Statistical analysis of the reproducibility and predictive capacity of MCTT HCE (TM) eye irritation test, a me-too test method for OECD TG 492

Abstract

We aimed to conduct additional statistical analysis of the reproducibility and predictive capacity of MCTT HCE (TM) eye irritation test (EIT), a me-too test method for OECD TG 492 with the data generated during the validation study in which 30 reference chemicals were tested in three repeated runs by three independent laboratories. We evaluated the within-laboratory reproducibility (WLR) and the between-laboratory reproducibility (BLR) through tabulation and graphs and presented the concordance of eye irritancy prediction with 95% Wilson's confidence intervals (CIs). Also, the analyses of the Intra-Class Correlation Coefficient (ICC) and Bland-Altman plot were applied to confirm the reproducibility and the comparability, referring to the validated methods of OECD TG 492. Kappa analysis was also performed to check the degree of agreement of the within- and between-laboratory reproducibility and agreement between MCTT HCE (TM) EIT and the reference methods. We calculated the predictive capacity via misprediction over total prediction method. The predictive capacity (sensitivity, specificity, and accuracy) was presented with 95% of Wilson's CIs. Also, bootstrap resampling was performed to express the 95% CI by the simple percentile methods for 30 chemicals and 141 reference chemicals additionally tested. Collectively, WLR (92.2%) and BLR (93.3%) met the criteria of the performance standards (WLR >= 90% and BLR >= 85%), and the results of ICC analysis and the Bland-Altman plot suggested an acceptable WLR and BLR and comparability to other validated methods of OECD TG 492. Also, the predictive capacity results for the 30 reference chemicals confirmed the good performance of the MCTT HCE (TM) EIT by satisfying the criteria of sensitivity, specificity, and the accuracy stated in the PS. The bootstrap resampling method showed a predictive capacity, sufficiently satisfying the criteria stated in the Performance Standards.