In this paper, we review the consecutive methodological steps to follow when assessing clusters of adverse pregnancy outcomes (APO) in the workplace and the decisions to be taken at each step. The example of clusters of APO reported in VDT users in 1979 to 1982 is taken to illustrate each point. It appears that the number of "expected-unexpected" clusters of APO in VDT users has been largely overestimated as compared to observations and that this might be due to inadequate choices in the models used.