Accuracy in Diagnosis of Celiac Disease Without Biopsies in Clinical Practice

TL;DR: Children can be accurately diagnosed with celiac disease without biopsy analysis, and a positive predictive value (PPV) above 99% in clinical practice is validated.

About: This article is published in Gastroenterology.The article was published on 2017-10-01 and is currently open access. It has received 163 citations till now. The article focuses on the topics: Pediatric gastroenterology & Predictive value of tests.

Summary

BACKGROUND & AIMS:

• The guidelines of the European Society of Pediatric Gastroenterology, Hepatology, and Nutrition allow for diagnosis of celiac disease without biopsies in children with symptoms and levels of immunoglobulin A against tissue-transglutaminase (TGA-IgA) 10-fold or more the upper limit of normal (ULN), confirmed by detection of endomysium antibodies (EMA) and positivity for HLA-DQ2/DQ8.
• The authors performed a large, international prospective study to validate this approach.
• Local centers recorded symptoms; measurements of total IgA, TGA, and EMA; and histopathology findings from duodenal biopsies.
• Secondary aims included determining the accuracies of various TGA tests and their reliability to predict CD if levels are 10xULN as well as the impact of HLA-typing, EMA-IgA, and type of symptoms on CD diagnosis without biopsies.

Study Design and Participants

• From November 2011 to May 2014, 33 pediatric gastroenterology units from 21 countries (Europe, Middle East) recruited consecutive patients younger than 19 years on a gluten-containing diet, with positive TGA results analyzed in their own or external laboratories.
• Exclusion criteria comprised refusal to duodenal biopsies, primary or secondary immunodeficiency, malignancy, or previous diagnosis of CD.
• Recruited patients were excluded from the analysis if local and central HLA results were unavailable, serum or histology slides were not provided for central assessment, biopsies were unreadable due to poor quality, total IgA was low, inclusion criteria were violated, or consent was withdrawn.

IMPACT

• More than 50% children and adolescents with celiac disease can be diagnosed without biopsies, avoiding the burden of upper endoscopy with anesthesia and saving health care costs.
• Abbreviations used in this paper: CD, celiac disease; CI, confidence interval; DGP, antibodies against deamidated gliadin peptides; EMA, endomysium antibodies; HLA, human leukocyte antigen;.
• Ig, immunoglobulin; T1DM, type 1 diabetes mellitus; TGA, autoantibodies against tissue-transglutaminase; PPV, positive predictive value; ULN, upper limit of normal.

Central Analyses

• All investigators performing central analyses were blinded toward available local and central results.
• Details and results on DGP-IgG tests are shown in the supplementary tables only.
• Standard curves were available for all tests.
• In patients with negative HLA status but positive central serology and histopathology, a third HLA-typing for rare risk alleles was performed from a new blood sample.
• The reference pathologist reported histology on provided slides (hematoxylin-eosin and CD3þ immunostaining), including Marsh-Oberhuber-staging.

Central Diagnosis

• The final central diagnosis for each patient was (1) proven CD, (2) no CD, or (3) inconclusive case.
• CD was proven if HLA-DQ2/DQ8, local TGA-IgA, and local and/or central EMA-IgA were all positive, and both local and reference pathologists reported at least Marsh 2 staging.
• Patients not meeting these criteria were initially considered as unclear and histopathology was revised as described previously.
• The diagnostic committee reviewed each unclear case and voted in a Delphi process .
• If this did not allow a clear diagnosis, cases were finally regarded as inconclusive.

Criteria for CD Diagnosis Without Biopsies

• For local and central TGA levels, the multiple of the respective ULN was calculated and stratified into high positive (10xULN) or low to moderate positive (>1 to <10xULN).
• For tests with a given gray zone, the lower bound was used as ULN.
• To evaluate whether the nonbiopsy approach would contradict the final central diagnosis, the authors considered the combination of high local TGA, positive local EMA-IgA, positive central HLA status, and symptoms.
• Furthermore, the diagnostic accuracies of high central TGA (10xULN) for each included commercial kit alone and in combinations with HLA status, EMA results, and symptoms were calculated against the central diagnosis as reference.

Study Oversight

• The study was approved by the ethics committees of each participating center.
• Written informed consent was obtained by legal guardians and patients as appropriate for age.
• The study was cofunded by industry (EUROIMMUN Medizinische Labordiagnostika AG; Eurospital, Trieste, Italy; INOVA Diagnostics; R-Biopharm, Darmstadt, Germany; Phadia/ Thermo Fisher; Dr. Schär GmbH, Apolda, Germany) and nonprofit organizations (ESPGHAN, AOK Bayern health insurance, and CD patient organizations from Denmark, Finland, Germany, Hungary, Italy, the Netherlands, and the United Kingdom).
• Funding partners were not involved in study design, recruitment, data collection, analysis, and interpretation or writing of the manuscript.
• All authors had access to the study data and reviewed and approved the final manuscript.

Statistical Analyses

• With 701 participants, the study had 80% power at 5% significance level to detect a PPV of more than 97% for most test scenarios.
• When sequential test design was considered (by ADDPlan Software, Cologne, Germany), the needed number increased to 357.
• The interim analysis with the first 200 patients showed that the proportion of cases potentially qualifying for omitting biopsies with local parameter ranged between 50% and 65%.
• Sensitivity, specificity, PPVs, and positive likelihood ratios for different scenarios (TGA 10xULN alone and in combination with other criteria) were calculated with 95% confidence intervals (CI) using binominal distribution (Copper-Pearson CI).
• All statistical analyses were done by B.F. and K.W. using SAS 9.3 (SAS Institute Inc., Cary, NC).

Results

• In 29 patients, local TGA-IgA was negative at time of biopsy, but all had positive TGA-IgA before referral (Supplementary Table 9 ).
• For the other 9 patients without central DNA sample, local HLA-typing was available and considered for analysis.
• Sixty-four patients had tentatively started a gluten-free diet before the diagnostic workup of CD; 32 of those within 12 months before biopsy.

Diagnostic Accuracy in Clinical Practice

• Using the central diagnosis as reference, the diagnostic accuracies of local TGA-IgA 10xULN in combination with other criteria are shown in Table 2 .
• HLA results did not improve accuracies (scenario 4).
• If only malabsorption symptoms would qualify (39.3% of the patients, scenario 6), the PPV increased to 100%.
• E Based on results from Eu-Gen-typing for 697 patients, on EUROarray for 1 patient and for local HLA typing results for 9 patients.
• If malabsorption and/or EMA-IgA were included in the diagnostic decision, no false positives were found (scenarios 9-12).

Diagnostic Accuracy of Central Serology Evaluations

• If malabsorption symptoms were considered for the decision, or if inconclusive cases were excluded, no false positive was found.
• For the DGP-IgGs 10xULN, the specificity was high (1 false positive) but sensitivity was low .

Discussion

• The results of their prospective multicenter diagnostic evaluation study ProCeDE show that the ESPGHAN nonbiopsy approach allows a correct diagnosis of CD.
• In their study, 10 different TGA-IgA tests were used by the local laboratories of the 32 centers, with only 4 patients with high TGA-IgA levels 10xULN being false positive.
• 1, 33, 34 A positive EMA result as obligatory criterion for the nonbiopsy approach is still debated.
• In only 22 patients, the reason for not being recruited was refusal toward biopsy, which may bear a risk for bias but does overall minimally influence the proportion of children qualifying for the nonbiopsy approach.

Figures

Table 2.Diagnostic Accuracies With 95% CIs to Diagnose CD Based on Local TGA-IgA Tests in Combination With Other Criteria, Either Considering Inconclusive Cases as No CD (Scenarios 1–6, n ¼ 707) or Excluding Inconclusive Cases (Scenarios 7–12, n ¼ 691)

Figure 2. PPV with 95% CI (gray shaded) for CD diagnosis for ea and 1 TGA-IgG test (T9), all with calibration curve–based result limit of normal according to the manufacturer’s instructions (all Table 3 for the names and manufacturers of each test.

Table 3.Specifications of Central Serology Tests

Figure 1. Flow chart of eligible, recruited, and excluded patients and central diagnosis of final cohort (n ¼ 707); for the nonbiopsy approach, local serology results have been considered. In total, 96 patients were excluded; of those, 36 due to nonevaluable histopathology and 60 for other reasons.

Frequently asked questions

Q1. What are the contributions in "Accuracy in diagnosis of celiac disease without biopsies inclinical practice" ?

Accuracy in Diagnosis of Celiac Disease Without Biopsies in Clinical Practice Katharina Julia Werkstetter, Ilma Rita Korponay-Szabó, Alina Popp, Vincenzo Villanacci, Marianna Salemme, Gabriele Heilig, Søren Thue Lillevang, Maria Luisa Mearin, Carmen Ribes-Koninckx, Adrian Thomas, Riccardo Troncone, Birgit Filipiak, Markku Mäki, Judit Gyimesi, Mehri Najafi, Jernej Dolin sek, Stine Dydensborg Sander, Renata Auricchio, Alexandra Papadopoulou, Andreas Vécsei, Peter Szitanyi, Ester Donat, Rafaella Nenna, Philippe Alliet, Francesca Penagini, Hélène Garnier-Lengliné, Gemma Castillejo, Kalle Kurppa, Raanan Shamir, Almuthe Christine Hauer, Françoise Smets, Susana Corujeira, Myriam van Winckel, Stefan Buderus, Sonny Chong, Steffen Husby, and Sibylle Koletzko, on behalf of the ProCeDE study group 

Q2. How many children in the study will benefit from the nonbiopsy approach?

At least 50% of affected children in clinical practice will benefit from this nonbiopsy approach, which reduces burden and risks of endoscopy and anesthesia while saving costs for health care systems. 

Q3. How many patients had started a gluten-free diet before the diagnosis of CD?

Sixty-four patients had tentatively started a gluten-free diet before the diagnostic workup of CD; 32 of those within 12 months before biopsy. 

Q4. How many different TGA tests were used in the ProCeDE study?

In the ProCeDE study, 9 different TGA tests were centrally used; 7 of them reliably predicted CD with a PPV of 100% with titers 10xULN and at even lower levels. 

Q5. What is the purpose of this study?

Their Prospective Celiac Disease Diagnostic Evaluation study (ProCeDE) aimed to evaluate in a multicenter setting whether this nonbiopsy approach allows a correct diagnosis in clinical practice with a positive predictive value (PPV) above 99% when all required conditions are fulfilled. 

Q6. What is the reason for the nonbiopsy approach?

There is some concern that the nonbiopsy approach may result in clinically relevant missed comorbidities, such as gastroesophageal reflux disease, eosinophilic esophagitis, or Helicobacter pylori infection–related complications.39 

Q7. What was the main analysis of accuracies?

For main analysis of diagnostic accuracies, all inconclusive cases were considered as no CD, or were excluded in a subsample analysis. 

Q8. What was the procedure for reorienting the cuttings?

Ifspecimens were nonevaluable, the paraffin-embedded biopsy blocks were requested for reoriented cuttings and blindly evaluated, including morphometry. 

Q9. What is the simplest way to diagnose celiac disease?

Diagnosis based on level of TGA-IgA 10-fold or more the ULN, a positive result from the EMA tests in a second blood sample, and the presence of at least 1 symptom could avoid risks and costs of endoscopy for more than half the children with celiac disease worldwide. 

Q10. How many patients were excluded from the nonbiopsy approach?

Ninety-six patients were excluded, 36 due to nonevaluable histology and 17 due to low total IgA (Figure 1; Supplementary Tables 5, 6, and 8). 

Q11. How many patients with CD had positive results?

Only 2 of 645 patients with CD had initially a negative HLA status, both were later reliably identified as having initially falsenegative HLA results. 

Q12. What was the PPV for each central diagnosis?

Central diagnosis in the final cohort (n ¼ 707) was proven CD in 645 (91.2%), no CD in 46 (6.5%), and inconclusive case in 16 (2.3%) patients (SupplementaryTable 11).