The myelodysplastic syndromes (MDS) are a group of malignancies with poor prognosis and obscure etiology. To better understand the distribution of MDS in the population and help generate etiologic hypotheses, we assessed potential clustering in the incidence of MDS in the state of Connecticut using population-based cancer registry data that recently became available. A significant spatial clustering was identified. The most likely area with a high incidence of MDS included 46 census tracts near the west border of Connecticut, and the ratio of observed/expected cases was 2.84. The P value associated with this cluster was 0.0001. Although no temporal clustering was indicated, a space-time analysis identified a cluster in the central south of Connecticut from March 2002 through August 2003 (P=0.008). This is the first analysis of potential clustering in the incidence of MDS using population-based data. If the intriguing finding on spatial clustering is supported by future studies with larger sample sizes and/or in other geographic areas, it would be extremely interesting to explore the "causes" of clustering, which may help shed light on the etiology of MDS.