It is logical to begin type 2 insulin therapy with an injection of an intermediate-acting or a long-acting insulin at bedtime, but one should treat to target, i.e. aim at fasting glycaemias lower than 1.20 g/l to obtain an HbA(1c) close to 7%. Nevertheless, basal insulin therapy does not prevent progression to insulin-secretory deficiency. If necessary, recourse should be made to multiple-injection protocols, taking into account postprandial hyperglycaemia. For every level of HbA(1c), the suppression of postprandial hyperglycaemia, 1 point of HbA(1c) can be gained in theory, whereas reducing the fasting glycaemia to values of less than 1.10 g/l reduces HbA(1c) to close to 7%, whatever the initial level of HbA(1c). However, when a diabetic is clearly not controlled, the preprandial acting use of rapid analogues allows the fasting glycaemia to be improved significantly. Inversely, an early treatment with basal insulin, by correcting glucotoxicity, can also decrease postprandial hyperglycaemia. Many industry-sponsored studies comparing insulin therapy regimens show annoying biased interpretations of results. It does not seem pertinent to compare a single injection with two or even three injections, nor to compare an efficient titration with an inefficient titration or to eliminate oral drugs, in particular sulphonylureas combined with a basal insulin. If premix insulins can give satisfactory results in patients who maintain a sufficient residual insulin-secretion, we think it would be preferable to adopt the basal-prandial regimen and a step-by-step escalating therapy. The first stage consists in combining oral therapy with an injection of NPH insulin or a long-acting analogue at bedtime, aiming at a fasting glycaemia of less than 1.20 g/l. In the next stages, a single injection of rapid-acting insulin analogue is added each time. The main advantage of this regimen is to fix a target adapted to each injection and, as a result, to facilitate forced titration of the doses.