Immunoglobulin A tissue transglutaminase offers a noninvasive way to detect celiac disease, but new research suggests that its sensitivity may be overestimated and that it may not be an effective screening test, at least in asymptomatic individuals. The reason comes down to verification bias, wherein a technique appears to have higher sensitivity and lower specificity because individuals who screen positive are more likely to have their disease confirmed by a follow-up small-bowel biopsy while those who screen negative are unlikely to have a follow-up biopsy that could reveal missed celiac disease.
Dr. Marisa Stahl
“The issue with verification bias is that only the patients that screen positive on that index test are going to be getting the reference test, so there’s probably a good chance that if they screen positive when they go to that reference test they’ll also be positive. What you’re missing from when you’re calculating sensitivity is, what about the ones that are negative on the index test? Would they have been positive on that reference test? That’s not even coming into your calculation because they’re not getting that reference test,” said Marisa Stahl, MD, a physician and researcher at the Children’s Hospital Colorado Center of Celiac Disease in Aurora. Dr. Stahl was not involved in the meta-analysis, but commented on it in an interview.
The only way to fully correct for this bias is to conduct both IgA tissue transglutaminase (tTG) testing and small bowel biopsy on a complete or random sample of patients and compare the sensitivity and specificity of IgA tTG with the preferred method small-bowel biopsy. However, this is rarely done. Instead, when the U.S. Preventive Services Task Force concluded that evidence was insufficient for IgA tTG testing for celiac disease, it relied on a 2016 comparative effectiveness review of nine studies that estimated sensitivity at 92.6% and specificity at 97.6%. USPSTF remained noncommittal because of inadequate evidence surrounding the balance of benefit and harms of screening for celiac disease in asymptomatic individuals.
In the current meta-analysis, Isabel Hujoel, MD, of the Mayo Clinic, Rochester, Minn., and colleagues tested whether the studies used by USPSTF may have overestimated sensitivity because of verification bias. In a report in the Journal of Clinical Gastroenterology, they reviewed those same nine studies to see the potential impact of verification bias. They rated each individual study as being at high, low, or unclear risk of verification bias and found five they considered to be high risk.
To reveal the impact of small-bowel biopsy referral rates on sensitivity and specificity, the researchers reviewed a separate set of nine retrospective and prospective studies to determine the frequency of referral for both IgA tTG–positive patients (positive referral rate) and IgA tTG–negative patients (negative referral rate), which were 79.2% and 3.6%, respectively.
The researchers then used these values to recalculate the sensitivities and specificities in the five original studies considered high risk for verification bias, then pooled those adjusted values with the remaining, unadjusted values from the studies considered low or unclear risk of bias. The new overall values were 57.1% sensitivity (95% confidence interval, 35.4%-76.4%) and 99.6% specificity (95% CI, 98.4%-99.9%).
“The reported sensitivity and specificity of IgA tTG ... are substantially biased due to a lack of adjustment for verification bias. Specifically, adjusting for verification bias decreases the sensitivity of IgA tTG from 92.5% to 57.1%, with a drop in the lower limit of the 95% CI to 35.4%, and an increase in the specificity from 97.9% to 99.6%, The low estimated sensitivity of IgA tTG raises concern on the accuracy of this test and supports performing a systematic review that accounts for verification bias. ... After adjusting for verification bias, the estimated sensitivity of IgA tTG falls to the point where the serologic marker may no longer be clinically useful as a screening test,” the authors wrote.
The numbers came as a bit of a shock to Dr. Stahl because the sensitivity was so much lower than has been traditionally accepted. “But the more important concept from the paper is that the sensitivity is probably lower than what we oftentimes reference, and we should think more about the population of patients that could potentially screen negative and still have celiac disease,” she said. Although there is no literature to back this up at this time, Dr. Stahl also believes that this may be more common in adults, who have a higher incidence of seronegative Celiac disease.
The issue isn’t restricted to celiac disease. Verification bias can also affect the sensitivity and specificity values from other index screens that are followed by invasive reference tests, like occult blood and colonoscopy or hepatitis C serology and liver biopsy. “A lot of times you ethically cannot put everyone through the [more invasive] reference test, so it definitely applies to other tests we screen for in GI. When we’re quoting numbers and doing systematic reviews and meta-analyses, we should be accounting for those biases,” said Dr. Stahl.
No source of funding was disclosed. The authors declared that they have nothing to disclose. Dr. Stahl consults for Evo-Endo.