to more severe or chronic forms of the disease 5,6 and avoid The objective of this study is to present and validate a the occurrence of new episodes, including some cases of clinical scale for the diagnosis of drug-induced liver injury fulminant hepatic failure. 7 (DILI). Five components were selected to be included in the Because there are no specific markers or tests for DILI, the scale: temporal relationship between drug intake and the ondiagnosis is usually based on circumstantial evidence. It set of clinical picture, exclusion of alternative causes, extraherelies on a great deal of speculation by the clinician, the patic manifestations, rechallenge or accidental re-exposure, collection of a detailed pharmacological history, the estaband previous report in medical literature. The relative imporlishment of a consistent relationship between drug intake tance of each component was weighed, and arbitrary scores and the onset of the clinical picture, and the exclusion of were attributed. The probability of the diagnosis of DILI was alternative causes. Clinical improvement following drug expressed as a final score, which could vary from 06 to 20.
withdrawal is another element that may indicate a drug-Content validity, criterion validity, construct validity, and inrelated cause, although there is an increased recognition of ter-rater reliability were studied. To analyze validity and reliaprolonged cholestatic hepatitis, sometimes lasting for more bility, a random sample of 50 cases of suspected DILI was than 6 months, after drug withdrawal. 9 Rechallenge is fredrawn from a series of 120 cases reported to our unit. The quently considered to be the most reliable test in the diagnoclassification of the 50 cases by three experts in DILI was sis of suspected cases of DILI, 10 but it is clearly dangerous used as the external standard in the study of criterion validity.
and is best avoidable. 7 Agreement between the scale and the standard, and agreement
This complex process of diagnosis usually requires an exbetween two independent raters (inter-rater reliability) was perienced clinician who is deeply aware of the critical compoanalyzed by weighted k coefficient. There was agreement benents to be weighed for an accurate diagnosis, including a tween the scale and the standard in 42 cases (84%) with a good knowledge of the published literature. The translation weighted k coefficient of 0.90. A good discriminatory capacity of such experience and subjective clinical judgment into a of the scale was found when construct validity was studied.
quantitative measurement to define, by more objective crite-Agreement between raters was observed in 86% of the cases, ria, the probability of an adverse event being related to a corresponding to the weighted k of 0.93. In conclusion, the drug, is of major importance. The complexity of the clinical clinical scale was shown to have a high-level of validity and diagnosis has led to attempts to improve in vitro diagnostic inter-rater reliability as well as a good discriminatory capacity tests to detect toxic drug metabolites or drug hypersensibetween different levels of probability. These data suggest tivity. In spite of significant improvements in the underthat the scale is suitable for use in clinical practice and may standing of the metabolic and immunological basis of contribute to overcome the difficulties in the process of causal-DILI, 18 the diagnosis still is dependent predominantly ity assessment in DILI. (HEPATOLOGY 1997;26:664-669.) on clinical criteria. For several years, we have used a clinical scale similar to that proposed by Be ยดnichou and Danan 19, to Drugs are estimated to account for approximately 5% of estimate the probability that a particular drug is involved in cases of jaundice 1,2 and 10% of cases of acute hepatitis admitcausing liver injury. ted to hospitals. 3 Drug-induced liver injury (DILI) is proba-
The aim of this study was to describe that clinical scale bly underdiagnosed because of the low level of awareness and to present data to validate the scale, including data on of this condition and because the differential diagnosis is content validity, criterion validity, construct validity, and frequently troublesome. The precise and early diagnosis of inter-rater reliability. DILI, including the identification of the offending drug in cases in which more than one drug is under suspicion, is of
PATIENTS AND METHODS
clear importance. This early detection may prevent evolution Cases of Drug-Induced Liver Injury
Fifty cases of DILI were randomly selected from a series of 120 suspected cases reported to our unit for clinical opinion and immu-Abbreviation: DILI, drug-induced liver injury. nological study. The clinicopathologic pattern of injury was classi-From the Faculty of Medicine of Lisbon, Department of Medicine 2, Clinical Immufied in three types, namely hepatocellular, cholestatic, and mixed, nology, Lisbon, Portugal. using criteria adopted from Danan et al. 21 The 50 cases were pre-