Ethnic Variations in Duodenal Villous Atrophy Consistent with Celiac Disease in the United States

Abstract

Background & Aims

Celiac disease is a common disorder with a worldwide distribution, though the prevalence among different ethnicities varies. We aimed to measure the prevalence of duodenal villous atrophy among patients of different ethnicities throughout the United States (US).

Methods

We performed a cross-sectional study of all patients who had duodenal biopsies submitted to a national pathology laboratory between January 2, 2008 and April 30, 2015. The prevalence of villous atrophy was calculated for the following ethnicities, using a previously published algorithm based on patient names: North Indian, South Indian, East Asian, Hispanic, Middle Eastern, Jewish, and other Americans.

Results

Among all patients (n=454,885), the median age was 53 years and 66% were female. The overall prevalence of celiac disease was 1.74%. Compared to other Americans (n=380,163; celiac disease prevalence 1.83%), celiac disease prevalence was lower in patients of South Indian (n=177, 0%; P=.08), East Asian (n=4700, 0.15%; P=<.0001), and Hispanic (n=31,491, 1.06%; P=<.0001) ethnicities. Celiac disease was more common in patients from the Punjab region (n=617, 3.08%) than in others from North India (n=1195, 1.51%; P=.02). The prevalence of celiac disease among patients of Jewish (n=17,806, 1.80%; P=.78) and Middle Eastern (n=1903, 1.52%; P=.33) ethnicities was similar to that of other Americans. Among Jewish individuals (n=17,806), the prevalence of celiac disease was 1.83% in Ashkenazi persons (n=16,440) and 1.39% in Sephardic persons (n=1366; P=.24).

Conclusions

Among patients undergoing duodenal biopsy, individuals from the Punjab region of India constitute the ethnic group in the US with the highest prevalence of villous atrophy consistent with celiac disease. Compared to other Americans, villous atrophy prevalence on duodenal biopsy is significantly lower among US residents of South Indian, East Asian, and Hispanic ancestry.

Introduction

Celiac disease (CD) is an immune-based disorder triggered by the consumption of gluten in genetically susceptible people who are subject to as-yet unidentified environmental triggers. (1) A recent study found that the overall prevalence of CD in the general population of the United States (US) is 0.7%, which is equal to approximately 1.8 million Americans. (1) When initially characterized, CD was thought to be a disease of Caucasian Europeans, though it is now recognized as one of the most common genetic disorders with a worldwide distribution. However, the prevalence in different ethnicities varies. (2) The prevalence of CD among Europeans is thought to be about 1-1.5% (2) with a similar estimated prevalence of about 1.1% in the adult Israeli population (3) and 1.2% in the United Arab Emirates, (4) while the disease appears to be less common in Indonesia, (2) South Korea, (2) and the Philippines (2), which may be related to the lower consumption of wheat in those populations. A retrospective study from the Northern part of India reported a significant increase in the prevalence of CD in the past decade. (5) In one study of ethnic minorities with biopsy-proven CD at a pediatric clinic in Canada (n=54), South Asians were found to comprise a significant majority (81%) of the ethnic minorities with CD. (6) Celiac disease in the Asia-Pacific region is considered to be under-diagnosed, though there are expectations for this to change. (7)

Few studies have investigated racial and ethnic variation of CD prevalence in the US. African-Americans and Hispanics undergoing upper endoscopy are less likely to be biopsied than Caucasians; therefore, CD may be under diagnosed in these populations. (8) One serologic screening study that estimated the prevalence of CD in the US population found the disease to be predominantly present in non-Hispanic Caucasians and less common among Hispanics and non-Hispanic African-Americans. (1) There is also uncertainty regarding whether the female predominance observed in European studies of CD (9-11) applies to different ethnicities in the US.

In this study, we aimed to measure the prevalence of duodenal villous atrophy (the histologic hallmark of CD) among different ethnicities throughout the US. Using a large pathology database of duodenal samples from endoscopic procedures performed by US physicians and diagnosed by a central group of pathologists, we sought to quantify the prevalence of CD among individuals of different ethnic backgrounds, all of whom underwent duodenal biopsy. We also aimed to determine whether the gender distribution in CD differed between these ethnic groups.

Methods

Data source

We used a large national pathology database of subjects who underwent esophago-gastro-duodenoscopy (EGD) with duodenal biopsy between January 2, 2008 and April 30, 2015 in endoscopy centers distributed throughout the US. The mucosal biopsy specimens were evaluated and reported by a single group of gastrointestinal pathology fellowship-trained histopathologists at three different laboratories of Miraca Life Sciences. Pathologists participate in daily consensus conferences and each reviews specimens from multiple different states. All data were derived from pre-existing records. No direct contact with either patients or health care providers was made and no individual patient information was revealed. All patient records were de-identified before being analyzed.

Ethnicity categories

A series of computer algorithms based on first and last name analysis were used to categorize patients by ethnicity. This method of ethnic classification, modified from similar existing models, (12,13) and described in detail in a recent publication, (14) was first validated by a progressive process, which consisted of adjusting the algorithms against lists of persons of known ethnicity until the specificity was greater than 95%. This level of specificity compares favorably to that of self-reported ethnic classification (15,16) and is substantially more accurate than the assignment of ethnicity by visual inspection as determined by the IC codes used in the United Kingdom. (17) The last validation step, specific for this cohort of patients, included pre-arranged visits to medical practices where substantial numbers of patients of different ethnicities were recruited and had telephone interviews with practice managers. These visits and interviews, aimed at determining the level of coincidence between the ethnic categories assigned by our algorithm and the ethnicities recorded by the practices, revealed an essentially perfect concurrence. Using this approach, patients were stratified into the following ethnicities: North Indian (with further subdivision into Punjabis or Other North Indian), South Indian, East Asian, Hispanic, Middle Eastern, Jewish (with further subdivision into Ashkenazi or Sephardic), and Other Americans. The latter group served as a reference and included individuals (mostly Caucasians and African-Americans) not specifically associated with any of the other ethnic groups. Patients with a combination of names that suggested more than one ethnicity (3.7%) were classified as undetermined and excluded from further analysis.

Celiac disease

We calculated the prevalence of CD among each of the ethnic groups described above. Patients were considered to have CD if duodenal biopsies showed villous atrophy. We then calculated the prevalence of various degrees of villous atrophy: partial villous atrophy (corresponding to Marsh 3A) and subtotal or total villous atrophy (Marsh 3B and Marsh 3C). (18)

Statistical analysis

The distributions by age, gender, and ethnicity were calculated and expressed as a percentage of the total study population. The prevalence of CD among different ethnicities was compared using the chi square test, with the group “Other Americans” serving as a reference for all comparisons. Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated using logistic regression. We then recalculated ORs and 95% CIs, adjusting for age and gender. Because gastric colonization with H. pylori varies by ethnicity (19) and the presence of H. pylori correlates inversely with CD, (20) we then additionally adjusted for H. pylori status using a multivariate model restricted to those individuals who had a concurrent gastric biopsy.

The prevalence of CD between the genders was compared overall and then stratified by ethnicity. We used logistic regression to measure the association between female gender and CD using ORs and 95% CIs; we then adjusted for age and H. pylori status.

We used SAS (Cary, NC) version 9.4 for all analyses. All reported p values are 2-sided. The Institutional Review Board of Columbia University Medical Center deemed this “non-human subjects research” since the data were stripped of all identifiers before being provided to the investigators.

Results

During the study period, there were 458,256 unique individuals with duodenal biopsies. We excluded 11 for likely erroneous age (recorded as over 99 years). Additionally, we excluded 2,931 patients whose biopsies showed duodenal neoplasia and 429 patients whose biopsies showed Giardia lamblia. The remaining 454,885 patients served as our study population. Demographic information and histologic findings are summarized in Table 1. The median age was 53 years and the majority of patients (75%) were older than 40 years; 66% were female. The most common indications for duodenal biopsy were gastroesophageal reflux disease, dyspepsia/epigastric pain, anemia, and diarrhea (see Table 1). CD was diagnosed in 7,928 patients, equivalent to 1.74% of those who underwent duodenal biopsy. The prevalence villous atrophy consistent with CD varied by indication for biopsy; it was lowest (1.25%) among those with gastroesophageal reflux disease, and highest (2.04%) among those with diarrhea.

Table 1.

Demographics and histologic findings of patients undergoing duodenal biopsy (n=454,885)

	N (%)
Age
0-19	17353 (3.81)
20-39	95610 (21.02)
40-59	173267 (38.09)
60+	168655 (37.08)
Gender
Male	153145 (33.69)
Female	301404 (66.31)
Ethnicity
Undetermined	16833 (3.70)
Other Americans	380163 (83.57)
North Indian	1812 (0.40)
Punjabis	617 (34.05)
Other North Indians	1195 (65.95)
South Indians	177 (0.04)
East Asians	4700 (1.03)
Hispanics	31491 (6.92)
Middle Eastern	1903 (0.42)
Jewish	17806 (3.91)
Ashkenazi	16440 (92.33)
Sephardic	1366 (7.67)
Indications for biopsy*
Dyspepsia/epigastric pain	71,815 (16)
Anemia	68,663 (15)
Diarrhea	79,393 (17)
Weight loss	35,227 (8)
Gastroesophageal reflux disease	178,073 (39)
Other	108,014 (24)
Not listed	81,633 (18)
Celiac disease	7928 (1.74)
Concurrent gastric biopsy	375448 (82.54)
H. pylori	36405 (9.70)

^*Total is greater than 100% due to patients having multiple indications listed.

Table 2 shows the prevalence of villous atrophy consistent with CD by ethnicity. Compared to the prevalence of CD among Other Americans (1.83%), the lowest prevalence of CD was found among patients identified as South Indians (0/177, OR and CI not calculable), East Asians (0.15%, OR 0.08, 95% CI 0.04-0.17, p <0.0001), and Hispanics (1.06% OR 0.58, 95% CI 0.52-0.64, p <0.0001). These comparisons were essentially unchanged when CD was subdivided into partial villous atrophy and subtotal/total villous atrophy and when ORs were adjusted for age, gender, and H. pylori status (data shown in Table 2). Among North Indians, there was a trend toward higher prevalence (2.04%) when compared to Other Americans (1.83%) that did not reach statistical significance (OR 1.41, 95% CI 0.99-2.00, p=0.06).

Table 2.

Prevalence of villous atrophy consistent with CD by ethnicity and stratified by degree of villous atrophy

Ethnicity	Celiac disease (%)	OR	95% CI	p value	ORa	95% CI	p value	ORb	95% CI	p value
Celiac Disease
Other Americans	6943 (1.83)	1.00	[ref]	[ref]	1.00	[ref]	[ref]	1.00	[ref]	[ref]
North Indians	37 (2.04)	1.12	0.81-1.55	0.49	1.09	0.79-1.51	0.60	1.41	0.99-2.00	0.057
South Indian	0 (0.00)	NC	NC	NC	NC	NC	NC	NC	NC	NC
East Asians	7 (0.15)	0.08	0.04-0.17	<0.0001	0.08	0.04-0.17	<0.0001	0.12	0.06-0.25	<0.0001
Hispanics	334 (1.06)	0.58	0.52-0.64	<0.0001	0.57	0.51-0.64	<0.0001	0.71	0.63-0.80	<0.0001
Middle Eastern	29 (1.52)	0.83	0.58-1.20	0.33	0.81	0.56-1.16	0.25	1.01	0.67-1.53	0.96
Jewish	320 (1.80)	0.98	0.88-1.10	0.78	0.99	0.89-1.11	0.90	1.04	0.91-1.19	0.61
Partial villous atrophy
Other Americans	3410 (0.90)	1.00	[ref]	[ref]	1.00	[ref]	[ref]	1.00	[ref]	[ref]
North Indians	15 (0.83)	0.92	0.55-1.54	0.76	0.91	0.55-1.51	0.72	1.30	0.78-2.16	0.32
South Indian	0 (0.00)	NC	NC	NC	NC	NC	NC	NC	NC	NC
East Asians	5 (0.11)	0.12	0.05-0.28	<0.0001	0.12	0.05-0.29	<0.0001	0.17	0.07-0.40	<0.0001
Hispanics	220 (0.70)	0.78	0.68-0.89	0.0003	0.78	0.68-0.89	0.0003	0.96	0.82-1.11	0.60
Middle Eastern	19 (1.00)	1.11	0.71-1.75	0.64	1.10	0.70-1.73	0.68	1.40	0.85-2.30	0.18
Jewish	159 (0.89)	1.00	0.85-1.17	0.99	1.01	0.86-1.18	0.94	1.06	0.88-1.28	0.52
Subtotal/Total villous atrophy
Other Americans	3533 (0.93)	1.00	[ref]	[ref]	1.00	[ref]	[ref]	1.00	[ref]	[ref]
North Indians	22 (1.21)	1.31	0.86-2.00	0.21	1.26	0.83-1.92	0.28	1.51	0.94-2.45	0.09
South Indians	0 (0.00)	NC	NC	NC	NC	NC	NC	NC	NC	NC
East Asians	2 (0.04)	0.05	0.01-0.18	<0.0001	0.05	0.01-0.19	<0.0001	0.07	0.02-0.28	0.0002
Hispanics	114 (0.36)	0.39	0.32-0.47	<0.0001	0.38	0.32-0.46	<0.0001	0.45	0.36-0.56	<0.0001
Middle Eastern	10 (0.53)	0.56	0.30-1.05	0.07	0.54	0.29-1.00	0.05	0.62	0.29-1.30	0.21
Jewish	161 (0.90)	0.97	0.83-1.14	0.73	0.99	0.84-1.16	0.86	1.01	0.83-1.23	0.94

^a= adjusted for age and gender;

^b=adjusted for age, gender, and H. pylori status

NC = Not calculated due to insufficient number of patients with celiac disease

Of the 1,812 patients with North Indian origin, 617 were Punjabis; 19 (3.08%) of Punjabi patients had villous atrophy consistent with CD. The prevalence of CD was significantly higher in Punjabis (3.08%) than that in Other North Indian patients (3.08% vs 1.51% (18/1,195); p=0.02) Among Jewish individuals (n=17,806), the prevalence of CD was 1.83% (301/16,440) in Ashkenazi subjects and 1.39% in Sephardic subjects (19/1,366, p=0.24).

The distribution of villous atrophy consistent with CD by gender and ethnicity is shown in Table 3. Although 5,338 (67%) of the patients with CD were female, this apparent majority was due to the fact that females comprised 66% of all individuals undergoing duodenal biopsy and the prevalence of CD was nearly identical in males and females (1.7% and 1.8%, respectively). The similar prevalence of CD between genders was present across all ethnicities, although there was a non-significant trend toward female predominance in North Indian, Hispanic, Middle Eastern, and Jewish patients. (Table 3)

Table 3.

Distribution of villous atrophy consistent with CD by Gender and Ethnicity

Ethnicity	Number with celiac disease (%)	OR	95% CI	p value	ORa	95% CI	p value	ORb	95% CI	p value
Other
Overall	6936 (1.83)
Men	2290 (1.79)	1.00	[ref]	[ref]	1.00	[ref]	[ref]	1.00	[ref]	[ref]
Women	4646 (1.84)	1.03	0.98-1.08	0.25	1.02	0.97-1.08	0.36	0.99	0.93-1.05	0.71
North Indian
Overall	37 (2.04)
Men	15 (1.62)	1.00	[ref]	[ref]	1.00	[ref]	[ref]	1.00	[ref]	[ref]
Women	22 (2.49)	1.55	0.80-3.01	0.19	1.54	0.79-2.98	0.21	1.49	0.73-3.04	0.27
Hispanic
Overall	333 (1.06)
Men	82 (0.90)	1.00	[ref]	[ref]	1.00	[ref]	[ref]	1.00	[ref]	[ref]
Women	251 (1.12)	1.26	0.98-1.61	0.07	1.27	0.99-1.63	0.07	1.22	0.92-1.61	0.16
Middle
Eastern
Overall	29 (1.53)
Men	14 (1.31)	1.00	[ref]	[ref]	1.00	[ref]	[ref]	1.00	[ref]	[ref]
Women	15 (1.81)	1.39	0.67-2.89	0.38	1.40	0.67-2.91	0.3744	1.30	0.57-2.97	0.54
Jewish
Overall	319 (1.79)
Men	99 (1.55)	1.00	[ref]	[ref]	1.00	[ref]	[ref]	1.00	[ref]	[ref]
Women	220 (1.93)	1.26	0.99-1.59	0.063	1.26	0.99-1.60	0.063	1.33	0.997-1.77	0.053

^a= adjusted for age and gender;

^b=adjusted for age and H. pylori status

Figure 1 shows the prevalence of villous atrophy consistent with CD by age, stratified by ethnicity. The distributions were fairly even among the groups where CD was more prevalent. There was an increase in CD among Jewish and North Indian patients in the youngest age group (0-19 years), although comparisons of the ethnic groups in this age stratum did not yield statistically significant differences due to the low number of children with CD in these groups.

Figure 1.

Prevalence of villous atrophy consistent with celiac disease by age, stratified by ethnicity

Discussion

In our analysis of over 400,000 duodenal biopsies from a nationwide pathology database, we found that the prevalence of CD in those undergoing duodenal biopsy was lower in patients identified as South Indian, East Asian, and Hispanic when compared to Other Americans. North Indian patients identified with ancestry in the Punjab region had a significantly higher prevalence of CD on duodenal biopsy compared to all Other North Indian patients. There were no significant differences in prevalence of CD between Middle-Eastern and Jewish patients when compared with Other Americans.

In this population, the prevalence of CD was 1.74%, slightly more than double the prevalence reported in the screening studies. (1,21) Our study population consisted of patients undergoing duodenal biopsy for various indications, including symptoms clinically suggestive of CD. Significantly fewer Hispanic and East Asian patients were found to have CD, which is consistent with prior reports. (2,21-22) Susceptibility to CD is predominantly associated with the human leukocyte antigen HLA-DQ2, which varies geographically and is found in higher frequency in Western Europe and in portions of Africa and India. (23) In studies of CD in India, the prevalence of compatible HLA haplotypes is similar to those in Western countries, and does not vary substantially between regions. (24) Large regional variation in the wheat consumption in India, (24) is possibly a more significant reason to explain why cases of CD in India are primarily reported from Northern regions, with only isolated case reports from the rest of the country (25) and virtually no cases reported in Southern India, (24) in keeping with the findings of our study. Our finding of a higher prevalence of CD in patients with Punjabi ancestry is also consistent with previous reports. (26)

Our study population of patients undergoing duodenal biopsy was majority female, which is consistent with prior reports in this setting and elsewhere that women undergoing EGD are more likely to have duodenal biopsies than men. (8,27) However, we found that CD was equally prevalent among men and women undergoing duodenal biopsy, which was true in all ethnic groups studied. Several screening studies of CD in the US have shown that CD is equally prevalent among men and women, (28-30) but screening studies of children in the US (31) and elsewhere (10,11) have shown a female predominance. Regardless of whether gender affects the true prevalence of CD, women are more likely to be diagnosed with CD than are men. (32) Our findings support the notion that CD should be considered as a diagnosis in men as often as it is considered in women.

We found no significant difference in the prevalence of CD on duodenal biopsy between patients of Ashkenazi and Sephardic origin. While the high prevalence of inflammatory bowel disease in Ashkenazi Jews is well documented, we are not aware of any studies investigating the prevalence of CD in Sephardic versus Ashkenazi Jews. One study of the prevalence of CD among the adult Jewish population in Israel included only 850 subjects and did not differentiate between Ashkenazi and Sephardic ancestry. (3) Our comparison may have been limited by the small number of patients of Sephardic ancestry in study population.

This study has several strengths, including its large sample size and uniform reporting of histologic findings, as all biopsies were read and reported by a central group of pathologists with subspecialty training in gastrointestinal pathology, who practice in the same environment, use uniform diagnostic criteria and standardized diagnostic codes, and participate in daily consensus conferences where cases and diagnostic criteria are discussed. Upon review of the reporting of villous atrophy by different pathologists on the same specimen, there was good to excellent agreement for variable villous atrophy (Marsh 3a) and villous atrophy (Marsh 3b and 3c). As such, diagnosis of duodenal biopsies consistent with CD was very consistent across all pathologists. Pathology specimens came from multiple centers around the country; thus patients in our study population were representative of the US general population and allowed us to generate true prevalence data among patients undergoing duodenal biopsy. Although some geographic regions have a higher proportion of certain ethnicities, and it is indeed possible that certain pathologists see more patients of a certain group, this is unlikely to have biased our results. We found no distinct geographic predominance with regards to patients of Indian, Jewish, or Middle Eastern descent. As such, there were essentially equal chances that any pathologist interpreted biopsies from these ethnicities. The largest proportions of East Asian patients in our patient population are in New York, New Jersey, California, Alaska, and Hawaii. These five states have more than 20 pathologists who share the diagnostic work. Similarly, Hispanic patients are distributed almost equally in California, the Southwest (including Texas), and the Northeast. Therefore, it is extremely unlikely that all of the pathologists interpreting biopsies from these different states have a bias for a low rate of CD diagnosis.

Our study has several limitations. We were able to measure villous atrophy but not the clinical entity of CD. As we had no serologic data on patients with duodenal biopsies that showed villous atrophy, it is possible that some patients may have been misclassified as having CD, although even the most common cause of seronegative villous atrophy is still CD. (33) Nevertheless, some patients with alternative causes of villous atrophy (such as tropical sprue (33) or sprue-like enteropathy due to olmesartan (34)) would have been classified as having CD in this analysis. In particular, multiple studies have shown that tropical sprue is still the most common cause of malabsorption syndrome in India (35,36), while CD is emerging as a more important cause of malabsorption than previously thought (35-37). Still, such cases of tropical sprue and sprue-like enteropathy due to olmesartan are far less common than CD in the US. (38,39) Our study population only included those undergoing duodenal biopsy; thus our prevalence calculations do not include those patients who may be diagnosed with CD based on serology and symptoms alone, nor do they take into account undiagnosed CD. As ethnicity was derived based on a name-based algorithm, misclassification of ethnicity is possible. For example, the proportion of patients in our sample classified as Hispanic was 6.9%, far lower than the prevalence of 16.3% based on self-report in the 2010 U.S. Census. (40) However, such misclassification would bias our results towards the null, as it is unlikely that misclassification is differential by CD status. Therefore, it is possible that the prevalence of CD differs by ethnicity to a greater extent than reported in this study. Misclassification was mitigated in part by our excluding patients whose names were deemed ambiguous or dual-classified by our algorithm. Another limitation to the name-based algorithm is the lack of data on year of immigration to the US, which would help inform if and when dietary and other environmental exposures affect the risk of CD. While the national setting enhances the generalizability of our findings, the pathology specimens were submitted from private offices and ambulatory surgical centers and not from hospital-based endoscopy suites, raising the possibility that these data are not entirely representative of the US population.

In conclusion, we found that in the US, the prevalence of CD in those undergoing duodenal biopsy is significantly lower among patients of South Indian, East Asian, and Hispanic descent. Among patients of North Indian descent undergoing duodenal biopsy, CD is significantly more common in those from the Punjab region than in all other patients from North India. Patients of Jewish and Middle Eastern ethnicity had CD prevalence similar to that of other Americans. Men and women had a similar prevalence of villous atrophy on duodenal biopsy, regardless of ethnicity. These findings may have clinical relevance to gastroenterologists across the US and may aid in their diagnostic practices.

Abbreviations used in this article

CD

celiac disease

EGD

esophagogastroduodenoscopy

OR

odds ratio

CI

confidence interval

US

United States