Nasal melanoma.

are arguments for using both. In prevention, the absolute risk is a better measure as it gives a direct measure of the number of cases that will be prevented. The relative risk may be a better measure in clinical practice as it can easily be transformed to the aetiological fraction (EF = (RR-I)/ RR) and thus gives a measure of the chance that the cancer of a certain patient is caused by asbestos. From a more theoretical epidemiological standpoint the best measure depends on whether the risk due to asbestos is multiplicative or additive compared with the background risk. For a multiplicative risk the incidence rate (IR) is IR = IRo* f(exposure), where f is a function not dependent on the background incidence (IR0). If the risk is additive an absolute risk is more appropriate as IR = IRO + IRasb, where IRsb is the incidence rate caused by exposure. A relative risk may at a first glance be preferred as the risk for lung cancer caused by asbestos is usually expressed as: SMR = 1 + a* dose where a is a constant. This relation does not seem to fit the data of Raffin et al, however, and the only measure of "dose" in their paper is employment time. The relative risk is certainly not related in linear fashion to employment time if all lung cancers are considered (1 9, 1-4, and 19 for <1, 1-4, and > 5 years respectively). The group with 1-4 years employment time is small and there are large confidence intervals for the risks especially when stratified according to histological type. Thus there seems to be little justification to restrict the analyses to a multiplicative model. There is an increased risk in the group with <1 year employment time. This raises the question about comparability between the exposed and reference groups. A dose-response model may also consider time since last exposure as some data indicate that the risk of lung cancer decreases some years after the exposure of asbestos has ceased.' A different risk according to time from onset of exposure may depend on the different growing rates of the tumours. Anaplastic carcinoma grows faster than squamous cell carcinoma, which grows faster than adenocarcinoma.2 The importance of the finding of a higher RR for adenocarcinoma in persons with a long time since onset of exposure compared with persons with other histological types of tumour does not necessarily mean that only adenocarcinoma are caused by exposure to asbestos. My conclusion is that the