This review assessed diagnostic accuracy of tests for coeliac disease in adults who presented with abdominal symptoms in primary care or similar settings and concluded that Immunoglobulin-A anti-tissue transglutaminase antibodies and Immunoglobulin-A anti-endomysial antibodies showed good diagnostic performance. The authors' conclusions appear to reflect the evidence, but there may have been some bias in the included studies and review process.

To assess diagnostic accuracy of tests to identify coeliac disease in adults who presented with abdominal symptoms in primary care or similar settings.

MEDLINE and EMBASE were searched up to December 2009 for articles published in English, Dutch, German or French. Search terms were available online. Reference lists of relevant studies, reviews and guidelines were searched manually.

Diagnostic studies that used a cohort design, and nested case-control designs that included a control group of patients with functional gastrointestinal problems, were eligible for inclusion if they were of adults in primary care or open-access outpatient clinic settings. The prevalence rate of gastrointestinal symptoms was required to be 50% or more. Prevalence of coeliac disease was required to be less than 15%. Eligible studies had to confirm coeliac disease using small-bowel biopsy and histology and assess gastrointestinal symptoms or serum antibody tests alone or in combination. Studies of hospitalised patients and studies where a diagnostic 2x2 table could not be reconstructed were excluded from the review.

Where reported, median age of patients in included studies ranged from 33 to 56 years. Prevalence of coeliac disease ranged between 2% and 13%. Index tests varied among studies, but included family history, gastrointestinal symptoms, symptoms of diarrhoea, weight loss or anaemia, irritable bowel syndrome and various serum antibody tests.

One reviewer screened citations for inclusion and a second reviewer independently checked eligibility of assessed abstracts. Any disagreements were resolved through discussion.

Two reviewers independently assessed studies for quality using the QUADAS (Quality Assessment of Diagnostic Accuracy Studies) tool. Studies were assessed on 11 items scored as positive (no bias), negative (potential bias) or insufficient information. Disagreements were resolved by consensus.

Two reviewers independently extracted the data required to reconstruct a 2x2 table and calculate sensitivity, specificity and likelihood ratios along with 95% confidence intervals (CIs). Where 2x2 tables contained 0, 0.5 was added to each cell. The authors did not state how discrepancies were resolved.

Where four or more studies on a specific index test showed no evidence of statistical heterogeneity, sensitivity, specificity and likelihood ratios, and their 95% CIs were pooled using bivariate analyses. Statistical heterogeneity was defined as non-overlapping 95% CIs of both sensitivity and specificity and differences in point estimates among the studies of more than 20%. Where pooling of the results was not possible, ranges were presented as part of a narrative synthesis.

Subgroup analyses and sensitivity analyses were undertaken to determine whether certain factors explained variation in performance.

Sixteen studies (n=6,085) were included in the review: 13 cohort (three of which were retrospective) and three nested case-control studies. Sample sizes ranged from 79 to 2,000 patients. Four studies performed well on quality and scored a positive assessment on at least eight of 11 items. The other studies suffered either from some form of bias or (in many cases) items could not be scored due to insufficient information.

Performance of individual abdominal symptoms varied among studies and it was not possible to pool the results. Diarrhoea had a sensitivity of 0.27 to 0.86 and specificity of 0.21 to 0.86 (six studies). Results for other abdominal symptoms were reported in the review with likelihood ratios close to one, which indicated poor performance.

Studies that assessed the diagnostic value of immunoglobulin A or G antigliadin antibodies (IgA-AGA or IgG-AGA) also reported variation and results could not be pooled. Sensitivity ranged between 0.46 and 0.87 for IgA-AGA (six studies) and between 0.25 and 0.93 for IgG-AGA (five studies) and specificity ranged between 0.70 and 0.98 for IgA-AGA and between 0.80 and 0.99 for IgG-AGA.

IgA anti-endomysial antibodies (EmA) showed pooled sensitivity of 0.90 (95% CI 0.80 to 0.95; eight studies) and specificity of 0.99 (95% CI 0.98 to 1.00; eight studies). Pooled estimates for IgA anti-tissue transglutaminase antibodies (IgA-tTG) (seven studies) showed sensitivity of 0.89 (95% CI 0.82 to 0.94) and specificity of 0.98 (95% CI 0.95 to 0.99).

Results from studies that assessed use of IgG-tTG and/or deamidated gliadin peptides (IgA-DGP) were reported in the review.

Subgroup analyses and sensitivity analyses were reported in the review.

IgA anti-tissue transglutaminase antibodies and IgA anti-endomysial antibodies showed good performance (high sensitivity and specificity) in diagnosing coeliac disease in adult patients who presented with chronic abdominal symptoms in primary care or other unselected populations; evidence in primary care populations was limited.

The review question was clear and supported by appropriate inclusion criteria. Two electronic databases were searched. Language restrictions were imposed and so language bias may have been introduced. The search did not appear to include any attempt to locate unpublished data and so potentially relevant studies may have been missed. Validity was assessed and most included studies did not perform well on quality. The authors undertook each stage of the process in duplicate, which minimised risk of reviewer error and bias. The authors used appropriate methods to synthesise data and presented a narrative synthesis where there was evidence of clinical and statistical heterogeneity among studies. The authors acknowledged the limitations with study designs and limited power of subgroup analyses. The authors' conclusions appear to reflect the evidence, but interpretation of the conclusions should take into account the quality of the included studies and potential that relevant papers may have been missed.

Practice: The authors stated that gastrointestinal symptoms alone were insufficient for diagnosing coeliac disease. Patients with coeliac disease may present with symptoms other than abdominal symptoms. For optimal diagnosis of coeliac disease, diagnostic testing should be considered in patients who present with other symptoms or health conditions that may indicate coeliac disease (such as iron-deficiency anaemia, infertility, type 1 diabetes, Down syndrome and reduced bone density).

Research: The authors stated that future research should investigate the performance of a diagnostic algorithm using sequential serological testing in patients with chronic or refractory abdominal symptoms in primary care, given the dependence of diagnostic performance on prevalence and spectrum of disease. The authors stated that a diagnostic randomised trial was required to compare costs of different diagnostic strategies and their effects on treatment decisions and subsequent patient outcomes.

Grant 945-06-001 from the Netherlands Organisation for Health Research and Development, the Hague, the Netherlands.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.