Abstract
Background
Associations of filaggrin (FLG) variants with asthma and rhinitis have been shown to be modulated by eczema status. However, it is unknown whether allergic sensitization status modifies this association. The aim of this study was to determine whether FLG variants need eczema and/or allergic sensitization as a necessary component to execute its adverse effect on coexisting and subsequent asthma and rhinitis.
Methods
Repeated measurements of asthma, rhinitis, eczema, and allergic sensitization (documented by skin prick tests) at ages 1, 2, 4, 10, and 18 years were ascertained in the Isle of Wight birth cohort (n = 1,456). FLG haploinsufficiency was defined as having at least the minor allele of R501X, 2282del4, or S3247X variants. Log binomial regression models were used to test associations and statistical interactions.
Results
FLG variants increased the risk of asthma (RR = 1.39, 95% CI: 1.06 – 1.80) and rhinitis (RR = 1.37, 95% CI: 1.16 – 1.63). In delayed effect models, ‘FLG variants plus allergic sensitization’ and ‘FLG variants plus eczema’ increased the risk of subsequent asthma by 4.93-fold (95% CI: 3.61 – 6.71) and 3.33-fold (95% CI: 2.45 – 4.51), respectively, during the first 18 years of life. In contrast, neither eczema nor allergic sensitization in combination with FLG variants increased the risk of later rhinitis.
Conclusions
Allergic sensitization and eczema modulated the association between FLG variants and asthma, but not rhinitis. Results of our study imply that the mechanisms and pathways through which FLG variants predispose to increased risk of asthma and rhinitis may be different.
Keywords: filaggrin variants, asthma, rhinitis, allergic sensitization, eczema, interaction
Introduction
The burden associated with allergic disorders, including eczema, asthma, and rhinitis is substantial and of public health importance. Prior investigations indicated that a complex interplay between genetic, immunological, and environmental factors contributes to the development and maintenance of such allergic manifestations. Until recently, etiologic research focused on the role of immune dysregulation in the pathogenesis of allergic disorders; however, a shift in the research paradigm towards understanding the contribution of a defective epidermal barrier rapidly emerged after the discovery of loss-of-function variants in the filaggrin gene (FLG) [1–3]. Filaggrin haploinsufficiency, defined as partial or complete loss of filaggrin (filament-aggregating protein) protein, is associated with the development of an impaired epidermal barrier that is characterized by increased allergen penetration and water permeability [4].
Thus far, FLG variants are the most replicated and strongest genetic risk factor for eczema [5]. Also, FLG variants are considered to be associated with asthma and rhinitis [6]. Of importance is that the association between FLG variants and asthma is stronger in the presence of eczema [5,6]. However, this possible effect modification by eczema was not demonstrated for the association between FLG variants and rhinitis [6]. Mechanisms that underlie the association of FLG variants with asthma and rhinitis are not well understood since FLG is not expressed in the upper or lower airway epithelium [7,8]. A proposed pathway is that cutaneous sensitization, facilitated by FLG variants, may lead to local and systematic inflammation at distant organs (i.e., lung and nasal tissues) [2,3].
While the concept that eczema leads to asthma in children who have a loss-of-function variants in the FLG gene is attractive, research in this area has neglected the role of allergic sensitization, defined as the propensity to produce immunoglobulin E (IgE) antibodies responses to environmental and food antigens [9]. Only one study based on the German Multicenter Allergy Study (MAS) birth cohort addressed allergic sensitization and reported an interaction between FLG variants and food sensitization in the pathogenesis of asthma among children with eczema [10]. Hence, further investigations on whether the associations of FLG variants with asthma and rhinitis are modified by eczema and/or allergic sensitization are needed so that preventive efforts can be directed to either clinical management of eczema or allergic sensitization or both.
We hypothesized that allergic sensitization, rather than eczema status, modifies the association of FLG variants with asthma and rhinitis. Therefore, statistical interactions (FLG variants × allergic sensitization and FLG variants × eczema) were tested to determine whether FLG variants plus allergic sensitization and/or FLG variants plus eczema jointly increased the risk of asthma and/or rhinitis. To this end, analyzing longitudinal data from the Isle of Wight birth cohort covering childhood and adolescence prospectively enabled us to determine whether FLG variants need allergic sensitization or eczema as a necessary component to execute its adverse effect on coexisting and subsequent asthma and rhinitis.
Methods
Study Design and Participants
An unselected whole population birth cohort (n = 1,536) was recruited in 1989 in the Isle of Wight, UK, to prospectively study the natural history of allergic conditions. After exclusion of adoptions, perinatal deaths, and refusal for follow-up, 1,456 (95%) children were enrolled, with follow-up assessments conducted at 1, 2, 4, 10, and 18 years of age. Ethics approvals were obtained from the Isle of Wight Local Research Ethics Committee (NRES Committee South Central – Southampton B) at recruitment and for the subsequent follow-ups (06/Q1701/34). At each follow-up, validated questionnaires, including the International Study of Asthma and Allergy in Childhood (ISAAC) questionnaire [11], were completed on allergic disorders plus demographic attributes and exposures to environmental factors.
Phenotypes
In all assessments of the Isle of Wight birth cohort, eczema was defined as chronic or chronically relapsing, itchy dermatitis lasting more than 6 weeks with characteristic morphology and distribution [12], following Hanifin and Rajka criteria [13]. For asthma, at the 1, 2, and 4-year follow-ups, the medical investigator determined the presence of asthma based on wheeze frequency over the last 12 months and treatment given for asthma or asthma related symptoms. At the 10 and 18 year follow-ups, asthma was defined as having “ever had asthma” and either “wheezing or whistling in the chest in the last 12 months” or “current treatment for asthma”, using ISAAC questionnaire [11]. Rhinitis was defined by a positive response to: ‘In the past 12 months have you had a problem with sneezing, or a runny or a blocked nose when you did not have a cold or the flu?’ [14] Since the 1-year and 2-year follow-up data on eczema, asthma, and rhinitis were collected in a relatively small time window, we combined them for analytic purposes (reported as 1-or-2 years).
To determine allergic sensitization status, skin prick testing (SPT) at ages 1 and 2 years was performed on children with any symptoms of eczema, asthma, or rhinitis. We combined SPT results for ages 1 and 2 years since they occurred within a short time period and will henceforth refer to this as SPT at 1-or-2 years. At 4, 10, and 18 years, regardless of symptoms, SPT was performed on most children attending the research center to a standard battery of common allergens (ALK-Albello, Horsholm, Denmark). Inhalant allergens tested were house dust mite, cat, dog, Alternaria alternata, Cladosporium herbarium, grass pollen mix, and tree pollen mix. Food allergens tested were cows’ milk, soya, hens’ egg, peanut and cod. Positive and negative controls were included. Allergic sensitisation was defined by having a positive SPT to at least one allergen test with mean wheal diameter of 3 mm greater than the negative control. Since allergic sensitization is a dynamic rather than a completely stable phenotype, we used the concurrent and/or preceding status and thus allowed the risk to change over time.
FLG Genotyping
Blood and/or saliva samples were collected at ages 10 and/or 18 years from which genomic DNA was isolated. DNA samples were interrogated using GoldenGate Genotyping Assays (Illumina, Inc, SanDiego, CA) on the BeadXpressVeracode platform (Illumina, Inc, SanDiego, CA) per Illumina’s protocol. Individuals carrying the minor allele for at least one of the FLG variants R501X, 2282del4, or S3247X were classified as having filaggrin haploinsufficiency. The R2447X variant was genotyped too, but none of the study participants carried the minor allele. Detailed information on genotyping is provided by Ziyab et al. [15].
Statistical Analysis
Since the proposed effect modifiers (i.e., allergic sensitization and eczema) and the two outcomes (i.e., asthma and rhinitis) were measured repeatedly at ages 1-or-2, 4, 10, and 18 years, we applied the generalized estimating equation (GEE) method with first-order autoregressive covariance matrix to account for the correlated observations and the within-child effect of the repeated measurements [16]. Risk ratios (RR) and their 95% confidence intervals (95% CIs) were estimated by applying log-binomial regression models using the GENMOD procedure in SAS 9.3 (SAS, Gary, NC, USA). To determine whether the RRs of the association of FLG variants with asthma and rhinitis differ according to the presence or absences of the proposed effect modifiers, we evaluated the association in the total study sample and sub-samples based on the status of the possible effect modifiers.
Interaction terms, on a multiplicative scale, were used to test the additional effect of two co-occurring risk factors on the health outcomes above and beyond their individual effects. Also, to evaluate whether the effect of FLG variants across levels of the effect modifiers were statistically significantly different, we included interaction terms in separate regression models (model 1: FLG variants × eczema and model 2: FLG variants × allergic sensitization). To decipher which interactive effect was more pronounced, a regression model including both interaction terms (model 3: FLG variants × eczema and FLG variants × allergic sensitization) was evaluated. Henceforth, we refer to models 1, 2, and 3 as the ‘concurrent models’ since the effect modifiers and the outcomes coexisted at the same time. In the case of a possibly statistically significant interaction term (Pinteraction < 0.1), the “combined effect”, referring to the joint impact of two individual risk factors plus their interaction on the occurrence of the outcome, was estimated. We repeated the previous analyses while stratifying by sex to determine if sex-specific effects existed. In all GEE models, sex and age at follow-up were included as potential confounders.
In addition to the concurrent models, we tested the interactions in ‘delayed effect’ models to investigate whether the time-order of risk and response supports the concurrent model findings. The delayed effect analysis helps to determine whether the interaction effects contribute to the developmental process of asthma and rhinitis. To this end, we analyzed whether FLG variants interact with preceding allergic sensitization or eczema on the risk of subsequent asthma and rhinitis. For example, the interaction of FLG variants with allergic sensitization status at age 1-or-2 years on the risk of asthma at age 4 years (FLG variants × SPT results at 1-or-2 years → asthma at 4 years) was evaluated. The prevalence of the clinical phenotype (e.g., asthma at 4), which was used in the delayed effect models, included new occurrences and persistence of the disease. Therefore, we constructed three delayed periods (1-or-2 to 4 years, 4 to 10 years, and 10 to 18 years) to determine if the presence of both FLG variants and preceding allergic sensitization or preceding eczema influences the positive transition, defined as the change in disease status from disease-free to diseased, and the persistence of the disease in two consecutive follow-ups [17].
Results
Description of Study Population
Of the 1,456 children enrolled in the study, 1,377 were followed-up at age 1-or-2 years, 1,214 at 4 years, 1,368 at 10 years, and 1,309 at 18 years. The period prevalence ranged between 11.9% to 14.2% for eczema, 14.3% to 17.7% for asthma, and 5.6% to 35.8% for rhinitis (table 1). At age 1-or-2 years, SPTs were performed on symptomatic children (n = 515), of which 20.6% had at least one positive SPT response. At all other ages, the majority of participates underwent SPTs with the proportion of positive SPTs being 19.4% at 4 years, 26.9% at 10 years, and 41.4% at 18 years of age. Genotype frequencies of FLG variants (table 1) were concordant with Hardy-Weinberg equilibrium.
Table 1.
Characteristics of study population
| Attributes | % (n/total) |
|---|---|
| Sex | |
| Male | 51.2 (786/1536) |
| Eczema at | |
| 1-or-2 years | 14.2 (196/1377) |
| 4 years | 11.9 (145/1214) |
| 10 years | 13.7 (186/1359) |
| 18 years | 12.3 (161/1307) |
| Asthma at | |
| 1-or-2 years | 14.3 (197/1377) |
| 4 years | 14.9 (181/1214) |
| 10 years | 14.7 (201/1368) |
| 18 years | 17.7 (231/1305) |
| Rhinitis at | |
| 1-or-2 years | 15.8 (217/1377) |
| 4 years | 5.6 (65/1214) |
| 10 years | 22.6 (308/1362) |
| 18 years | 35.8 (468/1309) |
| FLG variants* | |
| R501X | 4.1 (47/1161) |
| 2282del4 | 4.6 (54/1168) |
| S3247X | 1.6 (18/1165) |
| Combined | 10.3 (118/1150) |
| Allergic sensitization† at | |
| 1-or-2 years | 20.6 (106/515) |
| 4 years | 19.6 (192/982) |
| 10 years | 26.9 (279/1036) |
| 18 years | 41.4 (353/853) |
Analyses were conducted using the combined carrier frequency of 10.3%.
Presence of allergic sensitization was defined by having on or more positive skin prick test result at the respective assessment.
FLG Variants and Asthma
In the total study sample, FLG variants increased the risk of repeated occurrence of asthma from 1-or-2 to 18 years of age (RR = 1.39, 95% CI: 1.06 – 1.80; fig. 1). In the presence of eczema, the effect size increased (RR = 1.56, 95% CI: 1.10 – 2.21); in the absence of eczema, the magnitude of the association between FLG variants and asthma was reduced and lost statistical significance (RR = 1.26, 95% CI: 0.93 – 1.69). This association was noticeably modified when stratifying based on the presence (RR = 1.49, 95% CI: 1.14 – 1.94) or absence (RR = 0.76, 95% CI: 0.48 – 1.21) of allergic sensitization.
Fig. 1.
Concurrent association of FLG variants with asthma in total study sample and sub-samples: longitudinal analysis covering 1-or-2 to 18 years of age. The prevalence of asthma among those with filaggrin loss-of-function (FLG LOF) variants and those with wild-type genotype (FLG WT) is shown. RR: Risk ratio; 95% CI: 95% confidence interval; k= number of repeated measurements.
To determine whether the observed differences in the associations were statistically significant, statistical interactions on a multiplicative scale were evaluated in concurrent models. There was no evidence for an interaction between FLG variants and eczema on the risk for asthma (interaction effect RR = 1.18, 95% CI: 0.78 – 1.79, Pinteraction = 0.429; fig. 1). However, a statistically significant interaction was found between FLG variants and allergic sensitization on the risk for asthma (interaction effect RR = 2.00, 95% CI: 1.22 – 3.28, Pinteraction = 0.006; fig. 1). The combined effect of FLG variants and allergic sensitization increased the risk of asthma by 3.63-fold (95% CI: 2.81 – 4.70). Furthermore, to determine which interactive effect is more important, we simultaneously included both interaction terms in one regression model. Results showed that the interaction between FLG variants and allergic sensitization remained statistically significant after adjusting for the interactive effect between FLG variants and eczema (data not shown). However, the latter did not gain statistical significance.
We retested the results of the aforementioned concurrent models, using delayed effect models. In these models we used the preceding cumulative SPTs results (or eczema status), i.e. all SPTs results (or eczema status) from assessments that are prior to the outcome assessment, when testing the interaction between FLG variants and allergic sensitization (or eczema) on the risk of subsequent asthma. Repeated measurements analysis demonstrated a statistically significant interaction between FLG variants and preceding allergic sensitization on the risk of subsequent asthma (RRinteraction = 1.58, Pinteraction = 0.013). The results of the delayed effect models show that the combined effect of preceding allergic sensitization with FLG variants increased the risk of asthma in the next exam at each delay period (fig. 2a). For instance, the presence of both FLG variants plus positive SPT at 1-or-2 and/or 4 years increased the risk of asthma at 10 years (RR = 7.22, 95% CI: 4.84 – 10.80). Although results of the concurrent model (fig. 1) did not show an interaction between FLG variants and eczema, the delayed effect models demonstrated interactive effects between FLG variants and preceding eczema on the risk of subsequent asthma (repeated measurements analysis: RRinteraction = 1.96, Pinteraction = 0.006; fig. 2b). The repeated measurement analysis shows that the combined effect of FLG variants and preceding eczema, on average, increased the risk of later asthma during the first 18 years of life (RR = 3.33, 95% CI: 2.45 – 4.51).
Fig. 2.
a) Delayed effect models exploring the individual and combined effects of FLG variants and preceding allergic sensitization status (documented by skin prick tests (SPTs)) on asthma development. Independent effects of FLG variants (SPT − & FLG LOF), positive SPT (SPT + & FLG WT), and their combined effect (SPT + & FLG LOF) in the development of asthma are shown for different ages. We modeled the effect of preceding SPT results on later asthma. For instance, having at least one positive SPT between 1-or-2 to 10 years was used as a predictor of asthma at 18 years. b) Delayed effect models exploring the individual and combined effects of FLG variants and preceding eczema status on asthma development. Independent effects of FLG variants (Eczema − & FLG LOF), eczema (Eczema + & FLG WT), and their combined effect (Eczema + & FLG LOF) in the development of asthma are shown for different ages. We modeled the effect of preceding eczema results on later asthma. For instance, having at least one positive eczema diagnosis between 1-or-2 to 10 years was used as a predictor of asthma at 18 years. SPT +: positive SPT, SPT −: negative SPT; Eczema +: positive eczema diagnosis, Eczema −: negative eczema diagnosis; FLG LOF: filaggrin loss-of-function; FLG WT: filaggrin wild-type genotype; n = number of children; k= number of repeated measurements. The asterisk (*) and the double dagger (‡) indicate that the interaction effect (a: SPT × FLG; b: Eczema × FLG) is possibly statistically significantly different from the null value of 1 (*: PInteraction < 0.1; ‡: PInteraction < 0.05); suggesting interaction on multiplicative scale. The dagger (†) indicates that the risk estimate (RR) of the given column is statistically significantly (P < 0.05) different from the column without both risk factors (a: SPT − & FLG WT; b: Eczema & FLG WT). The presence of both an asterisk (*) or double dagger (‡) and dagger (†) signifies that the combined effect of the two risk factors is above and beyond their individuals effects.
To further understand whether the observed interactions between FLG variants and preceding allergic sensitization and eczema influence new occurrence and/or persistence of asthma, we calculated the proportions of positive transitions (defined as the change in disease status from disease-free to diseased in two consecutive follow-ups) and persistence among those with and without both risk factors (table 2). The proportion of positive transition was more frequent among those with FLG variants and allergic sensitization as compared to those without both risk factors across all transition periods. For instance, in the presence of both FLG variants and allergic sensitization, 5/16 (31.3%) of those without asthma at age 4 years developed asthma at age 10 years; whereas, only 23/532 (4.3%) of those without both risk factors developed asthma at 10 years. Furthermore, the presence of both risk factors influenced the persistence of asthma (table 2). For example, 18/21 (85.7%) of participants with both risk factors who had asthma at ages 10 years continued to have asthma at 18 years, as compared to 25/44 (56.8%) among those without the two risk factors. Similarly, the presence of both FLG variants and preceding eczema influenced the positive transition and persistence of asthma.
Table 2.
Prevalence, positive transition, and persistence of asthma stratified based on FLG variants and preceding allergic sensitization and eczema status: comparing proportions of those with both risk factors to those without the risk factors
| Asthma % (n/total) | |||
|---|---|---|---|
| Prevalence | Age 4 years | Age 10 years | Age 18 years |
| FLG WT & SPT − | 18.4 (57/310) | 7.9 (47/595) | 10.3 (62/604) |
| FLG LOF & SPT + | 50.0 (9/18) | 60.0 (18/30) | 54.8 (23/42) |
| p-value* | 0.003 | < 0.001 | < 0.001 |
| FLG WT & Eczema − | 11.9 (86/725) | 11.2 (91/811) | 16.0 (118/739) |
| FLG LOF & Eczema + | 52.0 (5/6) | 51.5 (17/33) | 51.3 (20/39) |
| p-value* | < 0.001 | < 0.001 | < 0.001 |
| Positive transition† | 1-or-2 to 4 years | 4 to 10 years | 10 to 18 years |
| FLG WT & SPT − | 8.5 (19/223) | 4.3 (23/532) | 6.7 (37/550) |
| FLG LOF & SPT + | 50.0 (6/12) | 31.3 (5/16) | 23.8 (5/21) |
| p-value* | < 0.001 | < 0.001 | 0.014 |
| FLG WT & Eczema − | 8.5 (53/635) | 7.0 (44/632) | 9.3 (59/639) |
| FLG LOF & Eczema + | 42.1 (8/19)) | 29.4 (5/17) | 14.3 (3/21) |
| p-value* | < 0.001 | 0.006 | 0.437 |
| Persistence‡ | 1-or-2 to 4 years | 4 to 10 years | 10 to 18 years |
| FLG WT & SPT − | 43.7 (38/87) | 38.7 (24/62) | 56.8 (25/44) |
| FLG LOF & SPT + | 50.0 (3/6) | 92.9 (13/14) | 85.7 (18/21) |
| p-value* | 0.542 | < 0.001 | 0.026 |
| FLG WT & Eczema − | 36.7 (33/90) | 44.7 (38/85) | 68.3 (56/82) |
| FLG LOF & Eczema + | 83.3 (5/6) | 80.0 (12/15) | 94.4 (17/18) |
| p-value* | 0.031 | 0.023 | 0.037 |
FLG LOF: filaggrin loss-of-function; FLG WT: filaggrin wild-type genotype; SPT −: negative skin prick test result; SPT +: positive skin prick test result.
p-values comparing the proportions across the two exposure groups were derived from fisher’s exact test.
Positive transition refers to the change in disease status from asthma-free to asthma in two consecutive assessments.
Persistence refers to the proportion of individuals who had asthma at two consecutive assessments
FLG Variants and Rhinitis
FLG variants were associated with increased risk for rhinitis from age 1-or-2 to 18 years in the total study sample of repeated measurements (RR = 1.37, 95% CI: 1.16 – 1.63; fig. 4). In both the presence (RR = 1.57, 95% CI: 1.23 – 2.00) and absence (RR = 1.28, 95% CI: 1.04 – 1.57) of eczema, the association between FLG variants and rhinitis was statistically significant. However, we observed a statistically significant association between FLG variants and rhinitis only in the presence of allergic sensitization (RR = 1.34, 95% CI: 1.17 – 1.53), and not in its absence (RR = 1.13, 95% CI: 0.82 – 1.55; fig. 3).
Fig. 4.
a) Delayed effect models exploring the individual and combined effects of FLG variants and preceding allergic sensitization status (documented by skin prick tests (SPTs)) on rhinitis development. Independent effects of FLG variants (SPT − & FLG LOF), positive SPT (SPT + & FLG WT), and their combined effect (SPT + & FLG LOF) in the development of rhinitis are shown for different ages. We modeled the effect of preceding SPT results on later rhinitis. For instance, having at least one positive SPT between 1-or-2 to 10 years was used as a predictor of rhinitis at 18 years. b) Delayed effect models exploring the individual and combined effects of FLG variants and preceding eczema status on rhinitis development. Independent effects of FLG variants (Eczema − & FLG LOF), eczema (Eczema + & FLG WT), and their combined effect (Eczema + & FLG LOF) in the development of rhinitis are shown for different ages. We modeled the effect of preceding eczema results on later rhinitis. For instance, having at least one positive eczema diagnosis between 1-or-2 to 10 years was used as a predictor of rhinitis at 18 years. SPT +: positive SPT, SPT −: negative SPT; Eczema +: positive eczema diagnosis, Eczema −: negative eczema diagnosis; FLG LOF: filaggrin loss-of-function; FLG WT: filaggrin wild-type genotype; n = number of children; k= number of repeated measurements. The dagger (†) indicates that the risk estimate (RR) of the given column is statistically significantly (P < 0.05) different from the column without both risk factors (a: SPT − & FLG WT; b: Eczema & FLG WT).
Fig. 3.
Concurrent association of FLG variants with rhinitis in total study sample and sub-samples: longitudinal analysis covering 1-or-2 to 18 years of age. The prevalence of rhinitis among those with filaggrin loss-of-function (FLG LOF) variants and those with wild-type genotype (FLG WT) is shown. RR: Risk ratio; 95% CI: 95% confidence interval; k= number of repeated measurements.
To test if the observed heterogeneous effects across the levels of the potential effect modifiers are statistically different, we tested multiplicative statistical interactions in concurrent models. There was no evidence for an interactive effect neither between FLG variants and eczema (interaction effect RR = 1.15, 95% CI: 0.83 – 1.60, Pinteraction = 0.409; fig. 3) nor between FLG variants and allergic sensitization (interaction effect RR = 1.07, 95% CI: 0.76 – 1.51, Pinteraction = 0.698; fig. 3) on the risk for rhinitis. Concordant with results of concurrent models, the delayed effect models did not show any interactions between FLG variants and preceding allergic sensitization or eczema on the risk of subsequent rhinitis (fig. 4).
Sex Differences and Time Trends
Separate analyses for each sex were performed and the results for boys and girls were similar (data not shown) and in agreement with the results of analyzing both sexes together in the repeated measurement (GEE) analysis. Furthermore, at each follow-up we analyzed boys and girls separately to determine whether the effects differ across both time and sex. The obtained results indicated that neither age of exam nor sex influenced our results.
Discussion
The objective of this study was to determine whether eczema and/or allergic sensitization status act as effect modifiers for the association of FLG variants with asthma and rhinitis. In this study, FLG variants increased the risk of asthma and rhinitis during the first 18 years of life in the total study sample. In concurrent effect models, interaction between FLG variants and allergic sensitization resulted in a combined effect that increased the risk of coexisting asthma by 3.63-fold. Delayed effect models, which take the time order of risk factors and disease occurrences into account, supported the findings of the concurrent effect models regarding allergic sensitization and FLG variants. In addition, the delayed effect models showed that FLG variants interact with preceding eczema status on the development of subsequent asthma, indicating that both allergic sensitization and eczema act as effect modifiers for the association between FLG variants and asthma. In contrast, neither eczema nor allergic sensitization modified the association of FLG variants with rhinitis in a statistically significant manner.
The association between FLG variants and asthma among those with eczema was first reported by Palmer et al. [1]. Subsequently, several studies have replicated this association and added a possible association between FLG variants and rhinitis [5,6]. Results of our study further support the general agreement that FLG variants are associated with asthma in the presence of eczema. We further demonstrated that both preceding allergic sensitization and eczema modified the association between FLG variant and asthma through multiplicative interaction (fig. 2). These findings indicate that FLG variants need eczema or allergic sensitization to execute their adverse effects on asthma. The finding of an interaction between FLG variants and allergic sensitization in this study improves our understanding of the possible link between genetics of the epidermal barrier (i.e., FLG variants) and a respiratory disorder (i.e., asthma). Such an observation further supports the hypothesis that cutaneous sensitization priming, facilitated by FLG variants, may migrate to the airways and cause local and systematic inflammation [18]. Although prior studies have widely suggested such a pathway, the majority of previous investigations did not take the status of allergic sensitization into account. An exception is the study by Marenholz et al., which showed an interaction between FLG variants and food sensitization on the risk of asthma in children with eczema [10], however, their analytical sample was restricted to children with eczema.
In the delayed effect models we used the prevalence of asthma as the outcome, which does not distinguish between new occurrences and persistence of asthma. Therefore, in additional descriptive analysis, we demonstrated that the presence of both FLG variants and preceding allergic sensitization or preceding eczema influenced the new occurrences (positive transition) and persistence of asthma. Hence, our data suggest that the presence of both risk factors (‘FLG variants and preceding allergic sensitization’ or ‘FLG variants and preceding eczema’) plays an important role in the development and persistence of asthma.
Attempts using the composite of ‘atopic asthma’ phenotype were conducted to investigate whether allergic sensitization is the link between FLG variants and asthma. Inconsistent results for the association of FLG variants with ‘atopic asthma’ have been reported [19–22]. These inconsistencies could be attributed to the use of the composite ‘atopic asthma’ phenotype, which does not clearly define the reference (no risk) group. By taking this approach investigators tend to group ‘non-atopic asthmatic’ and ‘atopic without asthma’ participants together in the reference group. Therefore, when assessing the association of FLG variants with ‘atopic asthma’ one should be aware that the reference group could include atopic participants, which could result in distorting assessments. The advantage of our approach, using two separate risk factors, over the composite ‘atopic asthma’ phenotyping, is that we were able to estimate the additional effect due to interaction (FLG variants × allergic sensitization) that is above and beyond their independent effects. Hence, allowing us to determine whether the combined effect of both risk factors is more than just the multiplication of their independent effects.
There is conflicting evidence in the literature as to whether the association between FLG variants and rhinitis is modulated by eczema status. The majority of previous studies found a stronger association between FLG variants and rhinitis in the presence of coexisting eczema when compared to their association in the absence of concurrent eczema [6,20,23–25]. Another investigation reported an association between FLG variants and ‘persistent allergic rhinitis’ phenotype [21]. Controversially, however, a significant association has also been reported between FLG variants and the composite ‘allergic rhinitis’ phenotype adjusted for eczema status [19]. Our analyses did not reveal interaction between eczema (Pinteraction = 0.409) or allergic sensitization (Pinteraction = 0.698) with FLG variants on the risk of rhinitis in the repeated measurement analysis. Our results of an interactive effect between FLG variants and allergic sensitization on the risk of asthma and the lack of such interaction on the risk of rhinitis suggest that the pathway through which FLG variants predispose to asthma and rhinitis may be different. For instance, the expression of filaggrin protein was not detected in the human bronchial epithelium; however, it has been shown that filaggrin is expressed in the nasal vestibule [7,19].
Major strengths of our prospective, 18-year study are the repeated phenotyping, objective assessments of allergic sensitization status, and the low loss to follow-up (ranged from 5% to 17%). Moreover, the majority (80%) of the study participants were genotyped for FLG variants common among populations of European ancestry [26]. We showed, previously, that the genotyped study participants did not differ from the total cohort with regard to multiple characteristics [15]. Hence, there is no indication of selection bias that could pose a threat to the validity of our study. Misclassification of eczema cases is minimal since a high proportion of subjects showed typical manifestation of eczema in the usual locations (antecubital or popliteal fossae, ankles, face or neck for 97% at 1 year, 91% at 2 years, 75% at 4 years, 86% at 10 years and 76% at 18 years) [27]. To ensure reproducibility of results, the data was randomly divided into two equal parts and analyses were repeated (see supplementary material). Results of this post-hoc confirmatory step showed substantial consistency in the results with regard to interaction effect sizes across the two sub-samples and the total analytical sample.
Potential limitations are the definition of asthma and rhinitis symptoms in early life. Our asthma conclusion at ages 10 and 18 years followed the ISAAC criteria [11], which was at that time not available for assessments at age 1, 2, and 4 years. Although slightly different methods were used to define asthma, in a previous report we have shown that the minor change in asthma definition over time did not influence the validity of our asthma classification [17]. In addition, since it is difficult to differentiate between infectious and other forms of rhinitis in infancy, the elevated prevalence at 1-or-2 years might have been influenced by misclassifying ‘viral induced infectious rhinitis’ [14]. However, by applying repeated measurements analysis including assessments later in childhood and adolescence, we believe that the influence of the possible misclassification of rhinitis at 1-or-2 years on the overall results of the study is minimal, since similar results were obtained when we reran our analyses excluding the 1-or-2 year data. Moreover, since SPTs at age 1-or-2 years were performed on children with any symptom, we could show when excluding the 1-or-2 year results that such a selection resulting in underestimation of RRs might have been induced. Although proportions of SPT positivity at ages 1-or-2 years (20.6%) and 4 years (19.6%; SPT performed irrespective of symptoms) were similar, RRs associating allergic sensitization at age 1-or-2 years with outcomes (i.e., eczema, asthma, and rhinitis) at the same age tended to be smaller than RRs relating allergic sensitization at age 4 years with outcomes at the same age (data not shown), since also the reference group had some symptoms. Hence, the estimated RRs at age 1-or-2 might be underestimated; however, the extent of the possible selection bias is minimal since our results did not noticeably change when we excluded the 1-or-2 years follow-up data from the analyses.
The use of a slightly relaxed statistical significance threshold (PInteraction < 0.1) for detecting interaction-effects is further supported by the fact that majority of epidemiologic studies lack sufficient power to detect higher order-terms [28]. However, applying repeated measurements analyses improved the statistical power for detecting interaction effects. For instance, in delayed effect models with separate analyses for the three delayed periods (1-or-2 years to 4 years, 4 to 10 years, and 10 to 18 years), p-values for interaction terms between FLG variants and eczema on the risk of subsequent asthma were between 0.05 and 0.1; however, results of repeated measurements models combining the three delayed periods yielded interaction terms with p-values that are highly statistically significant (e.g., FLG variants × preceding eczema: Pinteraction = 0.006). Whereas testing the interaction effects in three delayed effect periods might have identified false-positive interaction effects due to a relaxed p-value of 0.1; the statistical significance observed in the repeated measurements analysis reduces the possibility of finding false-positive results.
In conclusion, our results indicate that allergic sensitization and eczema status are effect modifiers of the association between FLG variants and asthma. The combined effects of ‘preceding allergic sensitization and FLG variants’ and ‘preceding eczema and FLG variants’ increased the risk of subsequent asthma by 4.93-fold and 3.33-fold, respectively, during the first 18 years of life. These findings suggest that FLG variants need allergic sensitization or eczema to execute its adverse effects on asthma. In contrast, neither eczema nor allergic sensitization status statistically significantly modulated the association between FLG variants and rhinitis. Hence, results of our study suggest that the mechanisms and pathways through which FLG variants, representing impaired epidermal barrier, predispose to increased risk of asthma and rhinitis may be different. Future studies confirming our reported observations and exploring the differential etiological pathways underlying the development of asthma and rhinitis are needed to better identify and stratify those who share similar risk characteristics.
Supplementary Material
Acknowledgments
We thank participants and their families who have helped us with this project over the last two decades. We would like to acknowledge the help of all the staff at The David Hide Asthma and Allergy Research Centre in undertaking the 18 year and previous assessments of 1989 Isle of Wight birth cohort. The 18 year assessment of the IOW birth cohort and future analyses were funded by grants from the National Institute of Health, USA (R01 HL082925 and R01AI091905).
Contributor Information
Ali H. Ziyab, Email: ali_ziyab@hotmail.com.
Wilfried Karmaus, Email: karmaus1@memphis.edu.
Hongmei Zhang, Email: hzhang6@memphis.edu.
John W. Holloway, Email: j.w.holloway@soton.ac.uk.
Susan E. Steck, Email: stecks@mailbox.sc.edu.
Susan Ewart, Email: ewart@cvm.msu.edu.
Syed Hasan Arshad, Email: S.H.Arshad@soton.ac.uk.
References
- 1.Palmer CN, Irvine AD, Terron-Kwiatkowski A, Zhao Y, Liao H, Lee SP, Goudie DR, Sandilands A, Campbell LE, Smith FJ, O’Regan GM, Watson RM, Cecil JE, Bale SJ, Compton JG, DiGiovanna JJ, Fleckman P, Lewis-Jones S, Arseculeratne G, Sergeant A, Munro CS, El Houate B, McElreavey K, Halkjaer LB, Bisgaard H, Mukhopadhyay S, McLean WH. Common loss-of-function variants of the epidermal barrier protein filaggrin are a major predisposing factor for atopic dermatitis. Nat Genet. 2006;38:441–446. doi: 10.1038/ng1767. [DOI] [PubMed] [Google Scholar]
- 2.Irvine AD, McLean WH, Leung DY. Filaggrin mutations associated with skin and allergic diseases. N Engl J Med. 2011;365:1315–1327. doi: 10.1056/NEJMra1011040. [DOI] [PubMed] [Google Scholar]
- 3.O’Regan GM, Irvine AD. The role of filaggrin in the atopic diathesis. Clin Exp Allergy. 2010;40:965–972. doi: 10.1111/j.1365-2222.2010.03522.x. [DOI] [PubMed] [Google Scholar]
- 4.Brown SJ, McLean WH. One remarkable molecule: Filaggrin. J Invest Dermatol. 2012;132:751–762. doi: 10.1038/jid.2011.393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Rodriguez E, Baurecht H, Herberich E, Wagenpfeil S, Brown SJ, Cordell HJ, Irvine AD, Weidinger S. Meta-analysis of filaggrin polymorphisms in eczema and asthma: Robust risk factors in atopic disease. J Allergy Clin Immunol. 2009;123:1361–1370. e1367. doi: 10.1016/j.jaci.2009.03.036. [DOI] [PubMed] [Google Scholar]
- 6.van den Oord RA, Sheikh A. Filaggrin gene defects and risk of developing allergic sensitisation and allergic disorders: Systematic review and meta-analysis. BMJ. 2009;339:b2433. doi: 10.1136/bmj.b2433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.De Benedetto A, Qualia CM, Baroody FM, Beck LA. Filaggrin expression in oral, nasal, and esophageal mucosa. J Invest Dermatol. 2008;128:1594–1597. doi: 10.1038/sj.jid.5701208. [DOI] [PubMed] [Google Scholar]
- 8.Ying S, Meng Q, Corrigan CJ, Lee TH. Lack of filaggrin expression in the human bronchial mucosa. J Allergy Clin Immunol. 2006;118:1386–1388. doi: 10.1016/j.jaci.2006.08.030. [DOI] [PubMed] [Google Scholar]
- 9.Busse WW. The role of allergy in disease. Immunol Rev. 2011;242:5–9. doi: 10.1111/j.1600-065X.2011.01036.x. [DOI] [PubMed] [Google Scholar]
- 10.Marenholz I, Kerscher T, Bauerfeind A, Esparza-Gordillo J, Nickel R, Keil T, Lau S, Rohde K, Wahn U, Lee YA. An interaction between filaggrin mutations and early food sensitization improves the prediction of childhood asthma. J Allergy Clin Immunol. 2009;123:911–916. doi: 10.1016/j.jaci.2009.01.051. [DOI] [PubMed] [Google Scholar]
- 11.Asher MI, Keil U, Anderson HR, Beasley R, Crane J, Martinez F, Mitchell EA, Pearce N, Sibbald B, Stewart AW, et al. International study of asthma and allergies in childhood (isaac): Rationale and methods. The European respiratory journal : official journal of the European Society for Clinical Respiratory Physiology. 1995;8:483–491. doi: 10.1183/09031936.95.08030483. [DOI] [PubMed] [Google Scholar]
- 12.Arshad SH, Karmaus W, Kurukulaaratchy R, Sadeghnejad A, Huebner M, Ewart S. Polymorphisms in the interleukin 13 and gata binding protein 3 genes and the development of eczema during childhood. The British journal of dermatology. 2008;158:1315–1322. doi: 10.1111/j.1365-2133.2008.08565.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hanifin JM, Rajka G. Diagnostic features of atopic dermatitis. Acta Derm Venereol Suppl (Stockh) 1980;92:44–47. [Google Scholar]
- 14.Kurukulaaratchy RJ, Karmaus W, Raza A, Matthews S, Roberts G, Arshad SH. The influence of gender and atopy on the natural history of rhinitis in the first 18 years of life. Clin Exp Allergy. 2011;41:851–859. doi: 10.1111/j.1365-2222.2011.03765.x. [DOI] [PubMed] [Google Scholar]
- 15.Ziyab AH, Karmaus W, Yousefi M, Ewart S, Schauberger E, Holloway JW, Zhang H, Arshad SH. Interplay of filaggrin loss-of-function variants, allergic sensitization, and eczema in a longitudinal study covering infancy to 18 years of age. PloS one. 2012;7:e32721. doi: 10.1371/journal.pone.0032721. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986;42:121–130. [PubMed] [Google Scholar]
- 17.Soto-Ramirez N, Ziyab AH, Karmaus W, Zhang H, Kurukulaaratchy RJ, Ewart S, Arshad SH. Epidemiologic methods of assessing asthma and wheezing episodes in longitudinal studies: Measures of change and stability. J Epidemiol. 2013;23:399–410. doi: 10.2188/jea.JE20120201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Heimall J, Spergel JM. Filaggrin mutations and atopy: Consequences for future therapeutics. Expert review of clinical immunology. 2012;8:189–197. doi: 10.1586/eci.11.100. [DOI] [PubMed] [Google Scholar]
- 19.Weidinger S, O’Sullivan M, Illig T, Baurecht H, Depner M, Rodriguez E, Ruether A, Klopp N, Vogelberg C, Weiland SK, McLean WH, von Mutius E, Irvine AD, Kabesch M. Filaggrin mutations, atopic eczema, hay fever, and asthma in children. J Allergy Clin Immunol. 2008;121:1203–1209. e1201. doi: 10.1016/j.jaci.2008.02.014. [DOI] [PubMed] [Google Scholar]
- 20.Henderson J, Northstone K, Lee SP, Liao H, Zhao Y, Pembrey M, Mukhopadhyay S, Smith GD, Palmer CN, McLean WH, Irvine AD. The burden of disease associated with filaggrin mutations: A population-based, longitudinal birth cohort study. J Allergy Clin Immunol. 2008;121:872–877. e879. doi: 10.1016/j.jaci.2008.01.026. [DOI] [PubMed] [Google Scholar]
- 21.Poninska J, Samolinski B, Tomaszewska A, Raciborski F, Samel-Kowalik P, Walkiewicz A, Lipiec A, Piekarska B, Komorowski J, Krzych-Falta E, Namyslowski A, Borowicz J, Kostrzewa G, Majewski S, Ploski R. Filaggrin gene defects are independent risk factors for atopic asthma in a polish population: A study in ecap cohort. PloS one. 2011;6:e16933. doi: 10.1371/journal.pone.0016933. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Li M, Chen X, Chen R, Bao Y, Yao Z. Filaggrin gene mutations are associated with independent atopic asthma in chinese patients. Allergy. 2011;66:1616–1617. doi: 10.1111/j.1398-9995.2011.02713.x. [DOI] [PubMed] [Google Scholar]
- 23.Marenholz I, Nickel R, Ruschendorf F, Schulz F, Esparza-Gordillo J, Kerscher T, Gruber C, Lau S, Worm M, Keil T, Kurek M, Zaluga E, Wahn U, Lee YA. Filaggrin loss-of-function mutations predispose to phenotypes involved in the atopic march. J Allergy Clin Immunol. 2006;118:866–871. doi: 10.1016/j.jaci.2006.07.026. [DOI] [PubMed] [Google Scholar]
- 24.Ekelund E, Lieden A, Link J, Lee SP, D’Amato M, Palmer CN, Kockum I, Bradley M. Loss-of-function variants of the filaggrin gene are associated with atopic eczema and associated phenotypes in swedish families. Acta dermato-venereologica. 2008;88:15–19. doi: 10.2340/00015555-0383. [DOI] [PubMed] [Google Scholar]
- 25.Schuttelaar ML, Kerkhof M, Jonkman MF, Koppelman GH, Brunekreef B, de Jongste JC, Wijga A, McLean WH, Postma DS. Filaggrin mutations in the onset of eczema, sensitization, asthma, hay fever and the interaction with cat exposure. Allergy. 2009;64:1758–1765. doi: 10.1111/j.1398-9995.2009.02080.x. [DOI] [PubMed] [Google Scholar]
- 26.Akiyama M. Flg mutations in ichthyosis vulgaris and atopic eczema: Spectrum of mutations and population genetics. The British journal of dermatology. 2010;162:472–477. doi: 10.1111/j.1365-2133.2009.09582.x. [DOI] [PubMed] [Google Scholar]
- 27.Ziyab AH, Raza A, Karmaus W, Tongue N, Zhang H, Matthews S, Arshad SH, Roberts G. Trends in eczema in the first 18 years of life: Results from the isle of wight 1989 birth cohort study. Clin Exp Allergy. 2010;40:1776–1784. doi: 10.1111/j.1365-2222.2010.03633.x. [DOI] [PubMed] [Google Scholar]
- 28.Greenland S. Interactions in epidemiology: Relevance, identification, and estimation. Epidemiology. 2009;20:14–17. doi: 10.1097/EDE.0b013e318193e7b5. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.