||Designing a new gas turbine is a challenging task: complex physical mechanisms and multiple disciplines are coupled with a large design space and numerous often conflicting objectives. These attributes have led to the decomposition and fragmentation of the design process: starting from a preliminary engine design that sets the requirements and the limits for each engine component, each module is designed autonomously in several phases, using an ever higher level of detail, with the support of progressively higher fidelity tools. While improving the tractability of the design process, this approach has two important limitations: the decomposition can conceal important trade-offs between components, leading to sub-optimal overall designs, and the use of high fidelity tools is limited to the very last phases of the process, reducing the possibility of introducing decisive design changes. The structure of the design process often leads to conservative design decisions, dictated by previous experience rather than real physical constraints. This study concentrates on reducing the level of decomposition in the design of gas turbine compression systems, seeking to perform the simultaneous preliminary design optimisation of an IP and an HP compressor and of the inter-connecting s-shaped duct. CFD has been used to evaluate the duct performance, overcoming the lack of design and evaluation rules for annular curved ducts that has often led to conservative designs. Response surfaces have been used extensively to limit the increase in design time arising from the integration of codes with different levels of fidelity in a preliminary design environment. The results demonstrate how integrated optimisation can improve compression system design by reducing the development time and by improving overall performance when compared to that achieved through the isolated optimisation of individual components.