The Association With Cardiovascular Disease and Type 2 Diabetes in Adults With Atopic Dermatitis

A Systematic Review and Meta-analysis

J.P. Thyssen; A.-S. Halling-Overgaard; Y.M.F. Andersen; G. Gislason; L. Skov; A. Egeberg


The British Journal of Dermatology. 2018;178(6):1272-1279. 

In This Article

Materials and Methods

Inclusion and Exclusion Criteria

No protocol was registered prior to the beginning of the study (although a study protocol was developed in Danish). A priori, all studies investigating the association between AD and type 2 diabetes, stroke, hypertension, myocardial infarction, angina, coronary artery disease, heart failure and cardiovascular death were included in the systemic review. Studies on the association between AD and bacterial endocarditis were excluded. Studies were also excluded if there was no reference group comprising patients or individuals without AD, as were studies that specifically stated that type 1 diabetes was examined. As the vast majority of diabetes cases in adults are type 2, studies that examined diabetes as outcome were considered to be mostly type 2 diabetes cases. The search process is presented in Figure 1.

Figure 1.

Flow diagram.

Literature Search

We searched the medical databases PubMed, Embase and Web of Science using the following search term: '(atopic dermatitis or atopic eczema) AND (cardiovascular or heart or myocardial or stroke or hypertension or ischaemia or angina or diabetes or mortality or death)'. We chose not to use the broader terms 'eczema' or 'dermatitis' as these are nonspecific for AD. The search included articles in any language, from database inception to August 2017, and only studies in adults (age ≥ 18 years) were included. Reference lists of key articles and reviews were also screened.

Data Extraction

Two of the authors (J.P.T. and A.E.) independently screened all titles and/or abstracts prior to retrieving full–text articles. If data were duplicated in more than one study, the study with the most comprehensive data was included. Study quality in the quantitative analysis was assessed using the Newcastle–Ottawa scale (NOS).[11] Each study was awarded up to 10 points according to the adapted NOS.

Statistical Analysis

We chose to perform a meta–analysis on both crude data and adjusted data from included articles. For the meta–analysis on crude data, we only included studies that listed the number of total patients with AD, patients with AD with the examined outcome, total reference individuals and reference individuals with the examined outcome, respectively. For the meta–analysis on adjusted data, we took the fully adjusted estimates from the relevant articles. If multiple adjusted models were presented in an article, we chose the model that included smoking.

Statistical analyses were performed using StatsDirect version 3·0·192 (StatsDirect Ltd, Cheshire, U.K.). Odds ratios (ORs) with 95% confidence intervals (CIs) of cardiometabolic end points in individuals with AD compared with reference individuals were estimated. Specifically, these included myocardial infarction, diabetes, hypertension, angina pectoris and stroke. For each end point category the heterogeneity of included studies was assessed using Cochran's Q and I 2 statistics, and forest plots were constructed. The Cochran's Q is calculated as the weighted sum of squared differences between individual study effects and the pooled effect across studies, whereas the I 2 statistic describes the percentage of variation across studies that is due to heterogeneity rather than chance. Random effects models with DerSimonian–Laird methods were utilized for all analyses, and between–studies heterogeneity was found in all studied end–point categories. Consequently, for all analyses, P–values were < 0·05 for all Cochran's Q statistics, and all I 2 values were > 75%. Funnel plot progression of logarithmic ORs and standard errors were used to assess for publication bias. All statistical tests were two–sided, with a significance level of < 0·05.