DILIN Overview | Retrospective | Prospective | Ancillary Studies | References | Links

Retrospective Study

Executive Summary

Background and Rationale: Drug-induced liver injury (DILI) is the single most common reason for regulatory actions concerning drugs, including failure to gain approval for marketing, removal from the market place, and restriction of prescribing indications. DILI is also a significant cause of morbidity and mortality in many patient populations. Establishing a diagnosis of DILI is difficult due to the presence of other potential causes of liver injury, and the injury itself can range from mild and transient to severe and protracted, and even lead to acute liver failure. To stimulate and facilitate research into DILI, the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) has recently established the Drug-Induced Liver Injury Network (DILIN). One of the initial projects to be conducted by the network is to retrospectively establish a nationwide registry of patients who have suffered severe idiosyncratic liver injury associated with drugs (ILIAD), and to collect, immortalize and store serum, DNA, and lymphocytes from these patients (hereafter referred to as the "ILIAD protocol"). This ILIAD protocol will serve as a resource for subsequent mechanistic investigations of the basis for susceptibility to severe idiosyncratic DILI.

To Top

Specific Aims and Objectives: The primary goal of the ILIAD protocol is to create: (a) a clinical database consisting of individuals who have experienced severe DILI caused by four specific drugs, and the relevant clinical data concerning the episode of DILI; and, (b) to create a bank of biological specimens obtained from these individuals and a similar number of appropriate controls. Corresponding information from control subjects will also be collected. These biological specimens will be DNA, plasma, and immortalized lymphocytes. Immortalized lymphocytes will provide unlimited amounts of genomic DNA for study as well as living immune cells for phenotyping studies. A secondary goal of the ILIAD protocol is to maintain a registry of cases in the ILIAD database so that they may be recontacted in the future. It is expected that this will facilitate additional studies exploring the mechanisms of DILI.

To Top

Targeted Drugs: The initial drugs to be targeted in the ILIAD protocol are isoniazid, phenytoin, clavulanic acid / amoxicillin and valproic acid. The target drugs were chosen because they cause severe DILI at a high rate compared with other drugs, making our target enrollment for each drug (n = 50-100) attainable. In addition, these drugs are frequently administered to reasonably healthy patients not concurrently receiving other drugs more likely to be hepatotoxic, facilitating causation assessment.

To Top

Study Population: Cases are individuals who suffered drug-induced liver injury due to one of the targeted drugs. For INH, phenytoin, or clavulanic acid / amoxicillin, severe liver injury is defined as a documented serum total bilirubin > 2.5 mg/dl; for valproic acid, the criteria are compatible symptomatic clinical presentation that is severe enough to prompt hospitalization and evidence of liver dysfunction (INR > 1.5 or ALT > 3 X ULN, and/or characteristic liver biopsy). Subjects must be living and may have undergone liver transplantation, but the DILI episode must have occurred since January 1, 1994. Subjects will be excluded if they are not willing to have medical information and blood samples taken, or if they are unable to adequately give informed consent to participate in the study. Individuals less than 2 years old at the time of study enrollment are excluded due to blood volume requirements.

To Top

Control Subjects: A matched design is employed with each case individually matched with one control based on: (a) age at the time of the DILI event, (b) minimum duration of study drug therapy, and (c) clinical setting and/or indication for therapy with the study drug. The matching criteria for age and indication for therapy will be specific to each drug. For DILI due to isoniazid, both cases and controls will be drawn from persons who are HIV-negative and are taking isoniazid for treatment of latent tuberculosis. For DILI due to valproic acid or to phenytoin, controls will be drawn from patients in neurology clinics in the DILIN Clinical Centers who are taking valproic acid or phenytoin, respectively. For DILI due to clavulanic acid / amoxicillin, controls will be drawn from patients who are taking clavulanic acid / amoxicillin in adult and pediatric primary care clinics affiliated with the DILIN Clinical Center. For each of the four study drugs, selection of controls will only begin once a decision is reached that the minimum of 50 cases can be attained in this study.

To Top

Data Collection Protocol: Several phases of data collection are envisaged for this study. Basic demographic information, family information, and prior history of liver disease and other medical conditions will be gathered using a telephone or personal interview format. Detailed clinical information concerning the DILI episode will be abstracted from patient charts and medication records. This includes concomitant medical conditions and illnesses, laboratory and liver function tests, as well as serological and other assays. Finally, detailed information from the NIDDK Genetics Repository will be forwarded to the DCC and merged with the clinical information. These data will be forwarded to a DILIN Causality committee. It will weight the evidence and determine whether the liver injury was indeed caused by the medication.

To Top

Statistical Considerations: An important goal of this study is to generate hypotheses for consideration in future studies. Type-II error, i.e., failing to find an interesting association when there is one, is an important consideration. Thus, as a compromise between type-I and type-II error, α = 0.01 will be applied to declare any effect statistically significant. Because the suspected DILI mechanism is thought to be different for each drug, statistical analysis will be performed separately for each medication. The conditional logistic regression model will be the primary analytic model. The adjusted odds ratio, together with its confidence interval, will be estimated from the associated regression parameters.

To Top

Power Calculations: With 50 cases and controls, only strong risk factors for DILI can be detected with reasonable power. For example, when the prevalence is low, e.g., 0.2, an OR of approximately 5.5 or higher is required to declare statistical significance with adequate power. With 100 cases and controls, power is predictably better. With a prevalence of 0.2, an OR 3.5 or higher can be detected with adequate power.

To Top

Study Administration: The administrative and funding mechanism used to undertake this project is an NIH "Cooperative Agreement" (U01). The Steering Committee is the main governing body of the project. It is composed of the Principal Investigators of the field centers, the Principal Investigator of the Data Coordinating Center, and the NIDDK Project Scientist. All decisions are determined by majority vote. In addition, a number of subcommittees have been established and report to the main Steering Committee. A policy concerning ancillary studies has been established, and detailed procedures for proposing and conducting these studies have been established.

To Top