Brittle" diabetes is one of the most poorly defined terms of modem medicine: it is emotive, subjective and imprecise. None the less it is an attractively appropriate descriptive term when clearly defined, and could usefully replace the many synonyms, such as "superlabile", "oscillatory" and "unstable". Woodyatt [1] was the first to describe patients with diabetes as "brittle". He defined such patients as "... insulin-dependent diabetics whose control is so fragile that they are subject to frequent and precipitous fluctuations between hyperglycaemia and insulin reactions and in whom causes of instability have been excluded." Used correctly this definition has proved useful although it has often been grossly abused. Rosenbloom [2], when discussing brittle diabetes in children, narrowed the definition to "those children whose sensitivity to insulin is such that a dosage change of 10% will result in either ketonaemia or reactions". Colwell [3] stated that "unstable diabetes is characterised by irregular, unpredictable behaviour of glycaemia and glycosuria with all insulins, especially the depot varieties, even though the food supply and amount of exercise are constant." These definitions, although useful, are rather restrictive and exclude those patients with predominantly hyperglycaemic or hypoglycaemic instability, which often form two distinct clinical subgroupings. In addition, it is now possible to define the cause of instability in many "brittle" patients (even if treatment remains difficult), and this would take many cases outside Woodyatt's definition. For these reasons, a more useful definition is that proposed by Tattersall in 1977 [4]... "insulindependent diabetics whose lives are constantly disrupted by episodes of hypo-or hyperglycaemia, whatever the cause". The concept of life disruption is an important one, which usefully separates problematic from non-problematic unstable diabetic patients.