Iatrogenic hypoglycaemia is the limiting factor in the glycaemic management of diabetes with insulin, a sulfonylurea or a glinide (1). Hypoglycaemia is particularly common in patients whose diabetes is treated with insulin (2). Serious, clinically important hypoglycaemia (3) occurs frequently. For example, continuous glucose monitoring detected glucose concentrations less than 3.0 mmol/L (54 mg/dL) were found to occur every three or four days in a study of patients with type 1 diabetes mellitus (T1DM) using multiple daily injections of insulin (4).
That hypoglycaemia can kill has been known since the discovery of insulin in 1921. Collip and colleagues found that decrements in the blood glucose concentrations following injection of the pancreatic insulin extract into rabbits could be fatal and documented that death of the animals could be prevented by administration of glucose (5). Clinical colleagues of Banting and Best had patients with diabetes die from “hypoglycaemic reactions” (5).
There are now numerous reports of deaths associated with hypoglycaemia in patients with diabetes (e.g., 5-13). In addition to clinical hypoglycaemic death (5), the findings included: Increased mortality with severe (requiring the assistance of another person), symptomatic hypoglycaemia in type 2 diabetes mellitus (T2DM) (6), increased mortality with severe hypoglycaemia in patients with T2DM (7-9), increased cardiovascular and arrhythmic mortality with severe hypoglycaemia in insulin-treated patients with T2DM or impaired glucose tolerance (10), increased mortality with severe hypoglycaemia and seizure, coma, or both in type 1 diabetes mellitus (T1DM) (11), increased mortality in intensive care unit (ICU) patients with hypoglycaemia of less than 70 mg/dL (3.9 mmol/L) (12) and increased mortality in ICU patients and in hospital inpatients with hypoglycaemia of less than 70 mg/dL (3.9 mmol/L) (13). Given the fact that it is known that hypoglycaemia can kill experimental animals (5) and that hypoglycaemia was documented by continuous glucose monitoring at the time of death of a patient with T1DM (14), it is reasonable to conclude that these are causal associations.
Where reported (8,9) the risk of hypoglycaemia associated with death was stronger with shorter intervals between the detected episode of hypoglycaemia and death, consistent with a causal connection between hypoglycaemia and death. Obviously, the last detected episode of hypoglycaemia was not the cause of death since, in the absence of a continuous glucose monitoring record (14), the patient had to survive to report it. But, previous hypoglycaemia is a potent risk factor for subsequent, potentially fatal hypoglycaemia (1). The culprit is not the last detected episode of hypoglycaemia but rather a subsequent episode predicted by that last episode. The interval is not critical, although a shorter interval between the last detected episode and death suggests more frequent hypoglycaemia.
The conclusion that these are causal associations is supported by reports of hypoglycaemic mortality rates in series of patients with diabetes (11,15-19). Early reports indicated that 2% to 4% of deaths of patients with T1DM were the result of hypoglycaemia (15-17). However, more recent reports include hypoglycaemic mortality rates of 7% (18), 8% (11), and 10% (19) in childhood-onset (largely T1DM) diabetes. Indeed, the estimate of Skrivarhaug and colleagues (19) that 10% of deaths of Norwegian patients with childhood-onset T1DM were the result of hypoglycaemia may well have been an under estimate since another 15% of the deaths were listed as “sudden death” or “unexpected death,” categories in which the cause of death was ill-defined and might have been the result of hypoglycaemia and cardiac dysrhythmias as the authors suggested. One wonders if the higher hypoglycaemic mortality rates in the more recent reports (11,18,19) might be a clue to overtreatment of diabetes in recent years.
Primary brain death sometimes occurs in patients with diabetes who suffer prolonged, profound iatrogenic hypoglycaemia, but most hypoglycaemia mortality is probably the result of a fatal cardiac arrhythmia with secondary brain death. There is increasing evidence that hypoglycaemia is pro-arrhythmogenic (20-22). Holter monitoring during continuous glucose monitoring detected episodes of hypoglycaemia has documented runs of cardiac arrhythmias ranging from ventricular tachycardia (20) to bradycardia (21) and repolarization abnormalities have been identified in diabetes (22).
In the absence of large, long duration, prospective randomized trials it is not possible to establish causation definitively and severe hypoglycaemia clearly cannot be induced deliberately, in one arm, for both ethical and practical reasons. It has been argued that “confounding” may explain the association between mortality and hypoglycaemia, i.e., that a comorbidity (such as renal or liver disease, weight loss or cognitive impairment) confers both an increased risk of mortality and hypoglycaemia. Zoungas and colleagues (7) have speculated that confounding contributed to the association between mortality and severe hypoglycaemia in the ADVANCE trial. That was based on the association they observed between non-cardiovascular mortality (as well as cardiovascular mortality) and severe hypoglycaemia; they reasoned that death from respiratory, gastrointestinal or skin disorders was unlikely to be caused by hypoglycaemia. However, the conclusion that the associations between mortality and hypoglycaemia are causal is further supported by a systematic review, meta-analysis and bias analysis of studies involving 903,510 participants with T2DM, which concluded that comorbid severe illness alone may not explain these associations since comorbid illnesses would have had to be extremely strongly associated with both cardiovascular disease and severe hypoglycaemia (23).
Given that glycaemic goals in diabetes are a trade-off between glycaemic control and iatrogenic hypoglycaemia, it has been suggested that a reasonable individualized glycaemic goal is the lowest hemoglobin A1C that does not cause severe hypoglycaemia and preserves awareness of hypoglycaemia, preferably with little or no symptomatic or even asymptomatic hypoglycaemia at a given stage in the evolution of the individual’s diabetes (24). Parenthetically, the substantial relationship between a lower A1C level and a higher incidence of severe hypoglycaemia has been consistently documented in randomized controlled clinical trials in both T1DM (25,26) and T2DM (27-29). In these trials when patients with diabetes were randomly assigned to intensive glycaemic therapy and shown to have lower A1C levels or to more conventional glycaemic goals and shown to have higher A1C levels, the incidence of severe hypoglycaemia was 2- to 3-fold higher in each of the groups with the lower A1C levels. The frequency of hypoglycaemia was inversely related to the A1C level in both the original DCCT and the follow-up EDIC phase (25,26), although the slope was less steep in the EDIC phase. The extent to which the latter is the result of insulin analogues, improved insulin delivery, glucose monitoring, patient education, patient or caregiver skill or some other factor is not known.
In an extensive review, with the exception of a 15% reduction of non-fatal myocardial infarction, Rodriguez-Gutierrez and Montori (30) found no significant impact of tight glycaemic control of T2DM on outcomes important to patients—end stage renal disease/dialysis, renal death, blindness, clinical neuropathy, cardiovascular or all-cause mortality, stroke or amputation or peripheral vascular disease. They did find a 2- to 3-fold increase in severe hypoglycaemia during intensive therapy. The authors concluded that the overwhelming consensus in favour of tight glycaemic control to prevent complications needs to recalibrated. That tight glycaemic control did not reduce mortality in T2DM was also reported in an earlier meta-analysis (31). Some of these reservations could also be applied to T1DM. But, there is an association between mortality and substantial elevations in A1C levels in T1DM (32,33). In a 27-year follow-up of DCCT patients a rise in mortality above that of the general U. S. population began only with an A1C level greater than 9% (75 mmol/mol) (32). An analysis of a much larger data set, disclosed a similar finding (33). At 30 years of follow-up previous intensive glycaemic therapy (i.e., during the DCCT) did not reduce major cardiovascular events significantly, although the trend was in that direction (34). Given these data (32-34) one might conclude that the overwhelming consensus for intensive glycaemic therapy for the prevention of macrovascular complications (35), like that for microvascular complications (30), is stronger than the evidence to support it.
Clearly, we must carefully match therapeutic benefit and harm when we select a glycaemic goal in a patient with diabetes. Therapy with insulin is life-saving and prevents symptomatic hyperglycaemia in T1DM and many with advanced T2DM, but these benefits, like the prevention of macrovascular complications (32-34), do not require intensive glycaemic control. If we cannot convincingly document clinically important benefits of intensive glycaemic control we can advocate intensive glycaemic therapy only if the treatment is free from harm. But, many patients with diabetes are at risk for therapeutic harm. For example, those with hypoglycaemia-associated autonomic failure (including both defective glucose counterregulation and impaired awareness of hypoglycaemia), a history of severe hypoglycaemia, a long duration of diabetes, chronic kidney disease or malnutrition are at risk for hypoglycaemia (1). Exclusion of such patients would eliminate many patients with insulin-treated diabetes, perhaps most with T1DM (1,2), from intensive glycaemic therapy. Furthermore, since potential cardiovascular benefits develop over decades (34) one cannot anticipate benefit in patients with chronic vascular complications or other comorbidities with a short life expectancy. In these groups a less stringent glycaemic goal is indicated (36), perhaps an A1C level less than 8.5% (69 mmol/mol) (37) rather than less than 7% (53 mmol/mol) (35). In short, it is difficult to justify attempts to maintain A1C levels less than 7% (53 mmol/mol) in patients at high risk of harm or with no likelihood of benefit. Diabetes could be considered over-treated in such patients. At the very least, the risks and benefits should be explained to patients and their family before embarking on such an approach.
Acknowledgements: This manuscript was prepared without external support or assistance. Dr. Cryer has served as a consultant to Novo Nordisk in recent years. Dr. Heller has undertaken consultancy and worked on advisory boards on behalf of Eli Lilly, Novo Nordisk, Takeda and Boeringher Ingelheim for which his institution has received fees and he has received personal fees from Novo Nordisk, Astra Zeneca, Roche for work on speaker panels.
Philip E. Cryer, MD
Division of Endocrinology, Metabolism and Lipid Research (Campus Box 8127)
Washington University School of Medicine
660 South Euclid Avenue
St. Louis, Missouri 63110 U. S. A.
Simon R. Heller, DM, FRCP
Department of Oncology and Metabolism
University of Sheffield School of Medicine
Beech Hill Road
Sheffield S10 1UK