If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding author at: Department of Pediatrics, University of Catania, Via S. Sofia 78, 90124 Catania, Italy. Tel.: +39 095 378 17 82; fax: +39 095 378 23 85; mobile: +39 320 11 25 476.
Department of Pediatrics, Università Politecnica delle Marche, Ancona, ItalyDivision of Pediatric Gastroenterology and Nutrition and Center for Celiac Research, Massachusetts General Hospital for Children, and Celiac Program – Harvard Medical School, Boston, MA, USA
Celiac disease is an immune-mediated disorder triggered by gluten in genetically susceptible persons. Despite its detrimental effects on human health, it has not disappeared over time. The current evolutionary theory is that celiac disease is more common in areas reached later by agricultural revolution than in countries that started consumption of wheat earlier, due to negative selection caused by celiac disease.
We reviewed data on worldwide prevalence of celiac disease, wheat consumption, and frequencies of HLA-celiac-disease-predisposing-genotypes to investigate their mutual relationship. Studies assessing prevalence of celiac disease were identified through a MEDLINE search. Wheat consumption and frequencies of HLA-DQ2-DQ8 were obtained from Food and Agriculture Organization of the United Nations and allelefrequencies.net database. Correlations between celiac disease, wheat consumption, and HLA were analyzed by linear regression. We observed a significant correlation between wheat consumption and HLA DQ2 (p = 0.01) and the sum of DQ2 and DQ8 (p = 0.01) frequencies. Wheat consumption and HLA-DQ2 tend to co-localize in different continents. The correlation between the prevalence of celiac disease and either DQ2 and/or DQ8, or the product of DQ2 + DQ8*wheat consumption was not statistically significant.
Co-localization of gluten consumption and HLA-celiac-disease-predisposing-genotypes can be explained by positive selection of HLA-DQ2 genes in wheat-consuming areas, and “demic diffusion” of Middle East farmers into Europe.
Celiac disease (CD) is a systemic immune-mediated disorder triggered by dietary gluten in genetically susceptible persons, resulting from the interplay between environmental and genetic factors. The major component of the genetic predisposition, explaining about 40% of it, depends on some HLA-related class II genes: most CD patients are HLA-DQ2 positive (90%), while half of the remaining 10% are HLA-DQ8 positive. As concerns the environmental factors, the ingestion of gliadin and glutenin fractions of wheat gluten and similar alcohol-soluble proteins in other grains (rye and barley) triggers the development of the celiac enteropathy, as well as other disease manifestations [
In the past, CD was considered a rare disorder mostly affecting individuals of European origin. In 1978, Simoons formulated a theory on the evolutionary history of the disease. He postulated that the spreading of wheat consumption from the so-called “Fertile Crescent” in the Middle East along the Mediterranean countries to northern Europe exerted a negative selective pressure on genes predisposing to CD. The higher frequency of the HLA-B8 allele (at that time the only gene known to be associated with CD) and the resulting higher frequency of CD in north-western Europe were attributed to a lack of exposure to cereals until recent times [
The Simoons’ theory did not survive the new era of CD genetics and epidemiology. The HLA-B8 was spuriously associated with CD, basically due to linkage disequilibrium with the HLA-DQ2, the “true” CD-predisposing genotype. At the same time, serological screening studies showed that (a) CD is one the commonest lifelong disorders, affecting around 1% of the population worldwide and (b) CD is a common disorder in both north-western and south-eastern Europe despite some inter-country variability [
]. Despite its negative effects on human health, the CD phenotype has not disappeared over time, but is even increasing nowadays in areas with high level of both gluten consumption and CD-predisposing HLA-genotypes (so-called “evolutionary paradox of CD”) [
The analysis of (a) the geographical distribution of genes that predispose to CD, (b) wheat consumption and (c) CD prevalence, can help to understand the evolutionary history of this multi-factorial disorder. To this aim, in the present study we reviewed the data on worldwide prevalence of CD, wheat consumption, and frequencies of HLA DQ2 and DQ8 haplotypes. We combined the three geographical maps to investigate their mutual relationship.
2. Methods
2.1 Protocol
Before review we developed a protocol, including eligibility criteria, search strategies, criteria for study selection, methods for extracting related data, and methods for assessing study quality and statistical methodology.
2.2 Information sources, eligibility criteria, search, and study selection
The current worldwide distribution of CD, wheat consumption and frequency of HLA DQ2 and DQ8 haplotypes were the focus of our search.
Studies on the prevalence of CD were identified by searching electronic databases and scanning reference lists of articles, and by consultation with experts in the field. The search was applied to the Medline database using PubMed by combining search terms for prevalence (prevalence, epidemiology) with keywords for CD (or coeliac disease, coeliac, gluten intolerance). As far as the diagnostic algorithm, we included only studies defining CD as either (a) CD positive serology plus typical small intestinal biopsy findings (Marsh 3a-3c) or (b) CD positive serology (high level of IgA anti-transglutaminase and/or antiendomysial antibodies). All types of study design (e.g. cross-sectional, cohort, case–control, and case series), except case reports, were considered for inclusion in this review. Search results were limited to studies based on the screening of unselected (low-risk) child and adult samples. Studies on healthy blood donors were excluded because they differed from the general population in social, health-related, and gender variables. No publication date or publication status restrictions were imposed. Eligibility assessment of studies was performed independently in an unblinded standardized manner by the two authors. Disagreements between reviewers were resolved by consensus. All studies described in this review were published between 1950 (start of Medline) and August 2012.
Worldwide distribution data of wheat consumption levels were obtained from the Food and Agriculture Organization of the United Nations (FAO) database (http://www.fao.org). The duration of wheat culture was taken from the work of Ammerman and Cavalli-Sforza [
]. According to the duration of wheat consumption we divided the Middle-East and European countries into four groups from 1 (gluten-containing cereals introduced around 10,000 years ago) to 4 (gluten-containing cereals introduced 6000–4000 years ago). The frequency of HLA-DQ2 and HLA-DQ8 in different countries were obtained by the allelefrequencies.net database (http://www.allelefrequencies.net) [
], pilot tested it on three randomly selected included studies, and refined it accordingly. One author (E.L.) extracted the data from included studies, and the second author (C.C.) checked the extracted data.
2.4 Data items
Regarding the prevalence of CD, from each included study information was extracted on (1) study design; (2) characteristics of participants, including number, age, setting of enrolment, method of diagnosis of CD; and (3) prevalence of CD.
To evaluate the worldwide distribution of genes predisposing to CD we extracted the frequency of the allele DQB1*02 (DQ2), and the frequency of the haplotype DQA1*0301/DQB1*0302 (DQ8) from the allelefrequencies.net database. DQB1*02 positivity is the hallmark of the different CD-predisposing DQ2 haplo/genotypes, that is DQA1*05/DQB1*02 in trans or cis position (DQ2.5) and DQA1*02/DQB1*02 (DQ2.2).
To evaluate the worldwide wheat consumption we extracted data on wheat availability in the year 2009 for each country (expressed in grams per capita/day) from the FAO database.
2.5 Risk of bias
To ascertain the validity of the eligible studies, the study design, the size and representativeness of the study population (i.e. the presence of selection bias), the validity of outcomes (risk of confounding or bias), and the quality of the statistical analysis were taken into account. We assessed the methodological quality of included studies in accordance with the guidelines of the Cochrane Consumers and Communication Review Group, adapted for the current review concerning observational non-intervention studies. In all cases, the two authors independently assessed the quality of the studies included, with any disagreements resolved by discussion and consensus. Where necessary, study authors were contacted for additional information or for clarification of the study methods.
2.6 Statistical analysis
We calculated the prevalence of CD, the levels of wheat consumption, and the frequencies of HLA-DQ2 and HLA-DQ8 for different regions of the world. In case of multiple studies in the same country we calculated the weighted average. For each parameter we constructed a geographical map.
The correlations were studied by Spearman's correlation test and associations were further studied by multivariate linear regression analysis.
All differences were considered to be statistically significant at a 5% probability level.
Data were analyzed with SPSS 13.0 software (SPSS, Chicago, IL).
3. Results
3.1 Study selection and characteristics (CD prevalence)
Supplemental Figure 1 shows the flow diagram of study selection of studies on the prevalence of CD. Table 1 summarizes the studies reporting on the prevalence of CD [
]. Thirty-one publications were included, containing 38 CD prevalence estimates, of which 23 in children and 15 in adults, across 24 countries. The included studies involved 122,858 participants. None but one of the articles reported a power calculation to determine the population sample size.
Table 1Summary of included studies evaluating the prevalence of celiac disease in different countries.
Diagnosis of celiac disease was defined as: 1=serology positive for celiac disease (anti-transglutaminase antibodies) and histology compatible with celiac disease; 2=serology positive for celiac disease (anti-endomisium antibodies antibodies) and histology compatible with celiac disaese; 3=anti-endomisium antibodies positives.
* Diagnosis of celiac disease was defined as: 1 = serology positive for celiac disease (anti-transglutaminase antibodies) and histology compatible with celiac disease; 2 = serology positive for celiac disease (anti-endomisium antibodies antibodies) and histology compatible with celiac disaese; 3 = anti-endomisium antibodies positives.
Overall, the worldwide prevalence of CD among low-risk (general population) adults and children ranged from 0 to 5.6%, with a mean prevalence ratio of 1% (Supplemental Figure S2). In Europe the overall prevalence of CD was 1%, with large variations between countries [
]. In areas of the developing world, rates overlapping European figures, especially in North Africa (i.e., 0.5% in Egypt, 0.8% in Libya, and 0.6% in Tunisia) [
]. No CD prevalence studies were found from China or other Far Eastern countries.
3.3 Wheat consumption
Data on wheat consumption were available for 150 countries. In general, wheat consumption ranged between 21 and 564 g per capita/day (mean = 260) (Supplemental Figure S3).
3.4 HLA-DQ2 and DQ8 frequency
Data on HLA-DQ were available for 28 (DQ2) and 15 (DQ8) countries, respectively. Frequencies of HLA DQ2 and DQ8 ranged between 0 and 28% (mean = 14) and between 1 and 9% (mean = 4), respectively (Supplemental Figures S4 and S5).
3.5 CD prevalence, wheat consumption and HLA-DQ2 and–DQ8 frequency
The prevalence of CD and the frequency of DQ2 or DQ8 or DQ2 plus DQ8 were not significantly correlated (Rho = 0.4, p = 0.07; Rho = 0.3, p = 0.4). Likewise, there was no significant correlation between the nationwide prevalence of CD and the level of wheat consumption (Rho = 0.4, p = 0.06), and between the prevalence of CD and the product of DQ2 + DQ8 frequency * wheat consumption (Rho = 0.3, p = 0.3). On the other hand, a highly significant correlation between wheat consumption and the frequency of either HLA DQ2 (Fig. 1) or the sum of DQ2 and DQ8 (Fig. 2) was found. There was a trend of positive correlation between wheat consumption and HLA-DQ8 frequency, although this did not reach statistical significance (Rho = 0.5, p = 0.06). A significant correlation between the frequency of HLA DQ2 and the duration of wheat consumption according to the pattern of wheat culture spreading was found (Fig. 3). In the same four areas, the duration of wheat consumption was significantly correlated with the level of current wheat consumption (Fig. 4).
Fig. 1Correlation between the level of wheat consumption and the frequency of HLA-DQ2 in different countries of the world.
Fig. 3Correlation between the frequency of HLA DQ2 and the duration of wheat consumption in four different areas of Middle East and Europe. Countries are grouped according to the duration of wheat consumption from 1 (gluten-containing cereals introduced around 10,000 years ago) to 4 (gluten-containing cereals introduced 6000–4000 years ago). 1: Turkey, Iran, Israel; 2: Tunisia, Italy, Greece; 3: Germany; 4: Sweden, Finland, Ireland.
Fig. 4Correlation between the current level of wheat consumption and the duration of wheat consumption in four different areas of Middle East and Europe. Countries are grouped according to the duration of wheat consumption from 1 (gluten-containing cereals introduced around 10,000 years ago) to 4 (gluten-containing cereals introduced 6000–4000 years ago). 1: Turkey, Iran, Israel; 2: Tunisia, Italy, Greece; 3: Germany; 4: Sweden, Finland, Ireland.
A detailed analysis of within-continent distribution showed that gluten and HLA-DQ2 co-segregated not only in Europe/Middle East, but also in other Continents. In India wheat consumption is higher in Northern-Western states such as Punjab, Haryana, Delhi, Rajasthan, Uttar Pradesh, Bihar and Madhya Pradesh (so-called “celiac belt”). Likewise, the DQ2 allele frequency is higher in North (31.9%) than in South India (12.78% for Piramalai Kallars and 9% for Yadavas) [
] and low level of wheat consumption. The frequency of the DQB1*0201 allele is higher in the northern Chinese than in the southern Chinese populations, and the consumption of wheat in the northern area is higher than in the southern area, where rice is the staple diet [
The multivariable linear regression analysis did not show any significant effect of predisposing genes and wheat consumption on the prevalence of CD.
4. Discussion
Our study shows that CD is a common condition in different continents, including many developing areas, with a mean prevalence of 0.9% worldwide. We did not find a significant correlation between the prevalence of CD and predisposing HLA genes, a finding that can be explained, at least in part, by a certain degree of data heterogeneity. For example, the Australian and New Zealand HLA-DQ allele frequencies derived from individuals of Aboriginal and Maori origin, while studies in the same countries on CD prevalence were performed in individuals of prominently European origin. Although a trend of positive correlation between the prevalence of CD and the level of wheat consumption was identified, this also did not reach statistical significance, probably due to non-linear correlation between the level of gluten intake and the risk of CD development. A dose–effect relationship between gluten intake and development of CD has been previously shown, but only at very low gluten intakes (ranging between 50 and 500 mg of daily gluten) [
]. Although the worldwide variation of wheat consumption is high, the current intake is well above 500 mg per day in most countries, therefore minimizing the effect that the level of gluten intake may have on the prevalence of CD.
The analysis of correlation between wheat consumption, HLA-DQ2 and DQ8 frequency and CD prevalence had already been performed by Abadie and coworkers in their recent review paper [
]. However our work took this analysis a step forward and showed a previously unreported strong correlation between the geographical distribution of HLA CD-predisposing genotypes and wheat consumption. The paradoxical distribution of these two CD-causative factors is best exemplified by findings in the Europe/Middle East area. Domestication of gluten-containing cereals (particularly ancient varieties of wheat and barley) began approximately 10,000 years ago in Neolithic settlements in the north-eastern (Turkey, Iran, and Iraq) and in the south-western (Palestine, Syria, and Lebanon) regions of the so-called “Fertile Crescent” area. Cultivation of wheat and barley slowly spread northwest ward across Europe to reach Western countries, as Britain, “only” 4000 years ago [
]. The agriculture revolution paved the way to new diseases, such as CD. Since treatment with the gluten-free diet was not available in the ancient time, the CD geno/phenotype exerted a negative impact on human reproductive fitness and survival over thousand years. As a consequence of a prolonged negative selective pressure, CD and its predisposing genotypes should be less common in Middle East than in European countries (as originally theorized by Simoons) [
]. On the contrary, nowadays CD is equally common in UK and Turkey, and the frequency of HLA-DQ2 (the major CD-predisposing haplotype) is even higher in Turkey and Iran, i.e. countries consuming more wheat and for a longer duration of time, than in Finland and Ireland, i.e. countries consuming less gluten and for a shorter duration of time. This “evolutionary celiac paradox” is strengthened by the co-localization of gluten consumption and HLA-DQ2 we found in other areas of the world, e.g. Africa and India. To explain these counterintuitive findings, we suggest that the diffusion of HLA CD-predisposing genotypes and the diffusion of CD occurred in different times (Fig. 5). The currently high prevalence of CD is most likely a recent event, explained by recent changes in environmental conditions, e.g. related to the quantity and quality of ingested gluten, dough fermentation, human intestinal microbioma and pattern of infant nutrition [
Fig. 5The chain of events eventually leading to the current high prevalence of celiac disease: (1) increasing gluten intake in the early era of the agriculture revolution in Middle East and then Europe; (2) increased prevalence of HLA-DQ2 due to demic population diffusion and/or positive selection caused by adverse effects of wheat consumption on health, e.g. dental caries; (3) in modern time, e.g. the last centuries, other environmental changes triggered the surge of celiac disease in the general population worldwide (length, height and slant of the lines are arbitrary). KYA, kiloyear (thousand years); CD, Celiac disease.
On the other hand, the parallel geographical distribution of gluten intake and HLA-DQ2 and DQ8 genotypes could be explained by different mechanisms, which are not mutually exclusive:
(a)
a founder effect reflecting higher prevalence of DQ2/DQ8 in the early farmers that 10,000 years ago spread from Middle East and gradually colonized Europe determining a South/East–North/West genetic gradient (so called “demic diffusion” model of Cavalli-Sforza and Ammerman) between Middle East and Europe [
] (local effect). This slow “wave of advance” of a population of early farmers with a high frequency of DQ2 and DQ8 could lead to progressive dilution of their genetic background by mixing with the local European populations;
(b)
a positive selection of HLA CD-predisposing genotypes pushed by protection against deleterious conditions driven by the agricultural revolution (worldwide effect). An extensive search of the literature showed that only one disease is partially “protected” by an HLA-DQ, CD-predisposing haplotype (DQ2), that is dental caries. A recent Brazilian study showed that adolescents positive for HLA-DQ2 allele were less likely to have dental caries than their counterparts who were negative for this allele (OR = 0.33, CI 0.16–0.66) [
]. Interestingly, also CD patients show a reduced prevalence of dental caries when compared to controls, despite high incidence of dental enamel hypoplasia [
]. Streptococcus mutans, the principal agent responsible for oral fermentation of sugars leading to dental caries, is less abundant in the intestinal mucosa of CD children, regardless of inflammatory status [
]. The consumption of wheat is positively correlated to the development of dental caries whereas rice, the typical gluten-free cereal, shows less cariogenic potential [
]. In the context of the ancestral wheat-consuming society, when cure and prevention of dental caries were basically non-existent, some sort of genetically determined resistance to dental caries could well drive a positive selection of HLA-DQ2. It remains to be elucidated how DQ2 positivity protects from dental caries development, e.g. by accelerating the clearance of “sticky” (and possibly cariogenic) gluten peptides from the mouth and teeth surface. The DQ2-dental caries protection hypothesis needs to find confirmation by further population studies.
Our (b) explanation is more generalizable than (a), and could explain the high gluten/high DQ2 co-localization in different, apparently unrelated, areas of the world such as Europe, Africa and India. Thus, the highest prevalence of DQ2 among wheat consuming populations could be another example of evolutionary adaptation driven by the dietary changes imposed by the agricultural revolution, in analogy to lactase persistence in milk drinking populations [
]. In this context CD could simply represent the modern “dark side” of a genetic structure (DQ2 positivity) conferring resistance to a negative consequence of a wheat-based diet.
We are aware of the limitations of our geo-epidemiological analysis: (a) current wheat consumption does not perfectly reflect historical wheat consumption (although both historical consideration and our data suggest that a strong correlation does indeed exist); (b) FAO data reflect individual availability of wheat (including unconsumed leftover) instead of individual wheat intake; (c) heterogeneity of HLA data does not allow to differentiate high-risk (e.g. double copy of the DQB1*02 gene) from lower risk genotypes; (d) CD prevalence data are based on different diagnostic methods and different age groups; (e) regional differences in the consumption of CD-triggering grains different from wheat (rye and barley) have not been investigated; (f) the possible evolutionary role of non-HLA predisposing genes has not been evaluated. At least 39 non-HLA CD predisposing genes have been identified, most of them influencing the inflammatory and immune pathways. Recent studies showed consistent signs of positive selection for CD-associated non HLA alleles, i.e. IL12A, IL18RAP, and SH2B3 [
In conclusion, we suggest that the current high prevalence of CD is the last link in a chain of events started about 10,000 years ago after wheat domestication and diffusion from the Middle East. To explain the so-called “evolutionary celiac paradox” of co-localization of gluten consumption and HLA CD predisposing genotypes, we suggest a primary role for positive selection of the major CD-predisposing haplotype, that is HLA-DQ2, in wheat-consuming populations. The force driving HLA-DQ2 selection could have been protection against dental caries, a disease causing a tremendous negative impact on the health and reproductive fitness of ancient populations. The “DQ2-dental caries protection” hypothesis needs to find confirmation by further population studies.
Conflict of interest
Carlo Catassi served as consultant for Menarini diagnostics s.r.l., for Dr Shaer, and for Heinz Company. Elena Lionetti served as consultant for Heinz Company.
Appendix A. Supplementary data
The following are the supplementary data to this article:
00600-8&id=gr1.gif){kind=link}
00600-8&id=gr2.gif){kind=link}
00600-8&id=gr3.gif){kind=link}
00600-8&id=gr4.gif){kind=link}
00600-8&id=gr5.jpg){kind=link}
00600-8&id=mmc2.jpg){kind=link}
00600-8&id=mmc3.jpg){kind=link}
00600-8&id=mmc4.jpg){kind=link}
00600-8&id=mmc5.jpg){kind=link}