Type 2 diabetes has reached epidemic proportions worldwide. The resulting increase in chronic and costly diabetes related complications has potentially catastrophic implications for healthcare systems, and economies and societies as a whole. One of the key pathological factors leading to type 2 diabetes is insulin resistance (IR), which is the reduced or impaired ability of the body to make use of available insulin to maintain normal blood glucose levels. Diagnosis of developing IR is possible up to 10 years before the diagnosis of type 2 diabetes, providing an invaluable opportunity to intervene and prevent or delay the onset of the disease. However, an accurate, yet simple, test to provide a widespread clinically feasible early diagnosis of IR is not yet available. Current clinically practicable tests cannot yield more than a crude surrogate metric that allows only a threshold-based assessment of an underlying disorder, and thus delay its diagnosis. This thesis develops, analyses and pilots a model-based insulin sensitivity test that is simple, short, physiological and cost efficient. It is thus useful in a practical clinical setting for wider clinical screening. The method incorporates physiological knowledge and modelling of glucose, insulin and C-peptide kinetics and their pharmaco-dynamics. The clinical protocol is designed to produce data from a dynamic perturbation of the metabolic system that enables a unique physiologically valid assessment of metabolic status. A combination of a-priori information and a convex integral-based identification method guarantee a unique, robust and automated identification of model parameters. In addition to a high resolution insulin sensitivity metric, the test also yields a clinically valuable and accurate assessment of pancreatic function, which is also a good indicator of the progression of the metabolic defect. The combination of these two diagnostic metrics allow a clinical assessment of a more complete picture of the overall metabolic dysfunction. This outcome can assist the clinician in providing an earlier and much improved diagnosis of insulin resistance and metabolic status and thus more optimised treatment options. Test protocol accuracy is first evaluated in Monte Carlo simulations and subsequently in a clinical pilot study. Both validations yield comparable results in repeatability and robustness. Repeatability and resolution of the test metrics are very high, particularly when compared to current clinical standard surrogate fasting or oral glucose tolerance assessments. Additionally, the model based insulin sensitivity metric is shown to be highly correlated to the highly complex, research focused gold standard euglycaemic clamp test. Various reduced sample and shortened protocols are also proposed to enable effective application of the test in a wider range of clinical and laboratory settings. Overall, test time can be as short as 30 minutes with no compromise in diagnostic performance. A suite of tests is thus created and made available to match varying clinical and research requirements in terms of accuracy, intensity and cost. Comparison between metrics obtained from all protocols is possible, as they measure the same underlying effects with identical model-based assumptions. Finally, the proposed insulin sensitivity test in all its forms is well suited for clinical use. The diagnostic value of the test can assist clinical diagnosis, improve treatment, and provide for higher resolution and earlier diagnosis than currently existing clinical and research standards. High risk populations can therefore be diagnosed much earlier and the onset of complications delayed. The net result will thus improve overall healthcare, reduce costs and save lives.