introduces a caveat for the use of SD with traditional statistics. Kovatchev and Clarke17 report that the autocorrelation coefficients become insignificant at a time lag of approximately 1 h, such that CGM readings more than 1 h apart could be considered linearly independent.
Outcome studies about variability in the hospital generally confirm an association of variability, identified by at least some measures, with adverse outcomes. Although variability is associated with hyperglycemia and hypoglycemia, in multivariate analyses, variability has remained an independent predictor of adverse hospital outcomes.6,8 The outcomes have included nosocomial infection rate, hospital length of stay, and mortality.13 The relationship between glycemic variability and outcomes in the hospital may be stronger among patients not having known diabetes than among those with diagnosed diabetes and, depending on the study, may be attenuated or not discernable in the presence of diabetes.4,9–11 As a caution on interpretation, it is noted that a one-time correction of BG with insulin during the time frame of data collection may inflate the SD or CV, thus complicating comparisons of these metrics among patients who did or did not experience such one-time corrections of overall glycemia. Additionally, in the analysis of large groups, it is likely that patient factors should be considered that may acutely change overall glycemia and degree of insulin resistance, including underlying diagnosis, use of glucocorticoids, nutritional treatment plan, and choice of antihyperglycemic therapy.
Technology and newer treatments may be at hand that have the potential to reduce glycemic variability, thus permitting both design of randomized trials and ultimately therapeutic intervention to reduce variability.18–21 The expected opportunities for improvement include refinement of insulin-based treatment algorithms, hypoglycemia protocols, use of glucose monitoring technology to improve the effectiveness of insulin delivery systems,22 new insulins, and use of incretin-based therapy.23
Glycemic Targets and Rate of Correction
As is well appreciated, the strongest evidence showing causal relationship between hyperglycemia and complications of diabetes arose from randomized trials of the 20th century, including the Diabetes Control and Complications Trial Research Group (DCCT) and UK Prospective Diabetes Study (UKPDS), enrolling patients having type 1 diabetes (T1D) and type 2 diabetes (T2D), respectively.24,25 These trials demonstrated, in comparison to control groups receiving less intensive treatment, that there was reduction in the risk of microvascular complications for patients receiving more intensive glycemic management. Each intensively treated group experienced a greater event rate for hypoglycemia. Reductions in the microvascular complications in these studies are thought to result from exposure to hyperglycemia experienced over the long term, in time frames measured in years (also termed “glycemic exposure”). A separate question is whether immediate beneficial consequences of glycemic exposure, in the absence of a hyperglycemic emergency, also would be realized by hospitalized patients vulnerable to a different set of specific risks that potentially would be realized in the short term. Theoretically, the improvement of hyperglycemia and glycemic variability could reduce reactive oxygen species accumulation and inflammatory activation, resulting in both short-term and long-term benefits.
Glycemic Targets and Rate of Correction in the Ambulatory Setting
In the ambulatory setting, during initial correction of chronic hyperglycemia, a patient may experience painful neuropathic symptoms.26–30 Similarly, during intensification of therapy, early worsening of retinopathy may be observed. Risk factors for early worsening include the duration of diabetes, severity of the diabetic retinopathy at baseline, higher HbA1c, and magnitude of HbA1c reduction, such that specific cautions on monitoring have been recommended.31,32 The temporary problems of acute painful sensory neuropathy or early worsening of retinopathy are seen as a tolerable price to pay for the long-term benefits expected from improved glycemic control. In the ACCORD trial, however, in an older population with more cardiovascular disease than among subjects studied in the UKPDS, intensification of therapy of T2D was accompanied by increased mortality.33–35 With acknowledgment that mechanisms of harm would be different in each situation, and that comorbidities and concomitant therapies must be considered, these experiences are mentioned as analogies when discussing whether treatment guidelines for hyperglycemia in the inpatient setting might differ according to the presence or absence of diabetes. Furthermore, among patients with diabetes, one needs to consider whether the target and rate of correction of hyperglycemia might differ according to severity of chronic hyperglycemia before admission, as reflected by the HbA1c or other indicators of preadmission control, including other biomarkers, such as fructosamine, glycated albumin, home blood glucose monitoring, or CGM. As we enter the era of personalized medicine, it is quite possible that glycemic targets may differ based on the preexisting metabolic status of the patient.
Glycemic Targets and Rate of Correction among Subgroups of Hospitalized Patients
Many observational studies of hyperglycemia in the hospital setting include both patients with diabetes and patients not known to have diabetes. Overall, the findings show that hospital hyperglycemia is associated with adverse outcomes, including mortality (see Chapter 1).36 If diabetes can be discounted, then hospital hyperglycemia may be designated as stress hyperglycemia.37 In the critical care setting, randomized trials of strict glycemic control have not produced consistent evidence for benefit among mixed intensive care unit (ICU) populations, but overall these trials have demonstrated increased risk for hypoglycemia. In some but not all reviews or meta-analyses of interventional trials among critically ill patients, a signal suggesting benefit persisted according to subpopulation, with beneficial outcomes most readily observable among surgical patients.
Relationship of Outcomes to Hyperglycemia in the Presence of Diabetes
In some but not all observational studies of patients confirmed to have diabetes, the severity of hyperglycemia at the time of admission or during the course of hospitalization has been found to correlate with adverse outcomes. Among surgical patients, an increased risk of surgical site infection, myocardial infarction, stroke, and death, associated with the presence of diabetes, has been thoroughly reviewed.38 There have been exceptions to a general observation that severity of chronic hyperglycemia may be associated with adverse surgical outcomes. Among surgical patients with diabetes, however, most studies reporting glucose levels before elective surgery, or preoperative HbA1c, do find a relationship between severity of chronic hyperglycemia and adverse outcomes. Despite the lack of randomized controlled trials that might define any impact of preoperative correction of hyperglycemia, guidelines have been issued suggesting an upper limit of acceptable HbA1c between 8 and 9% before elective procedures.39 However, there is no true consensus on this point.
Over the years during which health-care-provider protocols for the reduction of perioperative hyperglycemia were increasingly utilized in patients with diabetes having cardiothoracic surgery, a downward trend was noted for adverse outcomes in the population having diabetes (see Chapter 1).36 During the same time frame, however, other aspects of care also changed. Although studied prospectively, much of the data in the cardiac surgery population with diabetes has been observational, such that a clear causal relationship between attainment of specific glycemic targets and improvement of outcomes cannot be inferred with confidence.
Outside of the ICU, an effective glycemic treatment protocol among surgical patients with diabetes also signaled a benefit. Some of the benefits included reduction of infection, reoperation, and mortality. A randomized trial among general surgical patients with T2D showed improvement of a composite end point of postoperative complications, including wound infection, pneumonia, bacteremia, and respiratory and acute renal failure.40
Among noncritically ill admissions with T2D having a cardiac or infection-related diagnosis, a recent retrospective analysis of the impact of glycemic control on hospital outcomes of 378 patients was divided into two groups according to mean point-of-care glucose for evaluation of adverse events, including death during hospitalization, ICU transfer, initiation of enteral or parenteral nutrition, line infection, new in-hospital infection,