Baurecht 2007

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Toward a major risk factor for atopic eczema:

Meta-analysis of filaggrin polymorphism data


Hansjörg Baurecht, MSc,a,b Alan D. Irvine, MD,c Natalija Novak, MD,d Thomas Illig,
PhD,e Bettina Bühler, MSc,b Johannes Ring, MD,b,f Stefan Wagenpfeil, PhD,a* and
Stephan Weidinger, MDb,f* Munich, Bonn, and Neuherberg, Germany, and Dublin, Ireland

Background: With an impressive series of replication studies, Clinical implications: These results underline the importance of
filaggrin (FLG) has become the gene with the most widely a genetically determined epidermal barrier disruption in AE.
replicated association to atopic eczema (AE). However, studies (J Allergy Clin Immunol 2007;120:1406-12.)
published to date demonstrate differences concerning study
design and strength of associations. Key words: Atopic eczema, filaggrin, meta-analysis, atopic
Objectives: We sought to provide a general and overall estimate dermatitis
of FLG effect sizes and to estimate allele and carrier
frequencies.
Methods: We searched Medline and Institute for Scientific Atopic eczema (AE) is one of the most common
Information Web of Knowledge databases for relevant studies inflammatory skin disorders and affects up to 20% of
and abstracts from professional societies that were published children and up to 10% of adults in developed countries.1
through June 30, 2007. Initially, we accounted for different It is firmly established that AE is under strong genetic con-
study types and evaluated an overall estimate for case-control trol.2 Its frequent occurrence with other atopic pheno-
and family studies. In a second step, we combined those 2 study types, including asthma and allergic rhinitis, suggests
types and used a random-effects analysis approach to calculate
shared genetic risk variants.3 In addition, it is highly prob-
overall odds ratios (ORs). Tests of asymmetry were applied to
able that there are organ- and disease-specific factors, as
detect potential publication bias.
Results: Nine studies that met the inclusion criteria were indicated by genome-wide scans.2
included in the meta-analysis. For the combined genotype Despite considerable efforts that resulted in numerous
(R501X or 2282del4), we found an overall OR of 4.09 (95% CI, proposed or refuted associations with AE susceptibility
2.64-6.33) from the case-control studies and a summary OR of genes, until recently, no strong and reproducible genetic
2.06 (95% CI, 1.76-2.42) from the family studies. determinant could be identified. The detection of filaggrin
Conclusion: The powerful effect of FLG variation on AE risk (FLG) as major gene for AE in 2006 was an important
exceeds that of any other investigated candidate gene for AE breakthrough.4,5 Subsequent studies have shown FLG to
thus far and makes FLG one of the strongest genes known to be the gene with the most widely replicated association
date for complex diseases.
to AE.6-13
FLG is located within the epidermal differentiation
complex (EDC) in a region on chromosome 1q21, with a
significant linkage signal to AE and psoriasis.14,15 The
From athe Institute for Medical Statistics and Epidemiology (IMSE), Technical
dermatology, and drug allergy

EDC is a cluster of genes and gene families encoding pro-


University Munich (TUM); bthe Division of Environmental Dermatology
Food allergy, anaphylaxis,

and Allergy, GSF-National Research Center for Environment and Health


teins involved in terminal differentiation of the epider-
and ZAUM-Center for Allergy and Environment, Technical University Mu- mis.16 The main function of FLG is to aggregate keratin
nich; cthe Department of Paediatric Dermatology, Our Lady’s Children’s filaments, leading to keratinocyte compaction and forma-
Hospital, Dublin; dthe Department of Dermatology, University of Bonn;
e
tion of the stratum corneum.16,17 It is thereby crucial for en-
the Department of Epidemiology, GSF-National Research Center for Envi-
gineering and maintaining the barrier function of the skin.
ronment and Health, Neuherberg; and fthe Department of Dermatology and
Allergy Biederstein, Technical University Munich (TUM), Munich. Two FLG polymorphisms, R501X and 2282del4, have
*These authors contributed equally to this work. been shown to exert minor allele frequencies of greater
Supported by grants from the German Ministry of Education and Research than 0.01 in the white European population.4,10
(BMBF)/National Genome Research Network (NGFN), project NUW- Both variants are in trans and biochemically equivalent,
S31T05. S.W. and S.W. are supported by research grants KKF-07/04 and
KKF-27/05 of the University Hospital ‘‘Rechts der Isar,’’ Technical
leading to a complete loss of FLG expression.10 They have
University Munich. been shown to cause ichthyosis vulgaris, a common inher-
Disclosure of potential conflict of interest: A. D. Irvine has consulting arrange- ited skin disorder of keratinization that is frequently asso-
ments with Novartis. The rest of the authors have declared that they have no ciated and shows many overlapping characteristics with
conflict of interest.
AE.5,10,18 Subsequently, a variety of case-control and
Received for publication May 16, 2007; revised August 29, 2007; accepted for
publication August 31, 2007. transmission studies firmly established an association
Available online November 5, 2007. between each of the 2 prevalent FLG null alleles and
Reprint requests: Stephan Weidinger, MD, Department of Dermatology and AE.4,6-13 In the meantime, additional less common FLG
Allergy, Biederstein, Technical University Munich, Biedersteiner Str. 29, variants associated with AE also in non-European pop-
80802 Munich, Germany. E-mail: weidinger@lrz.tum.de.
0091-6749/$32.00
ulations have been reported.10,19
Ó 2007 American Academy of Allergy, Asthma & Immunology Thus far, FLG is unambiguously the only gene robustly
doi:10.1016/j.jaci.2007.08.067 associated with AE, as pointed out by several overview
1406
J ALLERGY CLIN IMMUNOL Baurecht et al 1407
VOLUME 120, NUMBER 6

parents to an affected offspring, in a trio setting the ratio of


Abbreviations used transmitted A and untransmitted B alleles (T) to untransmitted A
AE: Atopic eczema and transmitted B alleles (U) can be interpreted as an OR according
EDC: Epidermal differentiation complex to Morris and Gardner,24 as follows:
FLG: Filaggrin T
HWE: Hardy-Weinberg equilibrium OR 5 :
U
OR: Odds ratio
The number T can be considered a binomial variable if condi-
tioned on the sum of the numbers of ‘‘discordant’’ pairs (T 1 U).
Thus with the D method, we derived a symmetric confidence interval
for log(OR) in family studies.
articles.20-23 However, association studies published thus ORs in case-control studies were calculated for FLG carriers ver-
far show striking differences concerning study design sus noncarriers if contingency tables were available. A symmetric
and strength of associations. To gain a better understand- 95% CI was calculated for log(OR) by using the normal approxima-
ing of the contribution of genetic variation in FLG to risk tion of the Pearson-Mantel-Haenzel statistic.25
of AE and to provide a more accurate estimate of effect In a second step we combined those 2 study types and estimated
sizes, we performed a meta-analysis that combines data an overall OR.
and attempts to pool the different study types. We estimated the overall effect of FLG null alleles with a ran-
dom-effects model, as described by DerSimonian and Laird.26 The
Furthermore, we estimated allele and carrier frequencies
test of homogeneity was performed according to the method of
in control groups. Cooper and Hedges.27 To detect a potential publication bias, funnel
plots were constructed, and the test for asymmetry by Egger et al28
was applied.
METHODS Main results of meta-analyses are visualized graphically in forest
and funnel plots. Forest plots show the estimated ORs and 95% CIs
Data extraction for risk of AE imparted by FLG variations for individual studies,
A computerized search of Medline (National Library of Medicine, group analyses, and combined analyses of all groups. Funnel plots
Bethesda, Maryland, 1966–2007) and the ISI Web of Knowledge– might provide a hint on publication bias.
Web of Science (April 2006–April 2007) was conducted by 3 Additionally, the frequency of the mutant alleles in R501X and
independent researchers (HB, BB, and SW) to identify published 2282del4 was estimated in control subjects or parents by using the
studies and abstracts on AE and FLG polymorphisms R501X and inverse variance method, as described in Thakkinstian et al.29
2282del4 from the first reported association4 up to now. We included Furthermore, where appropriate, we assessed Hardy-Weinberg equi-
all articles presenting original data of family or case-control studies librium (HWE) for each study’s control group or parents by use of the
on FLG and AE. The reference list of the retrieved articles was also x2 test. Finally, where available, we estimated the prevalence of
reviewed to identify publications on the same topic. polymorphisms among cases. For all analyses, we used the base-,
For estimating allele and carrier frequencies in control groups, we stats- and meta-packages in R 2.4.1.30,31
included all studies on FLG and AE.
In the meta-analysis we included all articles presenting a 3 3 2
contingency table of FLG polymorphisms or combined genotype ver- RESULTS
sus AE and all publications providing odds ratios (ORs) plus 95%
CIs. We focused on the association with AE because only few com-

dermatology, and drug allergy


Our literature search yielded 39 publications, of which
parable studies are available thus far for related traits.

Food allergy, anaphylaxis,


All studies taken into account presented estimated ORs uncor-
9 met the inclusion criteria for the meta-analysis. Table
rected for potential confounders or gave complete contingency tables I4,6-13 shows the study type, the population origin, and
of FLG polymorphisms versus AE. Therefore we used unadjusted the number of cases and control subjects or nuclear fami-
estimates to quantify the OR of FLG carriers versus noncarriers. lies. In addition, reported ORs and their corresponding
Adjusted ORs were only available for a small subset of studies 95% CIs are given. We did not identify abstracts with orig-
considered here. inal data to be included in the analysis.
We excluded all articles without data for the effect of the genetic Morar et al8 analyzed 2 different family panels and pre-
variations on AE risk. sented results for each panel separately, as well as for the
combined panels. In a sensitivity analysis we compared
Statistical analysis the results of our meta-analyses under consideration of
Because the 2 common FLG mutations R501X and 2282del4 are the results achieved in this combined panel and the results
in trans and have equivalent biologic effects,10 association analysis for the separate panels. Because no significant differences
was also performed for a combined genotype in most of the studies. could be detected (Table I) for the overall estimate, both
In other words, individuals carrying one mutated and one wild-type panels were subsequently considered as one cohort. In
allele were coded heterozygous, and individuals heterozygous for addition, Morar et al8 investigated the risk of disease in
each of the 2 polymorphisms or homozygous for 1 polymorphism
affected offspring and unaffected children from the family
were coded homozygous for the combined genotype. Because we
were interested in the overall magnitude of the FLG effect, we
panels in a case-control setting. Because these study set-
focused our analysis on this combined genotype. tings are not independent, this case-control approach was
In our initial approach, we accounted for different study types and not used for the meta-analyses across study types. The
evaluated an overall estimate for case-control and family studies study on an Irish case-control cohort by Palmer et al4
separately. Considering transmission of possible alleles A and B from was included in a more comprehensive analysis in
1408 Baurecht et al J ALLERGY CLIN IMMUNOL
DECEMBER 2007

TABLE I. Unadjusted ORs and 95% CIs of studies on the association of FLG variants with AE included in the meta-analysis

Combined
R501X 2282del4 genotype
Diagnosis of AE
Study Type Population n OR 95% CI OR 95% CI OR 95% CI based on:

Weidinger et al12 Fam Germany 476 3.82 2.02-7.68 2.43 1.60-3.76 2.73 1.89-4.01UK Working Party
Weidinger et al13 cc Germany 274/252 3.59 1.43-9.01 5.07 2.42-10.64 4.17 2.30-7.59UK Working Party
Marenholz et al7 Fam Europe 490* 1.73 1.16-2.61 2.13 1.61-2.84 2.13 1.68-2.72Hanifin and Rajka
cc Germany 188/319 6.65 2.43-18.22 2.42 1.09-5.38 3.73 1.98-7.03Hanifin and Rajka
Barker et al6 cc United Kingdom 163/1463 5.85 3.86-8.86 7.61 4.74-12.21 7.59 5.30-10.87
UK Working Party
Palmer et al4 cc Denmark 142/190 — — — — 2.49 1.26-4.93Hanifin and Rajka
cc Scotland 279/1008à — — — — 2.89à 2.04-4.11à
Questionnaire
Ruether et al9 Fam Germany 338 2.17 1.34-3.61 — — — — Hanifin and Rajka
cc Germany 272/276 3.59 1.80-7.05 7.1 3.41-14.78 — — Hanifin and Rajka
Morar et al8 cc§ Northern Europe# 426/564 2.55 1.50-4.34 1.93 1.23-3.03 2.03 1.46-2.81 —
Fam§ Northern Europe# 426 2.55 1.70-3.90 2.38 1.55-3.71 1.72 1.32-2.25Modified Hanifin
and Rajka
Famk Northern Europe# 148 2.40 1.15-5.02 3.00 1.60-5.62 1.94 1.27-2.95 —
Fam{ United Kingdom 278 2.62 1.58-4.33 1.88 1.02-3.44 1.58 1.12-2.24 —
Stemmler et al11 cc Germany 378/700 1.31 0.65-2.62 1.93 1.25-2.96 4.18 2.60-6.73 Hanifin and Rajka
Sandilands et al10 cc Ireland** 188/736 14.05 8.04-24.53 8.94 4.99-16.01 10.02àà 6.74-14.89 UK Working Party
Pooled estimate 1 95% CI cc 4.23 2.36-7.59 4.13 2.38-7.16 4.09 2.64-6.33 —
Fam 2.32 1.73-3.09 2.25 1.83-2.77 2.10 1.66-2.65 —
Total 3.51 2.24-5.50 3.62 2.43-5.41 3.58 2.43-5.26 —
Continental Europeans Total 2.61 1.81-3.78 2.82 1.99-4.00 2.97 2.28-3.87 —

Fam, Family study; cc, case-control study; Total, combined analysis of both study types.
*Four hundred ninety families including 903 children with eczema.
Children of mothers with asthma.
àPatients with asthma and eczema.
§Consists of individuals belonging to nuclear families of 2 panels.
kECZ1 panel of northern Europe (90% white, 1% Asian, and 4.3% mixed 1 others).
{MRCE panel (69% white, 14% South Asian, and 8% mixed 1 others).
#Inclusive United Kingdom.
**Irish study by Palmer et al4 is included in Sandilands et al.10
Morar (case-control) excluded.
ààOR includes 3 additional rare variants.

Sandilands et al,10 which we therefore used for the present sizes in Ireland/United Kingdom cohorts, the differences
analysis. remained nonsignificant (Table III).4-13,32-34
dermatology, and drug allergy

All studies included were conducted on western


Case-control studies
Food allergy, anaphylaxis,

European populations, in particular from Germany, Great


Britain, and Ireland. To date, 9 independent case-control studies on FLG and
For pooled analysis of carrier frequencies, cases and AE have been performed across western Europe, 8 of
control groups from 17 case-control studies and parents which provided data on both polymorphisms, showing
and offspring from 5 family studies were used. The an increased risk for AE ranging from 2.03 to 10.02 for
study by Stemmler et al11 was excluded from this anal- the combined allele (Table I and Fig 1).
ysis because of a distortion of HWE for R501X (Table After applying a random-effects model, an overall OR
II).4-13,32-34 In addition, the Irish AE case series within was estimated at 4.09 (95% CI, 2.64-6.33, P 5 9.2 3 10210).
the study by Palmer et al4 was excluded because these Observed heterogeneity between the studies was 87.1%
cases have later been included into the Sandilands (95% CI, 76.7%-92.8%). The test of asymmetry showed
study.10 Nine other studies were not considered because no significant deviation from the symmetry assumption
of incomplete or missing data. Among the control sub- (P 5 .973), thus not indicating any publication bias (Fig 2).
jects, no significant heterogeneity was observed. Allele For both null alleles, a strong overall FLG effect was
frequencies were estimated at 1.9% (95% CI, 0.3%- calculated with an OR of 4.23 (95% CI, 2.36-7.59; range,
3.5%) for R501X, 1.9% (95% CI, 0.5%-3.3%) for 1.3-14.1) for R501X and 4.13 (95% CI, 2.38-7.16; range,
2282del4, and 4.1% (95% CI, 1.4%-6.8%) for the com- 1.9-8.9) for 2282del4, respectively. Tests of asymmetry
bined genotype (Table II).4-13,32-34 In contrast, striking were not significant (P 5 .6237 and P 5 .3988), indicating
differences were observed concerning the carrier no publication bias.
frequencies in cases between Ireland/United Kingdom After excluding the study of Stemmler et al11 because of
ranging from 22.9% to 45.2% and continental Europe distortion of HWE concerning the R501X allele, applying
ranging from 15.8% to 22.9%. Because of low sample a random-effects model, we then estimated an overall OR
J ALLERGY CLIN IMMUNOL Baurecht et al 1409
VOLUME 120, NUMBER 6

TABLE II. Minor allele frequencies of FLG variants (R501X, 2282del4, and combined genotype) and HWE test results
observed in studies on FLG published until April 2007

R501X 2282del4 Combined genotype


HWE HWE HWE
Study n* MAF P value n* MAF P value n* MAF P value

Weidinger et al12 Fam Germany 930 2.8% .371 952 5.5% .238 930 8.2% .319
Weidinger et al13 cc Germany 251 1.2% .848 250 1.8% .772 249 3.0% .624
Marenholz et al7 Fam Europe 970 2.8% NS 958 6.6% NS – 9.3% NS
cc Germany 319 0.8% .888 316 1.7% .753 314 2.5% .643
Barker et al6 cc United Kingdom 1463 2.9% .102 1463 1.7% .515 1463 4.6% .248
Palmer et al4 cc Ireland** 189 3.2% .652 186 1.1% .882 186 4.3% .540
cc Denmark – – – – – – 190 4.2% .233
cc Scotland 1008à 3.0% .026 1008 1.9% .542 1008à 4.9% .831
pop§ France 158 2.5% – 155 1.0% – – – –
pop§ Ireland/Scotland 242 2.8% – 257 1.0% – – – –
Ruether et al9 Famà Germany – – – – – – – – –
Ccà Germany – 2.2% – – 1.7% – – – –
Morar et al8 cc Northern Europe# – – NS – – NS – – NS
Fam Northern Europe# – – NS – – NS – – NS
Fam United Kingdom – – NS – – NS – – NS
Stemmler et al11 cc Germany 700à 1.6% 8.0 3 1026 678 3.5% .831 630c 2.2% .207
Sandilands et al10 cc Ireland 736 1.3% .723 736 1.3% .723 736 3.9% .918
Zhao et al32§ cc Ireland 654 1.5% .691 654 1.6% .676 654 3.1% .408
cc United Kingdom 1463 2.9% .102 1463 1.7% .515 1463 4.6% .248
Hüffmeier et al33§k cc Germany 369 1.8% NS 360 2.8% NS – – –
Smith et al5§ pop Ireland – 4.1% – – 0.5% – – – –
pop Scotland – 2.1% – – 1.2% – – – –
pop United States{ 124 2.4% – 133 1.1% – – – –
Gruber et al34§ pop Austria – 1.4% – – 1.4% – – 2.7% –
Pooled allele frequency (95% CI) 1.9% (0.3% to 3.5%) 1.9% (0.5% to 3.3%) 4.1% (1.4% to 6.8%)
Continental European 1.7% (0% to 4.0%) 2.8% (0.2% to 5.3%) 4.1% (0% to 8.5%)

MAF, Minor allele frequency; Fam, family study; cc, case-control study; NS, not significant; –, information not available; pop, population-based cohort.
*Number of control subjects in case-control–based studies, number of parents in family-based studies, and total number in population-based studies.
Calculated with stated genotyping rate of 0.99 and 0.977 for R501X and 2282del4, respectively.
àNot pooled because of violation of HWE or missing number of successfully genotyped control subjects.
§Control subjects from studies on related phenotypes.
kCalculated with stated genotyping rate of 0.981 and 0.957 for R501X and 2282del4, respectively.
{European Americans.
#Inclusive United Kingdom.

dermatology, and drug allergy


**Not pooled because of inclusion in Sandilands et al.10

Food allergy, anaphylaxis,


of 5.16 (95% CI, 3.02-8.81) and 4.13 (95% CI, 2.38-7.16) Combined analysis of all studies
for R501X and the combined genotype, respectively. Combining family-based and case-control studies, the
Tests of asymmetry remained not significant, indicating test of asymmetry remains not significant (P 5 .1994, Fig
no publication bias. 2), although ORs in family studies tend to be lower with
smaller CIs than in case-control studies. Because of the in-
Family studies terdependence, the case-control studies from Morar et al8
The 5 family studies evaluated showed ORs from 1.58 were not included. We also excluded the Irish case-control
to 2.73 for the combined genotype. These rather homo- study in Palmer et al4 because the respective individuals
geneous estimates result in an overall OR of 2.10 (95% CI, were included in a subsequent study by Sandilands et al.10
1.66-2.65; P 5 2.0 3 1029) by applying a random-effects The estimated overall OR for a random-effects model
model (Fig 1). A fixed-effects model revealed similar re- was 3.51 (95% CI, 2.24-5.50) for R501X, 3.62 (95% CI,
sults (OR, 2.06; 95% CI, 1.76-2.42). The heterogeneity 2.43-5.41) for 2282del4, and 3.58 (95% CI, 2.43-5.26;
test revealed no significant deviation from the homogene- P 5 3.7 3 10210) for the combined genotype.
ity assumption (P 5 .1364). Further investigation of Having further investigated Table I and Fig 1, the
R501X and 2282del4 revealed an overall OR of 2.32 OR estimates appeared lower for continental European
(95% CI, 1.73-3.09) and 2.25 (95% CI, 1.83-2.77) by populations. Therefore we calculated a pooled estimate
applying a random-effects model. Heterogeneity measures of studies conducted in continental Europe only.
showed no significance (P 5 .2198 and P 5 .8435, Applying a random-effects model, we observed an esti-
respectively). mate of 2.61 (95% CI, 1.81-3.78) for R501X, 2.82 (95%
1410 Baurecht et al J ALLERGY CLIN IMMUNOL
DECEMBER 2007

TABLE III. Carrier frequencies of FLG variants (R501X, 2282del4, and combined genotype) in cases and control subjects for
case-control studies or offspring and parents for family studies in studies on FLG published until April 2007

R501X 2282del4 Combined genotype


Carrier frequency Carrier frequency Carrier frequency
Control Control Control
subjects/ Cases/ subjects/ Cases/ subjects/ Cases/
Study parents offspring parents offspring parents offspring

Weidinger et al12 Fam Germany 5.7% 8.8% 10.9% 15.8% 16.0% 22.8%
Weidinger et al13 cc Germany 2.4% 8.1% 3.6% 15.9% 6.0% 21.1%
Marenholz et al7 Fam Europe – – – – – 22.9%
cc Germany 1.6% 9.6% 3.5% 8.0% 5.1% 16.7%
Barker et al6 cc United Kingdom 5.6% 25.8% 3.4% 20.9% 8.8% 42.3%
Palmer et al4 cc Irelandk 6.4% 38.5% 2.2% 26.9% 8.6% 55.8%
cc Denmark – – – – 7.9% 17.6%
cc Scotland 5.8%* – 3.8% – 9.3%* 22.9%
pop France – – – – – –
pop Ireland/Scotland – – – – – –
Ruether et al9 Fam Germany – – – – – –
cc Germany – – – – – –
Morar et al8 cc Northern Europe§ – – – – – –
Fam Northern Europe§ – – – – – –
Fam United Kingdom – – – – – –
Stemmler et al11 cc Germany 2.9%* 3.7% 6.8% 12.3% 4.3%* 15.8%
Sandilands et al10 cc Ireland 2.6% 27.1% 2.6% 19.1% 7.6% 45.2%
Zhao et al32 cc Ireland 3.1% n.AE 3.2% n.AE 6.3% n.AE
cc United Kingdom 5.6% n.AE 3.4% n.AE 8.8% n.AE
Hüffmeier et al33 cc Germany – n.AE – n.AE – n.AE
Smith et al5 pop Ireland – n.AE – n.AE – n.AE
pop Scotland – n.AE – n.AE – n.AE
pop United Statesà – n.AE – n.AE – n.AE
Gruber et al34 pop Austria – n.AE – n.AE – n.AE
Pooled allele frequency (95% CI) 3.5% (0.9% 9.0% (3.1% 3.8% (1.5% 14.7% (8.0% 8.0% (4.3% 22.9% (16.4%
to 6.0%) to 15.0%) to 6.2%) to 21.4%) to 11.7%) to 29.3%)
Continental European 2.8% (0% 6.4% (0.7% 5.7% (0.9% 12.6% (5.0% 8.2% (0.2% 22.9% (15.1%
to 6.8%) to 12.1%) to 10.4%) to 20.3%) to 14.2%) to 30.7%)

Fam, family study; cc, case-control study; –, information not available; pop, population-based cohort; n.AE, other trait than AE.
*Not pooled because of distortion of the HWE.
Control subjects from studies on FLG with related phenotypes.
àEuropean American.
dermatology, and drug allergy

§Inclusive United Kingdom.


Food allergy, anaphylaxis,

kNot pooled because of inclusion in Sandilands et al.10

CI, 1.99-4.00) for 2282del4, and 2.97 (95% CI, 2.28-3.87) independently conferred by both null polymorphisms,
for the combined genotype. Between the studies, there was with an estimated overall OR of 3.51, 3.62, and 3.58 for
no evidence for significant heterogeneity (P 5 .074). R501X, 2282del4, and the combined genotype, respec-
tively. Thus the powerful effect of FLG variation on AE
risk exceeds that of any other investigated candidate
DISCUSSION gene for AE thus far.2 Comparison with effects observed
for candidate genes of other complex diseases (eg
In this study we have conducted the most comprehen- ADAM33 and asthma35 or TCF7L2 and diabetes36) shows
sive assessment of currently available data on the effect of that FLG is also one of the strongest known genes for com-
the 2 most common FLG polymorphisms in European plex diseases in general. With values of 57% and 49% for
populations on AE. Our comprehensive meta-analysis is breast cancer by age 75 years, BRCA1 and BRCA2 muta-
based on an impressive series of replication studies that tions show penetrance rates comparable with FLG and
have followed the initial publication from Palmer et al4 AE.37 However, in contrast to FLG, BRCA mutations
in early 2006. Although these studies exhibited a consid- account for less than 5% of all breast cancer cases.38
erable heterogeneity concerning study design and the Although frequencies of at least one of the null alleles
magnitude of the genetic effect, they have firmly estab- were comparable among control subjects, strikingly
lished FLG as major gene for AE. Our analysis demon- higher carrier frequencies were observed in British and
strates that across all studies there is a high risk Irish case collections (22.9% to 45.2%) along with a
J ALLERGY CLIN IMMUNOL Baurecht et al 1411
VOLUME 120, NUMBER 6

FIG 1. Association of FLG mutations R501X and 2282del4 (combined genotype) with increased AE risk. Forest
plots showing the ORs and respective 95% CIs for the different studies (given by first author, year, and ethnic-
ity of the study population) included in the meta-analysis are shown: A, case-control studies; B, family studies;
C, combined analysis. The rhombus designates the overall estimate of all studies. The consistently higher AE
risk conferred by FLG null alleles in all studies is reflected by a high overall estimated OR. GER, Germany;
UK, United Kingdom; DEN, Denmark; SCO, Scotland; IRL, Ireland; cc, case control study.

FIG 2. Funnel plots showing the ORs and 95% CIs for the association between FLG mutations R501X and
2282del4 (combined genotype) and AE observed in case-control studies (A), family studies (B), and both study
types (combined analysis; C). Tests of asymmetry showed no significant deviation from the symmetry
assumption (Fig 2, A, P 5 .973; Fig 2, B, P 5 .953; Fig 2, C, P 5 .199), thus not indicating publication bias.

dermatology, and drug allergy


Food allergy, anaphylaxis,
higher disease risk (OR, 2.03-10.02) compared with substantially, and the clinical characteristics of cases and
studies from mainland Europe (15.8% to 22.9%). control subjects were often not sufficiently specified.
Background minor allele frequencies for R501X are lower Some groups of control subjects were population
in continental Europe, which partially explains this dif- based,4,9,13 whereas others were hospital based.11 Most
ference. Other possible explanations might include vari- studies did not control for known risk factors for AE. In
ation in disease spectrum, such as differences in severity addition, some studies lack information regarding the
and age ranges or differences concerning unknown envi- assessment of Hardy-Weinberg proportion in control
ronmental factors. subjects, such as the study by Stemmler et al,11 in which
Our findings must be taken with caution for a number of reanalysis showed that R501X appears to be not in HWE.
reasons. First, these estimates are obtained by pooling A rather homogeneous estimate of the FLG risk on AE
despite heterogeneous study designs (eg, case-control and could be observed for the family studies, possibly reflect-
family-based studies). Family-based studies include pro- ing the known additional level of control for occult genetic
bands who are designated as having AE as a result of being admixture, which might confound case-control studies.
examined for the purpose of completing the study, but the However, it has to be considered that the constructed
eczema might not be so severe as to need medical ORs24 are not directly comparable with those obtained
attention. In contrast, case-control studies deal with unre- from case-control studies.
lated individuals who, in the majority of cases, have We suggest that despite the issues discussed, this large
sufficiently severe AE to require attendance to secondary study has established FLG polymorphisms as an extraor-
care. Thus these studies cannot be directly compared. dinarily strong marker of AE disease risk because we
Second, the selection of cases and control subjects varied found no evidence of bias. We identified and considered
1412 Baurecht et al J ALLERGY CLIN IMMUNOL
DECEMBER 2007

major methodological issues of the various studies, and 13. Weidinger S, Rodrı́guez E, Stahl C, Wagenpfeil S, Klopp N, Illig T, et al.
Filaggrin mutations strongly predispose to early-onset and extrinsic
significant results were observed in all of the studies per-
atopic dermatitis. J Invest Dermatol 2007;127:724-6.
formed thus far, as well as in the pooled analysis. Given 14. Cookson WO, Ubhi B, Lawrence R, Abecasis GR, Walley AJ, Cox HE,
that additional prevalent variants have recently been et al. Genetic linkage of childhood atopic dermatitis to psoriasis suscep-
described,10 the current estimate of overall contribution tibility loci. Nat Genet 2001;27:372-3.
of FLG to AE is likely an underestimate, with additional 15. Mischke D, Korge BP, Marenholz I, Volz A, Ziegler A. Genes encoding
structural proteins of epidermal cornification and s100 calcium-binding
recurrent polymorphisms likely to be revealed in specific proteins form a gene complex (‘‘epidermal differentiation complex’’)
populations. on human chromosome 1q21. J Invest Dermatol 1996;106:989-92.
Clearly, well-defined and large-scale, population-based 16. Candi E, Schmidt R, Melino G. The cornified envelope: a model of cell
and longitudinal studies are needed to more precisely death in the skin. Nat Rev Mol Cell Biol 2005;6:328-40.
17. Presland RB, Dale BA. Epithelial structural proteins of the skin and oral cav-
work out the contribution of FLG polymorphisms to AE
ity: function in health and disease. Crit Rev Oral Biol Med 2000;11:383-408.
risk, to evaluate their role in other atopic diseases, and to 18. Sandilands A, O’Regan GM, Liao H, Zhao Y, Terron-Kwiatkowski A,
determine their effect in the general population. In addi- Watson RM, et al. Prevalent and rare mutations in the gene encoding fil-
tion, because it has been shown that FLG polymorphisms aggrin cause ichthyosis vulgaris and predispose individuals to atopic der-
are not solely responsible for the significant genetic link- matitis. J Invest Dermatol 2006;126:1770-5.
19. Nomura T, Sandilands A, Akiyama M, Liao H, Evans AT, Sakai K, et al.
age signal to the EDC region on chromosome 1q21,8 fur- Unique mutations in the filaggrin gene in Japanese patients with ichthyosis
ther investigation of this complex locus is likely to identify vulgaris and atopic dermatitis. J Allergy Clin Immunol 2007;119:434-40.
additional genetic variants with effects on AE. 20. Hudson TJ. Skin barrier function and allergic risk. Nat Genet 2006;38:
399-400.
21. Irvine AD. Fleshing out filaggrin phenotypes. J Invest Dermatol 2007;
We thank Domenique Ludwig for correcting the manuscript. 127:504-7.
22. Irvine AD, McLean WHI. Breaking the (un)sound barrier: filaggrin is a
major gene for atopic dermatitis. J Invest Dermatol 2006;126:1200-2.
23. McGrath JA. Second cite. Clin Exp Dermatol 2006;31:826-8.
REFERENCES
24. Morris J, Gardner M. Epidemiological studies. In: Altman D, Machin D,
1. Williams H, Robertson C, Stewart A, Aı̈t-Khaled N, Anabwani G, Bryant T, editors. Statistics with confidence. Vol. 2. Bristol: BMJ Book;
Anderson R, et al. Worldwide variations in the prevalence of symptoms 2005. p. 57-72.
of atopic eczema in the international study of asthma and allergies in child- 25. Altman D, Machin D, Bryant T, Gardner M, editors. Statistics with con-
hood. J Allergy Clin Immunol 1999;103:125-38. fidence—confidence intervals and statistical guidelines. 2nd ed. Bristol:
2. Morar N, Willis-Owen SAG, Moffatt MF, Cookson WOCM. The genet- BMJ Books; 2005.
ics of atopic dermatitis. J Allergy Clin Immunol 2006;118:24-36. 26. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin
3. van Beijsterveldt CEM, Boomsma DI. Genetics of parentally reported Trials 1986;7:177-88.
asthma, eczema and rhinitis in 5-yr-old twins. Eur Respir J 2007;29: 27. Cooper H, Hedges L. The handbook of research synthesis. Newbury Park
516-21. (CA): Russell Sage Foundation; 1994.
4. Palmer CNA, Irvine AD, Terron-Kwiatkowski A, Zhao Y, Liao H, Lee 28. Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis
SP, et al. Common loss-of-function variants of the epidermal barrier detected by a simple, graphical test. BMJ 1997;315:629-34.
protein filaggrin are a major predisposing factor for atopic dermatitis. 29. Thakkinstian A, McEvoy M, Minelli C, Gibson P, Hancox B, Duffy D,
Nat Genet 2006;38:441-6. et al. Systematic review and meta-analysis of the association between
5. Smith FJD, Irvine AD, Terron-Kwiatkowski A, Sandilands A, Campbell beta2-adrenoceptor polymorphisms and asthma: a huge review. Am J
LE, Zhao Y, et al. Loss-of-function mutations in the gene encoding Epidemiol 2005;162:201-11.
filaggrin cause ichthyosis vulgaris. Nat Genet 2006;38:337-42. 30. R Development Core Team. R: A Language and Environment for Statisti-
6. Barker JNWN, Palmer CNA, Zhao Y, Liao H, Hull PR, Lee SP, et al. cal Computing. R Foundation for Statistical Computing. Vienna, Austria
dermatology, and drug allergy

Null mutations in the filaggrin gene (FLG) determine major susceptibility 2006. Available at: http://www.R-project.org. Accessed October 12, 2007.
Food allergy, anaphylaxis,

to early-onset atopic dermatitis that persists into adulthood. J Invest 31. Guido Schwarzer. Meta: Meta-Analysis. R package version 0.8-2. Available
Dermatol 2007;127:564-7. at: http://umfragen.sowi.uni-mainz.de/CRAN/. Accessed October 12, 2007.
7. Marenholz I, Nickel R, Rüschendorf F, Schulz F, Esparza-Gordillo J, 32. Zhao Y, Terron-Kwiatkowski A, Liao H, Lee SP, Allen MH, Hull PR,
Kerscher T, et al. Filaggrin loss-of-function mutations predispose to phe- et al. Filaggrin null alleles are not associated with psoriasis. J Invest
notypes involved in the atopic march. J Allergy Clin Immunol 2006;118: Dermatol 2007;127:1878-82.
866-71. 33. Hüffmeier U, Traupe H, Oji V, Lascorz J, Ständer M, Lohmann J, et al.
8. Morar N, Cookson WOCM, Harper JI, Moffatt MF. Filaggrin mutations Loss-of-function variants of the filaggrin gene are not major susceptibil-
in children with severe atopic dermatitis. J Invest Dermatol 2007;127: ity factors for psoriasis vulgaris or psoriatic arthritis in German patients.
1667-72. J Invest Dermatol 2007;127:1367-70.
9. Ruether A, Stoll M, Schwarz T, Schreiber S, Fölster-Holst R. Filaggrin 34. Gruber R, Janecke AR, Fauth C, Utermann G, Fritsch PO, Schmuth M.
loss-of-function variant contributes to atopic dermatitis risk in the popu- Filaggrin mutations p.r501x and c.2282del4 in ichthyosis vulgaris. Eur
lation of northern Germany. Br J Dermatol 2006;155:1093-4. J Hum Genet 2007;15:179-84.
10. Sandilands A, Terron-Kwiatkowski A, Hull PR, O’regan GM, Clayton 35. Blakey J, Halapi E, Bjornsdottir US, Wheatley A, Kristinsson S, Upma-
TH, Watson RM, et al. Comprehensive analysis of the gene encoding fil- nyu R, et al. Contribution of ADAM33 polymorphisms to the population
aggrin uncovers prevalent and rare mutations in ichthyosis vulgaris and risk of asthma. Thorax 2005;60:274-6.
atopic eczema. Nat Genet 2007;39:650-4. 36. Cauchi S, El Achhab Y, Choquet H, Dina C, Krempler F, Weitgasser R,
11. Stemmler S, Parwez Q, Petrasch-Parwez E, Epplen JT, Hoffjan S. Two et al. TCF7L2 is reproducibly associated with type 2 diabetes in various
common loss-of-function mutations within the filaggrin gene predispose ethnic groups: a global meta-analysis. J Mol Med 2007;85:777-82.
for early onset of atopic dermatitis. J Invest Dermatol 2007;127:722-4. 37. Chen S, Parmigiani G. Meta-analysis of BRCA1 and BRCA2 penetrance.
12. Weidinger S, Illig T, Baurecht H, Irvine AD, Rodriguez E, Diaz-Lacava J Clin Oncol 2007;25:1329-33.
A, et al. Loss-of-function variations within the filaggrin gene predispose 38. Anglian Breast Cancer Study Group. Prevalence and penetrance of
for atopic dermatitis with allergic sensitizations. J Allergy Clin Immunol BRCA1 and BRCA2 mutations in a population-based series of breast
2006;118:214-9. cancer cases. Br J Cancer 2000;83:1301-8.

You might also like