| Summary: | The aim of this study was to assess by a stepwise multivariate discriminant analysis the value of four current serum tumour markers - carcinoembryonic antigen (CEA), carbohydrate antigen (CA) 50 and CA 242 and tissue polypeptide antigen (TPA) - and a new serum tumour marker, tissue polypeptide specific antigen (TPS), in the diagnosis of pancreatic cancer. The serum values were measured in a prospective series of patients with jaundice, with unjaundiced cholestasis and with a suspicion of chronic pancreatitis or a pancreatic tumour (n = 193). There were 24 patients with a cancer of the pancreas and two patients with a cancer of the papilla of Vater in this series. Our results showed that CA 50 (P < 0.001) and TPA (P < 0.01) were the best marker tests in predicting pancreatic malignancy. Also, the TPS (P = 0.07) and CA 242 (P = 0.08) tests showed marginally significant independent discriminating power, while the CEA test did not (P = 0.12). In order to sum up the contributions of different markers, a diagnostic score (DSI) was developed. The discrimination function was: DS1 = CA 50 x 1.75 + TPA x 0.62 + TPS x (-0.37) + CA 242 x (-1.21). The sensitivity of DS1 in detecting pancreatic cancer was 36% with a specificity of 90% and an efficiency of 82%. When the combination of CA 50 and TPA was used as a test, the discrimination function (DS2) was: DS2 = CA 50 x 0.69 + TPA x 0.67. The sensitivity of DS2 was 44% with a 88% specificity and an efficiency of 82%. According to this analysis, the further advantage gained by a computer-aided scoring system seems to be limited, since despite the considerably high specificity and efficiency its sensitivity remained low. In the present analysis the best combination in diagnosing pancreatic cancer was the combination of CA 50 and TPA.
|
|---|