Two-phase response-dependent designs have been widely used in modern cohort studies to define sampling strategies which yield efficient estimators while respecting budgetary constraints. Large cohort studies are platforms that support a series of biomarker studies addressing distinct but sometimes related scientific questions. We consider the design of two-phase studies which exploit the information available from previous two-phase studies conducted in the same platform. We focus on the setting where the current study examines the relationship between a previously studied biomarker and a new outcome - the work is similar in spirit to the analysis of secondary outcomes in case-control studies. The phase I sample is defined to contain outcomes and auxiliary covariate data, along with the exposure variable in individuals selected in phase II of the earlier two-phase studies. We construct the optimal phase II subsampling scheme based on such phase I samples to achieve maximum efficiency.