To evaluate the prognostic significance of a bone scan index (BSI) based on the weighted proportion of tumor involvement in individual bones, in relation to other factors and to survival in patients with androgen-independent prostate cancer. Baseline radionuclide bone scans were reviewed in 191 assessable patients with androgen-independent disease who were enrolled onto an open, randomized trial of liarozole versus prednisone. The extent of skeletal involvement was assessed by scoring each scan using the BSI and independently according to the number of metastatic lesions. The relationship of the scored bone involvement to other known prognostic factors was explored in single- and multiple-variable analyses. In single-variable analyses, the pretreatment factors found to be associated with survival were age (P = .0446), performance status (P = .0005), baseline prostate-specific antigen (P = .0001), hemoglobin (P = .0001), alkaline phosphatase (P = .0002), AST (P = .0021), lactate dehydrogenase (P = .0001), and treatment (P = .0098). The extent of osseous disease was significant using both the BSI (P = .0001) and the number of lesions present (P = .0001). In multiple-variable proportional hazards analyses, only BSI, age, hemoglobin, lactate dehydrogenase, and treatment arm were associated with survival. When the patient population was divided into three equal groups, with BSI values of < 1.4%, 1.4% to 5.1%, and > 5.1%, median survivals of 18.3, 15.5, and 8.1 months, respectively, were observed (P = .0079). The BSI quantifies the extent of skeletal involvement by tumor. It allows the identification of patients with distinct prognoses for stratification in clinical trials. Further study is needed to assess the utility of serial BSI determinations in monitoring treatment effects. The BSI may be particularly useful in the evaluation of agents for which prostate-specific antigen changes do not reflect clinical outcomes accurately.