(granulomatous hepatitis vs cholestasis from dilantin) see article below The Roussel Uclaf Causality Assessment Method (RUCAM) is the most widely used instrument. This instrument has 7 domains and provides a semi-quantitative scaled score ranging from -8 to + 14. The RUCAM domains and weighting were developed from a consensus opinion of an expert panel in 1990 and “validated” from a group of 49 published DILI cases which had been rechallenged and compared to 28 controls (2,3). Determination of the serum ALT to alkaline phosphatase ratio allows for the categorization of cases into hepatocellular, cholestatic, and mixed liver injury patterns for subsequent scoring. Limitations of the RUCAM include ambiguous instructions for use, reliance on rechallenge, and lack of evidence supporting the weighting and selection of domains (1). The Clinical Diagnostic Scale (CDS) is a simplified 5 domain instrument which initially was proposed for patients with a predominance of extrahepatic immunoallergic features (4). However, the CDS was inferior to the RUCAM in assigning causality in a large cohort of Spanish patients (5). In addition, both instruments performed poorly in the most severe cases of DILI that resulted in death, transplant, or prolonged cholestasis mostly due to the lack of dechallenge data. Therefore, the first generation causality assessment instruments have limited utility due to their attempt to classify all DILI cases using the same methodology and a lack of primary data to support the selection and weighting of component domains. Since DILI as a disease entity is highly variable between drugs, patient populations (i.e. valproate in children vs adults), and even within individual patients given the same drug (granulomatous hepatitis vs cholestasis from dilantin), it is unlikely that a single causality assessment instrument will be accurate and reliable in all patients with DILI from all agents unless there is dynamic weighting of component variables. In addition, mathematical methods to account for missing data will need to be incorporated. Bayesian approaches to adverse drug reactions that utilize mathematical calculations to develop prior and posterior probabilities of causality in individual cases have been proposed but require the accumulation of large datasets of cases and development of computerized algorithims (6). |