PPARγ-modulators, a class of anti-diabetic drugs as represented by thiazolidinediones (TZD), have been associated with cardiovascular risks in type-2 diabetes in humans but a similar liability has not been demonstrated in preclinical models. This gap between clinical and preclinical observations may reflect the lack of a translational model for cardiac safety assessment because preclinical efficacy for glycemic control for PPARγ-modulators is routinely conducted in animals with diabetic background while drug safety study is performed in young and health animals with little risk of heart failure, in contrast to the complex pathophysiological conditions of patients subjected to the treatment of TZDs. Therefore, some key steps are important to address this translational gap. First, it is essential to use an appropriate translational model that mimics most of human pathophysiology for the assessment of cardiovascular safety for TZDs. Second, it calls for the discovery of a translational biomarker (most likely a collection of biomarkers due to multiple risk factors contributed to the complex disease) to be able to sensitively detect the disease progression and in response to therapy. Specific examples are provided in this review for the use of a rodent model of myocardial infarction-induced heart failure to address the cardiac safety concern in response to chronic treatment of rosiglitazone. Multiple biomarkers, including physiological, biochemical, pharmacogenomic and imaging biomarkers, were applied to assess the cardiovascular risk in this heart failure model. The data and strategic approach are discussed from translational medicine perspectives.