A temporal cluster of cases may result from exposure to a new carcinog
en, but it may also occur under stable conditions or as result of anot
her factor, such as changes in medical practice. In this study, we use
d several complementary methods to detect a cluster and to conduct an
initial investigation regarding its cause. The analyses included three
stages, and were based on the time interval between consecutive diagn
oses. First, we applied the sets monitoring technique to detect cluste
ring. Following the detection (alarm), a confirmatory analysis was con
ducted in order to confirm or refute it as a true alarm. We applied th
is analysis to the first five cases diagnosed subsequent to the alarm.
With a graphical display of the temporal pattern of the diagnoses, we
considered the possible causes leading to the cluster. The data inclu
ded registered diagnoses of leukaemia cases made from 1960 to 1990 amo
ng residents of Ashkelon, Israel. Clustering was detected for acute ly
mphatic leukaemia and for chronic myeloid leukaemia. We found that the
temporal patterns of the diagnoses consistent with the possibility th
at these two clusters are due to exposure to a common local carcinogen
. We cannot however, rule out the possibility that these clusters are
due to unknown factors. (C) 1997 by John Wiley & Sons, Ltd.