Figure 3.2—Among patients having diabetes and admitted to intensive care units, odds ratios for hospital mortality are shown comparing those with higher (>7%) versus lower (≤7%) HbA1c, according to time-weighted average of blood glucose control in mMol/L.
Source: From Egi et al.64
Among patients with diabetes, it may be necessary to differentiate between those admitted with chronic hyperglycemia and those experiencing acute worsening of hyperglycemia upon admission, whose physiology might more closely resemble that of a person having stress hyperglycemia. The abnormalities of physiology that might cause stress hyperglycemia also may exist in a patient with diabetes. Among patients with diabetes, little published information examines outcomes separately for those with or without exacerbation of preadmission glycemic control, which might be taken as the equivalent of stress hyperglycemia among patients with diabetes.
It must be acknowledged that exacerbation of hyperglycemia in diabetes is not readily definable in a quantitative manner that would permit comparison to stress-induced hyperglycemia among people without diabetes. When HbA1c results are considered, hyperglycemia detected during admission may have the greater adverse prognostic significance among those patients with diabetes who formerly had the more satisfactory control.64–67 Although the numbers of patients examined so far has been small, a great step forward is taken by those studies that refer to HbA1c to identify acute changes of glycemia, including rapid reduction or stress-induced exacerbation of hyperglycemia among patients with diabetes, and their relationship to outcomes (Table 3.3). In a study of mortality among patients having known diabetes admitted with pneumonia, the 30-day mortality was not predicted by admission glucose, hypoglycemia during hospitalization, or HbA1c, but rather was predicted by mean glucose during the admission.65 In an observational study of 329 pyogenic liver abscess patients with diabetes, the HbA1c alone did not predict adverse outcomes. Among 131 patients having available both an HbA1c within 1 week of admission and also an evaluable admission BG, the “glycemic gap” (admission BG minus estimated average BG) was shown to be associated with adverse outcomes. In a receiver operating characteristic (ROC) analysis, a glycemic gap cutoff value of 72 mg/dL was optimal for predicting adverse outcomes.66 One center has shown that among patients with diabetes, peak glucose predicts adverse ICU outcomes for those having lower HbA1c, but not for those having higher HbA1c (see Figure 3.3).67 The study of 1,000 critically ill patients showed significantly higher mortality for those having diabetes with HbA1c >7%, but no relationship to peak hyperglycemia. Among the patients with HbA1c <7% and those having stress hyperglycemia, the mortality risk increased according to severity of peak hyperglycemia in the first 48 h. In a multivariable analysis, however, these relationships no longer were seen, and only the Acute Physiology and Chronic Health Evaluation II (Apache II) score was independently associated with mortality.67 In a search for metrics that might be used to assess the importance of exacerbation of both acute and chronic hyperglycemia, the stress hyperglycemia ratio (SHR) has been defined by one group as the ratio of admission glucose divided by estimated average glucose derived from HbA1c.68 From a study population of 4,691 admissions of adult nonobstetric patients, in a multivariable analysis containing glucose, SHR, and other defined variables, higher values for the SHR, but not for glucose, were found to be independently associated with critical illness (either in-hospital death or admission to the ICU). From the finding of a single elevation of admission HbA1c, the authors did not diagnose diabetes but rather identified patients as having background hyperglycemia.68 None of these studies attempted to infer the likely presence of stress-induced exacerbation of hyperglycemia by the company it might keep, such as increased lactate levels.
Table 3.3—Glycemic Gap: Hospital Complications among Patients Confirmed to Have Diabetes, in Relation to Acute Hospital Hyperglycemia
*Glycemic gap = difference between HbA1c-derived estimated average glucose (eAG) and observed plasma glucose measured at initial presentation.
See also a discussion of “background hyperglycemia” in relation to acute hospital hyperglycemia.68
Figure 3.3—Hospital mortality versus acute glycemia according to preadmission HbA1c in critically ill patients. Open circles, no diabetes or stringently controlled (HbA1c <6%); open squares, diabetes (HbA1c 6–<7%); closed diamonds, (HbA1c ≥7%).
Source: From Plummer et al.67
Attempts have been made to infer the prevalence of diabetes from sampling of HbA1c upon admission to the hospital.69 Even in the ambulatory setting, many intercurrent conditions may affect the HbA1c, so as to invalidate its utility in diagnosing diabetes or assessing glycemic control. Factors such as anemia, transfusions, renal disease with or without use of erythropoietin, and other conditions may be even more prevalent among hospitalized patients. Although we believe it is premature to embrace specific targets based on the absence of diabetes or admission HbA1c in the presence of diabetes, it is noted that Marik and Egi have suggested a spectrum of therapeutic targets for each of six categories of ICU patients.1,70
The Paradox of the Protective Effect of Diabetes: Fact or Artifact?
When the evidence is considered that diabetes may be a risk factor for some adverse outcomes, but that a given level of hyperglycemia is more strongly associated with harms for patients not having diabetes than for those with diabetes, then a paradox would appear to be presented.38,71 The paradox arises from comparison of outcomes at comparable bands of hyperglycemia between those hyperglycemic patients with or without diabetes. Kotagal et al. examined possible explanations for the seeming paradox and have suggested that one likely contributing factor, among several suggested explanations, could be that insulin therapy is more consistently and better delivered in patients with a known diagnosis of diabetes and that insulin therapy is less well tolerated when delivered to those not having diabetes.38,57 Another interpretation acknowledged by Kotagal et al.38 and others9,67 deserving consideration here could be that habituation to and tolerance of chronic hyperglycemia may occur for patients with diabetes, in contrast to intolerance that may be experienced by patients rendered acutely hyperglycemic who formerly did not have diabetes.47 Another highly plausible possibility could be that a given level of hyperglycemia for a patient not having diabetes may signify greater physiologic stress than for a patient with diabetes. In fact, some of the hyperglycemic group members cataloged as not having diabetes may in fact have had unrecognized diabetes.38 HbA1c, for example, is often falsely low in patients who have chronic illness and it may be impossible to have an accurate glycemic history for those patients.72
In describing the effect of hyperglycemia upon outcomes among patients with diabetes, we are reluctant to accept the terms “paradox” or “protective,” because we lack knowledge of the appropriate comparative groups of patients. Those who would advocate more lenient targets for patients who have diabetes with hyperglycemia generally have argued that the hyperglycemia may be protective or that acute reduction may be poorly tolerated. Given that stress hyperglycemia is recognized to have prognostic significance among people who do not have diabetes, it seems desirable to recognize stress-induced exacerbation of hyperglycemia among patients with diabetes. We suggest that outcomes for the following groups be considered separately, not only looking at observational studies but also, using effective technologies, eventually randomizing with the intent to treat to specific targets:
• No diabetes, no stress hyperglycemia
• No diabetes, stress hyperglycemia
• Diabetes, lacking stress-induced