Childhood type 1 diabetes may increase the risk of atopic dermatitis

Childhood type 1 diabetes may increase the

risk of atopic

dermatitis*

C.-H. Lin,

C.-C. Wei,

C.-L. Lin,

W.-C. Lin

and C.-H. Kao

Atopic dermatitis (AD) is a common chronic inflammatory skin disease, and the immune response observed during the disease course is characterized by T-helper (Th)2-dependent inflammation.1 By contrast, type 1 diabetes mellitus (T1DM),

an autoimmune disease targeting pancreatic islet beta cells, is mediated mainly by a Th1 response. Therefore, an inverse association has been theorized between T1DM and AD because the Th1/Th2 immune responses of these two diseases are mutually inhibitory.2–5

Some studies have reported that AD is associated with a lower risk of T1DM,6,7 and that AD can be protective against

childhood T1DM.8 However, a retrospective, case-controlled

study in which 920 Danish children with T1DM were compared with 9732 children without diabetes did not demonstrate that the early development of AD reduces the risk of T1DM or that a propensity for T1DM reduces the risk of

early-onset AD.9 Similarly, Gazit et al. and Stromberge et al. did

not find a significant difference in the prevalence of atopy among patients with T1DM vs. the general population. They

suggested that the conventional Th1/Th2 paradigm is an oversimplified explanation, on the basis of clinical and laboratory

parameters, for the complexity of the immune response in autoimmune and atopic diseases.10,11 Conversely, Stene and

Nafstad reported a strong positive association between T1DM and asthma,12 and Kero et al. found that asthma was more

common in children with T1DM than in those without

T1DM, suggesting the coexistence of Th1 and Th2 immunemediated disorders.13

T1DM and atopic disease, we hypothesized that children with T1DM may have an increased risk of subsequent AD compared with healthy children. To test this hypothesis, we conducted a population-based, retrospective cohort study using the National Health Insurance Research Database (NHIRD).

Materials and methods

Data source

The Taiwan National Health Insurance (NHI) programme has offered comprehensive, universal health insurance to all residents of Taiwan since 1995, and covers > 99% of residents

(http://nhird.nhri.org.tw/en/). The Bureau of National Health Insurance authorized the National Health Research Institutes to create the NHIRD for public research use. In this study, we used the Registry of Catastrophic Illnesses Patient Database (RCIPD), Longitudinal Health Insurance Database 2000 (LHID 2000) and Registry of Beneficiaries, all of which are part of the NHIRD.

The RCIPD contains health claims data for the treatment of catastrophic illness and includes 30 categories of diseases requiring long-term care. If insured people have major diseases, such as cancer or T1DM, they can apply for a catastrophic illness certificate. To reduce the financial hardship

associated with catastrophic illnesses, the NHI programme exempts beneficiaries from obligations to the NHI-defined catastrophic illnesses. The LHID 2000 contains the claims data of 1 million people randomly sampled from the 2000 Registry for Beneficiaries of the NHIRD. No considerable difference is observed in sex, age or healthcare costs between cohorts in the LHID 2000 and all insurance enrolees, as reported by the NHI in Taiwan. These datasets are linked using the encrypted unique personal identification number to obtain the longitudinal medical history of each individual. This study was

approved by the Research Ethics Committee of China Medical University (CMU-REC-101-012). The diagnoses and procedures are coded in the format of the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM).

included patients younger than 18 years who were newly diagnosed with T1DM (ICD-9-CM codes 250.x1 and 250.x3) between 1998 and 2011. The date of the diagnosis of T1DM was set as the index date for initiating the measurement of the follow-up person years. We excluded patients whose date of birth and sex were missing from the records, and those with pre-existing AD (ICD-9-CM code 691). The non-T1DM cohort was frequency matched to the T1DM cohort at a ratio of approximately 4 : 1 by sex and index date; the same study period and exclusion criteria were applied.

The baseline comorbidity history, including asthma (ICD-9-CM code 493), allergic rhinitis (472_0 and 477), hypertension (401–405), hyperlipidaemia (272), depression (296_2, 296_3, 300_4, 301_12, 309_0, 309_1 and 311), anxiety (300) and

obesity (278), was determined for each patient. All study participants were followed up until AD diagnosis, death, loss to

follow-up, withdrawal from the NHI programme or 31 December 2011.

Statistical analyses

Sociodemographic factors, including sex, age, urbanization level (level 1 and 4 being the most and least urbanized,

respectively), parental occupation (white-collar jobs, blue-collar jobs and others), frequency of medical visits (per year)

and comorbidities were compared between the T1DM and non-T1DM cohorts using the v2-test and t-test for categorical

and continuous variables, respectively.

To estimate the cumulative incidence of AD in the T1DM and non-T1DM cohorts, we performed survival analysis using the Kaplan–Meier method, with significance determined by a log-rank test. The incidence rates (per 1000 person years)

were estimated according to sex, age, urbanization level, parental occupation, comorbidity and follow-up time specific to

both cohorts. To examine the risk of AD in the T1DM cohort relative to that in the non-T1DM cohort, hazard ratios (HRs)

with 95% confidence intervals (CIs) were estimated using univariate and multivariate Cox proportional hazards regression

models. Multivariate Cox models were adjusted for sex, age, urbanization level, parental occupation, frequency of medical

visits and comorbidities of asthma, rhinitis, hypertension,

hyperlipidaemia, depression, anxiety and obesity. In addition, the dose–response effect on the risk of AD was assessed

according to the number of emergency room (ER) visits and/ or hospitalizations for T1DM.

All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, U.S.A.). A P value < 0_05 was considered statistically significant.

Results

In this study, 3386 and 12 725 children were included in the T1DM and non-T1DM cohorts, respectively, and 53_6% of the children with T1DM were girls. The mean ages of the T1DM and non-T1DM cohorts were 10_6 _ 4_5 and

11_2 _ 4_6 years, respectively (Table 1).

Children in both cohorts tended to reside in urbanized areas (60_5% and 56_0%), and the parents of slightly more than half of the children had white-collar jobs (55_5% and 57_3%). The mean frequencies of medical visits for the T1DM cohort and non-T1DM cohort were 24_0 _ 12_7 times per year and 11_1 _ 9_9 times per year, respectively. The T1DM cohort was more likely to have hypertension, hyperlipidaemia, depression, anxiety and obesity than the non-T1DM cohort. The mean follow-up times were 6_68 _ 3_99 years and 6_83 _ 3_98 years

in the T1DM and non-T1DM cohorts, respectively. Figure 1 shows the cumulative AD incidence curve for the two cohorts; the curve is significantly higher in the T1DM cohort than in the non-T1DM cohort (log-rank test, P = 0_01). The overall incidence rates of AD in the T1DM and non-T1DM cohorts were 3_31 and 2_35 per 1000 person years, respectively. After adjustment for sex, age, urbanization level, parental occupation, frequency of medical visits and comorbidities of asthma, rhinitis, hypertension, hyperlipidaemia, depression, anxiety and obesity, the T1DM cohort showed an increased risk of AD compared with the non-T1DM cohort [adjusted HR (aHR) 1_76, 95% CI 1_29–2_39)] (Table 2). Children with T1DM had a significantly increased risk of AD vs. children without T1DM, for both girls (aHR 1_66, 95% CI 1_09–2_52) and boys (aHR 1_89, 95% CI 1_19–3_02). Among children

aged ≤ 8 years, children with T1DM had a significantly higher risk of AD than children without T1DM (aHR 1_68, 95% CI 1_09–2_60).

Compared with children without T1DM, those with T1DM who had the second (aHR 2_79, 95% CI 1_59–4_91) and third (aHR 2_54, 95% CI 1_19–5_44) highest urbanization levels had a significantly higher risk of AD. Analyses specific to parental occupation revealed that the T1DM cohort had a significantly higher risk of AD than the non-T1DM cohort for

parental white-collar jobs (aHR 1_99, 95% CI 1_31–3_04). When stratified by comorbidities, children with T1DM who had no comorbidity had significantly higher risk of AD than those without T1DM (aHR 2_83, 95% CI 1_85–4_34). The

incidence of AD decreased with follow-up time, and the aHR of AD was significantly higher in the first follow-up year

(aHR 2_47, 95% CI 1_22–5_01).

Table 3 shows the joint effect of asthma and rhinitis on the risk of AD. Relative to the patients without asthma and without rhinitis, those with both asthma and with rhinitis were at

a higher risk of AD (aHR 1_45, 95% CI 1_03–2_03). Furthermore, as shown in Table 2, children with T1DM with no

comorbidity had a significantly higher risk of AD than those without T1DM. Therefore, we can exclude the influence of asthma and rhinitis in both groups.

Table 4 shows the HRs of AD risk associated with the frequency of ER visits and hospitalizations in patients with

T1DM. Compared with children without T1DM, those with T1DM who had more than two ER visits for their diabetes had a higher aHR of 30_1 (95% CI 18_7–48_5) for AD. Meanwhile, those in the T1DM cohort who had been hospitalized

more than twice for their diabetes had an aHR of 70_3 (95% CI 45_6–114_5) for AD.

Discussion

To the best of our knowledge, this is the first nationwide, population-based, retrospective cohort study to demonstrate an increased risk of AD among children with T1DM, particularly in children younger than 8 years without comorbidity,

Several epidemiological investigations have examined the

association between T1DM and atopic disease; however, conflicting results have been obtained. Therefore, Cardwell et al.

performed a meta-analysis by conducting a systematic review of published epidemiological studies, and their study revealed

a small but considerable reduction in the prevalence of asthma in children with T1DM, but not for other atopic diseases.3

However, most of these published epidemiological studies were case–control studies that relied on patient questionnaires. Some objective laboratory investigations have reported that IgE levels, specific IgE antibodies to allergens and skin prick tests in patients with T1DM are similar to those in healthy controls, and that autoimmune Th1 diseases such as T1DM can coexist with atopic diseases with a Th2-mediated response, suggesting that the Th1/Th2 paradigm is oversimplified.10,11,14 The new

paradigm indicates that additional lymphocyte subsets, such as Th17 cells and regulatory T cells (Tregs), may be involved, because Th17 cells exert a substantial positive effect on both atopy and autoimmunity, whereas Tregs exert an inhibitory effect.2,15

In contrast to published reports of an inverse association

between T1DM and AD, our results showed a positive relationship between T1DM and AD. This phenomenon can be

explained by Th1 cells mediating inflammation during the chronic stages of AD,16 which is associated with the extent of

the immediate hypersensitivity in a skin test to an allergen in children with T1DM, suggesting that Th1 cytokine secretion may be either pro- or anti-inflammatory in the same autoimmune disease.17,18

We observed a significantly higher risk of subsequent AD among children with T1DM than in healthy children, particularly in those younger than 8 years, and when the follow-up

time was < 1 year. The results also revealed that children with T1DM without comorbidities had a higher risk of AD than those without T1DM, excluding the potential influence of comorbidity on subsequent AD occurrence. AD commonly begins in infancy or early childhood. It affects predominantly infants and young children. In addition, AD is generally considered

to be one of the first manifestations in the atopic

march. When children develop, children with AD have higher risks of developing allergic rhinitis and asthma. The International Study of Asthma and Allergies in Childhood reported

that the prevalence of AD was 5–20% in children aged 6–7 and 13–14 years.19 Illi et al. reported that nearly half of cases

of early-onset AD within the first 2 years of life were in complete remission by age 3 years.20 These facts may explain the

increased incidence densities of AD for subjects aged ≤ 8 years compared with subjects aged > 8 years in both cohorts, and the greater risk of AD for those aged ≤ 8 years in the T1DM cohort compared with the non-T1DM cohort.

The incidence and risk of AD were higher within the first year of T1DM diagnosis in the T1DM cohort than in the control cohort. We adjusted for the frequency of T1DM-related

medical visits to remove surveillance bias. This finding may imply that common genetic factors, environmental triggers and immunological mechanisms contribute to the development of both disorders. In this study, city districts and townships where subjects were registered were grouped into four urbanization levels based on population density (people per km2). Level 1 indicates the most urbanized areas, and level 4

indicates the least urbanized areas. Allergy is a multifactorial disease, and multiple environmental factors may contribute to the development of AD. For example, access to medical care, a more allergenic home environment, smaller family size, more outdoor pollution and allergen exposure, increasing diversity

of foods consumed, modernized housing and increased industrialization and urbanization are all potential contributing factors

to AD.21,22 In this study, children with T1DM in the

second and third highest urbanization levels had a significantly higher risk of AD. However, we did not find a dose-dependent effect across urbanization-level strata in either group.

This suggests that subsequent factors associated with urbanization level (population density) may also influence the incidence

of AD.

Furthermore, children with T1DM with more ER visits or hospitalizations for diabetes had a higher risk of subsequent

AD than those with fewer visits or hospitalization. This is possibly

because patients with poor blood glucose control, whose pancreatic islet beta cells are diminished such that the inflammation

has not yet completely resolved, may go against the Th1/Th2 paradigm.23 Also, patients with T1DM are more

prone to develop frequent skin infections, leading to skin barrier dysfunction; thus, they have a higher risk of subsequent

AD.24

Another explanation for T1DM possibly increasing the risk of AD is that both diseases have the same risk factor of obesity. Verbeeten et al. recently reported a considerable association between childhood obesity, body mass index or

percentage weight for height, and increased risk of T1DM.25

Furthermore, Silverberg et al. reported prolonged obesity in early childhood as a risk factor for AD.26 However, in our

study, we can exclude obesity as a cofounding factor because the T1DM cohort showed an increased risk of AD compared with the non-T1DM cohort after adjustment for comorbidities including obesity (Table 2).

The main strength of this study is that it used a large population database with minimal selection bias to show the association between T1DM and the risk of subsequent AD. The NHIRD covers a highly representative sample of Taiwan’s general population because the reimbursement policy is universal and operated by a single buyer, the government in Taiwan. All insurance claims should be scrutinized by

medical reimbursement specialists and peer reviewed according to the standard diagnostic criteria in the study. Therefore, the diagnoses of AD and T1DM in this study were

highly reliable.

Our study has some limitations. Firstly, the diagnosis of

T1DM and AD was based on ICD codes, hence detailed laboratory data and clinical information were lacking, including

genetic/environmental data, body mass index, skin prick test, radioallergosorbent test, IgE level, specific IgE and severity of AD. Indeed, the diagnosis of AD depends mainly on clinical history, not on laboratory data, and we consider that the physician in Taiwan can make the diagnosis of AD accurately.

Secondly, regardless of the large sample sizes, the incidence of AD was relatively small at the end of follow-up in this study. Small incidence limits the statistical analyses, particularly regarding covariate adjustments and subgroup analyses, such

as for patients with and without several ER visits and hospitalizations. However, the strong association seen in this study

sends an important message to populations in other areas with a high prevalence of AD. Thirdly, the T1DM cohort had a

greater history of asthma, rhinitis, hypertension, hyperlipidaemia, depression and anxiety. Their average number of

clinic visits was higher than in the non-T1DM cohort. This might cause detection bias because the atopic manifestations could be more easily diagnosed due to frequent medical visits and examinations. Fourthly, the rates of T1DM and AD might

have been underestimated because the NHI data excluded selfpaying patients, and AD may be missed in patients with mild

symptoms who do not see a doctor.

In conclusion, our study indicates that T1DM may increase the risk of subsequent AD, and that this risk is higher especially in those younger than 8 years, and when the follow-up

time is < 1 year. Further study of other ethnic populations and future research to explore the role of Th1-mediated T1DM contributing to the development of Th2-mediated AD is warranted.