Overall Accuracy of the GastroPanel® Test in Diagnosis of Stomach Disorders: A Pooled Analysis of Three Clinical Validation Studies Using Random-effects Bivariate Model and Hierarchical Multilevel Logistic Regression Models

Background/Aim: The serological biomarker panel GastroPanel®, developed by Biohit Oyj (Finland), has gained increasing global acceptance in diagnosis of dyspeptic patients. The diagnostic accuracy of this test has only been confirmed for (atrophic gastritis) endpoint, but not for the whole spectrum of its target disorders. Patients and Methods: Three published GastroPanel-validation studies on gastroscopy referral patients (n=961) were pooled to estimate the overall DA of the GastroPanel test in detection of healthy stomach and of the whole spectrum of its target gastric disorders: i) Helicobacter pylori-associated gastritis, ii) atrophic antrum gastritis, iii) atrophic corpus gastritis, and iv) atrophic pan-gastritis. Random-effects bivariate model (REBM) and hierarchical summary receiver operating characteristic (HSROC) model were used to calculate the overall accuracy indicators of the GastroPanel test. Results: The overall agreement between GastroPanel results and classification by the Updated Sydney System was 0.877 (95% CI=0.86-0.89). The summary sensitivity, specificity, positive likelihood ratio, negative likelihood ratio and diagnostic odds ratio for the five diagnostic categories of GastroPanel were 0.62 [95% confidence interval (CI)=0.33-0.87], 0.98 (95% CI=0.92-0.99), 26.6 (95% CI=9.2-77.0), 0.37 (95% CI=0.16-0.83) and 73 (95% CI=22-245), respectively. Both REBM and HSROC models provided an identical summary AUC of 0.960 (95% CI=0.93-0.97). The post-test probability was shown to be 90% for a correct GastroPanel diagnosis in a population with 25% pre-test probability of these target conditions. Conclusion: The overall diagnostic accuracy of the GastroPanel test in detecting its target disorders of the stomach was 96%.