Chapter  104:  Celiac Disease

General Definitions

1. Classical CD. The most commonly described form. It describes patients with the classical features of intestinal malabsorption who have fully developed gluten-induced villous atrophy and the other classic histological features. These patients present because of GI symptoms, and are identified as CD sufferers through the investigation of these symptoms. This group can also be said to have symptomatic CD.

2. Atypical CD. Appears to be one of the most common forms. These patients generally have little to no GI symptoms, but seek medical attention because of another reason such as iron deficiency, osteoporosis, short stature, or infertility. These patients generally have fully developed gluten-induced villous atrophy. Because these patients are “asymptomatic” from the GI perspective, if their atypical CD feature is not recognized, they may be difficult or impossible to distinguish from “true” silent (asymptomatic) CD patients.

3. Silent CD. A very common form of CD. Refers to patients who are asymptomatic but are discovered to have fully developed gluten-induced villous atrophy after having undergone serologic screening or perhaps an endoscopy and biopsy for another reason. These patients are clinically silent, in that they do not manifest any clear GI symptoms or associated atypical features of CD such as iron deficiency or osteoporosis. These patients can be confused with atypical CD if their atypical features are not recognized in an early stage. As well, Fasano et al.¹⁵ have shown that many of these patients do not manifest fully developed villous atrophy.

4. Latent CD. Represents patients with a previous diagnosis of CD that responded to a GFD and who retain a normal mucosal histology upon later re-introduction of gluten. Latent CD can also represent patients with currently normal intestinal mucosa who will subsequently develop gluten-sensitive enteropathy.

5. Refractory CD. For the purpose of this review, patients with refractory CD are patients with true CD and villous atrophy (i.e., not a misdiagnosis) who do not, or no longer, respond to a GFD. Although the most common reason for failure to respond to a GFD is dietary indiscretion or unknown exposure to gluten, refractory CD also occurs in patients on a GFD who have developed a complication such as ulcerative-jejunoileitis, or enteropathy-associated lymphoma. Patients with refractory CD do not necessarily have positive serology for CD. Refractory CD was reviewed in the context of the requested objectives.

In order to utilize the above definitions, there needs to be clear and valid histological criteria for the diagnosis of CD. The histological patterns, particularly the more mild lesions, are not specific for CD and can be seen in a variety of other disorders (Table 1, Appendix A). To help standardize the histological criteria for the diagnosis of CD, several scoring systems have been developed. The classic Marsh criteria,¹ and its modification by Rostami,¹⁶ are presented in Table 2 (Appendix A). The revised ESPGAN criteria⁴ use histological, serological and clinical criteria (Table 3, Appendix A).

Histological

From the preceding discussion in the methodological consideration section it is clear that current histological criteria using a cut-off grade to define CD have important shortcomings. We therefore adopted an open histological definition of CD when selecting a study for inclusion, as long as the authors' explicitly stated or described the criteria used to define CD (see inclusion criteria below). However, with the help of the TEP, we defined a “standard” histological definition of CD as a biopsy grade showing a modified Marsh IIIa or greater. This definition was NOT used as an inclusion/exclusion criterion, but simply to frame our results and to allow for the evaluation of the effect of different histological criteria on the performance of the various CD tests.

The choice of biopsy criteria and/or histological grade “cut-off” used to define CD has important implications for the interpretation of the studies of serology, HLA, and biopsy. It is recognized that some patients with CD may have Marsh I or II lesions, and by definition patients with latent CD have Marsh 0 lesions. However, as emphasized by Marsh,¹ and as is discussed further below, in order to correctly interpret these early lesions, prospective follow-up studies are required, and an individual patient follow-up and documented response to gluten withdrawal would be required to firmly establish the diagnosis of CD.

The practical importance of the histological definition is evident from our preliminary review of articles that demonstrated considerable heterogeneity in the histological criteria used within the studies to define CD. Some used strict definitions, whereas, others accepted milder grade lesions. Furthermore, since the existence of latent CD and some silent CD without fully developed histology is now recognized, a study that aims to assess the sensitivity and specificity of biopsy itself in CD needs to use a design that incorporates the most sensitive and specific serologic and HLA tests available. The biopsy and serology should be performed simultaneously, with patients having discordant test results being further evaluated. Those with normal biopsy and positive serology would have to be followed over time to see if they have a latent form of CD. Conversely, patients with positive biopsies and normal serology would have to demonstrate improvement in histology on a GFD, and ideally, certification of relapse by biopsy with reintroduction of gluten. This type of study design was sought in order to address the objective of the sensitivity and specificity of biopsy.

Populations

1. Unselected general population. The unselected general population implies a representative sample of a given population, such as a random sample of healthy blood donors or healthy school children. Some unselected populations are better than others for determining the true prevalence or incidence of CD. For example, blood donors are required to have normal hemoglobin and no iron deficiency, and therefore may underestimate the true numbers of patients with CD.

2. Suspected CD. Patients with suspected CD include patients with GI symptoms, such as diarrhea or symptomatic malabsorption, who are being investigated for the possibility of CD. These patients are typically undergoing other investigations in addition to being worked-up for CD.

3. High-risk populations. High-risk populations include populations with an expectedly higher prevalence of CD. Such populations include asymptomatic family members of patients with CD, patients with type I diabetes where identified CD would likely be silent or latent, and populations such as those with iron deficiency or osteoporosis where identified CD would be in the atypical CD classification.

HLA DQ2/DQ8

The HLA DQ2 haplotype represents the occurrence of HLA class II heterodimer alleles DQA1*0501 and DQB1*0201. These typically occur in a cis position as HLA DR3-DQ2 or in a trans position as HLA DR5/DR7-DQ2. The HLA DQ8 haplotype DQA1*0301/DQB1*302 typically occurs in association with DR4.

Search Strategy

A series of searches were performed by National Library of Medicine staff in support of the literature review for CD. Strategies were developed using the guidelines supplied by the UO-EPC, and were divided into the five questions posed by AHRQ. All searches were limited to human studies published in English language journal articles. The specific strategies used for each search are located in Appendix B.

1. What is the sensitivity and specificity of the following tests for CD:

   1. EMA

   2. human tTG IgA antibodies

   3. AGA EMA

   4. HLA DQ2/DQ8

   5. small bowel biopsy

   Searches were run in the MEDLINE® and EMBASE databases for each of the five tests. With the exception of the search for small bowel biopsy, a reference to CD or its synonyms was not a requirement for retrieval in order to obtain the widest possible information on these tests. Because of their complexity, a separate search was run for each test, then the results combined into one Pro-Cite file and duplicates eliminated. Individual case reports and letters to the editor were also removed.

   The MEDLINE® searches were run in October 2003 for the year 1966 forward and yielded a total of 2885 citations, with a follow-up search for HLA DQ2 and DQ8 performed in November 2003 that yielded an additional 390 citations. The EMBASE searches were run in December 2003 for the year 1974 forward and yielded a total of 1,046 citations after duplicates to MEDLINE® were removed.

2. What is the prevalence and incidence of symptomatic and clinically silent CD in the general population and in the following identified high-risk populations:

   1. patients with an affected family member

   2. type 1 diabetes mellitus

   3. IDA

   4. osteoporosis

   Searches were run in the MEDLINE® and EMBASE databases. The MEDLINE® search was performed in October 2003 for the year 1966 forward and retrieved a total of 1,584 citations. The EMBASE search was run in December 2003 for the year 1974 forward and yielded 467 citations after duplicates to the MEDLINE® retrieval were removed. Individual case reports and letters to the editor were also removed from both searches.

3. What is the association between CD and GI lymphoma?

   Searches were run in the MEDLINE® and EMBASE databases. The MEDLINE® search was performed in October 2003 for the year 1966 forward and retrieved a total of 230 citations. The EMBASE search was run in December 2003 for the year 1974 forward and yielded 97 citations after duplicates to the MEDLINE® retrieval were removed. Individual case reports and letters to the editor were also removed from both searches.

4. What are the expected consequences of testing for CD in the following populations:

   1. patients with symptoms suggestive of CD

   2. asymptomatic, at-risk populations

   3. general population

   Searches were run in the MEDLINE®, EMBASE, PsycINFO, AGRICOLA, CAB, and Sociological Abstracts databases. In order to obtain the widest possible retrieval, all articles on screening for celiac and its synonyms were included, not just those discussing consequences.

   The MEDLINE® search was performed in October 2003 for the year 1966 forward and retrieved a total of 917 citations. The EMBASE (1974 forward), PsycINFO (1840 forward), AGRICOLA (1970 forward), CAB (1972 forward), and Sociological Abstracts (1963 forward) database searches were run in December 2003 and yielded a combined total of 204 citations after duplicates to the MEDLINE® retrieval were removed. Individual case reports and letters to the editor were also removed from both searches.

5. What interventions are effective for promoting or monitoring adherence to a GFD?

   Searches were run in the MEDLINE®, EMBASE, PsycINFO, AGRICOLA, CAB, and Sociological Abstracts databases. Because of the small number of citations retrieved, a few selected articles discussing adherence to dietary limitations for other conditions were included. The MEDLINE® search was performed in October 2003 for the year 1966 forward and retrieved a total of 152 citations. The EMBASE (1974 forward), PsycINFO (1840 forward), AGRICOLA (1970 forward), CAB (1972 forward), and Sociological Abstracts (1963 forward) database searches were run in December 2003 and yielded a combined total of 168 citations after duplicates to the MEDLINE® retrieval were removed. Individual case reports and letters to the editor were also removed from both searches.

Some citations fulfilled the criteria of more than one celiac objective. Duplicates within each celiac objective were electronically removed. The obtained citations were uploaded into an internal web-based review system (SRS) for online collaborative citation screening and abstraction. Articles passing the first level screen were retrieved in full for further screening (see below).

Reference lists of included studies, book chapters, and narrative or systematic reviews retrieved after having passed the first level of relevance screening, were manually searched to identify additional unique references. Through contact with content experts, and the TEP, attempts were made to identify other studies not identified by the search.

Study Selection and Eligibility Criteria

Study selection was performed using three levels of screening with increasingly more strict criteria to ensure that all relevant articles were captured (Table 1). Each celiac objective had its own selection criteria for each level of screening and, as discussed previously, each celiac objective was treated as a separate sub-review. Following a calibration exercise, two reviewers independently screened all studies using the SRS web-based system. This system allows automatic identification of review disagreements. Any disagreements were resolved by the two reviewers by consensus; rarely, a third reviewer was used to break an impasse. The specific screening questions for each screen level are included in Appendix C.

Level 1 broad screening. Level 1 screening was used to identify any potentially relevant citation, based on review of the title, abstract and key words. For each objective, the SRS system displayed the corresponding task order questions alongside the citation details. Reviewers answered a broad question of whether the citation potentially related to the current objective. Furthermore, the SRS system was set-up in such a way that articles which were identified in one celiac objective silo, that could also be relevant to another objective, could be identified and moved/copied to the other silo. The review team was divided up so that two members could be simultaneously reviewing each objective.

Level 2 refined screening. Potentially relevant articles identified at level 1 were obtained in full for level 2 screening. Again, using the SRS system with the actual articles on hand, reviewers selected articles that related to each of the specific objectives. The reviewers were asked to err on the side of inclusion for this level, and to classify articles as “original” or “review”. Original articles meeting level 2 inclusion also had basic demographic data—such as screening test used, celiac definition, and study population identified—recorded into the SRS system.

Level 3 final screening. Level 3 screening identified articles that specifically allowed for the answering of the task order questions. These articles fulfilled the final inclusion/exclusion criteria, allowed actual extraction of the required data, and did not have fatal methodological flaws.

Important articles answering a stated objective but not meeting inclusion criteria (i.e., containing potential threats to internal validity), were presented and discussed in the discussion section.

Data Abstraction

For each objective, a detailed and standardized data abstraction form was developed with the assistance of content experts and the TEP panel. The data abstraction forms included baseline study characteristics as well as questions allowing for the abstraction of all relevant study results and characteristics. The electronic data extraction forms began with basic study and patient demographic questions that were common across the five sub-review forms. These included reviewer name, author name, publication year, publication type, study design type, and basic study population demographics such as race, age, gender, and type of CD population. The extraction forms then moved to specific questions geared at extracting data to answer the respective objective's questions. The individual data abstraction forms are included in Appendix C.

Celiac 1 (sensitivity and specificity) data abstraction form. Separate data abstraction forms were developed for serology, HLA, and the biopsy sub-questions. Two-by-two tables were used to abstract data on sensitivity and specificity, and to determine positive and negative predictive values and the prevalence of CD in the tested population. The biopsy studies were quite heterogeneous, and did not allow for direct numeric extraction of data.

Celiac 2 (prevalence and incidence) data abstraction form. For this objective, the data extraction form included questions for detailing the screened study population, the number of individuals screened, the number of CD cases identified and how CD was confirmed. For incidence studies, the comparison population and time period were recorded.

Celiac 3 (lymphoma) data abstraction form. In addition to the basic demographic, and study design data, the extraction form contained fields for the extraction of risk data linking GI lymphoma to CD. Types of data sought were prevalence and incidence of lymphoma in CD in the setting of comparison data from a control population. Fields for extracting standardized incidence, morbidity, and mortality ratios were included.

Celiac 4 (consequences of screening) data abstraction form. The extraction forms for this objective included text fields to detail the consequences of testing for CD. The form contained fields that identified the specific consequence of testing which was addressed by the study, as well as a data field to report the study findings. The general field approach was chosen to allow extraction of the expected varied data for this objective.

Celiac 5 (monitoring and promoting adherence) data abstraction form. For this objective, standard demographic data was collected, as well as the methods used to monitor adherence to a GFD, the response of those measures to the diet, and the correlation of serological methods with biopsy findings. Space was provided to detail the sensitivity and specificity of the monitoring method when that data was available. For the objective of promoting adherence to a GFD, a text-based form was used to allow the extractor to describe the intervention and the results of its use.

Electronic forms. The abstraction forms were developed in Microsoft Excel to allow for electronic data entry and recording, and to allow exporting the evidence table data into Microsoft Word. For each celiac objective, data abstraction was conducted by one reviewer and verified by another. The extracted data was further verified by one of the principal investigators.

Quality Assessment

The quality of reporting of diagnostic test studies was assessed using the QUADAS tool.¹⁹ This tool is the first to be published that allows for the assessment of the quality of studies of diagnostic tests. The instrument was developed using a Delphi procedure. The Delphi panel consisted of nine experts in diagnostic research who refined an initial list of items in four rounds, after which agreement was reached on the items to be included in the tool. The QUADAS tool consists of 14 questions that are answered “yes,” “no,” or “unsure.” The tool addresses the items individually and does not incorporate an overall quality score (Appendix D).

Cohort and case-control study reports were assessed using the Newcastle-Ottawa scale (NOS; Appendix D). The NOS is an ongoing collaboration between the Universities of Newcastle, Australia and Ottawa, Canada. It was developed to assess the quality of non-randomized studies with its design, content and ease-of-use directed to the task of incorporating the quality assessments in the interpretation of meta-analytic results. A “star system” has been developed in which a study is judged on three broad perspectives: the selection of the study groups; the comparability of the groups; and the ascertainment of either the exposure or outcome of interest for case-control or cohort studies, respectively. The goal of this project is to develop an instrument that provides an easy and convenient tool for quality assessment of non-randomized studies for use in a systematic review.

The inter- and intra-rater reliability of the NOS have been established. The face content validity of the NOS has been reviewed based on a critical review of the items by several experts in the field, who evaluated its clarity and completeness for the specific task of assessing the quality of studies to be used in a meta-analysis. Furthermore, the validity of the NOS criteria has been established by comparisons to more comprehensive but cumbersome scales. An assessment plan is being formulated for evaluating its construct validity, with consideration of the theoretical relationship of the NOS to external criteria and the internal structure of the NOS components.²⁰

Quality assessments of cross-sectional reports were assessed using a 19-item instrument adapted from Ophthalmology (Appendix D).²¹

We did not conduct any sensitivity analysis of quality assessments on the observational studies, as there is little by way of guidance to suggest what a poor quality study score would be based on for these assessment instruments.

One reviewer assessed the quality of an entire celiac objective to maintain internal consistency. Quality assessment was not performed under masked conditions.

Data Synthesis and Analysis

The data obtained from this review fell into several broad categories, which correspond in large part to the individual study objectives. These will be addressed in turn.

Data for the sensitivity and specificity of each serological marker was considered separately. In addition, studies were subdivided by the population age group (adults, children, mixed population), and by study design (case control, relevant clinical population/cohort).

Attempts were made to identify, explain, and minimize clinical and statistical heterogeneity in the included studies. Heterogeneity was assessed graphically by plotting receiver operator (ROC) curves for each of the included studies in a given analysis. A Pearson's Chi Square with n-1 degrees of freedom, where n represents the number of included studies in an analysis was calculated to assess statistical heterogeneity.

Pooled estimates were only calculated if clinically and statistically appropriate. In situations where pooling was not performed, a narrative systematic review was conducted.

There are several potential ways to pool the results of studies of diagnostic tests, each having both advantages and disadvantages. The simplest and most intuitive is to simply perform a weighted mean of the sensitivity and specificity for the studies in question. This method provides a pooled estimate that is easy to interpret by clinicians. Several other techniques involve the pooling of diagnostic odds ratios or likelihood ratios. These methods have the distinct disadvantage of difficulty in interpretation, and the inability to derive a pooled sensitivity or specificity from the resulting estimates. Lastly, one can use one of several methods to produce a summary ROC curve. The method described by Littenberg and Moses,^{22, 23} has the advantage of being able to produce a summary curve while taking into account a threshold effect. This can occur when different studies use different thresholds to define a positive test, or even from differences in labs using the same cut-off. To interpret summary ROC curves it is necessary to know the sensitivity or specificity of the test in question in the population in which it will be applied. Since neither of these values is estimable without conducting yet another diagnostic accuracy study for the given population, the clinical usefulness of using this method alone is limited.^{24, 25}

In order to produce clinically useful pooled statistics, we calculated a weighted mean of the sensitivity and specificity from those of the included study. For both sensitivity and specificity, this pooling relies on the assumption that the test statistic is the same in all of the included studies. For each pooled estimate, a 95% confidence interval (CI) was calculated using both a fixed and random effects model. The results of which were compared as a further test for heterogeneity. The pooled estimates for the sensitivity and specificity were also compared with a summary ROC curve calculated for the same group of studies as a second check of the estimates (summary ROC Curves are included in Appendix E).

The prevalence and incidence data from the Celiac 2 objective, and the CD-lymphoma data from the Celiac 3 objective, were anticipated to be quite heterogeneous considering the different, countries, age groups, and risk characteristics of the studied patients. Attempts were made to group studies of prevalence by age group, study population, and serological screening method. If the grouped studies did not show evidence of heterogeneity, pooled estimates of the prevalence were produced for that group of studies, otherwise a descriptive presentation of the data with a qualitative systematic review was conducted. Likewise, the outcome measures of the Celiac objectives 4 and 5 were presented in a qualitative systematic review, except in cases where it was possible to pool the sensitivity and specificity data as measures of monitoring of patients at various stages of recovery on a GFD.

Serology

Out of 3,982 citations identified by the search strategy for the Celiac 1 objective, 907 met level 2 screening criteria. Of these, 204 diagnostic test studies of one or more of the serological markers of interest (AGA, EMA, tTG) were identified. Sixty studies fulfilled the level 3 inclusion criteria (Appendix F; Evidence Table 1, Appendix I).^26–85 The most common reasons for failing level 3 inclusions were AGA studies conducted before 1990, studies utilizing an improper or an unbiopsied control group, or studies that did not give any description of the biopsy criteria defining CD. Five pairs of duplicate publications were identified.^{27, 28, 45, 46, 58, 65, 73, 74, 84, 86} Out of each duplicate pair, the study with the most complete data was abstracted,^{27, 45, 46, 58, 74} bringing the total of included unique studies to 55. The majority of these studies assessed more than one serological marker, and some studied more than one age group. Of the included articles, 20 were conducted in or included an adult population, 33 were conducted in a population of children, and eight in a mixed population of adults and children of varying proportions. The statements in this section that relate to mixed studies or studies in children and adults refer to these eight studies, and not to a sample that we pooled from different studies.

To minimize clinical and statistical heterogeneity, the included articles of a particular antibody test were divided into groups by age of the included population (adults, children, mixed), the study design (case control, or relevant clinical population/cohort), by antibody type (IgA or IgG), and by test methodology (e.g., monkey esophagus [ME] or human umbilical cord [HUC]). Within these groups, further differences in study population, country of origin, and biopsy definitions (especially whether or not mild grades without villous atrophy were included) were assessed systematically. Studies that reported using the ESPGAN criteria for the diagnosis of CD were categorized as including patients with some degree of villous atrophy. Other potential causes of heterogeneity such as the cut-offs used to define a positive test were assessed.

Two articles were identified that assessed the diagnostic value of various antibodies in children⁶⁴ and in mixed-age populations⁴⁰ with IgA deficiency. As well, one study enrolled biopsy-proven CD patients who were known to be EMA negative.⁶⁶ These studies were considered separately from the others. Studies of using antibodies in combination were also assessed separately.

Pooled statistical estimates (with 95% CIs) are provided for studies without clinical and statistical heterogeneity, and summary ROC curves for the studied antibodies are provided in Appendix E. Sensitivity analyses by study design did not show a significant difference except for the analysis of IgA-tTG-guinea pig (GP) in adults. Therefore, apart from studies of IgA-tTG-GP in adults, pooled estimates, when available, included data from both study designs.

AGA. The diagnostic characteristics of IgA were assessed in 35 studies and the diagnostic characteristics of IgG-AGA were assessed in 30 studies. Of the 35 IgA-AGA studies, 11 were conducted in an adult population,^{30, 33, 45, 50, 54, 61–63, 71, 77, 80} 21 in a population of children, ^{26, 27, 29, 31, 34, 36, 38, 42, 43, 50, 52, 56, 59, 60, 64, 67, 68, 83, 85, 87, 88} and five in a mixed population.^{27, 37, 40, 74, 75}

Of the 30 IgG-AGA studies, seven were conducted in an adult population,^{30, 33, 54, 62, 63, 71, 80} 19 were conducted in population of children,^{26, 27, 29, 31, 34, 36, 38, 42, 43, 50, 52, 58, 59, 64, 66, 68, 69, 83, 85} and five in a mixed population.^{27, 37, 40, 74, 75} Some studies provided data for more than one age group.

Some studies only provided summary statistics without the raw two-by-two table results,^{33, 34, 54, 58, 59, 69} however, the raw data was calculated from the presented sensitivity and specificity, and from the group sizes.

One study⁶⁶ was conducted in CD patients who were known to be IgA-EMA negative, and was not included in the main analysis. In this study of children, the sensitivity for IgA-AGA was 22% and the sensitivity for IgG-AGA was 33%, whereas, the specificity for IgA-AGA was 67% and the specificity for IgG-AGA was 58%; these values are considerably lower than those reported in other studies. Another two studies were conducted in patients with IgA deficiency.^{40, 64} The first demonstrated a sensitivity of 0% using IgA-AGA, but a sensitivity and specificity of 100% using IgG-AGA,⁴⁰ whereas the second showed a sensitivity of 0% with IgA-AGA, but a sensitivity of 100% and a specificity of 80% using IgG-AGA.

Despite clinical subdivision of the identified studies, significant heterogeneity was identified for each of the pooled AGA subgroup results (Tables 2 to 7). Heterogeneity can be visualized graphically in the ROC curves (Figures 2 to 4) and suggests that the heterogeneity is in part related to a serological test cut-off threshold effect. As well, two studies included CD patients with less than a Marsh IIIa grade;^{37, 45} these studies had lower than average sensitivities (61% and 67% for IgA-AGA) then that reported in other studies. The remaining heterogeneity likely represents a combination of the effects of different test kits, inter-lab variability, and differences in the study groups. For example, within the child population, two of the outlier studies were conducted in Turkey,^{26, 85} although apparently using standard methodology. Therefore, overall pooled estimates do not represent true summary statistics in these situations.

IgA-AGA. Despite the apparent heterogeneity, one can make some broad statements regarding the diagnostic value of AGA antibodies. IgA-AGA appears to offer fair to good performance in children (Table 2; Figure 2).^{26, 27, 29, 31, 34, 36, 38, 42, 43, 50, 52, 58, 59, 64, 66, 68, 69, 83, 85} Ten of the 19 studies demonstrated a sensitivity of IgA-AGA of greater than 80%, and six of the studies demonstrated a sensitivity of greater than 90%. However, nine studies demonstrated sensitivities of less than 80%. The specificity was greater than 80% in 15 of the 19 studies, and greater than 90% in 11 studies. Only four studies showed a specificity of less than 80%.

Ten studies assessed IgA-AGA in adults (Table 3; Figure 3).^{30, 33, 54, 62, 63, 71, 80} Five of the ten studies demonstrated sensitivities greater than 80%, and three of the studies demonstrated sensitivities of greater than 90%. However, four studies demonstrated sensitivities of less than 65%. The specificity was greater than 80% in eight studies and greater than 90% in three. Five studies had specificities between 80% and 90%, and only two studies had specificities less than 80%.

Among the studies that assessed IgA-AGA in a mixed population of adults and children,^{27, 37, 74, 75} two demonstrated poor sensitivities of less than 70% but with specificities between 90% and 92%, one demonstrated a sensitivity of 85% and a specificity of 85%, and the last demonstrated a sensitivity of 91% and a specificity of 98% (Table 4; Figure 4).

IgG-AGA. The seven studies of IgG-AGA in adults demonstrated considerably greater heterogeneity.^{30, 33, 54, 62, 63, 71, 80} The sensitivity ranged from 17% to 100%, with little study grouping. However, there was less variation in the reported specificities. Five of the seven studies demonstrated specificities greater than 80%, whereas, the remaining two studies had specificities of greater than 70%. (Table 5; Figure 5)

In contrast, among the 17 analyzed studies (non-IgA deficient) of IgG-AGA conducted in children,^{26, 27, 29, 31, 34, 36, 38, 42, 43, 50, 52, 58, 59, 68, 69, 83, 85} there seemed to be greater variability in the specificity than in the sensitivity (Table 6; Figure 6). Fifteen of the 17 studies demonstrated sensitivities that were greater than 80%, and six demonstrated sensitivities greater than 90%. Only two studies showed a sensitivity of less than 80%. In contrast, with regards to specificity, two groupings of studies become apparent. The first group consists of 11 studies, all of which had specificities greater than 79%, and except for one study, had sensitivities that were greater than 80%. In contrast, the second group of six studies all had specificities below 70%, and with the exception of one study, had sensitivities greater than 80%. (Tables and figures)

Four studies looked at IgG-AGA in a non-IgA-deficient mixed population of adults and children.^{27, 37, 74, 75} Two of these demonstrated sensitivities greater than 80%, one showed a sensitivity of 84%, whereas the second had a sensitivity of 96%. However, only the first study had specificity greater than 80%. In total, three of the four studies had specificities less than 80% (Table 7; Figure 7).

EMA

EMA—ME. The diagnostic characteristics of IgA-EMA-ME were assessed in 35 studies, and the diagnostic characteristics of IgG-EMA-ME were assessed in three studies. Of these included studies, 11 IgA-EMA-ME studies were conducted in adults,^{30, 32, 39, 51, 57, 63, 71, 77, 78, 80, 81} 17 in children,^{27, 35, 36, 38, 41, 44, 46, 51, 52, 55, 56, 58, 60, 69, 79, 82, 83} and five in a mixed population.^{27, 37, 40, 47, 75} Some studies provided data for more than one age group. One study in children provided data on two different populations (including different control groups).⁵⁵ IgG-EMA-ME was assesed in one adult population,⁶³ one child population,⁶⁶ but not in any of the mixed-population studies.

One study was conducted in a population of known CD patients who had previously tested negative for EMA. In this study, the sensitivity and specificity of IgG EMA-ME were both 100%;⁶⁶ the performance of IgA-EMA was not reported. Another study that included CD patients with less than a Marsh IIIa grade,³⁷ demonstrated a sensitivity of 88%. Some studies only provided summary statistics without the raw two-by-two table results,^{46, 58, 69} however, the raw data was abstracted based on the reported sensitivity and specificity, and the group sizes.

IgA-EMA-ME. Among the 11 studies of IgA-EMA-ME conducted in adults,^{30, 32, 39, 51, 57, 63, 71, 77, 78, 80, 81} the specificity of the test was 100% in all except one, which showed a specificity of 97.2% (Table 8; Figure 8). The sensitivity of the test showed some slight variation among the studies. One outlier study demonstrated a sensitivity of only 74%;⁷⁷ however, the authors found that in the remaining five of 19 CD patients who tested negative for EMA, three were IgA deficient. If these patients were excluded, the sensitivity rose to 88%. The authors also go on to say that they seem to have a high proportion of IgA-deficient subjects in their referral base. The remaining ten studies showed sensitivities of 89% or greater. In fact, five studies showed a sensitivity of 100%, one a sensitivity of 99%, and another a sensitivity of 97%. In all, eight out of the 11 showed a sensitivity of 95% or greater, matching the very high specificity of this test. There was no statistical heterogeneity for this analysis. The pooled estimates for the sensitivity and specificity along with their 95% CI values were 97% (95% CI: 95.7–98.5) and 99.6% (95% CI: 98.8–99.9), respectively.

Among the 18 studies that assessed IgA-EMA-ME in children,^{27, 35, 36, 38, 41, 44, 46, 51, 52, 55, 56, 58, 60, 69, 79, 82, 83} all but one outlier⁶⁹ were grouped together, and the sensitivities and specificities were both greater than 89% (Table 9; Figure 9). The outlier study demonstrated a sensitivity of only 74%, and also demonstrated low sensitivity for IgA-EMA-HU (see below).⁶⁹ The authors comment on the difficulties of interpretating immunofluorescence data as a likely explanation. Ten studies showed sensitivities greater than 95%, and except for one study with a sensitivity of 89%, the remaining seven studies had sensitivities between 90% and 95%. All these studies demonstrated specificities of 89% or greater, 16 had specificities greater than 90%, and 14 had specificities greater than 96%. There was no evidence of statistical heterogeneity in this analysis. The pooled sensitivity and specificity was 96.1% (95% CI: 94.4–97.3) and 97.4% (95% CI: 96.3–98.2), respectively.

Among the four studies in a mixed-age population that assessed IgA-EMA-ME,^{27, 37, 47, 75} all showed specificities of greater than 98% (Table 10; Figure 10). However, these studies showed some variation in the reported sensitivities. One study reported a very low sensitivity of 75%.⁴⁷ Two other studies showed a sensitivity of 86% and 88%, respectively, whereas the last showed a sensitivity of 98%.

IgG-EMA-ME. Only two studies meeting our inclusion criteria assessed IgG-EMA-ME, one in adults (Table 11),⁶³ and one in children (Table 12).⁶⁶ In the single adult study,⁶³ the sensitivity of the test was found to be 39%, whereas, the specificity was 98%. In a case-control study design, Picarelli et al. studied 30 IgA-EMA-negative children suspected of having CD.⁶⁶ Of these 30 children, 18 were subsequently found to have CD by duodenal biopsy and nine of the 18 were found to be IgA deficient. In this highly selected population, the reported sensitivity and specificity of IgG-EMA-ME were both 100%.

EMA—HU. IgA-EMA-HU was assessed in 13 studies. Six of these studies were conducted in adults,^{45, 49, 54, 57, 61, 70, 89} five in children,^{36, 53, 55, 69, 70} and two in a mixed population.^{72, 74} One study provided summary statistics without the raw two-by-two table results,⁶⁹ however the raw data was calculated from the reported sensitivity and specificity and the group numbers. One study provided data on two different populations (including different control groups).⁵⁵

IgG-EMA-HU was not assessed in any of the studies meeting our inclusion criteria.

Two studies included CD patients (both adult and children) with less than a Marsh IIIa grade, and reported IgA-EMA-HU sensitivities of 87% and 100%.⁴⁵

⁷⁰

IgA-EMA-HU. Six studies in adults assessed IgA-EMA-HU (Table 13; Figure 11).^{45, 49, 54, 57, 61, 70, 89} In all six, the specificity was reported to be 100%. There was, however, variability in the reported sensitivities, which ranged from 87% to 100%. Three studies demonstrated sensitivities between 87% and 89%, two between 90% and 95% and one showing a sensitivity of 100%. There was no observed statistical heterogeneity for this analysis. The pooled sensitivity and specificity was found to be 90.2% (95% CI: 85.9–93.4) and 100% (95% CI: 99.1–100), respectively.

Five studies with six separate child populations assessed IgA-EMA-HU (Table 14; Figure 12).^{36, 53, 55, 69, 70} Four of the six studies were grouped together and revealed sensitivities between 94% and 100%, and specificities of 100%. Of the two outliers,⁹⁰ one showed a sensitivity of 100% and a specificity of 77%. The other study,⁶⁹ was an outlier in other analyses, and demonstrated a sensitivity of 46% and a specificity of 96%. The authors comment on difficulties of interpretation of the immunofluorescence as a likely explanation. After accounting for this study, there was no statistical heterogeneity documented for sensitivity. The pooled sensitivity for this analysis was 96.9% (95% CI: 93.5–98.6). A pooled specificity for this analysis was not calculated, but is likely close to 100% given that four of the five grouped studies demonstrated a specificity of 100%.

Two studies assessed IgA-EMA-HU in a mixed-age population (Table 15; Figure 13).^{72, 74} In both these studies, the specificity was 100% (95% CI: 97.5–100) and the sensitivity 93% (95% CI: 88.1–95.4).

tTG antibodies

tTG—GP liver. The diagnostic characteristics of IgA-tTG-GP were assessed by ELISA in nine studies, and the diagnostic characteristics IgG-tTG-GP assessed by ELISA in three studies. Of the IgA-tTG-GP studies, five were conducted in adults,^{30, 32, 39, 45, 70} five in children,^{35, 41, 52, 70, 83} and four in a mixed population.^{47, 72, 74, 76} One study provided separate data for more than one age group.⁷⁰

Of the IgG-tTG-GP studies that met the inclusion criteria, none were in adults or children, although two studies were in a mixed population.^{72, 76}

Two studies included CD patients with less than a Marsh IIIa grade.^{45, 70} These studies demonstrated sensitivities of 81% and 95% for IgA-tTG-GP.

IgA-tTG-GP. In the analysis of IgA-tTG-GP in adults, five studies grouped themselves by study design.^{30, 32, 39, 45, 70} The two cohort studies (relevant clinical population)^{30, 39} both showed sensitivities of 100%, and specificities of 92% and 98%, respectively. On the other hand, the three case-control studies^{32, 45, 70} demonstrated high specificities (97% to 98%), but sensitivities of only 81% to 88% (Table 16; Figure 14). This analysis did not show statistical heterogeneity, but the differences by study design were striking, so a pooled estimate for sensitivity was not performed. The pooled specificity was 95.3% (95% CI: 92.5–98.1).

The analysis of IgA-tTG-GP in children showed very little variability in either the sensitivity, or specificity (Table 17; Figure 15). Among these five studies,^{35, 41, 52, 70, 83} the sensitivities ranged from 89% to 96%. The specificities were all greater than 92%, with three studies showing specificities greater than 96%,^{41, 52, 83} and two studies having a sensitivity of 100%.^{35, 70} The pooled estimates of the sensitivity and specificity were 93.1% (95% CI: 88.8–95.9) and 96.3% (95% CI: 93.1–98.0), respectively (Table 17; Figure 15)

Among the studies of mixed-age groups,^{47, 72, 74, 76} there was one outlier study with a sensitivity of only 84% but a specificity of 100% (Table 18).⁷² The specificities of the remaining studies were all greater than 94% (Table 18; Figure 16), and the sensitivities were between 92% and 95%. Heterogeneity was detected in the estimates of sensitivity, but not for specificity. The pooled specificity was 95.4% (95% CI: 92.7–97.2).

IgG-tTG-GP. Two studies in a mixed-age population assessed IgG-tTG-GP (Table 19; Figure 17).^{72, 76} The specificities in both studies were greater than 98%, but the sensitivities were 23% and 62%, respectively.

tTG - human recombinant (HR)

IgG-tTG-HR. The diagnostic characteristics of IgA-tTG-HR were assessed by ELISA in ten studies, and the diagnostic characteristics IgG-tTG-HR were assessed by ELISA in two studies. Of the IgA-tTG-HR studies, three were conducted in adults,^{39, 49, 54} three in children,^{52, 79, 83} and three in a mixed population.^{40, 72, 75}

Of the IgG-tTG-HR studies, two were conducted in a mixed population (Table 20),^{40, 72} but none were conducted in adults or children. One study was conducted in IgA-deficient patients and is described below.⁴⁰

Two studies included CD patients with less than a Marsh IIIa grade.^{45, 70} These studies demonstrated sensitivities of 81% and 95% for IgA-tTG-GP.

One study was conducted in a mixed-age population of patients with known IgA deficiency.⁴⁰ In this study, the sensitivity of IgA-tTG-HR was 0%, wheras, the sensitivities and specificities of IgG-tTG-HR were 100% and 80%, respectively.

IgA-tTG-HR. Three studies assessed IgA-tTG-HR in an adult population (Table 21; Figure 18).^{39, 49, 54} There was very little variability in the reported values for the sensitivities and specificities. The sensitivities were 100% in two studies, and 95% in the other. The specificities were 100% in two studies, and 97% in another. The pooled estimates of the sensitivity and specificity were 98.1% (95% CI: 90.1%–99.7%) and 98.0% (95% CI: 95.8–99.1), respectively.

Among the three studies in children (Table 22; Figure 19),^{52, 79, 83} the sensitivities were 96% in two studies and 95% in one. The specificities were 100% in two studies, and 96% in one. The pooled estimates of the sensitivity and specificity were 95.7% (95% CI: 90.3–98.1) and 99.0% (95% CI: 94.6–99.8), respectively.

Only two studies assessed the IgA-tTG-HR in a mixed-age population without IgA deficiency (Table 23; Figure 20).^{72, 75} The sensitivities and specificities were 92% and 100%, respectively, for the first study, and 91% and 96%, respectively, for the second. The pooled estimates of the sensitivity and specificity were 90.2% (95% CI: 86.4–93.0) and 95.4% (95% CI: 91.5–97.6), respectively.

Overall, these studies demonstrated a specificity of close to 100% and sensitivity in the range of 90% to 96%.

IgG-tTG-HR, IgA deficient. Only one study of IgG-tTG-HR, conducted in an IgA-deficient population, was identified.⁷² In this study, the sensitivity and specificity of IgG-tTG-HR was 68% and 100%, respectively.

Mixed-antibody combinations. Several studies were identified that tested different antibodies in combination. Six studies in children assessed the use of IgA- and IgG-AGA (Table 24).^{34, 42, 48, 50, 59, 85} When either of these tests were positive, the resulting sensitivities ranged from 83% to 100%, and the specificities ranged from 71% to 99%. One study, that apparently used similar methodologies, had the lowest sensitivity (83%) and specificity (36%) of the group.⁸⁵ When the same authors tested the antibodies under the requirement of both tests being concordant, the sensitivity fell, as would be expected, to 50%, and the specificity rose to 67%.⁸⁵ Three adult studies were identified that used IgA- and IgG-AGA in an either/or protocol (Table 24).^{33, 50, 78} As was observed in the studies of children, significant between-study differences existed, making pooled estimates inappropriate. Nonetheless, in these studies the sensitivity ranged from 77% to 100%, while the specificity ranged from 90% to 97%.

One study in a mixed-age population assessed the use of a combination of IgA- and IgG-tTG-HR antibodies (Table 25).⁷² In this study, the sensitivity when either test was positive was 98.5%, while the specificity remained high at 100%. Another study in children assessed the combination of IgA-AGA and IgA-EMA-HU when either test was positive, and found a sensitivity of 100% and a specificity of 73% (Table 26).⁶⁹ This same study assessed the same antibodies under the situation where both tests needed to be concordant. In this circumstance, the sensitivity remained 100% and the specificity rose to 93%.

In general, combining tests when either test is positive tended to improve sensitivity at the cost of specificity, while a requirement for the tests to be concordant tended to improve specificity.

Prevalence of CD and the positive predictive value (PPV) and negative predictive value (NPV) of serology. The prevalence of CD in the tested populations is presented in Tables 2 to 26 for the individual studies, and in Table 27 for the pooled estimate for the analysis groups.

The minimum prevalence of CD in individual study populations was greater than 25% in most of the studied analysis groups (i.e., IgA-AGA, IgG-AGA, etc), except for ten analysis groups where the minimum prevalence was between 9% and 12%. In all the analysis groups, the maximum prevalence ranged from 30% to as high as 70%. The pooled prevalence for the analysis groups was predominantly between 30% and 45%.

In assessing the IgA-EMA and IgA-tTG analysis groups, the pooled prevalence ranged from 33% to 46% except for the analysis of IgA-tTG-HR in adults, which showed a pooled prevalence of 16%. Figure 21 is a plot of the individual study prevalence versus the study's PPV, and suggests that below a CD prevalence of about 35% to 40%, the PPV of these IgA-based tests tends to drop from about 90% to 100%, to about 80% or less. As expected, Figure 22 demonstrates the reverse relationship, with the NPV being between 95% and 100% up to a CD prevalence of about 45%, and then dropping off.

HLA DQ2/DQ8

We identified 99 potentially relevant HLA articles that appeared to address HLA DQ2/DQ8 in a CD population (Appendix F).^{8–11, 15, 53, 54, 62, 91–100, 100–177} These studies were not designed to determine the diagnostic utility of DQ2 or DQ8 per se.

Of the identified studies, 54 allowed estimation of the prevalence, sensitivity or specificity of HLA DQ2/DQ8 in the studied population.^{8–10, 15, 53, 54, 93, 100, 109, 120, 134–177} In one study, DQ2 data could not be reliably extracted.¹⁶⁹ The authors of one study⁹ explicitly stated that the patients used were the same as in two of their other publications.^{8, 93} In two other publications by the same authors,^{9, 10} the patients appear to be different and the authors do not indicate that they used patients from a previous study. However, the possibility that these two studies^{9, 10} share a subset of patients cannot be excluded. Another two studies addressing different topics but with extractable HLA data, also appeared to have used the same patients.^{53, 136} In cases of duplicate publications, the studies with the greater number of patients were used.^{9, 136}

The study designs and strictness of CD diagnosis in these articles varied, as did the inclusion of a control group. Most of the CD cases were diagnosed based on the ESPGAN criteria, although in some studies CD was diagnosed based on serology and then in most cases later confirmed by biopsy.^{15, 109, 120, 160, 161, 164, 168, 170, 172, 177} Nine of the studies were classified as cross-sectional studies,^169–177, 32 were case-control studies,^{8–10, 53, 100, 120, 134–159} and 12 were mixed cross-sectional/case-control studies or could be considered as diagnostic cohort studies.^{15, 54, 109, 160–168} Four of the mixed design studies^{109, 164, 166, 178} used screen-negative patients as the control group, whereas the rest used a control group that was separate from the screened population. The study populations were also variable. The case-control studies used known CD cases compared with variously defined CD negative controls.

Seven studies used relatives of CD patients,^{158, 161, 164, 166, 169, 172, 177} four used a population with Down Syndrome,^{109, 134, 160, 170} two used a population with type I diabetes,^{165, 173} and one used a mixed group of patients with CD including some with Down's and others with diabetes.¹⁷³ The mixed-design/cohort studies used patients suspected of CD on clinical grounds or subjects who belonged to a high-risk group, such as type 1 diabetics or first-degree relatives of patients with CD. The remaining articles used a screened healthy population or another specific group.

The articles with extractable data stated the frequency of HLA DQ2, and to a lesser extent the frequency of HLA DQ8, in their CD group. The cross-sectional studies did not include a control group. Only the frequency as a surrogate of sensitivity was available. None of the case-control or mixed-design studies calculated the sensitivity or specificity of HLA DQ2 or DQ8. However, these studies allowed us to derive estimates of these statistics from their results or tables. The considerable degree of clinical and methodological heterogeneity between the identified studies did no allow for statistical pooling of the results.

Two studies fulfilled our inclusion requirement of both cases and control groups undergoing intestinal biopsy (Evidence Table 2, Appendix I; Table 28).^{136, 152} The remaining studies had various control group types: unbiopsied, healthy controls, disease controls, or serology-negative controls. These studies provide useful information and are presented at the end of the HLA results section for reference.

The study by Iltanen et al.,¹³⁶ was conducted in a group of Finnish children to assess the density of gamma delta positive intraepithelial lymphocytes (γδ+ IELs) in: patients with CD by biopsy (ESPGAN); patients with suspected CD where the diagnosis was excluded by biopsy; and, in a group of biopsy-negative patients who underwent endoscopy for dyspepsia. The biopsy aspect of this study is presented in its respective section. In this study, HLA DQ2 was found in 19 of 21 (90.5%) of patients with CD as apposed to 29 out of 67 (29.9%) of the control patients. Elevated γδ+ IEL density was significantly associated with DQ2 positivity. The calculated diagnostic measures for this study are presented in Table 28. In this population, DQ2 demonstrated a high sensitivity of 90.5% but a relatively modest specificity of only 70%, which is understandable given that the control population had a fairly high frequency of DQ2 positivity. The prevalence of CD in the study population was 1:4.2 (or 24%). The PPV was 49% and the NPV was 96%, suggesting that a negative DQ2 test result provides the greatest diagnostic information.

Sacchetti et al.¹⁵² studied a group of Italian children suspected of having CD. Patients fulfilling the ESPGAN criteria were classified as having CD (n = 48 of 80), whereas, the remainder (n=32) were considered disease controls. The authors also used a second retrospectively defined group of known CD patients by ESPGAN criteria (n = 74), and a second group control of 180 unbiopsied healthy subjects. HLA DQ2 was determined in the CD group as a whole and in the two control groups, with the results presented in Table 28. In this study, the sensitivity of HLA DQ2 was 88.9% and the specificity was 81% for the comparison with the biopsied controls; the sensitivity of HLA DQ2 was 88.9% and the specificity was 73% for the comparison with the unbiopsied controls. Interestingly, in this study only 18.8% of the biopsy-negative controls were positive for HLA DQ2, whereas, 26.7% of the unbiopsied controls were HLA DQ2 positive. This difference accounts for the higher specificity seen for HLA DQ2 in the comparison with the biopsy-negative control group as compared with the comparison with the healthy controls. The prevalence of CD in the studied population was also quite high in both portions of this study (79% for comparison with biopsied controls and 51% for the comparison with unbiopsied controls). As such the PPV and the NPV of HLA DQ2 in this study were 95% and 62%, respectively. The difference in prevalence between this and the Iltanen study accounts for the differences seen in the PPVs and NPVs.

HLA all study data. The following section presents the data of the HLA studies that failed to be included on the basis that the control groups were not assessed with the gold standard test for CD (biopsy). These studies collectively provide useful information on the diagnostic value of HLA testing, but have to be interpreted with caution.

The prevalence of DQ2 and DQ8 in these studies is presented in Table 29, while the results of the diagnostic value of HLA DQ2 and HLA DQ8 are presented in Tables 30 and 31. Unfortunately, none of these studies were actual studies of the diagnostic value of HLA DQ2 or HLA DQ8 for the diagnosis or screening of CD. However, as presented in the Tables, the crude data was abstracted and the diagnostic characteristics were calculated. Significant clinical and statistical heterogeneity existed between these studies, making arithmetic pooling of the studies unjustified. Figure 24 and Figure 25 represent the plotting of each study's sensitivity (true positives) versus 1-specificity (false positives) to create a ROC presentation. The value of these figures lies in the global picture they represent regarding the results of each of the studies. Figure 24 demonstrates that the vast majority of the studies cluster together in a region where the sensitivity of HLA DQ2 is greater than 80%, with most studies lying above the 90% sensitivity mark. In contrast, these same studies have specificities in the range of 55% to 80%. Outlier studies are identified by author name. The best sensitivities and specificities were seen in two studies. The first, by Kaur et al.,¹⁶³ was a study from India where only 4.6% of the control population was positive for HLA DQ2. The second study, by Tighe et al.,¹⁴⁹ was conducted in a group of patients with CD and ethnically-matched control subjects from Rome, Italy. The prevalence of CD was quite high in the studied group (51%), and the frequency of HLA DQ2 in the control population of 12.2% was much lower than that observed in other Italian studies.

The remaining outlier studies were divided into a low-sensitivity/high-specificity group (Group 1), and a high-sensitivity/low-specificity group (Group 2). In the first case, all the studies were conducted in a non-Western European population. In particular, the worst performance of HLA DQ2 occurred in a study from Chile,¹³⁷ where the frequency of HLA DQ2 was very low in both the patients with CD and the control subjects. It is important to note, however, that not all non-Western populations deviated from the main cluster of studies. For example, Catassi et al.¹²⁰ found that 91% of Saharawi Arabs (Algeria) with CD carried HLA DQ2 compared with 38.9% of Saharawi controls. These values are similar to those seen in most Western populations. The second group all showed relatively poor specificity, although the sensitivity was preserved. As would be expected, the control groups of these studies were at high risk of having CD (relatives of CD^{158, 164, 166}), or were a population with a known higher frequency of HLA DQ2 (individuals with diabetes^{161, 165}). As such, the high frequency of HLA DQ2 in these control populations makes the specificity of HLA DQ2 rather poor.

The frequency of HLA DQ8 in Western European populations with CD varies from approximately 2.7% to 6% (Table 29). The frequency is slightly higher in studies from Italy, the UK, and France (5.6% to 8% of CD patients). The frequency of HLA DQ8 in a subset of patients who had HLA testing in a large American serology screening study for CD was 22%,¹⁵ which is quite a bit higher than that reported in the European studies.

A small group of studies allowed the estimation of the sensitivity and specificity of having either HLA DQ2 or DQ8. The results of these studies are presented in Table 33 and Figure 25. As can be seen in the figure, these studies confer a wide variation. Clearly, the sensitivity of using this strategy is quite high and is likely close to 100% in Western populations. The study by Balas et al.,¹⁵⁵ likely represents the closest to the truth, as this was a typical case-control design in patients with know CD compared with healthy controls. The Fasano et al. study¹⁵ represents the largest study and gives similar results to those obtained by Balas et al., however, the higher frequency of HLA DQ8 in their control group compared with other studies is of concern. Once again, the remaining studies can be grouped into high-specificity/low-sensitivity (Group 1) and high-sensitivity/low-specificity (Group 2). As was the case for HLA DQ2, Group 1 consists of two studies of non-Western populations, whereas, Group 2 represents studies with first-degree relatives and a study that used patients with diabetes as their control group.

Biopsy

Using epidemiologically appropriate eligibility criteria, our comprehensive literature search did not identify any studies that specifically addressed the question of the sensitivity or specificity of biopsy for the diagnosis of CD.

However we sought to obtain indirect evidence regarding the diagnostic performance of biopsy as a test for CD. Some data was available from those studies identified for other review objectives, such as the cross-sectional screening studies, the HLA DQ2/8 studies, and studies of IELs. We also sought studies of follow-up of biopsy negative patients suspected of CD, and studies of silent and latent CD. The findings from these studies are presented in the Discussion and in Appendix H.

Quality Assessment

Overall, the quality of the diagnostic studies assessed in the Celiac 1 objective was quite good (Appendix J, Table 1). However, 59% of the studies reported using a selected patient population that may not be representative of a clinically relevant population. This is likely related to study design. Only 11% of the studies reported on whether the reference test was reported without knowledge of the index test. We felt that this was not a major threat to the validity of the studies.

Incidence of CD in the General Population

The incidence of CD in North America and Western Europe was derived from studies from the following countries: US,^{128, 205} England,²⁰¹ Italy,²⁰², Sicily,²⁰³ Spain,²⁰⁴ Netherlands,²⁰⁰ Sweden,¹⁹⁵ Denmark,^{196, 197} and Finland (Evidence Table 3, Appendix I; Table 34).^{198, 199} In the report, crude incidence is defined as the number of new cases per 100,000 population-at-risk per year and cumulative incidence as the number of new cases per 1,000 live births; cumulative incidence is age-specific and its denominator reflects the total number of individuals from the same year of birth (i.e., birth cohort).

Incidence in children: The crude incidence of CD in children age 0 to 15 years varied from 2.15 to 51 cases per 100,000 patient years.^{193–195, 198, 201, 204} When reported, the relative risk (RR) of CD was greatest for the 0- to 2-year age group, as well as for women, and varied from 32.26 to 42.4^{193, 195, 204} and from 1.9 to 3.34,^{128, 193, 195} respectively. The cumulative incidence at age 5, when reported, varied between 0.089 and 9 cases per 1,000 live births.^{128, 196, 202, 203} (see Table 34).

The incidence of CD has been most studied in the Scandinavian countries, particularly Sweden,^{193, 195, 310–313} Denmark,^{196, 197, 313, 314} and Finland,^{194, 198, 199} where important disparities have been observed over time and between countries. Reports from these countries have the advantage of being derived from comprehensive prospective databases and from populations which are genetically fairly stable, shedding light on potential environmental causal exposures,^{195, 196} or on variations in practice patterns.

In Scandinavia, the highest incidences of CD in children were found in Sweden for the 0- to 2-year age group from 1987 to 1997, where an average of 198 new cases per 100,000 patient years (95% CI: 186–210) were observed.^{193, 195} This peak in incidence was followed by a rapid decline, observed during 1995-97, where incidences dropped to an average of 51/100,000 patient years (95% CI: 36–70). In contrast, the incidence of CD in children aged 2 to 4.9 years and 5 to 15 years was only slightly increased over the 1973-97 period, with a peak in 1996 of 33 cases (95% CI: 24–44) per 100,000 patient years and 10 cases (95% CI: 7–13) per 100,000 patient years for these respective age groups. A cohort effect was noted in that the cumulative incidences at 2 years of age for the children belonging to birth cohorts from 1984 to 1994 were on the gradual rise (up to 4.4 cases/1,000 births [95% CI: 3.8–5.0] for the 1993 cohort), while a progressive decline was observed for birth cohorts from 1994 to 1996 (down to 1.7 cases [95% CI: 1.3–2.1] per 1,000 births for the 1995 cohort). Most of these cases were symptomatic, so that these observations are unlikely to be due to changes in screening practices. Interestingly, these changes mirrored changes in the composition of infant formulas, with the highest values of a wheat/rye/barley exposure index during the years 1982-1994.

In contrast, the incidence of CD in Denmark, a neighbouring country, has been significantly lower and very stable from 1960 to 1988,¹⁹⁶ with an average incidence of 0.089/1,000 live births for that period.³¹³ A comparison of dietary exposures between Swedish and Danish children diagnosed with CD between 1972 and 1989 showed that by the age of 8 months, the Swedish diet contained more than 40 times more gliadin than the Danish diet.³¹³ In Finland, incidences have also been fairly stable, and have in fact decreased among infants but increased among older children.¹⁹⁸ However, these observations date back to 1984 and can therefore not be compared with the Swedish epidemics.

Spain has also seen an increased incidence of CD over the past 25 years, from 6.87 (95% CI: 5.26–8.83) cases/100,000/year in 1981-90 to 16.04 cases/100,000/year (95% CI: 12.99–19.59) in 1991-99,²⁰⁴ an observation that was correlated with an increased proportion of silent or atypical presentations at diagnosis (i.e., inferring a role for changes in clinical practice). The age at diagnosis also correlated positively with the age at which gluten was introduced in the diet.

The role of dietary exposure during infancy is also highlighted in studies from the UK, where recommendations on infant feeding, promoting breastfeeding and later introduction of starches, were published in 1974. Subsequent to these recommendations, there was a fall in the incidence of childhood CD;^{315, 316} however, this data is not presented in detail because we focused on reports from the past 15 years.

As opposed to the incidences derived from reported cases, the incidence observed from a prospective screening protocol are not subject to variations related to practice patterns and are obviously more comprehensive and accurate. Hoffenberg et al., from the US, conducted the only prospective CD screening study available to date.¹²⁸ Between December 1993 and September 1999, a total of 22,346 newborns in Denver, Colorado were screened for HLA genotypes associated with CD and type 1 diabetes. A representative sample of at risk HLA DRB1*03 positive infants were prospectively followed (n=987), for as long as the first seven years of life. Serological screening was performed at nine, 15 and 24 months of age, then yearly. Small bowel biopsies were recommended if the serology (tTG in most cases) was positive on two separate occasions, or in the presence of clinical suspicion. Between 1993 and 1999, 19 children were found to have evidence of CD, ten children had biopsy-confirmed CD, whereas, nine children had a positive tTG result at least twice. The mean age at presentation of evidence of CD was 4.6 years (range 2.6–6.5). Compared with HLA-DR3-negative children, the RR for evidence of CD was 5.6 (1.5–21, p=0.009) and 9.1 (1.7–48, p=0.003), for those expressing one and two HLA-DR3 alleles, respectively. The RR of CD in females was 3.34 (1–10.9, p=0.048) times that of males. Cognisant of the prevalence of HLA-DR mono- and heterozygotes among the same birth cohort, the authors calculated that by the age of 5, the estimated cumulative incidence of CD in the general population (defined as either biopsy-proven CD or persistently elevated tTG) was 9/1000 births (95% CI: 4–20), or 1:104 (1:49 to 1:221). This remarkably high cumulative incidence (i.e., twice that of the highest value among Swedish children at 4 years of age - 5.0 [95% CI: 4.4–5.7]¹⁹³) has to be interpreted in light of the fact that only ten out of the 19 cases had been biopsied; the remaining nine cases were diagnosed on the basis of a persistently elevated tTG titre, the PPV of which the same authors reported to be only 70% to 83%.³¹⁷ However, as mentioned above, these results are derived from an actual prospective and systematic screening intervention for CD, where asymptomatic cases would be detected. In all likelihood, there is therefore an important proportion of CD cases who remain undiagnosed during early childhood.

Incidence in adults: The crude incidence of CD in adults varied from lows of 1.27 in Denmark¹⁹⁷ and 3.08 in England, ²⁰¹ to a high of 17.2 cases per 100,000 patient years in Finland,¹⁹⁹ where specific efforts had been untaken to encourage screening for CD (see Table 34).

As has been observed for children, the incidence of CD in adults seems to have increased over the past 20 years.^{199, 201} This is largely explained by a change in practice patterns: physicians are more aware of the condition, its atypical manifestations and associated condition, while at the same time, serological testing has become widely available. There are therefore more diagnoses made on the basis of case-finding. This is reflected by the fact that the proportion of patients being diagnosed with CD in the absence of symptoms, or as a result of serological testing, has also increased.^{199, 201, 318–320}

In Finland over the period 1975-94, Collin et al.¹⁹⁹ have observed a ten-fold rise in the incidence of CD. The authors attributed this to the use of serologic screening (physicians were actively told to screen patients with type I insulin-dependent diabetes (IDDM), autoimmune thyroid disease, connective tissue diseases, women with infertility, patients with neurologic symptoms and first-degree relatives of CD patients), the routine performance of intestinal biopsies on all patients undergoing gastroscopy, and to the opening of open-access endoscopy clinics, creating the ability of all general practitioners to refer patients for gastroscopy.

In Italy, a gradual increase in the number of annual new CD diagnoses was observed between 1968 and 1992;^{318, 320} this increase correlated with an increased proportion of patients with subclinical presentations being identified.^{318, 320} Interestingly, despite the changing clinical presentation, there was no statistical difference between the histological grades at diagnosis.³²⁰

The incidence of CD in individuals of all ages varies from 1.0 in the Netherlands²⁰⁰ to 2.13 in Italy.²⁰² In Italy, the RR of CD in adults ranged from 0.11 in the >60 year group to 0.33 in the 16–39 year group, compared with children.²⁰² The RR of CD for females was 1.90 (95% CI: 1.48–2.45).²⁰²

In the US, the 30-year incidence (1960-90) for Olmstead County was 1.2 (95% CI: 0.7–1.6), and the incidence for 1980-90 was slightly higher at 1.7 (95% CI: not reported).²⁰⁵ This observation contrasts with the cumulative incidence of 9/1000 by age 5 reported by Hoffenberg from Denver, Colorado;¹²⁸ clearly, further knowledge of the epidemiology of CD in the US is required.

The point prevalence of CD can be calculated from registers of CD cases and the size of the population at risk; we found reports of such an observation in three of the included incidence studies.^{199, 199, 205} The point prevalence of CD was 21.8/100,000 in Olmstead County in 1991,²⁰⁵ 2.7/100,000 (95% CI: 11.0–14.5) in the Netherlands in 1992,²⁰⁰ and 204/100,000 (95% CI: 181–231) in Finland in 1994.¹⁹⁹ Of note, the later prevalence from Finland was observed in a community where intense efforts had been carried to screen the population at risk for CD.

Prevalence of CD in the General Population—Different Geographic and Racial/Ethnic Populations

Thirty-seven studies reported on the prevalence of CD in a general population (Evidence Table 4, Appendix I; Table 35). Three of these were conducted in the US,^206–208 16 in the Scandinavian countries,^{184–187, 209–219, 232}) eight in Italy,^{126, 182, 183, 220, 221, 223–225} five in the UK,^{188, 226–229} and five in other countries (Spain,²³⁰ Republic of San Marino,²²² the Netherlands,²³¹ Switzerland²³³ and Germany²³⁴). Several pairs of duplicate publications were identified including two triplets,^{182–188, 211, 213, 218, 220, 321} which brought the total number of included unique articles down to 30. The articles with the most complete data were used for the report.^{126, 206–234} Only seven studies were conducted in a child population,^{206, 209, 215, 220, 221, 223, 232, 233} but one large American study included separate data for both adults and children.²⁰⁶ All the included studies were conducted between 1992 and 2003. A summary of the included study characteristics is presented in Table 35. A breakdown of the included studies by screening test and age group is provided in Table 36.

The prevalence of CD by serology in the general unselected populations of North America and Western Europe, ranged widely from 152 per 100,000 (0.152% or 1:658) to 2,670 per 100,000 (2.67% or 1:37). The prevalence by biopsy ranged from 152 per 100,000 (0.152% or 1:658) to 1,870 per 100,000 (1.87% or 1:53). In four of the studies, a large proportion of the serology-positive subjects did not undergo biopsy.^{206, 216, 224, 232}

Among the included studies, there was no clear pattern relating prevalence to study age group, or in a consistent way to country, with large numbers of studies clustering around a prevalence range of 0.0025 to 0.014 by serology and 0.0025 to 0.010 by biopsy (Table 35; Figure 26, 27). In fact for prevalence by serology, the 50^th, 75^th, and 80^th percentiles occurred at a prevalence of 0.00637 (0.64%), 0.0117 (1.2%), and 0.0125 (1.3%), respectively, while by biopsy the 80^th percentile was at a prevalence of 0.0074 (0.74%) (Table 37; Figure 26, 27). Categorizing the studies by screening test and age group reduced the variability somewhat, but significant between study variation persisted. There were not enough studies to divide an analysis by screening test, age group, and country, simultaneously.

Among the studies conducted in the US,^206–208 the prevalence ranged from 0.00312 (0.312% or 1:320—only child population in this group) to 0.00949 (0.949% or 1:105). The largest of these, by Fasano et al.,³²² found a prevalence of CD in “not at risk” populations to be 0.95% in adults, 0.31% in children, and 0.75% overall (0.0075 or 1:133). This study included a predominately Caucasian population, although other ethnic groups were included (94% white; 3% black; 1.5% hispanic; 1% asian; 0.5% other). Not et al.²⁰⁸ found the prevalence by EMA confirmation of initial AGA testing to be 0.004 (0.4% or 1:250) in another predominately Caucasian population that also included other ethnic backgrounds (Caucasian [87%], African-American [11.5%], and Asian [1.5%]). Finally, Green et al.²⁰⁷ found a prevalence of 0.005 (0.5% or 1:200) in 1,749 patients undergoing upper endoscopy. The reason for the initial endoscopy in this study was not clearly described, and only those patients with endoscopic features suggestive of CD were biopsied, which may have underestimated the true prevalence of CD. The prevalence of CD among the six Italian studies was similar to that seen in the American studies, showing a range from 0.2% to 0.86%.^{126, 221, 223–225, 230} The prevalence of CD in other countries is presented in Table 35.

Only four studies demonstrated a prevalence of CD of greater than 0.015 (1.5%) (UK,³²³ Sweden,^{209, 219} Germany²³⁴), and an additional six showed a prevalence of between 0.010 (1.0%) and 0.015 (1.5%) (UK,²²⁸ Sweden,²¹⁶ Netherlands,²³² Ireland,²²⁹ Finland^{214, 215}). These studies would suggest a potentially higher prevalence of CD in these countries, though it should be kept in mind that other studies from these same countries showed a prevalence of less than 1.0%, including four studies from Sweden^{211, 213, 216, 217} (Figure 28). Only three of the eight studies conducted in a child population demonstrated a prevalence of CD of greater than 1.0% (Finland,²¹⁵ Sweden,²⁰⁹ Netherlands²³²).

Among the 30 included studies, there was a considerable amount of variation in the point estimates for the prevalence of CD both by serology and by biopsy due to differences in serological test strategies, biopsy definitions and patient sampling, making pooled estimates unreliable. To further explore the potential sources of variability in the observed prevalence of CD, we plotted the studies' prevalence versus its sample size (Figure 29). This scatter diagram visually illustrates the distribution of the prevalence of CD among the included studies. The study with the highest reported prevalence of CD (2.67%), was also the one with the smallest sample size of 150 healthy patients, and also included several other at-risk groups, which were the primary focus of that study.²³⁴ Overall, studies with the smallest sample sizes tended to produce both the highest and lowest prevalence of CD. Using an arbitrary cut-off of 1,600 patients to divide “small” and “large” sample size studies, the prevalence by serology ranged fairly evenly from 0.17% to 2.67% for the 13 small studies, while 12 of the 18 large studies were located within a range of 0.5% to 1.26% (one study did not provide prevalence by serology).