eTable. Association Between Atopic Dermatitis Diagnosis and Family History of Hypertension and Type 2 Diabetes Mellitus
Silverberg JI, Becker L, Kwasny M, Menter A, Cordoro KM, Paller AS. Central Obesity and High Blood Pressure in Pediatric Patients With Atopic Dermatitis. JAMA Dermatol. 2015;151(2):144-152. doi:10.1001/jamadermatol.2014.3059
Atopic dermatitis (AD) is associated with multiple potential risk factors for obesity and high blood pressure (BP), including chronic inflammation, sleep disturbance, and mental health comorbidity. Previous studies found associations between general obesity and AD. However, it is unknown whether AD is associated with central obesity and/or high BP.
To determine whether central obesity and high BP are increased in pediatric AD.
Design, Setting, and Participants
This case-control study performed in multicenter pediatric dermatology practices in the United States recruited 132 children (age range, 4-17 years) with active moderate to severe AD and 143 healthy controls from April 1, 2009, through December 31, 2012.
Diagnosis and severity of AD assessed by a pediatric dermatologist.
Main Outcomes and Measures
Body mass index, waist circumference, waist to height ratio, systolic BP, and diastolic BP.
Moderate to severe AD was associated with body mass index for age and sex of 97th percentile or greater (logistic regression; odds ratio [OR], 2.64; 95% CI, 1.15-6.06), International Obesity Task Force obesity cutoffs (OR, 2.38; 95% CI, 1.06-5.34), waist circumference in the 85th percentile or greater (OR, 3.92; 95% CI, 1.50-10.26), and waist to height ratio of 0.5 or greater (OR, 2.22; 95% CI, 1.10-4.50). Atopic dermatitis was associated with higher BP for age, sex, and height percentiles (systolic BP: OR, 2.94; 95% CI, 1.04-8.36; diastolic BP: OR, 3.68; 95% CI, 1.19-11.37), particularly a systolic BP in the 90th percentile or higher (OR, 2.06; 95% CI, 1.09-3.90), in multivariate models that controlled for demographics, body mass index and waist circumference percentiles, and history of using prednisone or cyclosporine. Atopic dermatitis was associated with higher systolic BP in Hispanics/Latinos (general linear model; β, .23; 95% CI, .04-.43) and Asians (β, .16; 95% CI, .03-.30). Severe to very severe AD was associated with systolic BP in the 90th percentile or higher (adjusted OR, 3.14; 95% CI, 1.13-8.70). Atopic dermatitis was associated with a family history of hypertension (adjusted OR, 1.88; 95% CI, 1.14-3.10) and type 2 diabetes mellitus (adjusted OR, 1.64; 95% CI, 1.02-2.68) but not obesity or hyperlipidemia.
Conclusions and Relevance
Moderate to severe pediatric AD may be associated with central obesity and increased systolic BP.
Atopic dermatitis (AD) is a chronic pruritic disorder of the skin that is a major health burden. The current US prevalence of AD in childhood of 10.7%1 has markedly increased from previous estimates of 2% to 3% before 1960 and 9% to 12% after 1970.2,3 Obesity has steadily increased during a similar course to near-epidemic proportions in the United States.4- 7 Obesity has now been reported to be associated with pediatric psoriasis,8,9 an inflammatory skin disease with a prevalence of approximately 1% in US children.
The parallel increases in prevalence of AD and obesity may suggest an association between obesity and AD. Previous epidemiologic10- 13 and clinical14,15 studies support an association between allergic disorders, including AD, and obesity. International epidemiologic studies12,13 found associations between obesity and self- or parental report of AD in childhood and/or adolescence. Little is known about whether risk factors for metabolic disease, such as central vs general obesity and elevated blood pressure (BP), are also increased in children with AD. We performed a prospective, clinical, case-control study that addressed whether AD is associated with central obesity and elevated BP in childhood.
The study was performed as an ancillary to a previously reported investigation of the association between obesity and psoriasis.8 The study was approved by the institutional review boards at each participating site. Parents and children, as required by each center’s institutional review board or ethics committee, provided written informed consent. Participants in this US multicenter, cross-sectional study were recruited from April 1, 2009, through December 31, 2012, from all children with AD who were seen at 7 dermatology referral centers. Inclusion criteria included age of 4 to 17 years and active moderate to severe AD. Children attending these dermatology centers for other skin problems (eg, nevi, molluscum contagiosum, warts, and acne) who were of similar age, sex, and ethnicity without skin or systemic inflammatory or allergic disease were serially recruited as controls. Siblings of any children included in the study were excluded. To minimize ascertainment bias, an attempt was made to recruit all AD patients serially at presentation, and no potential participant refused. Investigators completed a questionnaire with each patient or parent that addressed patient history, race/ethnicity, age, and sex. Patients with AD and healthy children without AD were recruited into parallel arms of that study and oversampled based on that study's primary power analysis. A post hoc power analysis was based on the Fisher exact test for the 275 patients recruited, with the use of 2-sided tests and an α level of .05. The null proportion of children with AD was assumed to be 4.3%, based on epidemiologic data revealing a 4.3% prevalence of moderate to severe AD in children and adolescents in the United States.16 The ability to discriminate an odds ratio (OR) of 2.5 or higher is associated with greater than 95% power.17 Deidentified data were compiled centrally, and statistical analysis was performed by 2 statisticians (J.I.S. and M.K.) at Northwestern University, Chicago, Illinois.
A dermatologist diagnosed AD using the Hanifin and Rajka criteria.18 Severity of AD was classified using a physician’s global assessment score as moderate, severe, and very severe. For all participants, current treatment or a history of treatment with phototherapy and/or systemic medications was recorded.
A verbal questionnaire was administered to parents of all patients to determine the following: history of asthma, hay fever, or food allergies; history of using prednisone, cyclosporine, or phototherapy; and history of obesity, type 2 diabetes mellitus, hypertension or high BP, and hyperlipidemia or high cholesterol in first-degree relatives (all yes/no responses).
Weight and height were measured, and body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. An age- and sex-adjusted BMI percentile was assigned using the modified least mean square estimation procedure from the 2000 Centers for Disease Control and Prevention (CDC) growth charts.19 Underweight, overweight, and obesity cutoffs for BMI percentile were defined according to the recommendations of the CDC (<5th, 5th-84th, 85th-94th, and ≥95th percentiles)19 and the European Childhood Obesity Group (ECOG) (<5th, 5th-89th, 90th-96th, and ≥97th percentiles).20 Overweight and obesity were further defined using the international cutoffs proposed by the International Obesity Task Force (IOTF) that correspond to BMIs of 25 and 30 in adulthood.21 Waist circumference (WC) was measured midway between the most inferior rib and the superior border of the iliac crest with an inelastic measuring tape. The WC percentiles were determined according to sex- and age-specific cutoffs22 and were divided into 3 percentile groups (<15th, 15th-84th, and ≥85th). Waist to height ratio (WHR) was considered as a continuous measure of risk, but for estimating the odds of excess central adiposity defined by the WHR, previously defined cutoffs were used to define high (≥0.50)23,24 and low (≤0.42)23,25 waist size. These cutoffs represent the 75th and 25th percentiles for WC of participants in this study.
Systolic BP (SBP) and diastolic BP (DBP) were measured using a single measure with a sphygmomanometer. A pediatric-sized cuff was applied for children, whereas an adult-sized cuff was used for adolescents. The SBP and DBP percentiles for age, sex, and height were determined26 and treated as continuous and ordinal (<50th, 50th-89th, and ≥90th percentiles) variables. Investigators were not masked to the dermatologic condition at the time of anthropometric and BMI measures.
All data processing and statistical analyses were performed using SAS statistical software, version 9.3 (SAS Institute Inc). Bivariate associations were tested using 1 of 3 approaches: (1) Wilcoxon rank sum tests for continuous variables, (2) χ2 tests for categorical or ordinal variables, or (3) Fisher exact test for categorical or ordinal variables with more than 20% of expected cell frequencies less than 5%.
Binomial logistic regression models were constructed to determine associations between AD prevalence and severity (dependent variables) and anthropometric measures. Because of the nonproportionality of the odds (score test, P = .01), AD severity was coded as a binary outcome variable (severe to very severe vs moderate). The independent (explanatory) variables were weight and height for age and sex percentiles (continuous and ordinal: <5th, 5th-84th, 85th-94th, and ≥95th percentiles), BMI for age percentiles (continuous and ordinal: <5th, 5th-84th, 85th-94th, and ≥95th percentiles and <5th, 5th-89th, 90th-96th, and ≥97th percentiles), IOTF BMI classification (normal weight, overweight, or obese), WC for age percentiles (ordinal: <15th, 15th-84th, and ≥85th percentiles), or WHR (continuous and ordinal: ≤0.41, 0.42-0.49, and ≥0.50). Crude (unadjusted) ORs were determined. Adjusted ORs were determined from models, including age (continuous), sex (binary), and race/ethnicity (categorical). Both 2- and 3-way statistical interactions among covariates were tested and included in the final models if P < .10 and OR modification by more than 20% was present on univariate analyses.
Several different modeling approaches were used to study the association between AD diagnosis and severity with SBP and DBP. Logistic regression models were fit predicting AD diagnosis or severity (severe to very severe vs moderate) as the dependent variables and SBP or DBP as the independent variables. Multivariate models controlled for age; race/ethnicity; sex; BMI and WC percentiles; and history of asthma, hay fever, food allergies; and prednisone, cyclosporine, or phototherapy use.
To test for interactions between AD and demographic factors or adiposity as predictors of BP, generalized linear models were constructed with SBP and DBP as the continuous dependent variables using an identity link function. Independent variables included age, sex, race/ethnicity, and binary indicators of excess adiposity (BMI ≥90th percentile, WC ≥85 percentile, or WHR ≥0.5), AD diagnosis, and a 2-way interaction term between them. Post hoc analyses were conducted for differences among the levels of one factor at a fixed level of the other factor. Adjusted β values were estimated for each combination of factors included in the interaction effects. Multivariate models controlled for age; race/ethnicity; sex; and history of asthma, hay fever, food allergies; and prednisone, cyclosporine, or phototherapy use.
No missing data were encountered in the study. Mixed models were constructed that included a random effect for the study center. The results of models with or without random effects were similar. Therefore, only the results of models without random effects are presented. A 2-sided P < .05 was taken to indicate statistical significance for all estimates. However, the multiple dependent tests performed in this study increase the risk of falsely rejecting the null hypothesis. Therefore, P values near .05 should be interpreted with caution.
A total of 132 patients with AD and 143 control patients without inflammatory skin disease were serially selected for study, and no potential participants refused. Patients with AD were significantly younger than controls (P = .03) but were matched with respect to sex (P = .88) and race/ethnicity (P = .37) (Table 1). Patients were diagnosed as having AD at a mean age of 26.4 months. Severity of AD was not associated with age (P = .37), sex (P = .23), or race/ethnicity (P = .35).
In models of height percentiles for age as a continuous variable, children with AD had significantly lower height than those without AD (logistic regression; OR, 0.36; 95% CI, 0.16-0.81) (Table 2). Review of distribution of height percentiles for age revealed that the healthy controls had above average height, which accounted for these differences. However, when height percentiles for age were modeled as an ordinal variable, no associations were found between AD and short stature (<5th percentile) (Table 2).
In the univariate and multivariate models, no significant differences were found in continuous or ordinal weight percentiles or continuous BMI and WC percentiles between children and adolescents with AD and controls. However, when BMI was modeled as an ordinal variable according to the classification of the ECOG, then AD was associated with BMI in the 97th percentile or higher (OR, 2.64; 95% CI, 1.15-6.06) but not BMI in the 5th percentile or 90th to 96th percentile. Similarly, AD was associated with obesity as defined by the IOTF cutoffs (OR, 2.38; 95% CI, 1.06-5.34) but not by the CDC criteria (OR, 1.64; 95% CI, 0.82-3.29). The associations remained significant in multivariate models that controlled for age, race/ethnicity, and sex. There was significant agreement between the ECOG and IOTF measures (weighted κ = 0.89; Bowker test of symmetry = 12.3, P = .007) and the CDC and IOTF measures (weighted κ = 0.90; Bowker test of symmetry = 9.0, P = .03).
Atopic dermatitis was associated with central obesity as judged by a WC in the 85th percentile or greater (OR, 3.92; 95% CI, 1.50-10.26) and a WHR in the 5th percentile or greater (OR, 2.22; 95% CI, 1.10-4.50). The associations remained significant in multivariate models. In contrast, no associations were found between any of the anthropometric measures and AD severity.
Overall, AD was associated with higher SBP and DBP for age, sex, and height percentile (SBP: OR, 2.96; 95% CI, 1.16-7.51; and DBP: OR, 2.72; 95% CI, 1.01-7.36) (Table 3 and Table 4). When the SBP percentile was modeled as an ordinal variable, AD was associated with an SBP in the 90th percentile or higher (OR, 2.06; 95% CI, 1.13-3.75). These associations remained significant in multivariate models that controlled for age, sex, race/ethnicity, BMI and WC percentiles, and history of prednisone or cyclosporine use. In contrast, AD was not associated with a DBP in the 90th percentile or higher. Moreover, children with AD were less likely to have SBP and DBP in the less than 25th percentile compared with those who did not have AD (SBP: OR, 0.35; 95% CI, 0.14-0.86; and DBP: OR, 0.26; 95% CI, 0.09-0.72), suggesting relative increases of SBP and DBP in AD even without extreme values. However, this association did not remain significant in multivariate models, particularly when controlling for adiposity.
Overall, race/ethnicity, current age, age at AD onset, duration of AD, and sex were not associated with SBP. However, a significant interaction was found between AD and race/ethnicity as a predictor of SBP. In particular, AD was associated with significantly higher SBP in Hispanics/Latinos (β, .23; 95% CI, .04-.43; P = .02) and Asians (β, .16; 95% CI, .03-.30; P = .02).
Overall, DBP was higher in boys (median, 73.8%; interquartile range [IQR], 33.6%) than in girls (median, 59.7%; IQR, 41.5%) (P < .001, Kruskal-Wallis test) and in Asians (median, 76.0%; IQR, 28.5%), African Americans/blacks (median, 73.4%; IQR, 32.1%), and Hispanics/Latinos (median, 72.3%; IQR, 50.3%) than in non-Hispanic whites (median, 60.2%; IQR, 37.7%) (P < .001), whereas current age, age at AD onset, duration of AD, and sex were not associated with DBP. In contrast with SBP, no interaction was found between race/ethnicity and AD as predictors of DBP.
Severe to very severe AD was associated with SBP in the 90th percentile or higher (adjusted OR, 3.14; 95% CI, 1.13-8.70) (Table 4). However, none of the other measures of BP were significantly associated with AD severity in univariate or multivariate models. All these associations remained significant in mixed models, including random effects from different investigators.
To determine whether the association between AD and SBP and/or DBP is related to excess adiposity, generalized linear models were constructed with interactions between central obesity and BP. A BMI in the 90th percentile or higher (β, .14; 95% CI, .02-.26) and a WHR in the 5th percentile or higher (β, .14; 95% CI, .04-.24) were associated with SBP in children and adolescents without AD (Table 5). A WHR in the 5th percentile or higher (β, .10; 95% CI, .01-.20) was associated with DBP in those without AD. However, obesity was not associated with higher SBP or DBP in children with AD.
In multivariate models, AD was associated with a family history of hypertension (adjusted OR, 1.88; 95% CI, 1.14-3.10) and type 2 diabetes mellitus (adjusted OR, 1.64; 95% CI, 1.02-2.68) but not obesity or hyperlipidemia (eTable in the Supplement). This association remained significant in multivariate models. Generalized linear models were constructed to determine whether the association between AD and SBP and/or DBP is related to a family history of hypertension. In multivariate models, a family history of hypertension was associated with SBP (β, .11; 95% CI, .02-.19) in children with AD but not in those without AD (β, .04; 95% CI, −.06 to .14). Diastolic BP was not associated with family history of hypertension in children with or without AD.
The present observational study uncovered several new associations related to metabolic issues in children with AD, particularly related to central obesity and elevated BP, criteria for metabolic syndrome. Although we also found associations of AD prevalence and severity with obesity, these associations have previously been described.10- 15
The associations of AD with increased SBP and DBP and of severe AD with high SBP, even after controlling for adiposity in children with AD, have not been previously described. Other inflammatory disorders have previously been reported to be associated with hypertension in children and adolescents with psoriasis,27 and hypertension and subclinical cardiovascular disease with lupus,28 dermatomyositis,29 and scleroderma30; as in these studies, administration of systemic corticosteroids as a possible confounding factor was ruled out in psoriasis27 and lupus.28 In contrast to our findings, a previous study31 of 521 Japanese adults found no association between hypertension and active and/or healed AD and, in fact, suggested that AD is associated with low rates of hypertension. Inherent differences in genetics (race/ethnicity) and environmental factors (socioeconomic, diet, and exposures) between the cohorts of US children and Japanese adults might explain these conflicting results. Indeed, we found that a family history of hypertension was associated with higher SBP in children with AD, suggesting that AD may result in higher BP in genetically predisposed individuals. Alternatively, there may be common risk factors that occur in families of children with hypertension that also predispose them to AD. Additional studies are needed to confirm the association between AD and hypertension and to determine the mechanisms of such association.
It is intriguing that increased SBP and DBP was only related to adiposity in healthy controls but not patients with AD. Furthermore, AD was associated with increased SBP and DBP overall, with more children having an SBP in the 90th percentile or higher and fewer children having an SBP or DBP in the less than 25th percentile. This finding suggests that AD may result in relative increases of BP that might go undetected because levels are not extremely high. High SBP has been related to chronic inflammation32 and sleep impairment,33,34 both features of AD. The long-term sequelae of increased BP are unknown in children, but it is possible that cumulative increases of BP are associated with cardiovascular disease later in life, similar to that observed in psoriasis.35 Additional studies are needed to identify the trigger(s) and mechanism(s) for increased BP in AD and determine the long-term effect on cardiovascular health.
The mechanism of association between obesity and AD remains unknown. Previous studies36,37 have found altered levels of adipokines in AD. A study of 27 children with AD and 46 healthy controls found higher serum levels of resistin and apelin but lower levels of visfatin in children with AD.36 Low serum adiponectin levels were reported to be associated with increased AD in a cohort of German school children.37 Together, these studies suggest that adipose tissue may directly influence the risk of AD. In addition, obesity is associated with numerous immune sequelae, including increased immune infiltration of adipose tissue, upregulation of inflammatory cytokines, and effects on cellular immunity.38- 45 The association between AD and central obesity, in particular, may be useful for guiding further research. Central obesity has previously been reported to have particularly harmful effects on a variety of medical disorders, including asthma, dyslipidemia, diabetes, coronary artery disease, and myocardial infarction.46- 52 Perhaps some of the lessons learned from those disorders may shed light on the effects of obesity on AD.
Among the strengths of this study are (1) prospective data collection, (2) a well-powered sample size of cases and controls, (3) the use of clinical assessment of AD and its severity by well-trained dermatologists, and (4) consideration of multiple anthropometric measures of adiposity. However, as with any study, there are potential limitations. The study was cross-sectional and unable to assess the direction of association among AD, adiposity, and elevated BP. The severity of AD was assessed using the investigator’s global assessment, which is less standardized than other AD scoring systems (eg, Eczema Area and Severity Index or Severity Scoring of Atopic Dermatitis). To address this, we constructed models that also controlled for random effects for different investigators and found no significant differences of results. Nevertheless, we welcome future longitudinal studies using standardized AD scoring systems to confirm the results of this study and determine the direction of associations. Indeed, the healthy controls had above-average height. To address this, we used SBP and DBP percentiles that adjusted for height and controlled for confounding factors in statistical models. Blood pressure was measured with a single recording. There may be confounding factors that underlie the association among AD, obesity, and BP. For example, significant differences were found for age between patients with AD and controls. To address potential differences between patients with AD and controls, age, race/ethnicity, sex, and treatment use were included in all multivariate models. Nevertheless, there is the potential for residual confounding because of a lack of information regarding the cumulative dose and duration of topical and systemic treatment use. There are other potentially confounding variables, such as socioeconomic status, that were not collected based on the a priori hypotheses of the study, which precludes any conclusions about the mechanism(s) of association. These are all important considerations for future confirmatory studies. The multiple dependent tests performed in this study increase the risk of falsely rejecting the null hypothesis. Therefore, P values in this study near .05 should be interpreted with caution. Finally, the associations of AD with obesity and elevated SBP point toward a higher risk for metabolic syndrome in pediatric AD. However, we did not record lipid levels and were therefore not able to determine whether children and adolescents with AD met criteria for metabolic syndrome. In the future, even larger studies with additional BP measurements should examine the association among AD, lipid levels, and metabolic syndrome.
Several important clinical points arise from this study. Children and adolescents with moderate to severe AD had a higher prevalence of central obesity that was unrelated to severity of AD, suggesting that they may benefit from routine screening for adiposity using BMI and WC. In addition, BP should be assessed in children with moderate to severe AD, even those who are a healthy weight with moderate disease and who are not receiving systemic therapy. Children with AD may require closer observation for the development of elevated BP, especially those of Hispanic/Latino and Asian race/ethnicity. Furthermore, children and adolescents with AD who have elevated BMI, WC, and BP may benefit from aggressive intervention for weight loss and lifestyle modification for the management of BP. Finally, AD appears to not be associated with overall vertical growth impairment in children and/or adolescents. Future prospective, longitudinal studies are needed to confirm these associations and the direction of association among AD, obesity, and elevated BP.
Accepted for Publication: August 12, 2014.
Corresponding Author: Jonathan I. Silverberg, MD, PhD, MPH, Department of Dermatology, Northwestern University Feinberg School of Medicine, Ste 1400, 680 Lakeshore Dr, Chicago, IL 60611 (JonathanISilverberg@gmail.com).
Published Online: December 23, 2014. doi:10.1001/jamadermatol.2014.3059.
Author Contributions: Drs Silverberg and Paller had full access to all the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.
Study concept and design: Silverberg, Becker, Paller.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Silverberg, Becker, Paller.
Critical revision of the manuscript for important intellectual content: Silverberg, Becker, Kwasny, Menter, Cordoro.
Statistical analysis: Silverberg, Kwasny, Cordoro.
Obtained funding: Paller.
Administrative technical or material support: Becker, Menter.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by the International Psoriasis Council (Dr Paller).
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Additional Contributions: The following study center principal investigators each contributed 6 or fewer patients to this study: Wynnis Tom, MD, Neil Korman, MD, April Armstrong, MD, and Craig Leonardi, MD. The International Psoriasis Council supported accrual of data and helped with data entry.