The presence of circulating tumor cells might be an indicator of hematogenous spread of tumor cells leading to extrahepatic metastasis. Messenger RNA (mRNA) expression of human albumin, as a liver specific cell marker, has been proposed for this purpose in hepatocellular carcinoma. We conducted a multicenter prospective study in 101 patients with biopsy-proven hepatocellular carcinoma followed-up every 3 months for 1 year or until death. At entry into the study, albumin mRNA was detected in the blood by reverse transcription-polymerase chain reaction (RT-PCR). At entry into the study, 45% of the patients had a positive albumin mRNA test, 53% a single tumor, 16% a portal or venous hepatic thrombosis, and 16% had proven metastasis. After 1 year, there was no significant difference in survival of patients with positive or negative albumin mRNA at entry (P = .16, log-rank test). When patients with metastasis at entry were excluded, again survival did not differ between the two groups (P = .20). Independent prognostic factors of survival were radical therapeutic procedures, metastasis, number of tumors, Child-Pugh score, and thrombosis, but not the albumin mRNA test. Taking the presence of metastasis as a reference, the specificity of the test was 56%, its sensitivity 50%, and its negative predictive value 85%. The present study shows that circulating albumin mRNA detected by means of RT-PCR fails to provide significant information in the diagnosis and prognosis of hepatocellular carcinoma. Further studies are needed to determine whether the use of specific tumor markers could have clinical relevance in this setting.