Type 2 diabetes mellitus is a progressive and complex disorder that is difficult to treat effectively in the long term. The majority of patients are overweight or obese at diagnosis and will be unable to achieve or sustain near normoglycaemia without oral antidiabetic agents; a sizeable proportion of patients will eventually require insulin therapy to maintain long-term glycaemic control, either as monotherapy or in conjunction with oral antidiabetic therapy. The frequent need for escalating therapy is held to reflect progressive loss of islet beta-cell function, usually in the presence of obesity-related insulin resistance. Today's clinicians are presented with an extensive range of oral antidiabetic drugs for type 2 diabetes. The main classes are heterogeneous in their modes of action, safety profiles and tolerability. These main classes include agents that stimulate insulin secretion (sulphonylureas and rapid-acting secretagogues), reduce hepatic glucose production (biguanides), delay digestion and absorption of intestinal carbohydrate (alpha-glucosidase inhibitors) or improve insulin action (thiazolidinediones). The UKPDS (United Kingdom Prospective Diabetes Study) demonstrated the benefits of intensified glycaemic control on microvascular complications in newly diagnosed patients with type 2 diabetes. However, the picture was less clearcut with regard to macrovascular disease, with neither sulphonylureas nor insulin significantly reducing cardiovascular events. The impact of oral antidiabetic agents on atherosclerosis--beyond expected effects on glycaemic control--is an increasingly important consideration. In the UKPDS, overweight and obese patients randomised to initial monotherapy with metformin experienced significant reductions in myocardial infarction and diabetes-related deaths. Metformin does not promote weight gain and has beneficial effects on several cardiovascular risk factors. Accordingly, metformin is widely regarded as the drug of choice for most patients with type 2 diabetes. Concern about cardiovascular safety of sulphonylureas has largely dissipated with generally reassuring results from clinical trials, including the UKPDS. Encouragingly, the recent Steno-2 Study showed that intensive target-driven, multifactorial approach to management, based around a sulphonylurea, reduced the risk of both micro- and macrovascular complications in high-risk patients. Theoretical advantages of selectively targeting postprandial hyperglycaemia require confirmation in clinical trials of drugs with preferential effects on this facet of hyperglycaemia are currently in progress. The insulin-sensitising thiazolidinedione class of antidiabetic agents has potentially advantageous effects on multiple components of the metabolic syndrome; the results of clinical trials with cardiovascular endpoints are awaited. The selection of initial monotherapy is based on a clinical and biochemical assessment of the patient, safety considerations being paramount. In some circumstances, for example pregnancy or severe hepatic or renal impairment, insulin may be the treatment of choice when nonpharmacological measures prove inadequate. Insulin is also required for metabolic decompensation, that is, incipient or actual diabetic ketoacidosis, or non-ketotic hyperosmolar hyperglycaemia. Certain comorbidities, for example presentation with myocardial infarction during other acute intercurrent illness, may make insulin the best option. Oral antidiabetic agents should be initiated at a low dose and titrated up according to glycaemic response, as judged by measurement of glycosylated haemoglobin (HbA1c) concentration, supplemented in some patients by self monitoring of capillary blood glucose. The average glucose-lowering effect of the major classes of oral antidiabetic agents is broadly similar (averaging a 1-2% reduction in HbA1c), alpha-glucosidase inhibitors being rather less effective. Tailoring the treatment to the individual patient is an important principle. Doses are gradually titrated up according to response. However, the maximal glucose-lowering action for sulphonylureas is usually attained at appreciably lower doses (approximately 50%) than the manufacturers' recommended daily maximum. Combinations of certain agents, for example a secretagogue plus a biguanide or a thiazolidinedione, are logical and widely used, and combination preparations are now available in some countries. While the benefits of metformin added to a sulphonylurea were initially less favourable in the UKPDS, longer-term data have allayed concern. When considering long-term therapy, issues such as tolerability and convenience are important additional considerations. Neither sulphonylureas nor biguanides are able to appreciably alter the rate of progression of hyperglycaemia in patients with type 2 diabetes. Preliminary data suggesting that thiazolidinediones may provide better long-term glycaemic stability are currently being tested in clinical trials; current evidence, while encouraging, is not conclusive. Delayed progression from glucose intolerance to type 2 diabetes in high-risk individuals with glucose intolerance has been demonstrated with troglitazone, metformin and acarbose. However, intensive lifestyle intervention can be more effective than drug therapy, at least in the setting of interventional clinical trials. No antidiabetic drugs are presently licensed for use in prediabetic individuals.