Antibody screening for celiac disease
Online posting: October 1, 1997
Published in print: January 27, 1998 (CMAJ 1998;158:171)
Re: Effectiveness of antigliadin antibodies as a screening
test for celiac disease in children, Lucie J. Chartrand and
associates, CMAJ 1997;157(5):527-33 [full text / résumé]
See response from: E.G. Seidman & L.J. Chartrand
In this interesting study the authors report that antigliadin
antibodies can be used to help in determining whether patients
with suspected celiac disease should undergo duodenal biopsy.
However, I would like to point out 2 methodologic limitations
that should be considered before their results are applied to
other patients.
First, the authors reported performing "negative controls," which
they described as assays with the anti-human immunoglobulin
antibodies but without serum. The optical densities from the
negative control wells were subtracted from the results obtained
with each patient's serum. However, this description refers to a
blank, not a negative control. A negative control would be used
to verify the absence of nonspecific binding of immunoglobulins
to the microtitre plate wells. A true negative control would have
involved serum from a normal subject, and its results would not
have been used in the calculation of the patients' results.
The method for selecting the best cut-off values is also
problematic. Using a receiver operating characteristics (ROC)
curve, the authors evaluated each possible cut-off point and
retained the point that gave the highest sensitivity with an
acceptable false-positive rate. It must first be noted that the
cut-off points reported are only valid for the assay conditions
in this study. If antibodies from another manufacturer were used
or the assay conditions were modified, another cut-off point
would have to be determined. A change in the specific activity of
the labelled anti-human immunoglobulins, which can occur from one
lot to another, would also lead to variation in the cut-off
point, even if all other assay conditions remained constant.
On purely statistical grounds, it should be noted that the
optimal cut-off points and the discriminative ability were
determined from the same cohort of patients. Such a design tends
to overestimate the performance of the marker. If the experiment
were repeated in another cohort according to the predetermined
cut-off points, a lower discriminative performance would be
expected because the cut-off points would not be optimal for the
new cohort. Furthermore, the 95% confidence intervals for
sensitivy (not reported) were large (65% to 95% for IgA
antigliadin antibody). Therefore, the results should be confirmed
prospectively in another cohort (with the predetermined cut-off
points) before antigliadin antibody testing is used to decide
whether duodenal biopsy is appropriate for patients with
suspected celiac disease. If the true sensitivity is only 70%, a
large proportion of affected patients (30%) would be denied a
diagnostic test and appropriate therapy.
Jacques Massé, MD, MSc
Biochemistry Division
Centre hospitalier universitaire de Québec
Québec, Que.
jmasse@upc.qc.ca
CMAJ email
CMAJ Readers' Forum