A-C, FLG-based comparisons with error bars. In FLG patients with homozygotes and compound heterozygotes, a lower proportion reported skin clearance (A) and a greater proportion reported steroid use (B) and frequent steroid use (C) compared with those with fewer than 2 loss-of-function alleles. D-E, TSLP-based comparisons with error bars. In rs1898671 homozygotes, a lower proportion reported TCI use (D) and frequent TCI use (E) compared with wildtype and heterozygous patients. TCI indicates topical calcineurin inhibitor.
In patients that were no longer using pimecrolimus and tacrolimus, those with the thymic stromal lymphopoietin rs1898671 single-nucleotide polymorphism were less likely to be using other treatments compared with those without the gene variant.
eTable 1. Likelihood of Self-Reported Outcomes Based on Patient FLG Genotype.
eTable 2. Likelihood of Self-Reported Outcomes Based on Patient TSLP Genotype.
Chang J, Mitra N, Hoffstad O, Margolis DJ. Association of Filaggrin Loss of Function and Thymic Stromal Lymphopoietin Variation With Treatment Use in Pediatric Atopic Dermatitis. JAMA Dermatol. Published online November 30, 2016. doi:10.1001/jamadermatol.2016.4467
In atopic dermatitis, do variations in FLG and TSLP genotype correspond to differences in treatment use over time?
In this cohort study of 842 children with atopic dermatitis, those with 2 FLG LOF alleles were less likely to experience skin clearance and more likely to use steroids, and TSLPrs1898671 homozygotes were less likely to use topical calcineurin inhibitors. Among all patients that had discontinued topical calcineurin inhibitors, those with the rs1898671 single-nucleotide polymorphism were more likely to have stopped all other treatment as well.
Treatment use and likely effectiveness was associated with genetic variation.
Atopic dermatitis (AD) is a common illness of childhood.
To determine whether variations in FLG and TSLP genotype corresponded to differences in treatment use over time.
Design, Setting, and Participants
This prospective cohort study recruited and followed a volunteer sample of 842 children enrolled in the Pediatric Eczema Elective Registry who provided saliva samples for DNA extraction for 10 years. Eligibility criteria included age 2 to 17 years, AD diagnosis without cancer, and prior pimecrolimus use. Participants were followed for an average of 7.6 years (approximately 6396 person-years); 138 patients (16.4%) had no missing data over 10 years of follow-up.
Evaluation of FLG and TSLP genotypes.
Main Outcomes and Measures
Self-reported outcomes of whether a child’s AD required the use of topical steroids, topical calcineurin inhibitors, or other medications within the past 6 months at 6-month intervals.
Overall, 842 children (mean [SD] age, 1.9 [2.7] years; 438 girls) were included in this study. Treatment use among patients with 0, 1, or 2 FLG loss of function (LOF) alleles was compared as well as those that were wildtype, heterozygous, or homozygous for the TSLPrs1898671 single-nucleotide polymorphism. Patients with 2 FLG LOF alleles were less likely to report skin clearance (odds ratio [OR], 0.20; 95% CI, 0.07-0.55) and more likely to use steroids (OR, 5.04; 95% CI, 1.91-13.31). TSLPrs1898671 homozygotes were less likely to report topical calcineurin inhibitor use (OR, 0.16; 95% CI, 0.06-0.42), and among all patients that had discontinued topical calcineurin inhibitors, those with the rs1898671 single-nucleotide polymorphism were more likely to have stopped all other treatment as well (OR, 0.45; 95% CI, 0.26-0.76). In all but 1 of our comparisons, no significant difference between wildtype and heterozygous patients were found.
Conclusions and Relevance
Treatment use and likely effectiveness was associated with genetic variation. Variation was limited to children with 2 FLG LOF alleles or TSLPrs1898671 homozygotes, with no significant difference observed between wildtype and heterozygous patients in the majority of the outcomes studied. Therefore, the key differentiating factor in our analyses was the number of FLG LOF alleles or TSLP SNPs rather than the absolute presence or absence of these variants. This may be an important consideration for future studies.
Atopic dermatitis (AD) is a chronic inflammatory skin disorder that commonly affects children, with the majority of cases presenting within the first few years of life1; AD pathogenesis likely involves a combination of genetic and environmental factors, particularly those related to barrier dysfunction and immune dysregulation.2 Multiple therapeutic options are available for AD, with topical steroids and calcineurin inhibitors serving as first-line and second-line treatments, respectively.3
Filaggrin is an epidermal peptide that plays a key role in maintaining the skin barrier. Loss-of-function (LOF) mutations in the FLG gene have been associated with both a risk of developing and the persistence of AD, and intragenic copy number variation within FLG has been shown to have a dose-dependent effect on AD risk.4,5 Similarly, genetic variants of the IL-7-like cytokine thymic stromal lymphopoietin (TSLP) have also demonstrated an association with AD severity and persistence.6 In humans, TSLP is an activator of CD11c+ dendritic cells and thereby triggers TH2-mediated allergic inflammatory responses. High levels of TSLP expression have been observed in AD skin lesions.7 In addition, variants of the TSLP gene appear to either promote or protect against the cutaneous inflammation of AD, perhaps a result of corresponding increases or decreases in TSLP protein activity. In patients with AD, the rs1898671TSLP single-nucleotide polymorphism (SNP) has been associated with less persistent disease8 and a decreased risk of developing eczema herpeticum.6
Using the Pediatric Eczema Elective Registry (PEER) as a source of longitudinal patient data, we sought to determine whether variation in FLG and TSLP genotype had an impact on self-reported treatment usage over time. Differences in these outcomes might then suggest underlying differences in treatment response across the FLG and TSLP groups. In this study, we compare patients with 0, 1, or 2 FLG LOF alleles, as well as those that are wildtype, heterozygous, or homozygous for the TSLPrs1898671 SNP.
Information regarding the PEER study population has been previously described.8- 11 The PEER is part of a US Food and Drug Administration postmarketing study evaluating the long-term safety of 1% pimecrolimus cream, a topical calcineurin inhibitor (TCI), for the treatment of AD. Eligibility criteria included: age from 2 to 17 years, AD diagnosis (without cancer), and prior pimecrolimus use (at least 42 of the past 180 days before enrollment). Written informed consent including, when appropriate, assent was obtained for all individuals as approved by the institutional review board of the University of Pennsylvania. All patients provided a saliva sample for DNA extraction. It was previously shown that no clinically important differences exist between participants who did and did not provide DNA samples.12 Survey data were collected through biannual questionnaires, typically completed by the parent of the enrolled participant. Race/ethnicity was assessed as a demographic factor using US census categories and was defined and classified by participants.
FLG genotypic assessments were completed for the 4 most common LOF mutations: R501X (rs61816761), 2282del4, R2447X, and S3247X.9 With multiple LOF variants assayed, as is common in the literature, FLG LOF status was jointly categorized as having any FLG LOF variant.5,9 Individuals could be wildtype, heterozygous for a single LOF allele, or grouped together if they were homozygous for a single LOF allele or heterozygous for 2 LOF alleles (compound heterozygous). Similarly, assessments of TSLP variant status were carried out with Illumina GoldenGate Genotype Assays (Illumina).8 With only 1 locus involved, individuals could be wildtype, heterozygous, or homozygous for the variant.
Initially, 5 self-reported outcomes were measured from survey responses to symptom-based and treatment-based questions: (1) skin clearance (“Has your child’s rash cleared completely at any time during the last 6 months?”); (2) TCI use (“Have you used Elidel [Valeant] cream to treat this child’s eczema during the last 6 months?” and “Have you used Protopic [Astellas] ointment to treat this child’s eczema during the last 6 months?”); (3) frequency of TCI use (“How often was Elidel cream used during the last 6 months?” and “How often was Protopic ointment used during the last 6 months?”); (4) steroid use (“During the last 6 months, have you used any prescription or over-the-counter topical corticosteroids?”); and (5) frequency of steroid use (“How often did you use topical corticosteroids during the last 6 months?”). Topical calcineurin inhibitor use was measured as any use of either pimecrolimus (Elidel) or tacrolimus (Protopic). For questions concerning the frequency of treatment use, patients chose from 4 responses: none, infrequent (<1.5 months in total), frequent (between 1.5 and 5 months in total), and continuous (daily or almost daily). Frequency of treatment use was then converted to a binary outcome (frequent treatment use) by coding “none” and “infrequent” responses as 0 and “frequent” and “continuous” responses as 1. In patients with different responses for frequent pimecrolimus and tacrolimus use, frequent TCI use was measured as the greater (more frequent) of the 2 values.
We then evaluated the subcohort of children who had discontinued TCI use. Specifically, we evaluated their use of any other treatments over time. This outcome was measured as any use of steroids, prescription creams or ointments (“Have you used a prescription cream or ointment to treat this child’s itchy rash/eczema during the last 6 months?”), or other prescription medications (“Have you used any other prescription medications for the treatment of this child’s eczema during the last 6 months?”) after stopping TCIs.
Demographic and other subject-level data were described as percentages or means as appropriate and presented according to FLG LOF and TSLP variant status. For each variable, differences among genotypes at enrollment were assessed using χ2 or ANOVA as appropriate. Drug usage was described by proportions and displayed graphically. To analyze variation in our outcomes over time, mixed-effects logistic regression were used. We used mixed-effects models to account for intrasubject correlation due to repeated measurements over time. We specified an independence working correlation for these models. While in the past we based our gene variant analyses on additive models, here we chose a genetic model based on grouping FLG wildtype and heterozygous individuals vs homozygotes and compound heterozygotes (Figure 1). Visit number was also included as a covariate in our models. All analyses were conducted using Stata, version 13.1 (StataCorp).
Of the 842 children with FLG genotype data, 770 were also genotyped for TSLPrs1898671 data (Table 1 and Table 2). With respect to our cohort of 842 patients, 52.0% were female (n = 438); 43.6% were black (n = 367); the mean (SD) ages at AD onset and PEER enrollment were 1.9 (2.7) and 7.2 (3.8), respectively; participants were monitored for a mean (SD) 7.6 (2.7) years or approximately 6396 person-years; at enrollment, 382 patients (45.4%) reported any period of skin clearance, 529 (62.8%) reported steroid use, and 294 (34.9%) reported frequent steroid use in the prior 6 months. Similar demographics were observed in the subset with TSLPrs1898671 data. In accordance with the PEER inclusion criteria, all patients were using TCIs (specifically, pimecrolimus) at the beginning of the study.
We used mixed-effects models to assess the relationship between certain self-reported outcomes and patients’ FLG and TSLP genotypes. In each, we grouped wildtype and heterozygous patients together and compared them against homozygous and compound heterozygous patients, thereby assessing the entire cohort for a given outcome using a single model. This grouping method is supported by our initial set of analyses, in which wildtype and heterozygous patients were statistically indistinguishable, and all significant differences were limited to comparisons involving homozygotes and/or compound heterozygotes (eTables 1 and 2 in the Supplement). Table 3 presents results from the simplified models, and Figure 1 plots the corresponding data. FLG LOF homozygotes and compound heterozygotes were less likely to report any period of skin clearance (odds ratio [OR], 0.20; 95% CI, 0.07-0.55) and more likely to report steroid use (OR, 5.04; 95% CI, 1.91-13.31) and frequent steroid use (OR, 3.18; 95% CI, 1.22-8.30). Patients with TSLPrs1898671 homozygotes were less likely to report TCI use (OR, 0.16; 95% CI, 0.06-0.42) and frequent TCI use (OR, 0.40; 95% CI, 0.19-0.84).
Next, we determined the proportion of those that had discontinued TCIs and were still using other treatments. Mixed-effects models revealed a significant difference in the use of other treatments when comparing those with and without the rs1898671TSLP variant (ie, wildtype vs heterozygotes or homozygotes) but not when comparing heterozygotes and homozygotes (eTable 2 in the Supplement). Therefore, we investigated a simpler model, this time grouping together rs1898671 heterozygotes and homozygotes. Figure 2 plots the corresponding data. In patients that had discontinued TCIs, those with the rs1898671 SNP were also less likely to be using any other treatments (OR, 0.45; 95% CI, 0.26-0.76).
To our knowledge, this is the first study specifically investigating the relationship between treatment use and patient FLG and TSLP genotype status in AD. FLG LOF homozygotes and compound heterozygotes were less likely to report any instance of skin clearance within the previous 6 months and more likely to use steroids for the treatment of their eczema. Though rs1898671 status did not affect rates of self-reported skin clearance in this cohort, we were able to demonstrate a lower rate of TCI use in rs1898671 homozygotes. In addition, among patients that had stopped using TCIs, those with the rs1898671 variant were less likely to be using any other forms of treatment. In nearly all FLG-based and TSLP-based comparisons, no significant differences were identified between wildtype and heterozygous patients.
As previously mentioned, filaggrin has been implicated as a key molecule in the pathogenesis of AD, with LOF variants in the FLG gene associated with increased disease risk and persistence.4,5 The deleterious impact of these mutations is supported by our observations of decreased skin clearance and increased steroid usage (a higher-potency treatment option than TCIs) in FLG LOF homozygotes and compound heterozygotes. With essentially equal proportions of TCI use observed across all FLG genotypes, the lower rate of skin clearance in homozygotes and compound heterozygotes might be explained by a limited effectiveness of TCIs in this subset of patients because the skin barrier dysfunction brought about by these mutations may not be directly addressed by TCIs. In addition, thymic stromal lymphopoietin is also thought to play an important role in AD, with increased TSLP expression and proinflammatory activity seen in AD skin lesions.7 The TSLP variant rs1898671 is associated with a protective effect in patients and may be tied to reductions in TSLP protein activity.6,8 In our comparisons, patients with rs1898671 homozygotes were less likely to report TCI use, and among patients that had stopped using TCIs, those with the variant were more likely to have discontinued all other treatment as well. Given these observations, the decline in medication use is unlikely to reflect an ineffectiveness of any specific treatment but rather an overall lack of need for therapeutic interventions. Though rs1898671 status did not have an effect on skin clearance over time within this cohort, patients with the SNP may still have had milder overall disease and therefore been less likely to require medications to manage their AD. Finally, wildtype and heterozygous patients were statistically indistinguishable across all but 1 of our comparisons, with significant differences largely limited to comparisons involving homozygous (and, in FLG-based comparisons, compound heterozygous) patients. Thus, it may be important for future studies to consider not only the presence but also the number of FLG LOF alleles and TSLP variants in statistical comparisons. Had we failed to do so here, no relationship between patient genotype and treatment use would have been identified due to the small sample size of homozygotes and compound heterozygotes.
Data collection remains the foremost limitation of this study, as all measures were of survey responses typically provided by the patient’s parent or primary caregiver. Issues involving response accuracy and recall bias are relevant here, though the latter may not be as important owing to the longitudinal nature of the study and that none of the individuals filling out the survey forms had any knowledge of the patient’s genotype. In addition, survey responses regarding the use of steroids, other prescription creams, and other prescription medications did not address the specific drug used. Although we were therefore unable to account for factors such as steroid strength in our comparisons, we were primarily interested in overall treatment use and less concerned with patient responses to individual medications. Improper or inadequate treatment provision could also be raised as a potential concern, but if present, we would expect this to have an equal effect across all genotypic groups. Finally, though the differences we report here are statistically significant, the power of our findings is limited by the relatively small number of homozygotes and compound heterozygotes present in our sample population.
The concept of precision medicine, which is often discussed in the context of genetic variability, emphasizes the individualization of health care by accounting for the unique personal attributes of each patient. Here, we identified variations in patient genotype that corresponded to differences in skin clearance and treatment usage over time. Our results suggest a potential role for patient genotyping in assisting clinical decision-making in the treatment of AD.
Variations in FLG and TSLP genotype were associated with differences in self-reported skin clearance, TCI usage, and steroid usage. FLG LOF homozygotes and compound heterozygotes were less likely to report skin clearance and more likely to report steroid usage, while TSLPrs1898671 homozygotes were less likely to report TCI usage. In a comparison among those that had discontinued TCIs, those with the rs1898671 variant were more likely to be stopping all other forms of treatment as well. Together, these results suggest that TCIs may be less effective in FLG LOF homozygotes and compound heterozygotes and that the rs1898671 SNP may have a protective effect against AD. In addition, as the majority of these findings were obtained from comparisons involving homozygous (and, in FLG-based comparisons, compound heterozygous) patients, the number of LOF alleles or SNPs was a more important differentiating factor than the absolute presence or absence of these variants. Therefore, future studies should consider not only the presence but also the number of FLG LOF or TSLP variants in their analyses.
Corresponding Author: David J. Margolis, MD, PhD, Departments of Dermatology and Biostatistics and Epidemiology, University of Pennsylvania Perelman School of Medicine, 901 Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104 (firstname.lastname@example.org).
Published Online: November 30, 2016. doi:10.1001/jamadermatol.2016.4467
Author Contributions: Mr Chang and Dr Margolis had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Chang, Margolis.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Chang, Margolis.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Chang, Mitra, Margolis.
Administrative, technical, or material support: Ole, Margolis.
Conflict of Interest Disclosure: None reported.
Funding/Support: This study was funded by the National Institute of Arthritis Musculoskeletal and Skin Diseases (NIAMS) (grant R01-AR0056755) to Dr Margolis. The PEER study is funded by a grant from Valeant Pharmaceuticals.
Role of the Funder/Sponsor: The funder/sponsor 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 decision to submit the manuscript for publication.