Context:

Celiac disease (CD) is a disorder of small bowel malabsorption. It is characterized by mucosal inflammation, villous atrophy and crypt hyperplasia that occur upon exposure to gluten, and clinical and histological improvement with withdrawal of gluten from the diet. The classical presentation of CD has now been shown to be less common than silent or atypical presentation, in which patients do not have intestinal symptoms. Untreated CD is associated with multiple important short- and long-term complications including nutritional derangements, anemia, reduced bone density, as well as intestinal lymphoma. In the vast majority of patients, CD is effectively treated with dietary modifications that eliminate gluten. Mounting evidence suggests that CD is actually considerably more common than previously believed and, therefore, this disorder warrants consideration for screening of at-risk patients, as well as possibly the general population.

Objectives:

To conduct a comprehensive systematic review on five areas of CD: (1) sensitivity and specificity of serological tests; (2) prevalence and incidence of CD; (3) CD associated lymphoma; (4) consequences of testing for CD; and, (5) interventions for the promotion and monitoring of adherence to a gluten-free diet (GFD).

Data Sources:

Staff of the National Library of Medicine performed a series of searches in support of the literature review of CD. Searches were run in the MEDLINE® (1966 to Oct 2003) and EMBASE (1974 to Dec 2003) databases for each of the five objectives and their respective sub-objectives separately.

Study Selection:

Study selection for each objective was performed using three levels of screening with predetermined increasingly more strict criteria to ensure that all relevant articles were captured. Following a calibration exercise, two reviewers independently screened all studies using a web-based system allowed automatic identification of review disagreements. These disagreements were resolved by consensus.

Data Extraction:

For each CD objective, a detailed and standardized data abstraction form was developed. For each 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 assessments were performed using specific instruments for each of the included study types.

Data Synthesis:

The data obtained from this review fell into several broad categories, which correspond in large part to the individual study objectives. Data for the sensitivity and specificity of each serological marker was considered separately, and studies were further divided according to the age group of the study population. Attempts were made to identify, explain, and minimize clinical and statistical heterogeneity in the included studies. 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 qualitative systematic review was conducted.

To produce clinically useful pooled statistics, a weighted mean of the overall sensitivity and specificity from the included studies was calculated, along with 95% confidence intervals (CIs). The pooled estimates for the sensitivity and specificity were compared with a summary receiver operating characteristic (ROC) curve, calculated for the same group of studies as a second check of the estimates.

Results/Conclusions:

This report has provided a systematic review of five broad areas (and corresponding sub-areas) of CD. Perhaps one of the most important findings of this report is the significance of how one chooses to define CD in the era of serological testing, and how this apparently clear-cut task has profound implications on all the results presented in this report. Specifically, can CD be diagnosed solely on the basis of serology? Is some degree of villous atrophy necessary for a diagnosis of CD. These questions have important implications downstream of the diagnosis as well. For example, do CD patients without symptoms or villous atrophy have the same risk of complications as those with villous atrophy. Is serological improvement on a GFD sufficient to reduce CD complications, or must there be documented histological improvement, and what degree of histological improvement is necessary?

The results of the Celiac 1 objective suggest that in the era of EMA and tTG antibody testing, AGA antibody testing in both children and adults has a limited role. The sensitivity and specificity of EMA and tTG are quite high (over 95% for sensitivity, and close to 100% for specificity), as are their positive and negative predictive values; however, one has to be aware that the reported diagnostic parameters are taken from studies in which the prevalence of CD was, for the most part, much higher than that seen in usual clinical practice. The positive predictive values reported for these tests will certainly not be as high as that reported when these tests are used to screen the general population. The bulk of the evidence on the diagnostic characteristics of these tests was derived from studies that defined CD as having at least some degree of VA.

HLA DQ2/DQ8 testing appears to be a useful adjunct in the diagnosis of CD. The test has high sensitivity (in excess of 90%–95%), however, since approximately 30% of the general population, and an even higher proportion of “high-risk” subjects (e.g., diabetics and family members) also carry these markers, the specificity of this test is not ideal. The greatest diagnostic utility of this test appears to be its negative predictive value.

Biopsy itself, when used with a strict cut-off requiring villous atrophy, appears to have high specificity, but poor sensitivity. Using a lower grade cut-off clearly improves sensitivity, but because of the wide differential of causes of histological lesions similar to Marsh I to IIIa, the specificity suffers. The use of histomorphometric measures such as quantification of gamma delta positive intraepithelial lymphocytes (γδ+ IELs) are likely to allow for the use of lower grade cut-offs, while maintaining reasonable specificity. Ultimately, a trial utilizing multiple diagnostic tests in an attempt to capture as many CD patients in a clinically-relevant population as possible, along with a time dimension such as a response to a GFD or gluten challenge, is required to fully assess the diagnostic characteristics of biopsy alone. This type of study would be able to characterize the false-positive and false-negative rates, provided that all studied patients are followed forward in time.

The included prevalence studies demonstrated important differences between the studies including, execution, tests for prevalence assessment, and patient sampling. Thus, results have to be interpreted in the light of some of the limitations that have been identified regarding the diagnostic performance of the tests for CD. Nonetheless, the results of this report suggest that CD is a very common disorder with a prevalence in the general population that is likely close to 1:100 (1%). Several high-risk groups with a prevalence of CD greater than that of the general population have been identified and include: those suspected of having CD; family members of CD patients; type I diabetics; and, those with iron-defiency anemia (IDA) or low bone mineral density (BMD). Additionally, the review identified many other high-risk groups, including those with Down Syndrome, short stature, and infertility, to name a few. Their inclusion was however, beyond the scope of this report

The results of this report confirm that, apart from a few limitations, there is a strong association between CD and GI lymphoma. The report identified standard incidence ratios (SIR) for lymphoma that ranged from 4 to 40, and standard mortality ratios (SMR) that ranged from 11 to 70. A diagnostic delay—in particular a diagnosis of CD in adulthood as apposed to in childhood—is associated with poorer outcomes. Fortunately, several studies suggest that adherence to a GFD reduces the risk of lymphoma in CD patients.

The consequences of testing for CD in at-risk and symptomatic patients appears to be more straightforward, since these patients appear to be more compliant with a GFD and would be expected to benefit from this intervention. The data is less clear for asymptomatic screen-identified patients, particularly those who have truly silent CD and/or don't have fully-developed villous atrophy. On the one hand the outcome of such patients has not been extensively studied, and on the other hand compliance with a GFD appears problematic, particularly for those diagnosed in adulthood.

Finally, no specific interventions have been identified that promote adherence to a GFD, but education of patients and family members about CD and about the intricacies of a GFD, and participation in local celiac societies, has been shown to improve compliance. Although somewhat controversial, biopsy monitoring of adherence to a GFD appears to be important, since improvement in histological grade has been associated with improved BMD, IDA, and nutritional status. The serological markers appear to be adequate for detecting gross dietary indiscretion, and respond to a gluten challenge, but appear to have poor sensitivity for detecting lesser degrees of dietary indiscretion, and inadequately correlating with histological improvement at least in the short-term. It should, however, be noted, that we could not identify a controlled study that objectively determined the level of histological improvement that would be associated with improved outcomes, and this is an area for future study. Nonetheless, based on this report it would appear that follow-up biopsy, at least 1 year after a GFD in adults to document improvement of the histological grade, would be valuable.