Recently, we have proposed a model-driven approach for enforcing fine-grained access control (FGAC) policies when executing SQL queries. More concretely, we have defined a function SecQuery() that, given an FGAC policy S and a SQL select-statement q, generates a SQL stored-procedure SecQuery(S, q), such that: if a user u with role r is authorised, according to S, to execute q based on the current state of the database, then calling SecQuery(S, q)(u, r) returns the same result as when u executes q; otherwise, if the user u is not authorised, according to S, to execute q based on the current state of the database, then calling SecQuery(S, q)(u, r) signals an error. Not surprisingly, executing the query q takes less time than calling the corresponding stored-procedure SecQuery(S, q). Here we propose a model-based methodology for optimising the stored-procedures generated by the function SecQuery(). The idea is to eliminate authorisation checks in the body of the stored-procedures generated by SecQuery(), when they can be proved to be unnecessary. Based on our previous mapping from the Object Constraint Language (OCL) to many-sorted first-order logic, we can attempt to prove that authorisation checks are unnecessary by using SMT solvers. We include a case study to illustrate and show the applicability of our methodology.