The share of specific match for each outcome category in the use of the WHO-CMU system was: “definitely” 1.0, “probable” 0.55, “possible” 0.60, “unlikely” 0.76 and “non-computable” 0.77. The specific concordance proportion for each outcome category used by the LCAT was: “definitely” 0.67, “probable” 0.55, “possible” 0.24, “unlikely” 0.56 and “non-computable” 0.29. We propose that improvements to LCAT may be warranted. These improvements should aim to reduce the potential of the information loss instrument, to focus on the greater convergence of the two “subjective” issues identified as sources of disagreement, and to provide guidance for answering questions in ambiguous situations. This study was a two-arm study of the AKI, which used a sample of cases identified as suspected ADR-related adr admissions during our hospital-based investigation. In part of the study, we measured the AKI obtained when the evaluators used the WHO-UMC, and we measured the AKI that was achieved when the evaluators used the LCAT in the same cases in the second arm of the study. A new aspect of our study design was to eliminate the need for a wash time between the two arms by using an incomplete and randomized design. Instead of each evaluator assessing each case using each method, our design used randomized blockers to randomly assign cases and evaluators causality assessment methods, so that each case was assessed by a new random half of the reviewers with one method and by the remaining evaluators with the second method. Results: The overall agreement between the experts was poor (kappa=0.20), although it was significantly different from chance and varied according to the degree of causality. For improbable, dubious, unclassifiable and plausible categories (kappa=0.03, 0.03, -0.01 and 0.13 respectively), it was lower than that of the extremes: excluded, probable and safe (kappa= 0.40, 0.32 and 0.30 respectively). Methods: A random sample of 150 active action pairs was produced. For each couple, five chief experts had to assess, independently of each other, the probability of a cause from 0 to 1 with a visual analogue scale (VAS) of 100 mm.
For the analysis, these probabilities were divided secondarily into seven levels of causality: excluded (0-0.05); unlikely (0.06-0.25); doubtful (0.26-0.45); non-evaluable/unclassifiable (0.46-0.55); plausible (0.56-0.75); probable (0.75-0.95); and some (0.95-1). The concordance between the five experts was assessed using kappa coefficients (Kappa). Note: Mouton JP, Mehta U, Rossiter DP, Maartens G, Cohen K (2017) Interrater agreement of two adverse drug reaction causality assessment methods: A randomised comparison of the Liverpool Adverse Drug Reaction Causality Assessment Tool and the World Health Organization-Uppsala Centre Monitoring System. PLoS ONE 12 (2): e0172830. doi.org/10.1371/journal.pone.0172830 context and objective: Assessing the causality of drug-induced liver damage (DILI) remains a challenge and looks forward to the development of reliable hepatotoxicity biomarkers. None of the various available algorithms used to assess the causality of DILI have been generally accepted as gold standards. This study was carried out in order to examine the concordance between the different causality assessment scales when covering DILI. .