One of the longest‐running controversies in medicine concerns the aims of diabetes treatment. The question debated for 80 years has been whether the clinician should just relieve symptoms, or try to achieve the much more difficult objective of near‐physiological normality as measured by an absence of glycosuria and/or normal blood sugar levels.

At the beginning of World War One, most clinicians and physiologists thought the severity of diabetes was inversely proportional to the number of functioning islets of Langerhans. Hyperglycaemia, it was hypothesized, stressed the surviving islets and led to a downward spiral of increasing glandular fatigue and hyperglycaemia. The aim of undernutrition was to rest the damaged tissue in the hope of promoting a return of functional efficiency and possibly regeneration. Most experts stressed that rest of the islets could only be achieved by abolishing glycosuria and restoring normal blood sugar levels.

The first clinical use of insulin in 1922 led to astonishing improvements in the health and strength of patients with diabetes and the concept of pancreatic rest seemed to be confirmed when some regained such carbohydrate tolerance that after weeks or months they could reduce the dose of insulin without developing glycosuria. Initially there were expectations that insulin would allow the islets of Langerhans to recover completely, so that diabetes was cured. Most physicians insisted that the best chance of preserving what pancreatic function remained was biochemical normality. It was also contended that patients who had normal blood sugar levels were more healthy than those without and had fewer ‘complications’. The complications in question were mainly infective, since specific diabetic tissue damage was not recognized until the late 1930s. The toll of microvascular complications (retinopathy and nephropathy) in those whose lives had been saved by insulin did not become apparent until the late 1930s and early 1940s, when it generated an often acrimonious debate about whether they were due to the metabolic disorder or an associated phenomenon.

Liberalization of diet in patients taking insulin began in 1926 and by 1930 it was clear that patients who were prescribed 200 g of carbohydrate per day felt better and more energetic than those on the old regimens of 50 g or less per day. Even these more liberal diets were measured but, in the early 1930s some paediatricians, feeling that a strict measured diet was psychologically damaging, experimented with ‘free’ or unmeasured diets. Most still tried to achieve the best possible metabolic control whereas the New York physician, Edward Tolstoi, allowed his patients to have any degree of hyperglycaemia provided they felt well and did not lose weight. He claimed that in units which aimed for ‘chemical control’ microvascular complications were as, or even more, frequent than in patients on a free diet. The debate remained unresolved because, as one expert put it, ‘one deals perforce with individuals whose diabetes has been under varying degrees of inadequate control and speaks of the bset of them as having good control’. Indeed, until the introduction of haemoglobin A1 in 1976 there was no way of assessing long‐term glycaemic control except by relatively subjective criteria and repeated random blood glucose measurements.

In 1960 an attempt was made, in patients with newly diagnosed maturity onset diabetes, to find out whether treatment with tolbutamide or phenformin was better or worse than insulin. This study of over 1000 patients in 12 university clinics (the University Group Diabetes Program or UGDP) produced its first report in 1970 and led to what …