Seventy six cohort studies (n = 8,053 participants) were included; no controlled trials were identified. The age range of study participants was two months to 88 years; 44 studies were of children (up to 18 years). Participants were either symptomatic or had various risk factors for coeliac disease; two studies were in patients with diabetes mellitus. Eighteen studies were classified as clearly described cohorts with explicit selection methods and full quality assessment reported for these studies.
No evidence of threshold effect was observed in any data set. However, there was significant between-study heterogeneity in measures of test accuracy, which may have been due to differences in aspects of study quality, tests and their execution or the study populations.
All antibody tests assessed showed reasonable overall accuracy, as indicated by the area under the SROC curve, which ranged from 0.845 for IgG AGA (well-described studies) to 0.992 for IGA EMA (all included studies).
IgA EMA (pooled positive likelihood ratio of 43.64, 95% CI: 28.70 to 66.37 and pooled negative likelihood ratio of 0.09, 95% CI: 0.06 to 0.13) and IgA TTG (17.00, 95% CI: 11.19 to 25.84 and pooled negative likelihood ratio of 0.11, 95% CI: 0.06 to 0.21) appeared to have the best overall performance for both ruling in and ruling out disease (all included studies).
IGA ARA had a high pooled positive likelihood ratio (27.98, 95% CI: 9.53 to 82.12), but only a moderate pooled negative likelihood ratio (0.25, 95% CI: 0.14 to 0.46) with highly variable sensitivity estimates, making it a poor candidate for ruling out disease (all included studies).
Pooled likelihood ratios for IgA AGA and IgG AGA indicated that these tests were only moderately useful for either ruling in or ruling out disease.
No summary estimates were calculated for IgG ARA, IgG EMA, or IgG TTG as there were insufficient studies, but data from individual studies indicated that sensitivities were low for all three tests.