Primary biliary cirrhosis (PBC) is a chronic cholestatic liver disease with a variable clinical course. Identification of serologic markers associated with increased risk of liver failure would assist in management of PBC patients. The objective of this study was to identify antinuclear antibody (ANA) markers that may be used to predict PBC outcome. Indirect immunofluorescence was used to identify ANAs in 492 PBC patients. chi2 and Kaplan-Meier analyses were used to examine the association between ANAs and liver failure. A greater percentage of ANA-positive, compared to ANA-negative, PBC patients developed liver failure (41% vs 25%, P = .005). The presence of anti-centromere antibodies was associated with liver failure (anti-centromere antibody positive vs negative, 58% vs 33%, P = .001). The time to liver failure was shorter in ANA-positive, compared with ANA-negative, patients (log rank score 5.8, P = .02). After 8.9 years (the median follow-up for patients without liver failure), 68% of ANA-positive and 81% of ANA-negative patients were free of liver failure. Anti-centromere antibodies were also associated with a shorter time to liver failure (log rank score 8.4, P = .004). After 8.9 years, 52% of anti-centromere antibody positive and 74% of anti-centromere antibody negative patients were without liver failure. ANAs in general, and anti-centromere antibodies in particular, are associated with liver failure in PBC. PBC patients with ANAs may be candidates for treatment with experimental therapies to prolong the interval between diagnosis and liver failure. ANA-negative patients, who appear to have a relatively benign clinical course, should perhaps be treated with ursodeoxycholic acid alone.