Liver biopsy is an invasive diagnostic method for detecting liver involvement (LI) in Hodgkin's disease (HD). The aim of this retrospective study was to determine and evaluate a method for restricting liver biopsy to a subset of patients. Between 1988 and 1994, a total of 2,016 patients with HD were treated within the HD4-6 study protocol of the German Hodgkin's Lymphoma Study Group (GHSG). We investigated the predictive power of abdominal ultrasound (US) and computed tomography (CT), as well as of various clinical factors related to LI, using univariate and multivariate methods. LI occurred in 4.9% of all patients (99/2,016) and in 3.0% of those who, if LI were disregarded, would have been included in clinical stages I and II. In multivariate analysis the presence of LI was significantly associated with splenic involvement or infradiaphragmatic involvement, absence of mediastinal involvement, serum alkaline phosphatase (SAP) level over 230 units/l, and age over 40 years. We used these factors to define a risk score for LI. LI is very rare in patients who would otherwise be in clinical stages I or II, but knowledge of LI is important because it has therapeutic consequences. With our risk score, liver biopsy is indicated for approximately one quarter of these patients otherwise in clinical stages I or II.