3 Serum diagnostic tests for pancreatic cancer JONATHAN M. RHODES CHI KONG CHING DEFINING AN APPROPRIATE ROLE Despite the availability of several serological tests for pancreatic cancer their adoption into clinical practice has been slow. These tests have been criticized because they fail to detect pancreatic cancer at a treatable stage and tend to be less specific in the presence of jaundice. Before assessing the individual tests we should decide what role we can reasonably expect them to perform. EARLY DIAGNOSIS--AN UNATTAINABLE ROLE The poor overall results of current treatment for pancreatic cancer (Carter, 1990) contrast with a 5-year survival of 37% following resection of tumours with a diameter under 2 cm (Tsuchiya et al, 1986). This suggests that better results might be achieved if patients could be identified earlier. On the other hand, the size of the tumour does not always correlate with biological behaviour. Thus only 44% of patients with tumours of diameter <2 cm had stage 1 disease (Tsuchiya et al, 1986) and long-term survival can occasionally occur without resection (Gudjonsson, 1987). Furthermore if severe pain is present the tumour is likely to be inoperable (Williamson, 1988) and if the patient is jaundiced serological tests are probably going to be unnecessary and inappropriate. Thus, the test required is one that can diagnose patients with small tumours who are not jaundiced and who either have vague discomfort or are asymptomatic. The requirements for such a screening test can be calculated. The inci- dence of pancreatic cancer in Western countries is around 10 per 105 of the population (Conrath, 1986). The prognosis of pancreatic cancer is so poor that the prevalence can be regarded as approximately the same as the incidence. If screening is restricted to the population aged over 40 years this prevalence could probably be doubled to 20 per 105. If such a population of 105 were screened with a test that had a sensitivity of 100% and a specificity of 99%, there would be 1020 positive tests of which only 20 would be true positives and the remaining 1000 would be false positives. All these patients would then have to be subjected to some other test with an even higher Baillibre's Clinical Gastroenterology-- 833 Vol. 4, No. 4, December 1990 Copyright (~ 1990, by Bailli~re Tindall ISBN 0-7020-1470-2 All rights of reproduction in any form reserved