Summary
Background
Childhood eczema is variable in onset and persistence.
Objectives
To identify eczema phenotypes during childhood, and their associations with early‐life environmental and genetic factors.
Methods
In this study of 5297 children from a multiethnic population‐based prospective cohort study, phenotypes based on parent‐reported physician‐diagnosed eczema from age 6 months to 10 years were identified using latent class growth analysis. Information on environmental factors was obtained using postal questionnaires. Four filaggrin mutations were genotyped and a risk score was calculated based on 30 genetic variants. Weighted adjusted multinomial models were used for association analyses.
Results
We identified the following five eczema phenotypes: never (76%), early transient (8%), mid‐transient (6%) and late transient (8%) and persistent eczema (2%). Early transient and persistent eczema were most common in first‐born children, those with a parental history of eczema, allergy or asthma and those with persistent wheezing [range of odds ratio (OR): 1.37, 95% confidence interval (CI) 1.07–1.74 and OR 3.38, 95%CI 1.95–5.85]. Early transient eczema was most common in male children only (OR 1·49, 95% CI 1·18–1·89). Children with late transient or persistent eczema were more often of Asian ethnicity (OR 2·04, 95% CI 1·14–3·65 and OR 3·08, 95% CI 1·34–7·10, respectively). Children with early, late transient and persistent eczema more often had a filaggrin mutation or additional risk alleles (range OR: 1.07, 95%CI 1.02–1.12 and OR 2.21, 95%CI 1.39–3.50). Eczema phenotypes were not associated with maternal education, breastfeeding, day care attendance and pet exposure.
Conclusions
Five eczema phenotypes were identified in a multiethnic paediatric population with limited differences in risk profiles, except for sex and ethnicity.
What's already known about this topic?
Two previous studies in longitudinal birth cohorts identified four and six different eczema phenotypes, predominantly in children of European ethnicity.
What does this study add?
Five eczema phenotypes were identified in a multiethnic paediatric population using latent class growth analysis.
Children with early transient and persistent eczema were most often first‐born children and had persistent wheezing, filaggrin mutation or additional risk alleles.
Previously known eczema risk factors had limited differentiating capabilities for eczema phenotypes, except for the association of early transient eczema with male children, and late transient and persistent eczema with Asian ethnicity.
Short abstract
https://doi.org/10.1111/bjd.18575 available online
Childhood eczema is a major common chronic health problem with a prevalence of up to 25%.1 The age of onset and the persistence of eczema during childhood vary. There is a need to define more detailed eczema phenotypes and to understand their specific underlying risk factors in order to make better predictions about the natural course of eczema and to prevent the onset and worsening of eczema. Defining eczema as a dichotomous trait is an oversimplification. Eczema phenotypes that take into account the age of onset and persistence over time may enable better identification of specific environmental exposures and genetic mechanisms that might play a role in the development of eczema.2, 3, 4 Previous studies suggest that higher maternal education, non‐European ethnicity, having older siblings, shorter duration or nonexclusivity of breastfeeding, day care attendance and no pet exposure are associated with an increased risk of childhood eczema.5, 6, 7, 8, 9, 10, 11 In addition, loss‐of‐function mutations in the gene encoding filaggrin (FLG), an indispensable protein for epidermal differentiation and maintenance of an optimal skin barrier, are well known to be associated with eczema.12 Furthermore, genome‐wide association (GWA) studies identified 31 variants that were associated with childhood eczema.13, 14, 15 It is unclear whether these early‐life exposures and genetic variants are related to the various eczema phenotypes. Two previous longitudinal birth cohorts identified different eczema phenotypes with sex, parental history of eczema, asthma or allergies, breastfeeding, pet exposure, FLG mutations, genetic risk score, asthma and other allergenic comorbidities as determinants.15, 16 However, those cohorts consisted predominantly of children of European ethnicity, did not take repeated measurements of eczema into account in the cluster analysis, or only explored the risk of early‐life exposures and genetic variants in unadjusted analyses.
Therefore, in order to predict and prevent the natural course of eczema, we aimed to identify eczema phenotypes in a multiethnic population‐based prospective cohort study among 5297 children. We further examined the associations of socioeconomic and lifestyle exposures in early‐life and genetic risk factors with the identified eczema phenotypes.
Patients and methods
Design
This study was embedded in the Generation R Study, a population‐based prospective cohort study from early fetal life onwards.17 Written informed consent was obtained from parents or legal guardians. Of 7893 live‐born children participating after birth, those without data on physician‐diagnosed eczema available for at least three time points were excluded, leaving a total of 5297 children for the current analyses.
Eczema definition
Information on eczema was obtained from parental questionnaires at age 6 months, and at ages 1, 2, 3, 4 and 10 years (response rates 72–76%). Physician‐diagnosed eczema was defined as a positive response to the question ‘Was your child diagnosed with eczema in the last 6 months/last year by a physician?’ (no/yes).7 This question was adapted from the core questionnaire of the International Study of Asthma and Allergies in Childhood (ISAAC).18
Early‐life environmental exposures
Information on parity (nulliparous/multiparous), maternal education (primary or secondary school/higher than secondary school), and parental history of eczema, allergy or asthma (no/ yes) was available from parental questionnaires obtained at enrolment. Information on the sex of the child was obtained from midwives and hospital records. Ethnic origin (European/non‐European) of the child was based on the parents’ country of birth according to Statistics Netherlands.17 Postnatal questionnaires provided information on breastfeeding (never/ever) at 2, 6 or 12 months after birth, pet exposure at ages 2 months and 6 months (no/yes, exposure to cat, dog, rodent or bird at home) and day care attendance at age 12 months (yes/no). Questionnaires adapted from the ISAAC study were used to determine wheezing at ages 1–6 years (no/yes).18 Wheezing patterns were classified based on time of onset and persistence as ‘never’, ‘early’ (wheezing at age ≤ 3 years only), ‘late’ (wheezing at ages > 3–6 years only), or ‘persistent wheezing’ (wheezing at age ≤ 3 years and at age > 3–6 years) for children with data on wheezing available for at least two time points.19
Genetic risk factors
The most prevalent FLG mutations in European populations (2282del4, R2447X, R501X and S3247X) were genotyped by modified Taqman allelic discrimination assays using previously described primers.20, 21 Children without any mutant alleles were classified as wild‐type. As we observed only two cases of homozygous FLG mutations, we created a combined FLG genotype (no mutation/one mutation or more).22 A recent and large GWA study identified and replicated 30 single‐nucleotide polymorphisms (SNPs) that were associated with childhood eczema.13 Information on these SNPs was available from the GWA screening performed on DNA isolated from cord blood leucocytes or, in a small minority of children with missing cord blood samples, at age 6 years using the Illumina 670K platform.17 Genotype data were imputed for all polymorphic SNPs into the 1000 Genomes panel. A genetic risk score for each individual was calculated by summation of the number of eczema‐increasing risk alleles (between 0 and 2 for each SNP) across all SNPs.13
Statistical analysis
Firstly, we compared characteristics of individuals included in our study with those who were not included by using independent sample t‐test, Mann–Whitney U‐test and Pearson's χ2‐tests. Secondly, eczema phenotypes were identified using latent class growth analysis based on parent‐reported physician‐diagnosed eczema data. This type of analysis assumes that a number of different latent classes exist in the study population, which describe the variation of observed responses over time, and clusters individuals with similar patterns while taking into account correlations between measurements from the same individual.23, 24, 25 Details on model selection are provided in the Supporting Information. For comparison with previous studies on eczema subgroups, we performed longitudinal latent class analysis to identify eczema phenotypes.15, 16 Thirdly, we examined the associations of early environmental exposures and genetic risk factors with the identified eczema phenotypes using weighted mutually adjusted multinomial regression models. Multiple imputation using chained equations was used to impute missing values of environmental exposures (range 0–30% per variable). Twenty completed datasets were created and the results were pooled using Rubin's rules.26 Physician‐diagnosed eczema, FLG genotype and the calculated genetic risk score were not imputed as they could not be appropriately predicted from the available data.27 Finally, in order to examine the associations between different ethnicities and eczema phenotypes in more detail, we divided ethnicity into European (European, American or Oceanian), Mediterranean (Turkish or Moroccan), Asian (Asian, Indonesian, Surinamese–Hindustani or Surinamese–mixed) and African (African, Dutch–Antillean or Surinamese–Creole) subgroups based on similarities in skin type and cultural background.11, 28 All measures of association are presented as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). Latent class analyses were performed using Mplus (version 7·11) for Windows (Muthén and Muthén, Los Angeles, CA, U.S.A.), imputation and weighted multinomial regression analyses were performed using the packages ‘mice’ (version 2·46·0)29 and ‘nnet’ (version 7·3·12) in R version 3·4·3,30 respectively.
Results
Parental and child characteristics
Maternal and child characteristics are presented in Table 1. The prevalence of eczema declined from 16% (n = 662) at age 6 months to 7% (n = 347) at age 10 years. Individuals who were not included in the current analyses had somewhat less favourable socioeconomic and environmental factors (Table S1; see Supporting Information).
Table 1.
Characteristics of children and their mothers after multiple imputation
| Individuals (N = 5297) | |
|---|---|
| Parental characteristics | |
| Parity (nulliparous) | 3096 (59) |
| Maternal education (higher) | 3009 (57) |
| History of eczema, allergy and asthma (yes, at least one parent) | 3189 (60) |
| Child characteristics | |
| Sex (male) | 2632 (50) |
| Ethnicity (non‐European) | 1359 (26) |
| Breastfeeding (ever) | 4885 (92) |
| Day care attendance (yes) | 3145 (59) |
| Pet exposure (yes) | 2038 (39) |
| Wheezing pattern | |
| Never | 2887 (55) |
| Early | 1495 (28) |
| Late | 261 (5) |
| Persistent | 655 (12) |
| FLG genotype (≥ 1 mutations)a | 247 (8) |
| Genetic risk score, mean ± SDa | 31 ± 3·45 |
| Physician‐diagnosed eczemaa | |
| 6 months | 662 (16) |
| 1 years | 637 (13) |
| 2 years | 719 (14) |
| 3 years | 442 (9) |
| 4 years | 378 (8) |
| 10 years | 347 (7) |
Data are presented as n (%) unless stated otherwise. Values are based on 20 imputed datasets. aData on FLG genotype, genetic risk score and physician‐diagnosed eczema were not imputed. Data were missing for FLG genotype [n = 2186 (41%)], genetic risk score [n = 1880 (36%)], and physician‐diagnosed eczema at 6 months [n = 1569 (30%)], 1 year [n = 730 (14%)], 2 years [n = 457 (9%)], 3 years [n = 726 (14%)], 4 years [n = 789 (15%)] and 10 years [n = 1418 (27%)].
Eczema phenotypes
In children with data on physician‐diagnosed eczema available from at least three time points (n = 5297), latent class growth analysis identified the model with five eczema phenotypes as the best fit (Fig. 1, Tables S2, S3; see Supporting Information). The five eczema phenotypes were described as never eczema (76%), early transient eczema (8%), mid‐transient eczema (6%), late transient eczema (8%) and persistent eczema (2%). Similar results were observed in children with data on physician‐diagnosed eczema available at all six time points (n = 1975) (Table S2; see Supporting Information). To compare our results with previous studies, we used longitudinal latent class analysis and identified that the model with three, not five, eczema phenotypes was the best fit for our data (Table S3, Fig. S1a, b; see Supporting Information). The three phenotypes were similar in pattern to the never, early transient and persistent eczema phenotypes identified using latent class growth analysis.
Figure 1.
Eczema phenotypes trajectories from latent class growth analysis.
Early environmental exposures and eczema phenotypes
No major differences in the magnitude or the direction of the effect estimates were observed between analyses with imputed data and analyses with complete cases only. Here, we present only the results based on imputed data. Nulliparity and parental history of eczema, allergy or asthma were positively associated with early transient and persistent eczema compared with never eczema and reference groups (OR range 1·37, 95% CI 1·07–1·74 and OR 2·01, 95% CI 1·20–3·36, respectively) (Table 2). Boys were significantly more likely to have early transient eczema (OR 1·49, 95% CI 1·18–1·89), while non‐European ethnicity was associated with late transient (OR 1·35, 95% CI 1·03–1·78) and persistent eczema (OR 1·74, 95% CI 1.09–2.79). Children with late‐onset wheezing had higher risks for early transient and persistent eczema (OR 2·65, 95% CI 1·67–4·20 and OR 3·63, 95% CI 1·71–7·71, respectively) than children classified as ‘never wheezing’ and ‘never eczema’. Children with persistent wheezing more often had early and late transient, and persistent eczema (OR 2·67, 95% CI 1·94–3·69; OR 1·89, 95% CI 1·36–2·65 and OR 3·50, 95% CI 2·03–6·04, respectively) than children classified as ‘never wheezing’ and ‘never eczema’. No other significant associations of early environmental exposures with eczema phenotypes were observed.
Table 2.
Associations of early environmental exposures with eczema phenotypes
| Early transient eczema, n = 439 | Mid‐transient eczema, n = 306 | Late transient eczema, n = 412 | Persistent eczema, n = 122 | |
|---|---|---|---|---|
| Environmental exposure model, n = 5297 | ||||
| Parity (nulliparous) | 1·37 (1·07–1·74) | 1·38 (0·98–1·95) | 1·10 (0·88–1·39) | 1·65 (1·06–2·55) |
| Maternal education (higher) | 1·04 (0·80–1·36) | 1·11 (0·76–1·61) | 1·04 (0·80–1·35) | 0·70 (0·44–1·12) |
| Parental history of eczema, allergy or asthma (yes) | 1·71 (1·29–2·25) | 1·14 (0·79–1·64) | 1·20 (0·93–1·55) | 2·01 (1·20–3·36) |
| Sex (male) | 1·49 (1·18–1·89) | 0·98 (0·71–1·35) | 0·83 (0·66–1·04) | 1·21 (0·80–1·83) |
| Ethnicity (non‐European) | 1·02 (0·76–1·36) | 1·04 (0·69–1·58) | 1·35 (1·03–1·78) | 1·74 (1·09–2·79) |
| Breastfeeding (ever) | 0·90 (0·58–1·39) | 0·83 (0·45–1·53) | 0·91 (0·59–1·40) | 0·69 (0·34–1·39) |
| Child day care(yes) | 1·06 (0·81–1·39) | 1·22 (0·82–1·80) | 1·21 (0·92–1·58) | 1·49 (0·88–2·50) |
| Pet exposure (yes) | 0·88 (0·68–1·14) | 0·82 (0·57–1·19) | 1·10 (0·85–1·42) | 0·67 (0·41–1·11) |
| Wheezing pattern (early) | 1·20 (0·88–1·62) | 1·05 (0·71–1·55) | 1·14 (0·85–1·52) | 0·87 (0·47–1·58) |
| Wheezing pattern (late) | 2·65 (1·67–4·20) | 1·23 (0·54–2·80) | 1·23 (0·68–2·25) | 3·63 (1·71–7·71) |
| Wheezing pattern (persistent) | 2·67 (1·94–3·69) | 1·27 (0·74–2·17) | 1·89 (1·36–2·65) | 3·50 (2·03–6·04) |
Values are pooled odds ratios with their 95% confidence intervals. All environmental exposure were entered simultaneously in the model. Reference groups are ‘never eczema’ phenotype group (n = 4018), and multiparous, primary education, no parental history of eczema, allergy or asthma, female sex, never breastfeeding, no day care attendance, no pet exposure or never wheezing groups. Bold values indicate statistical significance at the α = 0·05 level.
Genetic risk factors and eczema phenotypes
Children with one or more FLG mutations had increased risks of early and late transient eczema (OR 2·21, 95% CI 1·40–3·49 and OR 2·09, 95% CI 1·30–3·34, respectively) compared with children without FLG mutations and those classified as ‘never eczema’ (Table 3). For each additional risk allele in the genetic risk score, children had increased risks of early transient and persistent eczema (OR 1·08, 95% CI 1·03–1·13 and 1·09, 95% CI 1·01–1·18, respectively). The size and the direction of the effect estimates of the associations of genetic risk factors with eczema phenotypes remained similar when we additionally adjusted for all early environmental exposures. Also, the size and the direction of the effect estimates of the associations of early environmental exposures with eczema phenotypes remained similar, although some were attenuated to nonsignificant when we additionally adjusted for FLG genotype and the genetic risk score (Table S4; see Supporting Information).
Table 3.
The associations of genetic risk factors with eczema phenotypes and three separate sensitivity analysis investigating the associations between early environmental exposures, genetic risk factors and/or ethnicity and eczema phenotypes
| Early transient eczema | Mid‐transient eczema | Late transient eczema | Persistent eczema | |
|---|---|---|---|---|
| Genetic risk factor model, n = 2981 | n = 258 | n = 177 | n = 235 | n = 76 |
| FLG genotype (≥ 1 mutations) | 2·21 (1·40–3·49) | 1·52 (0·73–3·15) | 2·09 (1·30–3·34) | 1·68 (0·69–4·09) |
| Genetic risk score (per additional allele) | 1·08 (1·03–1·13) | 1·06 (0·99–1·13) | 1·02 (0·98–1·06) | 1·09 (1·01–1·18) |
| Environmental exposure model and genetic risk factors (i), n = 2981 | n = 258 | n = 177 | n = 235 | n = 76 |
| FLG genotype (≥ 1 mutations) | 2·21 (1·39–3·50) | 1·56 (0·75–3·24) | 2·02 (1·26–3·24) | 1·80 (0·73–4·47) |
| Genetic risk score (per additional allele) | 1·07 (1·02–1·12) | 1·05 (0·99–1·12) | 1·01 (0·97–1·06) | 1·09 (1·00–1·18) |
| Environmental exposure model in ethnic subgroups (ii), n = 5297 | n = 439 | n = 306 | n = 412 | n = 122 |
| Ethnicity (Mediterranean) | 0·78 (0·48–1·26) | 0·44 (0·18–1·05) | 1·12 (0·72–1·74) | 1·10 (0·50–2·40) |
| Ethnicity (Asian) | 1·27 (0·78–2·06) | 1·66 (0·91–3·01) | 1·83 (1·20–2·80) | 2·26 (1·11–4·60) |
| Ethnicity (African) | 1·35 (0·89–2·05) | 1·11 (0·60–2·06) | 1·49 (1·00–2·23) | 2·01 (1·06–3·79) |
| Environmental exposure and genetic model in ethnic subgroups (iii), n = 2981 | n = 258 | n = 177 | n = 235 | n = 76 |
| Ethnicity (Mediterranean) | 0·99 (0·53–1·86) | 0·57 (0·19–1·70) | 1·41 (0·79–2·49) | 1·00 (0·35–2·83) |
| Ethnicity (Asian) | 1·43 (0·75–2·73) | 1·95 (0·88–4·31) | 2·04 (1·14–3·65) | 3·08 (1·34–7·10) |
| Ethnicity (African) | 1·35 (0·77–2·37) | 1·68 (0·82–3·43) | 1·43 (0·81–2·50) | 1·80 (0·77–4·17) |
Values are pooled odds ratios with their 95% confidence intervals. The genetic risk factor model was adjusted for ethnicity only. The environmental exposure model was mutually adjusted for all environmental exposures as presented in Table 1, and additionally adjusted for (i) genetic risk factors, (ii) ethnic subgroups and (iii) for genetic risk factors and ethnic subgroups. Reference groups are the ‘never eczema’ phenotype group, and European ethnicity, or no FLG mutation group. Effect estimates for the association of early environmental exposures with eczema phenotypes additionally adjusted for the genetic risk factors and ethnicities subgroups are shown in Tables S4–S6 (see Supporting Information). Bold values indicate statistical significance at the α = 0·05 level.
Ethnicity and eczema phenotypes
After dividing ethnicity into more detailed subgroups in our early environmental exposure model, we observed that Asian and African ethnicity were positively associated with late transient (OR 1·83, 95% CI 1·20–2·80 and OR 1·49, 95% CI 1·00–2·23, respectively) and persistent eczema (OR 2·26, 95% CI 1·11–4·60 and OR 2·01, 95% CI 1·06–3·79, respectively), compared with ‘never eczema’ and European ethnicity (Tables 3, S5; see Supporting Information). When we additionally adjusted for genetic risk factors, only the associations of Asian ethnicity with an increased risk for late transient (OR 2·04, 95% CI 1·14–3·65) and persistent eczema (OR 3·08, 95% CI 1·34–7·10) remained (Tables 3, S6; see Supporting Information).
Discussion
Five eczema phenotypes were identified in a multiethnic paediatric population, which was followed from birth until age 10 years, based on age of onset and persistence of eczema. Several known risk factors for eczema were associated with distinct phenotypes, but no clear patterns emerged, which suggests that the previously known eczema risk factors have limited differentiating capacities for eczema phenotypes. Most of the associations were found in relation to early transient and persistent eczema. Early transient and persistent eczema were most common in first‐born children, those with a parental history of eczema, allergy or asthma and those with persistent wheezing. Early transient eczema was most common in male children only. Children with late transient or persistent eczema were more often of Asian and African ethnicity. Eczema phenotypes were not associated with maternal education, breastfeeding, day care attendance or pet exposure. Children with early and late transient and persistent eczema more often had a filaggrin mutation or additional risk alleles in the genetic risk score. Most effect estimates did not materially change when we adjusted our analyses for both environmental and genetic factors. The explanation of why early transient and persistent eczema phenotypes share several determinants is not clear, but both patterns are dominant around the age of 1 year. This may be an important age in relation to exposure to environmental factors and the expression of genetic predisposition in the maturation of the skin and the immune system leading to the development of eczema.31
In our study, never, early transient and late transient and persistent eczema showed a similar pattern as those identified in a cohort study among 1038 children followed from birth until age 6 years.16 Compared with a different study among 3652 and 9894 children followed from birth until age 11 years and 16 years,15 higher early eczema probabilities were observed with steeper resolving curves, and no phenotype was identified with an onset after age 6 years. There was a greater degree of similarity in the patterns when we used longitudinal latent class analysis. However, this analysis does not take repeated measurements into account, which we considered relevant in this study because eczema measurements were taken at different time intervals. The remaining discrepancy in number and pattern of phenotypes might be explained by the differences in follow‐up time (10 years compared with 6 years and 11–16 years in the other studies),15, 16 number of repeated eczema measurements (six measurements compared with seven and 10–12), eczema definition (physician‐diagnosed eczema vs. itchy rash on specific locations) and population characteristics (multiethnic vs. mostly European ethnicity).
The observations in this study support the view that nulliparity, parental history of eczema, allergy and asthma, late‐onset and persistent wheezing, FLG genotype and the genetic risk score based on previously identified SNPs are risk factors for childhood eczema.6, 12, 15, 16 The functions of many of these SNPs are not yet determined, but may be related to autoimmunity and skin barrier.13 A previous study suggests that children of Surinamese–Creole and Surinamese–Hindustani origin have an increased risk of eczema.22 In addition, this study showed that Asian and African children had an up to 3·6‐fold increased risk of late transient and persistent eczema compared with European children. Possible underlying mechanisms include differences in skin barrier properties, parental psychological distress, microbiome development and other genetic factors.11, 22, 32
In contrast to the literature, no associations were observed between breastfeeding and eczema phenotypes. However, most literature focused on the presence or absence of eczema and not on distinct phenotypes, or a univariate analysis was used.15 Even though favourable effects have been found between pet exposure and eczema, we did not find any effect between postnatal pet exposure and eczema phenotypes.8 This is in line with previous studies on eczema phenotypes.15, 16 Differences between our study and the previous literature might be due to the timing of the measurements, different distribution of risk factors or mild severity of included eczema cases as is illustrated by the low prevalence of FLG mutations in our multiethnic paediatric population.12
The number of eczema phenotypes is based on statistical fit and depends on clinical relevance. The three eczema phenotypes identified by latent class growth analysis might present a more useful model for clinical practice, because it makes a clearer distinction between transient and persistent eczema. All patients with eczema should receive optimal care, but it would be useful to identify children with a higher chance of developing persistent eczema, as they may well benefit from earlier, more aggressive, treatment.33 For future studies, it would be clinically relevant to know whether individuals with specific eczema phenotypes are more prone to develop other atopic diseases, such as asthma and/or food allergies. From an aetiological point of view, it is important to improve the identification of specific early‐life environmental exposures and genetic risk factors in the development and persistency of eczema. A sufficient number of cases of eczema and detailed information on endogenic factors are needed to compare immunological response, skin barrier defects and genetic predisposition in children with different eczema phenotypes.
The strengths of this study include the prospective population‐based design, multiethnic population with detailed information on eczema, early environmental exposures and genetic factors. Our eczema phenotypes model seems valid for a multiethnic population in an ever‐globalizing world. Latent class growth analysis is an objective method to identify classes within a population. More precise and unbiased effect estimates are obtained by using multivariate multinomial models based on imputed data. However, some methodological limitations to this study need to be considered. We assumed that data was missing at random. It is always a possibility that missing data is not missing at random, which might lead to biased estimates.34 Including children with at least 50% of the data observed ensures a more reliable model. Selection bias might be present if the associations of the selected environmental and genetic factors with eczema phenotypes were different for the children included and those who were not included in the analyses. Furthermore, we cannot rule out under‐reporting or misreporting of eczema. However, there is evidence that parent‐reported physician‐diagnosed eczema is sufficient for epidemiological research and eczema prevalence in our study is similar to that of the Dutch population.33, 35 No information was available to determine the severity of eczema or subtypes of eczema, which could have influenced the observed effect estimates and associations, and there is a possibility that residual confounding may have occurred. Also, a longer follow‐up period could influence the number and pattern of eczema phenotypes. The uncertainty of class assignment of children is only partially accounted for by using weights in the multinomial analyses. Moreover, the results were not adjusted for multiple testing. We extracted the SNPs from the most recent GWA study that included non‐European populations; however, most of the genetic risk factors, including FLG mutations, have been discovered in European populations.13 Although early transient and persistent eczema appear to share several determinants, most of the selected environmental and genetic factors did not strongly differentiate between the various eczema phenotypes. This may partly be explained by minimizing the number of categories per factor in order to maximize the number of factors required to maintain appropriate statistical models. Future large‐scale studies examining early‐life environmental exposures in more detail could provide better differentiation between eczema phenotypes.
In conclusion, five eczema phenotypes were identified in a multiethnic paediatric cohort followed from birth until the age of 10 years. Previously known eczema risk factors differentiated between the different phenotypes to a limited degree. Male sex and Asian and African ethnicity were differently associated with eczema phenotypes and therefore could be useful for prediction purposes. Further studies are needed to compare the trajectories of different eczema phenotypes and identify other potential predictive factors, ideally in a ‘hypothesis‐free approach’ as known risk factors are relatively poor discriminators, in order to improve eczema management and prevention strategies.
Author contributions
C.H., L.D., N.S.E. and T.N. contributed to the conception and design, acquisition of data, analyses and interpretation of the data, drafted the article, revised it critically for important intellectual content, and gave final approval of the version to be published. N.J.E., C.P., V.B., S.B., J.J., S.P. and J.F. contributed to the conception and design, acquisition of data, revised the drafted manuscript critically for important intellectual content, and gave final approval of the version to be published.
Supporting information
File S1 Supplemental methods.36
Table S1 Characteristics of children and their mothers of those included and those not included in the analyses.
Table S2 Model fit after latent class growth analysis.
Table S3 Model fit after longitudinal latent class analysis.
Table S4 Associations of early environmental and genetic factors with eczema phenotypes.
Table S5 Associations of early environmental and eczema phenotypes with ethnicity subdivision.
Table S6 Associations of early environmental and genetic factors with eczema phenotypes with ethnicity subdivision.
Fig S1. (a) Three eczema phenotypes after longitudinal latent class analysis. (b) Five eczema phenotypes after longitudinal latent class analysis.
Video S1 Author video.
Powerpoint S1 Journal Club Slide Set.
Acknowledgments
The Generation R Study is conducted by the Erasmus Medical Center in close collaboration with the School of Law and Faculty of Social Sciences of the Erasmus University Rotterdam, the Municipal Health Service Rotterdam area, Rotterdam, the Rotterdam Homecare Foundation, Rotterdam, and the Stichting Trombosedienst & Artsenlaboratorium Rijnmond (STAR‐MDC), Rotterdam. We gratefully acknowledge the contribution of children and parents, general practitioners, hospitals, midwives and pharmacies in Rotterdam. The Generation R Study is made possible by financial support from the Erasmus Medical Center, Rotterdam, the Erasmus University Rotterdam and the Netherlands Organization for Health Research and Development. The study received funding from the European Union's Horizon 2020 research and innovation programme (LIFECYCLE project, grant agreement no 733206; 2016). T.N. and L.D. received funding from Nestlé Skin Health – Galderma R&D (grant agreement no 35195; 2016). L.D. received funding from the European Union's Horizon 2020 cofunded programme ERA‐Net on Biomarkers for Nutrition and Health (ERA HDHL) [ALPHABET project (no 696295; 2017), ZonMw The Netherlands (no 529051014; 2017)]. We would like to thank Raquel Granell from the Bristol Medical School, Population Health Sciences from the University of Bristol, for advice on the statistical analyses.
Funding sources The Generation R Study is made possible by financial support from the Erasmus Medical Center, Rotterdam, the Erasmus University Rotterdam and the Netherlands Organization for Health Research and Development. The study received funding from the European Union's Horizon 2020 research and innovation programme (LIFECYCLE project, grant agreement no 733206; 2016). The project received funding from Nestlé Skin Health – Galderma R&D (grant agreement no 35195; 2016). L.D. received funding from the European Union's Horizon 2020 cofunded programme ERA‐Net on Biomarkers for Nutrition and Health (ERA HDHL) [ALPHABET project (no 696295; 2017), ZonMw the Netherlands (no 529051014; 2017)].
Conflicts of interest This study was funded by Nestlé Skin Health – Galderma R&D. C.P., V.B. and S.B.‐R. are employed by Nestlé Skin Health – Galderma R&D.
https://doi.org/10.1111/bjd.18575 available online
References
- 1. Odhiambo JA, Williams HC, Clayton TO et al Global variations in prevalence of eczema symptoms in children from ISAAC Phase Three. J Allergy Clin Immunol 2009; 124:1251–8.e23. [DOI] [PubMed] [Google Scholar]
- 2. Böhme M, Soderhall C, Kull I et al Filaggrin mutations increase the risk for persistent dry skin and eczema independent of sensitization. J Allergy Clin Immunol 2012; 129:1153–5. [DOI] [PubMed] [Google Scholar]
- 3. Henderson J, Northstone K, Lee SP et al The burden of disease associated with filaggrin mutations: a population‐based, longitudinal birth cohort study. J Allergy Clin Immunol 2008; 121:872–7.e9. [DOI] [PubMed] [Google Scholar]
- 4. Bisgaard H, Simpson A, Palmer CN et al Gene‐environment interaction in the onset of eczema in infancy: filaggrin loss‐of‐function mutations enhanced by neonatal cat exposure. PLOS MED 2008; 5:e131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Dom S, Droste JH, Sariachvili MA et al The influence of parental educational level on the development of atopic sensitization, wheezing and eczema during the first year of life. Pediatr Allergy Immunol 2009; 20:438–47. [DOI] [PubMed] [Google Scholar]
- 6. Strachan DP, Ait‐Khaled N, Foliaki S et al Siblings, asthma, rhinoconjunctivitis and eczema: a worldwide perspective from the International Study of Asthma and Allergies in Childhood. Clin Exp Allergy 2015; 45:126–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Elbert NJ, van Meel ER, den Dekker HT et al Duration and exclusiveness of breastfeeding and risk of childhood atopic diseases. Allergy 2017; 72:1936–43. [DOI] [PubMed] [Google Scholar]
- 8. Pelucchi C, Galeone C, Bach JF et al. Pet exposure and risk of atopic dermatitis at the pediatric age: a meta‐analysis of birth cohort studies. J Allergy Clin Immunol 2013; 132:616–22.e7. [DOI] [PubMed] [Google Scholar]
- 9. Hagerhed‐Engman L, Bornehag CG, Sundell J, Aberg N. Day‐care attendance and increased risk for respiratory and allergic symptoms in preschool age. Allergy 2006; 61:447–53. [DOI] [PubMed] [Google Scholar]
- 10. Loo EX, Shek LP, Goh A et al Atopic dermatitis in early life: evidence for at least three phenotypes? Results from the GUSTO Study. Int Arch Allergy Immunol 2015; 166:273–9. [DOI] [PubMed] [Google Scholar]
- 11. Kaufman BP, Guttman‐Yassky E, Alexis AF. Atopic dermatitis in diverse racial and ethnic groups ‐variations in epidemiology, genetics, clinical presentation and treatment. Exp Dermatol 2018; 27:340–57. [DOI] [PubMed] [Google Scholar]
- 12. 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] [PMC free article] [PubMed] [Google Scholar]
- 13. Paternoster L, Standl M, Waage J et al Multi‐ancestry genome‐wide association study of 21,000 cases and 95,000 controls identifies new risk loci for atopic dermatitis. Nat Genet 2015; 47:1449–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Paternoster L, Standl M, Chen CM et al Meta‐analysis of genome‐wide association studies identifies three new risk loci for atopic dermatitis. Nat Genet 2011; 44:187–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Paternoster L, Savenije OEM, Heron J et al Identification of atopic dermatitis subgroups in children from two longitudinal birth cohorts. J Allergy Clin Immunol 2018; 141:964–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Roduit C, Frei R, Depner M et al Phenotypes of atopic dermatitis depending on the timing of onset and progression in childhood. JAMA Pediatr 2017; 171:655–662. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Kooijman MN, Kruithof CJ, van Duijn CM et al The Generation R Study: design and cohort update 2017. Eur J Epidemiol 2016; 31:1243–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Asher MI, Keil U, Anderson HR et al International Study of Asthma and Allergies in Childhood (ISAAC): rationale and methods. Eur Respir J 1995; 8:483–91. [DOI] [PubMed] [Google Scholar]
- 19. van Meel ER, den Dekker HT, Elbert NJ et al A population‐based prospective cohort study examining the influence of early‐life respiratory tract infections on school‐age lung function and asthma. Thorax 2018; 73:167–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Irvine AD, McLean WH, Leung DY. Filaggrin mutations associated with skin and allergic diseases. N Engl J Med 2011; 365:1315–27. [DOI] [PubMed] [Google Scholar]
- 21. Kezic S, O'Regan GM, Yau N et al Levels of filaggrin degradation products are influenced by both filaggrin genotype and atopic dermatitis severity. Allergy 2011; 66:934–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Elbert NJ, Duijts L, den Dekker HT et al Role of environmental exposures and filaggrin mutations on associations of ethnic origin with risk of childhood eczema. The Generation R Study. Pediatr Allergy Immunol 2016; 27:627–35. [DOI] [PubMed] [Google Scholar]
- 23. Feldman BJ, Masyn KE, Conger RD. New approaches to studying problem behaviors: a comparison of methods for modeling longitudinal, categorical adolescent drinking data. Dev Psychol 2009; 45:652–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Muthen B, Muthen LK. Integrating person‐centered and variable‐centered analyses: growth mixture modeling with latent trajectory classes. Alcohol Clin Exp Res 2000; 24:882–91. [PubMed] [Google Scholar]
- 25. Jung T, Wickrama KAS. An introduction to latent class growth analysis and growth mixture modeling. Soc Pers Psy Compass 2008; 2:302–17. [Google Scholar]
- 26. Rubin DB. Multiple Imputation for Nonresponse in Surveys. Hoboken, NJ: John Wiley & Sons, 1987. [Google Scholar]
- 27. Sepúlveda N, Manjurano A, Drakeley C, Clark TG. On the performance of multiple imputation based on chained equations in tackling missing data of the African α3.7 ‐globin deletion in a malaria association study. Ann Hum Genet 2014; 78:277–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Voortman T, van den Hooven EH, Heijboer AC et al Vitamin D deficiency in school‐age children is associated with sociodemographic and lifestyle factors. J Nutr 2015; 145:791–8. [DOI] [PubMed] [Google Scholar]
- 29. van Buuren S, Groothuis‐Oudshoorn K. mice: Multivariate Imputation by Chained Equations in R. J Stat Soft 2011; 45:1–67. [Google Scholar]
- 30. Venables WN, Ripley BD. Modern Applied Statistics with S, 4th edn New York, NY: Springer, 2002. [Google Scholar]
- 31. Telofski LS, Morello AP 3rd, Mack Correa MC, Stamatas GN. The infant skin barrier: can we preserve, protect, and enhance the barrier? Dermatol Res Pract 2012; 2012:198789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Gupta VK, Paul S, Dutta C. Geography, ethnicity or subsistence‐specific variations in human microbiome composition and diversity. Front Microbiol 2017; 8:1162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Bruijnzeel‐Koomen CAFM, Poos MJJC. [Prevalentie van constitutioneel eczeem: Rijksinstituut voor Volksgezondheid en Milieu]. Available at: https://www.volksgezondheidenzorg.info/onderwerp/constitutioneel-eczeem/cijfers-context/huidige-situatie#node-prevalentie-van-constitutioneel-eczeem (last accessed 11 June 2019) (in Dutch).
- 34. Sterne JA, White IR, Carlin JB et al Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ 2009; 338:b2393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Silverberg JI, Patel N, Immaneni S et al Assessment of atopic dermatitis using self‐report and caregiver report: a multicentre validation study. Br J Dermatol 2015; 173:1400–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Henderson J, Granell R, Heron J et al Associations of wheezing phenotypes in the first 6 years of life with atopy, lung function and airway responsiveness in mid‐childhood. Thorax 2008; 63:974–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
File S1 Supplemental methods.36
Table S1 Characteristics of children and their mothers of those included and those not included in the analyses.
Table S2 Model fit after latent class growth analysis.
Table S3 Model fit after longitudinal latent class analysis.
Table S4 Associations of early environmental and genetic factors with eczema phenotypes.
Table S5 Associations of early environmental and eczema phenotypes with ethnicity subdivision.
Table S6 Associations of early environmental and genetic factors with eczema phenotypes with ethnicity subdivision.
Fig S1. (a) Three eczema phenotypes after longitudinal latent class analysis. (b) Five eczema phenotypes after longitudinal latent class analysis.
Video S1 Author video.
Powerpoint S1 Journal Club Slide Set.