The epidemiology of adverse hepatic reaction remains poorly documented. Controlled clinical trials have the advantages of close and prospective surveillance as well as a control group. However, the median number of subjects exposed to a new drug at the time of marketing is usually around 1500, with rarely more than 100 patients receiving the product for more than a year (Rawlins, 1995). This is clearly inadequate since around 30 000 treated subjects need to be observed to identify, with a power 0.95, at least one with drug hepatotoxicity when the incidence is 1 in 10 000 patient years (Stricher, 1992). The debate surrounding the initial approval and the recent withdrawal from the market of troglitazone high lights the realities of drug development and the need for postmarketing surveillance. In the clinical trials of troglitazone (representing a novel class of oral antihyperglycaemic agents), 1.9% of 2510 patients receiving the drug had elevated liver enzymes, two of which developed reversible jaundice (Watkins and Whitcomb, 1998). It took more than 3 years and 90 deaths or liver transplantations (in over a million patients treated), before the drug was withdrawn from the market (Lumpkin, 2000). Furthermore, clinical trials usually include selected patients and the findings may therefore not be generalisable to a wider population. Hence, in the United Kingdom and many other countries, postmarketing surveillance relies largely on spontaneous reporting (Rawlins, 1995), and data on adverse hepatic reactions have come most often from this source. Spontaneous reporting allows surveillance to continue throughout the life of a marketed drug when a large number of individuals have been exposed to the drug, and hence relatively rare adverse reactions have been recognised. However, only 10% of the serious and 2%-4% of non-serious reactions are usually reported (Rawlins, 1995). A relatively high rate of reporting may result from a high frequency of adverse reactions or may simply be due to the publicity or novelty of a new agent. One such ''apparent epidemic'' of flucloxacillin-induced jaundice in Australia (reporting 357 ADRs and 17 deaths) has been considered to be a reporting artefact (Devereaux et al., 1995; Roughead et al., 1999). In addition to the variability in reporting, identification of cases in a non-systematic way introduces significant inaccuracy to the data. In a recent survey in the United Kingdom, about half of the reported adverse hepatic reactions were classified as ''unrelated'' to the drugs under systematic evaluation (Aithal et al, 1999). A further difficulty with spontaneous reporting is that the denominator is usually unknown, although drug sales figures could be used to estimate the frequency of adverse reactions.
Record linkage studies connect information on drug exposure from prescription data with outcome, and have the advantages of prospective design and comprehensive identification of cases.
Established linkages such as the General Practice Research Database in the United Kingdom and Group Health Cooperative of Puget Sound in the United States have contributed valuable epidemiological information regarding drug-induced liver disease (Beard et al., 1986; Derby et al., 1993; Jick et al., 1999). However, most often the outcomes such as deaths, hospital admissions and discharge diagnoses, used in linkage studies are those that pertain only to the more serious reactions or those that occur while the patient is in the hospital. The latter underestimates the frequency of adverse hepatic reactions since acute hospital inpatient stays are usually shorter than the latent period (5 days to 3 months) of most types of drug-induced liver disease.
Case-control studies are particularly useful when the outcome is rare. In the field of drug-induced liver disease, they have been applied to hepatic tumors, industrial hepatotoxicity and the role of aspirin in Reye's syndrome (Farrell, 1994).
Was this article helpful?