Pedro Zapater1, José Such, Miguel Pérez-Mateo, José Francisco Horga. 1. Clinical Pharmacology Unit, University General Hospital of Alicante and Department of Pharmacology and Therapeutics, Faculty of Medicine, University Miguel Hernández, San Juan de Alicante, Spain. zapater_ped@gva.es
Abstract
OBJECTIVE: The diagnosis of drug-induced hepatotoxicity is based on circumstantial evidence and is often inaccurate. We have designed a method based on published data to assign causality to suspected cases of drug-induced hepatotoxicity. DESIGN: Forty-seven published cases of ticlopidine-induced hepatotoxicity were identified by a Medline-based literature search. Data regarding abnormal liver function in patients receiving ticlopidine were obtained from the only published placebo-ticlopidine clinical trial (the Canadian American Ticlopidine Study; CATS). Thus, we calculated the maximum number of expected hepatotoxicity cases in patients exposed to ticlopidine and those not exposed to the drug by means of the Poisson distribution. The calculated odds ratio was used as a prior odd for subsequent quantification, using a Bayesian-based approach, of individual ticlopidine-induced hepatotoxicity likelihood. Concretely, the prior odd is modified by several separate likelihood ratios: age; sex; AST level; ALT level; alkaline phosphatase level; total bilirubin level; latent period of adverse reaction appearance; and period of remission of adverse reaction. This methodology was applied to two new cases of suspected ticlopidine-induced hepatotoxicity. RESULTS: The prior probability of ticlopidine-induced hepatotoxicity derived from CATS data is 61.29%. This is in contrast with the 28.83% incidence rate of drug-induced liver alterations in the general population. Alkaline phosphatase levels and total bilirubin levels were six times the normal values among individuals with ticlopidine-induced hepatotoxicity than in the general population. They were the most relevant likelihood ratios of the Bayesian model to establish a high level of causality relationship between a hepatotoxicity event and ticlopidine use. CONCLUSIONS: The proposed method, which links information from clinical trials with the profile of clinical hepatotoxicity of a drug defined from published cases reported after a drug is marketed, can be a useful tool for drug postmarketing surveillance research.
OBJECTIVE: The diagnosis of drug-induced hepatotoxicity is based on circumstantial evidence and is often inaccurate. We have designed a method based on published data to assign causality to suspected cases of drug-induced hepatotoxicity. DESIGN: Forty-seven published cases of ticlopidine-induced hepatotoxicity were identified by a Medline-based literature search. Data regarding abnormal liver function in patients receiving ticlopidine were obtained from the only published placebo-ticlopidine clinical trial (the Canadian American Ticlopidine Study; CATS). Thus, we calculated the maximum number of expected hepatotoxicity cases in patients exposed to ticlopidine and those not exposed to the drug by means of the Poisson distribution. The calculated odds ratio was used as a prior odd for subsequent quantification, using a Bayesian-based approach, of individual ticlopidine-induced hepatotoxicity likelihood. Concretely, the prior odd is modified by several separate likelihood ratios: age; sex; AST level; ALT level; alkaline phosphatase level; total bilirubin level; latent period of adverse reaction appearance; and period of remission of adverse reaction. This methodology was applied to two new cases of suspected ticlopidine-induced hepatotoxicity. RESULTS: The prior probability of ticlopidine-induced hepatotoxicity derived from CATS data is 61.29%. This is in contrast with the 28.83% incidence rate of drug-induced liver alterations in the general population. Alkaline phosphatase levels and total bilirubin levels were six times the normal values among individuals with ticlopidine-induced hepatotoxicity than in the general population. They were the most relevant likelihood ratios of the Bayesian model to establish a high level of causality relationship between a hepatotoxicity event and ticlopidine use. CONCLUSIONS: The proposed method, which links information from clinical trials with the profile of clinical hepatotoxicity of a drug defined from published cases reported after a drug is marketed, can be a useful tool for drug postmarketing surveillance research.