Hyperglycaemia frequently occurs in the critically ill, in patients with diabetes, as well as those who were previously glucose-tolerant. The terminology ‘stress hyperglycaemia’ reflects the pathogenesis of the latter group, which may comprise up to 40% of critically ill patients. For comparable glucose concentrations during acute illness outcomes in stress hyperglycaemia appear to be worse than those in patients with type 2 diabetes. While several studies have evaluated the optimum glycaemic range in the critically ill, their interpretation in relation to clinical recommendations is somewhat limited, at least in part because patients with stress hyperglycaemia and known diabetes were grouped together, and the optimum glycaemic range was regarded as static, rather than dynamic, phenomenon. In addition to hyperglycaemia, there is increasing evidence that hypoglycaemia and glycaemic variability influence outcomes in the critically ill adversely. These three categories of disordered glucose metabolism can be referred to as dysglycaemia. While stress hyperglycaemia is most frequently managed by administration of short-acting insulin, guided by simple algorithms, this does not treat all dysglycaemic categories; rather the use of insulin increases the risk of hypoglycaemia and may exacerbate variability. The pathogenesis of stress hyperglycaemia is complex, but hyperglucagonaemia, relative insulin deficiency and insulin resistance appear to be important. Accordingly, novel agents that have a pathophysiological rationale and treat hyperglycaemia, but do not cause hypoglycaemia and limit glycaemic variability, are appealing. The potential use of glucagon-like peptide-1 (or its agonists) and dipeptyl-peptidase-4 inhibitors is reviewed.