This review critically appraises studies examining the association of novel factors with diabetes. We show that many of the most studied novel and apparently 'independent' risk factors are correlated with each other by virtue of their common origins or pathways, and that residual confounding is likely. Available studies also have other limitations, including differences in methodology or inadequate statistical analyses. Furthermore, although most relevant work in this area has focused on improving our understanding of the pathogenesis of diabetes, association studies in isolation cannot prove causality; intervention studies with specific agents (if available) are required, and genetic studies may help. With respect to the potential value of novel risk factors for diabetes risk prediction, we illustrate why this work is very much in its infancy and currently not guaranteed to reach clinical utility. Indeed, the existence of several more easily measured powerful predictors of diabetes, suggests that the additional value of novel markers may be limited. Nevertheless, several suggestions to improve relevant research are given. Finally, we show that several risk factors for diabetes are only weakly associated with the risk of incident vascular events, an observation that highlights the limitations of attempting to devise unified criteria (e.g. metabolic syndrome) to identify individuals at risk of both CHD and diabetes.