Serum Ca 19.9 levels The sensitivity, specificity, positive and negative predictive figure 2 values of abnormal Ca 19.9 levels for the diagnosis of pancreatic cancer were 91, 87, 91 and 87%, respectively (Table 1). Serum Ca 19.9 levels were normal or non interpretable due to cholestasis in 26 patients. Among them, 14 (54%) patients had serum KRAS2 mutations. Combination of both tests increased sensitivity to 98% with a negative predictive value of 96% for the diagnosis of pancreatic cancer (Table 1). Table 1 Accuracy of serum KRAS2 mutation detection, serum Ca19.9 levels, and both for the diagnosis of pancreatic cancer DISCUSSION In the present study, only the most frequent KRAS2 gene mutation G12D (aspartic acid) observed in pancreatic cancer (Iguchi et al, 1996; Tada et al, 1998; Castells et al, 1999; Watanabe et al, 1999) was analysed in the serum of patients and controls, in order to limit the cost of such test and to validate it in clinical practice.

Our study underlines the high feasibility of KRAS2 mutations analysis, as circulating DNA was obtained in sufficient quantities in all patients. Mean serum DNA concentration were 730ngml?1, which is comparable to that observed in the series of Mulcahy et al (1998). In the present study, detection of KRAS2 mutations in circulating DNA had a low sensitivity but a high specificity for the diagnosis of pancreatic cancer. The sensitivity (47%) was in agreement with previous studies (27 to 81%) (Mulcahy et al, 1998; Yamada et al, 1998; Castells et al, 1999; Porta et al, 1999; Theodor et al, 2000; Zambon et al, 2000), although the search for five other possible KRAS2 mutations in codon 12 was not performed.

Higher KRAS2 mutation prevalences have been reported in pancreatic or duodenal juice (63 to 87%), due probably to higher DNA tumour content in pancreatic juice as compared to plasma (Wilentz et al, 1998; van Laethem et al, 1998; Watanabe et al, 1999). Use of samples of pancreatic juice however requires invasive procedures (fine-needle aspiration during endoscopic ultrasonography Drug_discovery or endoscopic retrograde pancreatography). Three studies have reported a higher specificity of serum KRAS2 mutations compared to the current study (100 vs 87%), but their control groups included few patients and essentially healthy subjects (Mulcahy et al, 1998; Porta et al, 1999; Theodor et al, 2000). Since KRAS2 mutations have been reported in pancreatic tissue or juice from 6�C42% of patients with chronic pancreatitis (Furuya et al, 1997; Mulligan et al, 1999; L��ttges et al, 2000; Ha et al, 2001), and knowing that a part of this mutated DNA can be released into circulation (Yamada et al, 1998), the control group should include patients with chronic pancreatitis.