Prevalence of celiac disease in low and high risk population in Asia–Pacific region: a systematic review and meta-analysis

This systematic review and meta-analysis study was conducted to estimate the pooled prevalence of CD in low and high risk groups in this region. Following keywords were searched in the Medline, PubMed, Scopus, Web of Science and Cochrane database according to the MeSH terms; celiac disease, prevalence, high risk population and Asian-Pacific region. Prevalence studies published from January 1991 to March 2018 were selected. Prevalence of CD with 95% confidence interval (CI) was calculated using STATA software, version 14. The pooled sero-prevalence of CD among low risk group in Asia–Pacific region was 1.2% (95% CI 0.8–1.7%) in 96,099 individuals based on positive anti-tissue transglutaminase (anti-t-TG Ab) and/or anti-endomysial antibodies (EMA). The pooled prevalence of biopsy proven CD in Asia–Pacific among high and low risk groups was 4.3% (95% CI 3.3–5.5%) and 0.61% (95% CI 0.4–0.8%) in 10,719 and 70,344 subjects, respectively. In addition, the pooled sero-prevalence and prevalence of CD in general population was significantly higher in children compared with adults and it was significantly greater in female vs. male (P < 0.05). Our results suggest high risk individuals of CD are key group that should be specifically targeted for prevention and control measures, and screening may prove to have an optimal cost–benefit ratio.


Records after duplicate, English
language, year and region of publication removed (n=152) Full text articles assessed based on inclusion criteria (n=66) Studies included in the meta-analysis (n=61) Records excluded based on exclusion criteria (n=86) Records excluded based on not available the full text, high risk of bias and study population prior to January 1991 (n=5) Sero-prevalence   www.nature.com/scientificreports/ CD suggestion a pooled prevalence of biopsy-proven CD among not at-risk population to be 0.61% (95% CI 0.4-0.8%, I 2 = 84%, P < 0.01) (Fig. 3A). The I 2 test indicated significant heterogeneity among the studies.
Age and gender-based difference in the prevalence of biopsy-confirmed CD among not low risk group. In low risk population, eight studies reported prevalence of biopsy-confirmed CD in children, 12 studies in adults, and 1 study reported the prevalence in both adults and children combined. The pooled prevalence of CD in low risk children was 0.66% (95% CI 0.4-0.9%, I 2 = 82.3%, P < 0.001), in adults was 0.55% (95% CI 0.3-0.8%, I 2 = 82.5%, P < 0.001) and combined adults and children was 1.04% (95% CI 0.8-1.2%). The pooled prevalence of biopsy-confirmed CD among low risk population in children was significantly higher than adults (0.6% vs. 0.5%, P = 0.022). The I 2 tests indicated similar heterogeneity among the studies reporting CD prevalence for adults and children. Subgroups analyses for pooled prevalence of CD are presented in Table 4. Gender based prevalence of biopsy-confirmed CD in not at-risk group was reported in 15 studies. Pooled prevalence of biopsy-confirmed CD in not at-risk group in males and females were 0.53% (95% CI 0.3-0.7%, I 2 = 75.5%, P < 0.001) and 0.74% (95% CI 0.5-0.9%, I 2 = 48.3%, P < 0.001), respectively, and it was significantly higher in females than that in males (P = 0.04). The test of heterogeneity showed significant heterogeneity in the prevalence of CD in males and not in females (Table 4).
Geographical difference in the prevalence of CD in at risk group. Of all studies included in the present metaanalysis that reported the CD prevalence among at-risk population, 2 studies originated from Oceania, 21 from Middle-East, 3 from South-Asia, and 3 from East-Asia. The pooled prevalence of CD were 4.2% (95% CI 3.3-5.1%, I 2 = 1%, P < 0.001) in Oceania, 4% (95% CI 2.9-5.3%, I 2 = 75.2%, P = 0.002) in Middle-East, 7.1% (95% CI 1.3-16.7%, I 2 = 70.4%, P = 0.002) in South-Asia and 3.6% (95% CI 0.00-27.7%, I 2 = 90.7%, P = 0.017) in East-Asia (Table 4). The pooled prevalence of CD was significantly higher in high risk people in South-Asia as compared with the other region (P = 0.01). There were no other significant differences in the prevalence of CD in other regions. The I 2 test indicated significant heterogeneity among the studies in Middle-East, South-Asia and East-Asia.

Exploration of heterogeneity.
We performed meta-regression analysis to find the source of heterogeneity among the studies in low and high risk population ( Table 5). The year of the study, sample size of each study, age and gender of the subjects were used for exploration the heterogeneity. While there was an inverse association between age of the participant and the prevalence of CD in low risk group of patients (P = 0.02), no such association was observed in high risk population. In addition, we performed meta-analysis according to risk of bias for studies including low and moderate risk. Pooled prevalence of CD according to risk of bias and also heterogeneity test are shown in Table 6.
Publication bias. The result of Egger test showed presence of publication bias for studies conducted on healthy population (P = 0.009) and also showed presence of publication bias for studies conducted on at-risk Table 4. Subgroup analysis for pooled prevalence of CD in Asian-Pacific region among at-risk and not at-risk populations. DM1 Diabetes Mellitus type1, DS Down syndrome, IBD Inflammatory bowel disease, ATD autoimmune thyroiditis diseases, FDR first-degree relatives. www.nature.com/scientificreports/ population (P = 0.003). Funnel plots have shown asymmetric mood and confirmed presence of publication bias (Fig. 4).

Discussion
To our knowledge, this is the first meta-analysis to examine the prevalence of CD in the Asia-Pacific region and to compare it between low and high risk groups. Considering that two previous studies in this region have shown only the prevalence of CD in the general population 4,9 , our findings represent the best approximation of the pooled prevalence of CD in low and high risk groups according to age (adult and children), gender (male and female) and geographical categories (Oceania, Middle-East, East-Asia, and South-Asia) in the Asia-Pacific region.
Our results revealed that the pooled sero-prevalence of CD among general population was (1.2%), and the pooled prevalence of biopsy-confirmed CD in high risk and low risk groups was (4.3%) and (0.61%) respectively. So, the pooled prevalence of CD was significantly higher in high risk population compared to low risk subjects (P < 0.001). Sero-prevalence and biopsy-confirmed prevalence in Asian-Pacific countries varied from 0.06% in Turkey to (2.8%) in Saudi Arabia and (0.05%) in Japan and 1.4% in India, respectively. The analysis of CD prevalence within 4 geographical categories of Oceania, Middle-East, East-Asia, and South-Asia showed the highest prevalence of CD among low and high risk population was in the South-Asia (0.8%) and (7.1%), respectively. While, the highest sero-prevalence of CD was reported in Middle-East countries (1.4%).
Our findings suggest that CD is a much greater problem in the Asia-Pacific region than has previously been appreciated. The prevalence of CD in this region, both in low and high risk groups, was similar and comparable to its prevalence in Europe and the United States [62][63][64] . Our results showed that the pooled prevalence of CD among FDR (4.8%) in 1009 individual, were similar to those reported in previous studies in the US and Europe between (4.5%) and (10%) [64][65][66][67][68] . We found that the prevalence of CD in ATD patients was higher compared with the general   www.nature.com/scientificreports/ population (2.9% vs. 0.6%) and that the risk of CD can be increased by about 4-5 times in ATD subjects. This is slightly higher than global pooled prevalence of biopsy proven CD (1.6%) reported by Roy et al. 69 71 .
In addition, we evaluated the prevalence of CD in children and adults with symptoms associated with CD incudes, diarrhea and abdominal pain. The pooled prevalence of CD in patients with chronic diarrhea estimated 8.4% in the study based on two studies from Iran and China on 942 children 36,44 . According to the data presented in the study, CD is common among patients labeled as chronic diarrhea especially in children. Given that CD may be missed or diagnosed late in children with chronic diarrhea, immunological screening with the subsequent morphologic study of the small intestine is recommended to all patients with the chronic diarrhea syndrome to enable the early diagnostics of CD 72 .
Our analyses revealed a significant heterogeneity in prevalence of CD among low and high risk groups from different countries in Asia-Pacific region. To explore this heterogeneity we examined subgroups of studies such as year of study, sample size, age and gender. Meta-regression analysis has confirmed that CD prevalence in low risk groups decreased with age at testing and female gender. While, in high risk population did not found any association between age or gender and prevalence of CD in Asia-Pacific region. The prevalence of CD in high risk adults was significantly higher than in children, suggesting a link between the duration of gluten consumption and the development of an immune response to gluten. Therefore, heterogeneity was substantially reduced when sero-prevalence and prevalence of CD in not at-risk populations was calculated separately for men/women and adult/children. The heterogeneity reported in the prevalence of CD in this study is partly due to methodological differences between studies which include the type of diagnostic (serology/biopsy test) and study population (adults/children). It is likely that prevalence of CD also varies from country to country in Asia-Pacific, because of diverse dietary practices and prevalence of predisposing HLA-DQ2/HLA-DQ8 haplotypes in the general population 9,73 .
While the present study reports a pooled prevalence of CD in Asian-Pacific region among low and high risk population for the first time, this meta-analysis has a few limitations too. Studies on the prevalence of CD in general population are available only from 13 countries in this region. Therefore, the lack of population-based prevalence data from many countries (Azerbaijan, Kazakhstan, Turkmenistan, Kyrgyzstan, Tajikistan, Cambodia, Vietnam, Mongolia, Hong Kong, Sri Lanka, Myanmar, Maldives, Nepal, and Bhutan) in the Asia-Pacific region is a major limitation. Another limitation was the most studies in this region reported the prevalence of CD based on the serology and even if the biopsies were performed in seropositive individuals, only small proportion of patients underwent biopsies. So, we had to exclude a lot of studies based on our inclusion and exclusion criteria.
In conclusion, we have undertaken the first meta-analysis study in low and high risk population in the Asia-Pacific region. Our results suggest that CD is common in Asian-Pacific region and pooled sero-prevalence and prevalence of biopsy-confirmed CD in low risk groups was 1.2% and 0.6%, respectively, which is similar to Western countries. In addition, the prevalence of CD in high risk population was significantly higher than low risk group (4.3% vs. 0.6% P < 0.001). High risk individuals of CD are key group that should be specifically targeted for prevention and control measures, and screening may prove to have an optimal cost-benefit ratio.

Methods
We developed a protocol, including eligibility criteria, search strategies, criteria for study selection and methods for extracting data according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines 74 .
Search strategy. Previously published papers indexed in Medline (National Library of Medicine), Pub-Med, Scopus, Web of Science (Thomson Reuters; New York, USA), and Cochrane Library (Cochrane Collaboration; Oxford, United Kingdom) were searched for this systematic review and meta-analysis with the following MeSH terms and keywords: "Celiac diseases", "Coeliac disease" and "Prevalence" alone or combination. To find prevalence of CD among high risk population search strategy was based on the words of CD prevalence in patients with "diabetes mellitus type 1", "chronic diarrhea", "autoimmune thyroid disease", "Down syndrome", "inflammatory bowel disease", "dyspepsia", and "first-degree relatives with CD". Each one was cross-referenced with "Asia-Pacific region" and countries in this region such as Australia, New Zealand, India, Pakistan, Turkey, Iran, etc. The first recommendations for diagnosis of CD were published by the European Society for Pediatric Gastroenterology Hepatology and Nutrition (ESPGHAN) in 1990 75 , which we considered this year as a dividing year for well-defined diagnostic criteria for CD and other gluten-related disorders. All related articles published between January 1991 and March 2018 were, therefore, included in this review. The search for the studies was performed in English and this analysis did not include those without access to the full text. Moreover, in order to conclude the qualifying studies, all reference lists of relevant publications were also reviewed and the retrieved references were also disregarded due to duplication. To exclude unrelated studies with no eligibility requirements, the names, abstracts, as well as full texts were carefully read. Diagnostic criteria for CD. The diagnosis of CD was based on a combination of at least one positive celiacspecific serological tests such as anti-tissue transglutaminase (anti-t-TG) antibodies, anti-endomysial antibodies (EMA) and deamidated gliadin peptides (DGP) antibodies, anti-gliadin antibody (AGA) and all confirmation villous atrophy by duodenum biopsy according to Marsh classification 76 . In addition, studies that reporting the sero-prevalence of CD in the healthy population (having a positive t-TG, EMA and DPG antibodies without biopsy confirmation) were analyzed separately. reviewed all the full texts, and individually evaluated the articles based on pre-decided inclusion and exclusion criteria. Moreover, the risk of bias was calculated using the risk of bias tool for prevalence studies developed by Hoy et al. 77 . Based on this tool, studies were assessed for external and internal validity using a 10-point checklist and grouped into a low, moderate, or high risk of bias. The studies with a score of less than 6 were considered to have a high risk, 6 to 8 was considered a moderate risk, and 9 to 10 was considered a low risk of bias. The studies with a high risk of bias were excluded from the present meta-analysis. Disagreements between two authors were resolved by discussion. In case disagreements persisted, third author (RN.M.) reviewed the study and made the final decision. To increase the quality of the review, a blind method was used hiding the authors name and name of the journal.

Data extraction.
Information was extracted separately about the sero-prevalence and biopsy confirmed prevalence of CD in at-risk and not at-risk populations in adults and children. Information contained the name of the first author, year of publication, place of study, demographic characteristics of study participants including; number, sex and age, the type of serological tests and duodenal biopsy. Based on our inclusion criteria, finally 61 articles including 19 articles on CD prevalence in not at-risk population, 29 on CD prevalence among at-risk population and 13 articles on sero-prevalence of CD in not at-risk population in English language from January 1991 to March 2018, which reported the prevalence or sero-prevalence of CD in Asia-Pacific region, were entered in this study.
Pooled prevalence and sero-prevalence of CD. Only studies in which 50 percent or more of seropositive individuals (those with positive anti-tTG and/or AEA) underwent a biopsy were included to measure the pooled prevalence of CD. The 50 percent discontinuity value was chosen because we assumed that the real prevalence of biopsy-proven CD was wrongly reduced among the studies in which less than 50 percent of positive individuals were subjected to biopsy. For the estimation of pooled sero-prevalence only, studies in which less than 50% of seropositive individuals underwent a biopsy were included.
Statistical analysis. We obtained pooled prevalence and sero-prevalence of CD in not at-risk and at-risk population, separately. Pooled prevalence of CD was obtained based on the proportion of individuals with CD and its confidence interval in each study. Prevalence was calculated assuming binomial distribution. In addition, we calculated prevalence of CD for subgroups such as region or sex. CD prevalence between groups was compared using chi-square test. For all pooled prevalence, the random model was used. I 2 statistics was employed to evaluate heterogeneity among studies. I 2 value > 50% was denoted as high heterogeneity. We applied the fixed effect model when the data were homogeneous. When the cause of heterogeneity was not known, the random effect model was used. To explore the sources of heterogeneity, meta-regression analysis was done. Moreover, Begg's test was carried out for recognizing publication bias. All analyzes performed by STATA 14.0 (STATA Corp; College Station, Texas, USA) software and significant level was considered as 0.05.