Impact of interobserver disagreement on phenotype–genotype associations in Crohn's disease

• Published in: Inflammatory bowel diseases 2007-02, Vol.13 (2), p.156-163
• Main Authors: Oefferlbauer‐Ernst, Anna, Miehsler, Wolfgang, Eckmüllner, Otto, Travis, Simon, Waldhoer, Thomas, Dejaco, Clemens, Gangl, Alfred, Vogelsang, Harald, Reinisch, Walter
• Format: Article
• Language: English
• Subjects: Adult; classification; Crohn Disease - classification; Crohn Disease - genetics; Crohn Disease - pathology; Crohn's disease; Female; Genotype; Humans; interobserver agreement; Male; Observer Variation; Phenotype; phenotype–genotype association
• ISSN: 1078-0998; 1536-4844
• DOI: 10.1002/ibd.20016

Background: Nonvalidated definitions of disease‐related parameters in inflammatory bowel disease cause variations in diagnosis and disease classification. We determined interobserver agreement on applications of definitions of the Vienna Classification variables and computed the potential influence of misclassification on genotype/phenotype associations. Methods: Ten records of patients with Crohn's disease (CD) were independently evaluated by 19 observers using a standardized inflammatory bowel disease documentation system, which included the Vienna Classification. Interobserver agreement (IOA) was calculated as a percentage of the observers' agreement with a predetermined reference observer and by Cohen's kappa. Randomized reclassifications were then computed with 10,000 simulation runs using the IOA results and published NOD2/CARD15 gene status. A chi‐square independence test was calculated for each simulation run. Results: IOA for location and behavior was 70% (κ = 0.57) and 95% (κ = 0.91), respectively. IOA for location subgroups ranged from 48% to 88% and for behavior from 91% to 97%. By including the results of histopathology into the evaluation of location, the overall IOA increased significantly, to 80% (P = 0.019). Assuming a true genotype/phenotype association, the proportion of studies with nonsignificant findings (P > 0.05) because of the observed misclassification of location ranged from 13.3% to 63.8% and of behavior from 0.2% to 22.2%, depending on a study sample size of 500 or 150 patients respectively. Conclusions: We concluded that there is appreciable interobserver disagreement on the location of CD according to the original Vienna Classification that may obscure true genotype/phenotype associations. Definitions of disease parameters have to be validated before being used as the bases for classifications. (Inflamm Bowel Dis 2007)