Estimating the Impact of Verification Bias on Celiac Disease Testing

Abstract

Goal: The goal of this study was to estimate the impact of verification bias on the diagnostic accuracy of immunoglobulin A tissue transglutaminase (IgA tTG) in detecting celiac disease as reported by an authoritative meta-analysis, the 2016 Comparative Effectiveness Review (CER). Background: Verification bias is introduced to diagnostic accuracy studies when screening test results impact the decision to verify disease status. Materials and methods: We adjusted the sensitivity and specificity of IgA tTG reported by the 2016 CER with the proportion of IgA tTG positive and negative individuals who are referred for confirmatory small bowel biopsy. We performed a systematic review from January 1, 2007, to July 19, 2017, to determine these referral rates. Results: The systematic review identified 793 articles of which 9 met inclusion criteria (n=36,477). Overall, 3.6% [95% confidence interval (CI): 1.1%-10.9%] of IgA tTG negative and 79.2.2% (95% CI: 65.0%-88.7%) of IgA tTG positive individuals were referred for biopsy. Adjusting for these referral rates the 2016 CER reported sensitivity of IgA tTG dropped from 92.6% (95% CI: 90.2%-94.5%) to 57.1% (95% CI: 35.4%-76.4%) and the specificity increased from 97.6% (95% CI: 96.3%-98.5%) to 99.6% (95% CI: 98.4%-99.9%). Conclusions: The CER may have largely overestimated the sensitivity of IgA tTG due to a failure to account for verification bias. These findings suggest caution in the interpretation of a negative IgA tTG to rule out celiac disease in clinical practice. More broadly, they highlight the impact of verification bias on diagnostic accuracy estimates and suggest that studies at risk for this bias be excluded from systematic reviews.