BACKGROUND: For over 20 years the association between sarcoidosis and malignancy, particularly lymphoma and lung cancer, has been disputed with misclassification being the major concern. The aim of the present study was to analyse the incidence of malignancies in a cohort of patients with sarcoidosis by linkage to a nationwide population based cancer register. METHODS: The cohort comprised 254 patients followed for a median of 25 years until death, emigration, or 31 December 1992, whichever came first. The expected number of cancer cases was calculated using the annual age and sex specific cancer rates from the Danish Cancer Registry. RESULTS: Thirty six cancers were registered, three of which were misclassified as sarcoidosis, leaving 33 cancers compared with 23 expected (standardised incidence ratio (SIR) = 1.4; 95% CI 0.99 to 2.0). Five lung cancers were observed compared with 2.5 expected, yielding an SIR of 2.0 (95% CI 0.7 to 4.7). There was no incidence of lymphoma and only one case of leukaemia. There was a significant excess number of pharyngeal cancers based on two cases (SIR = 15.4; 95% CI 1.7 to 56). CONCLUSIONS: This study does not support the theory of an association between sarcoidosis and malignancy, and the main reason other studies have shown such an association is most likely to have been due to selection bias and misclassification.