included a random effect for the individual study (PROC GLIMMIX; SAS Institute, Inc). The complex weighting factors for all studies could not be combined given the varied sampling and study designs and were therefore not included in the generalized linear mixed models. Multivariate models included a history of or current eczema, asthma, and hay fever; age; race/ethnicity; household income; highest level of education in the household; US birthplace; number of persons in the home; use of prescription medication; and maternal/paternal height when available. We estimated least squares means, adjusted odds ratios, and 95% CIs. Our a priori hypothesis was that sleep disturbance in eczema is associated with growth retardation. The 2003-2004 and 2007-2008 NSCH assessed the number of nights per week with sufficient sleep. However, the distribution of this variable was bimodal and highly skewed; therefore, it was dichotomized into 0 to 3 vs 4 to 7 nights of sufficient sleep, similar to a previous study.³ The dichotomized variable for nights of sufficient sleep and a statistical interaction term between sufficient sleep and a history of eczema were included as covariates. Interactions were also tested between eczema and a history of asthma and/or hay fever or use of prescription medication. Estimates of effect size for significant interactions between 2 variables were conducted for the levels of one variable at a fixed level of the other variable. Least squares means and adjusted odds ratios (95% CIs) were estimated for each combination of factors included in the interaction effects.

We collected data on 264 326 children and adolescents and 83 511 adults, representing all age, sex, racial/ethnic, household educational attainment, and household income groups. The pooled US prevalence of eczema in childhood was 11.4% (95% CI, 11.3%-11.5%); in adults, it was 8.8% (95% CI, 8.6%-9.0%). Among the children with eczema, 23.0% (95% CI, 22.5%-23.5%) had a history of asthma and 31.9% (95% CI, 31.4%-32.4%) had hay fever and/or respiratory allergies; 13.1% (95% CI, 12.9%-13.4%) of adults with eczema had a history of asthma and 33.2% (95% CI, 31.4%-34.9%) had hay fever and/or respiratory allergies.

In generalized linear mixed models that controlled for random effects for different studies, prevalence of eczema was significantly associated with female sex, higher household income, more people in the household, a higher level of household educational attainment, and US birthplace in children and adults (eTable 3 in the Supplement). Eczema prevalence was significantly lower in Hispanic and Asian children and adults but significantly higher in those who self-identified as multiracial or race other than Hispanic or Asian. In contrast, eczema prevalence was significantly higher in black children but lower in black adults.

To model height percentiles across studies that had different levels of variance and skew, we performed Box-Cox transformations for pediatric height percentiles (λ = 0.50) and adult height (λ = −0.75). Eczema was not associated with significant differences of transformed height percentiles in any of the 8 studies of children and adolescents (survey linear regression) (Figure 1). Furthermore, we found no significant association between eczema and height percentiles in the pooled analysis using generalized linear mixed models of all 8 data sets (β = −0.002 [95% CI, −0.008 to −0.004]; P = .44).

Given that vertical growth and eczema activity may vary at different ages, generalized linear mixed models including 2 studies of children and adolescents aged 0 to 17 years (2003-2004 NSCH and 2005-2006 NHANES ), 1 study of adolescents aged 10 to 17 years (2007-2008 NSCH), and 5 studies of adolescents aged 12 to 17 years (NHIS for 2008-2012) were stratified by 2-year age intervals. Least squares mean height percentiles were significantly lower at ages 4 to 5 and 12 to 13 years, higher at ages 6 to 7 and 8 to 9 years, and not different at ages 10 to 11, 14 to 15, or 16 to 17 years (eFigure in the Supplement). Finally, eczema was not associated with significant differences in any of the 4 adult studies using transformed height distributions (Figure 1). Similar results were found in sensitivity analyses of the NHANES data that excluded children aged 2 to 9 years and participants from a household with 1 person.

To determine whether a subset of patients with eczema were at higher risk for very short stature, we constructed survey logistic regression models for each of the 8 pediatric and 4 adult studies. We found no significantly different odds of being at less than the 5th or less than the 25th percentile in people with eczema in any of the 12 studies, suggesting that overall eczema is not associated with very short stature (Table 1).

The 2007-2008 NSCH assessed for self-reported eczema severity in those children and adolescents who were reported to have eczema. The associations²⁰ and comorbidities^3,21 of moderate to severe eczema from this cohort were described previously. Moderate to severe eczema was associated with significantly lower transformed height percentiles compared with mild disease (β = −0.07 [95% CI, −0.12 to −0.02]; P = .004) (Figure 2). In contrast, mild eczema was not associated with lower height percentiles compared with no eczema (β = −0.001 [95% CI, −0.03 to 0.03]; P = .95). Moderate (β = 2.06 [95% CI, 1.28-3.33]; P = .003) and severe (β = 2.29 [95% CI, 1.17-4.48]; P = .02) eczema were associated with significantly higher odds of having less than 25th but not less than the 5th height percentiles (Table 2).

Several measures of sleep disturbance were reported in different studies. The 2003-2004 and 2007-2008 NSCH assessed for the number of nights with sufficient sleep, whereas the 2012 NHIS assessed for fatigue, daytime sleepiness, and insomnia in children. Overall, we found no significant interaction between eczema and insufficient sleep as predictors of height percentiles (P = .47) (Figure 2B). However, we found a significant interaction between eczema and nights of sufficient sleep, such that adolescents with eczema and 0 to 3 nights of sufficient sleep per week had significantly higher odds of less than the fifth height percentiles (Table 3). Only 1.3% of children with eczema had 0 to 3 nights of sufficient sleep and short stature, suggesting that only a small fraction of children with eczema are at higher risk for short stature. Furthermore, this interaction remained significant only at ages 10 to 11 years (P = .003) and not at other ages (P > .08 for all), suggesting that the association may be reversible. We found no significant associations between nights of insufficient sleep and having less than 25th height percentiles. Also, we found no significant interactions of eczema with fatigue, daytime sleepiness, or insomnia as predictors of overall height percentiles, less than the 5th height percentile, or less than the 25th height percentile (P ≥ .15). Finally, we found no significant interactions between eczema and history of asthma and/or hay fever (P = .44) or between eczema and use of prescription medication (P = .41).

The NHIS from 2008 through 2012 assessed for height in other members of the child’s household, including the mother and father (26 283 parents [22.5% of those cohorts]). As expected, pediatric height percentiles were significantly associated with maternal height (survey linear regression; β = 0.04 [95% CI, 0.03-0.04]; P < .001) and paternal height (β = 0.03 [95% CI, 0.02-0.04]; P < .001). Parental height was not associated with one’s child having eczema (survey logistic regression OR for the mother, 1.02 [95% CI, 0.98-1.06]; P = .31; survey logistic regression OR for the father, 1.03 [95% CI, 0.93-1.13]; P = .63). Furthermore, we found no interactions between a history of eczema and maternal or paternal height quartiles as predictors of the child’s height (P = .39 and P = .43, respectively), and a history of eczema did not modify the effects of parental height on the child’s height (P ≥ .09 for all).

In the present study, we analyzed data from 9 US population-based studies and found that, in general, eczema in childhood or adulthood is not associated with short stature. In a single study that assessed for eczema severity, albeit self-reported, moderate to severe eczema was associated with significantly shorter stature compared with mild disease. In contrast, mild disease was not associated with shorter stature compared with no eczema. Eczema accompanied by only 0 to 3 nights of sufficient sleep per week was associated with less than the fifth height percentile. This association occurred at ages 10 to 11 years and appeared to reverse at older ages. Finally, eczema did not modify the association between child and parental height. Together, these results suggest that, although eczema is not associated with short stature overall, a small subset of pediatric patients with moderate to severe disease may experience a negative effect on their growth. The association of short stature with eczema and sleep insufficiency at 10 to 11 but not 12 to 13, 14 to 15, or 16 to 17 years of age or in adulthood suggests that any growth impairment that may occur in adolescence secondary to eczema is related to sleep impairment and probably reversible.

Previous studies found conflicting results about whether eczema is associated with short stature. A previous case-control study of 35 British adults with childhood onset and 35 without AD¹¹ found no significant differences of stature related to a history of AD. Furthermore, growth impairment was not related to body surface area affected, potency of topical corticosteroids used, or comorbid asthma.¹¹ In contrast, an earlier study of 89 British children⁸ found that 10% had standing heights of less than the third percentile, with growth impairment being associated particularly with widespread eczema, comorbid asthma, and use of potent topical corticosteroids. Another study⁹ found that children with lesional body surface area of less than 50% had normal heights, whereas those with lesional body surface area greater than 50% were significantly shorter. However, those studies were unable to generate estimates of risk for short stature in the pediatric and adult population with eczema owing to small sample size and selection bias of being a single referral center for the management of eczema. A recent study of 2372 Korean elementary schoolchildren¹⁰ found that AD, particularly when associated with sleep disturbance, was associated with shorter stature. Similarly, we found that eczema with only 0 to 3 nights of sufficient sleep per week was associated with shorter stature. Together, the results of the present study and the above-mentioned studies suggest that severe eczema accompanied by sleep impairment may contribute to reversible growth impairment in childhood. Distinguishing between the effects of itch per se and sleep disturbances in eczema is difficult. The associations observed for 0 to 3 nights of sufficient sleep per week may represent a distinct subset of patients with AD who have comorbid sleep disturbances (approximately 15% of children with AD) or may be a proxy measure of very severe AD. Either way, self-reported insufficient sleep appears to be a useful approach to identifying those children and adolescents at higher risk for growth impairment. Future studies are needed to determine the particular role of sleep impairment in AD and its effects on stature.

The mechanisms underlying the association between severe eczema and short stature are not fully understood. Our results suggest that the association is largely due to the harmful effects of poor sleep. Growth hormone secretion during sleep plays an important role in vertical growth, and disorders that are associated with impaired sleep, such as obstructive sleep apnea, are associated with short stature.^22-25 However, children with eczema have a number of other known risk factors for growth impairment, including use of systemic corticosteroids²⁶ and perhaps topical corticosteroids,²⁷ depression and other psychological comorbidity,^5,6 and restrictive diets that may result in malnutrition.²⁸ We were not able to control for these risk factors in the present study. In addition, chronic inflammation from eczema per se may contribute toward growth impairment. Previous studies^29-31 found growth impairment in multiple chronic inflammatory diseases, including Crohn disease and hepatitis B and C virus infection. We found no significant effects for history of asthma and/or hay fever or use of prescription medications. Future studies are warranted to confirm the association between severe eczema and growth impairment using objective measures of disease severity and to determine the mechanisms thereof.

A recent study of adults from the 2012 NHIS found that sleep disturbance is a major driver of poor overall health status, lost workdays, and increased use of health care services in adults with eczema.¹⁵ Several other studies in children and adults^32-36 found that sleep disturbance is one of the most important factors to have a negative effect on health-related quality of life. The present study demonstrates that sleep disturbance may have additional harmful consequences by retarding growth. Together, these studies emphasize the need for more rigorous screening for sleep disturbances in patients with pediatric eczema and for developing safer and more effective interventions aimed at reducing sleep disturbance in childhood eczema. The interaction between eczema and the number of nights of insufficient sleep but not fatigue, daytime sleepiness, or insomnia suggests that nights of sufficient sleep may be a simple and useful patient-reported outcome of chronic sleep disturbance in eczema. Future studies are needed to validate the use of this outcome in eczema.

This study has several strengths, including the 9 different US population-based studies with prospective data collection and a very large, random-sampled, diverse population that allowed for subset analyses and controlling for a large number of potential confounding factors. However, this study has potential limitations. Eczema history was based on self-report in all studies and was not assessed clinically or verified with any diagnostic testing. However, self-report of eczema using different instruments has been validated and found to have good correlation with results of a clinical examination.^37,38 Caregiver-reported height was previously validated in the NSCH for children and adolescents 10 years or older.³⁹ None of the studies categorized the type of prescription medications used. Thus, we were not able to control for the use of high-dose topical and/or oral corticosteroids or other specific treatments that may affect stature adversely. Future longitudinal studies with objective measures of eczema severity and height measurement and with assessment of the specific medications used are needed to identify which subset of patients with eczema, if any, is predisposed to short stature.

Childhood eczema is not associated with short stature overall, although severe disease with prominent sleep disturbance is associated with higher odds of short stature in early adolescence. Future studies are warranted to better characterize sleep disturbances and other risk factors and mechanisms of growth impairment in eczema and to determine whether such impairment is reversible.

Accepted for Publication: August 27, 2014.

Corresponding Author: Jonathan I. Silverberg, MD, PhD, MPH, Department of Dermatology, Northwestern University, 680 Lakeshore Dr, Ste 1400, Chicago, IL 60611 (jonathanisilverberg@gmail.com).

Published Online: December 10, 2014. doi:10.1001/jamadermatol.2014.3432.

Author Contributions: Dr Silverberg had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Silverberg.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Silverberg.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Silverberg.

Administrative, technical, or material support: Paller.

Study supervision: All authors.

Conflict of Interest Disclosures: None reported.

Previous Presentation: The results of this study were presented in part at the Eighth Annual Meeting of the International Society of Atopic Dermatitis; May 22, 2014; Nottingham, England.