The optimization of similarity queries using metric access methods has been widely discussed in the last decades. Similarity queries consider one object as the query center, and retrieve objects that are either far up to a radius or the nearest ones. Another important retrieval operation, less studied so far, is the Aggregate Similarity Query, which retrieves objects with the smallest aggregate distances to a set of query centers. This paper presents the definition of Circumscription-Constrained Aggregate Similarity Queries, discusses performance issues involved, and proposes an optimization based on the triangle inequality to reduce the number of distance calculations and disk accesses needed to answer the queries.