Although the terms "metabolic memory" and "legacy effect" have been used to describe the prolonged benefits of good blood glucose control, the former is now recognized as a phenomenon related to the prolonged harm produced mainly by hyperglycemia. At least three randomized clinical trials (Diabetes Control and Complications Trial in type 1 diabetes, United Kingdom Prospective Diabetes Study and Steno-2 in type 2 diabetes) have demonstrated that patients treated intensively for a period of time have a lower risk of micro- and macrovascular complications that persists during subsequent follow-up, even after their tight control has relented and the levels of glycated hemoglobin in the conventionally treated group improve. The mechanisms are not fully understood but most probably relate to the physiopathology of vascular complications of diabetes, and in recent years a unifying theory has been emerging to understand them. The excess superoxide anion produced by the mitochondria in response to hyperglycemia leads through disturbances at the nuclear level to the accumulation of potentially harmful substances such as advanced glycated end-products, protein kinase C, and nuclear factor κB, which are directly implicated in the development of vascular complications in diabetes. These adverse effects are not reversed when the high blood glucose is corrected, and some may be permanent because of epigenetic changes. Some antidiabetes drugs and antioxidant substances have produced partial reversibility of the mechanisms involved in the metabolic memory at the experimental level, but probably the best strategy is to optimize the metabolic control as early as possible, even before diabetes is diagnosed.