Querying imperfect data received increasing attention in the area of database management. The complexity and the volume of imperfect data requires advanced techniques for efficient access, and satisfying user-queries in a reasonable response time. In this paper, we are particularly interested in evidential databases, i.e., databases where imperfection is represented through Dempster-Shafer theory. To answer userqueries in such databases, we propose a new index system based on a tree data structure (e-Tree) that is adapted to the complexity of the evidential data. The experiments done on our solution showed encouraging results.