8%). While serum CEA may be helpful in diagnosis of some ICC and in follow-up of patients who showed elevated titres prior to surgery, it is not sufficiently sensitive promotion info or specific for reliable diagnosis of the disease (Kawarada and Muzumoto, 1984; Wang et al, 1994). CA 19-9 currently is in wide use, particularly for detecting bile duct cancer in patients with primary sclerosing cholangitis (PSC) (Nichols et al, 1993; Ramage et al, 1995; Bergquist et al, 1998; Chalasani et al, 2000). Serum CA 19-9 concentrations also are elevated in 65�C80% of patients with ICC, and CA 19-9 has been considered the most sensitive serologic marker for diagnosis and follow-up of ICC (Kawarada and Mizumoto, 1990; Yamanaka et al, 1995; Kim et al, 1999; Uenishi et al, 2001).

However, when we used a cutoff value giving a specificity of 95% vs benign liver disease, CYFRA 21-1 had the highest diagnostic sensitivity for ICC (87.0%). Importantly, sensitivity of serum CYFRA 21-1 for HCC was low in this study (17.1%). Although the sensitivity of CA 19-9 was relatively high in our patients with ICC (60.9%), its specificity for ICC was limited; serum CA 19-9 was elevated in 33.5% of HCC patients. Serum CYFRA 21-1 therefore may be a particularly useful marker for distinguishing ICC from HCC in a radiologically demonstrated hepatic tumour. However, our study could not clarify whether screening for serum CYFRA 21-1 detects very early ICC, since only two patients had a stage I tumour. Apparent sensitivity of a tumour marker in a study is influenced by several factors, particularly numbers of patients in early or advanced stages of cancer.

Some investigators have indicated that serum CYFRA 21-1 may be useful for early detection of non-small-cell lung cancer, where CYFRA 21-1 elevations were common even in patients with early stage disease (Sugama et al, 1994; Takada et al, 1995). Prospective studies are needed to assess the place of CYFRA 21-1 in screening for ICC. When two or more tests are available for use in diagnosis, Cilengitide comparison of ROC curves often will indicate which is best. The diagnostic test with the ROC curve enclosing the largest area is most accurate. For example, a high degree of specificity and sensitivity of CYFRA 21-1 for diagnosis of non-small cell lung cancer was shown by ROC curve findings (Pujol et al, 1993; Stieber et al, 1993; Sugama et al, 1994; Takada et al, 1995; Nisman et al, 1998). Analysis of the AUCs clearly showed better discrimination between ICC and benign liver diseases for CYFRA 21-1 than for CEA or CA 19-9. These results indicated that serum CYFRA 21-1 should be a useful and reliable tumour marker for ICC.