Comparison of Stress Hyperglycemia and Diabetes
In studies that classify hyperglycemia according to presence or absence of chart history of diabetes or preadmission antihyperglycemic medication use, some hyperglycemic patients will have diagnosed or undiagnosed diabetes and others will have stress hyperglycemia. A study of burn patients showed greater length of stay for burn victims admitted with diabetes compared with those who had acute hyperglycemia.42 A meta-analysis of observational and interventional studies between May 2005 and May 2010 involving 12,489,574 critically ill patients, of whom 2,327,178 had diabetes, suggested increased mortality among patients with diabetes and admitted to the surgical ICU, with odds ratio (OR) for ICU mortality (95% confidence interval [CI]) being 1.48 (1.04 to 2.11), in-hospital mortality 1.59 (1.28 to 1.97), and 30-day mortality 1.62 (1.13 to 2.34). Otherwise, in this meta-analysis, the presence of diabetes showed no overall association with mortality.43 In the general critical care setting, a number of studies have found that diabetes itself is not a predictor of increased mortality.9,42,44–47 In many studies, hyperglycemia without a history of diabetes has been associated with outcomes worse than those outcomes observed among patients known to have diabetes.9,38,48–56 In a seminal report, although mean BG on admission of 223 patients with new hyperglycemia did not differ significantly compared with that of 495 patients with known diabetes, higher hospital mortality was observed in the new hyperglycemia group.48 Details of studies contrasting the impact of diabetes versus stress hyperglycemia upon outcomes are shown in Table 3.2.
Table 3.2—Outcomes Associated with Diagnosis of Diabetes, Compared to Stress Hyperglycemia, in Relation to Hyperglycemia
Some studies refine the comparison between stress hyperglycemia and diabetes by focusing on given levels of hyperglycemia, which may be associated with harms experienced among patients not having diabetes, and which may be associated with less risk among those known to have diabetes. Among 8,727 cardiac surgery patients, when the adjusted ORs for specific complications were considered at each of three bands of glycemia (good, moderate, or poor control) according to the presence or absence of diabetes, the difference in complication rates between those having or not having diabetes was greatest for those with poor BG control (peak glucose >250 mg/dL in the first 60 h postoperatively), such that the diabetes group had the lower rate for each type of several complications (see Figure 3.1).52
Figure 3.1—Prevalence of complications by blood glucose concentrations and diabetes status.
Source: With permission from Ascione et al.52
In a case-control single-center ICU study, after implementation of an insulin-based glycemic management protocol for BG >150 mg/dL, the rate of death in the hospital was 10% (95% CI, 9–12%) for patients without diabetes who required glycemic control, higher than the rate of 6% (95% CI, 4–7%) for patients with diabetes, or 5% (95% CI, 4–6%) for controls (P < 0.001). Mortality increased for patients without diabetes at mean blood glucose of 144 mg/dL and for patients with diabetes at 200 mg/dL (P < 0.001).49 In another mixed ICU study, when glucose readings were examined by several metrics among 4,946 admissions, the degree of hyperglycemia was found to be associated with mortality among the patients without diabetes but, at similar levels, not among the 728 patients with diabetes. Hyperglycemia showed no significant association with outcome in the patients with diabetes.53 Under some methods of analysis, hyperglycemia in the presence of diabetes even may seem to confer a “protective” effect when compared with similar degrees of hyperglycemia in the control population.38,52,55 In a large multicenter international observational study of 44,964 patients admitted to 23 ICUs with 12,880 patients identified as having diabetes, diabetes was associated independently with a slightly lower mortality rate, with OR (95% CI) of 0.93 (0.87–0.97); among patients with mean BG 80–110 mg/dL, diabetes was independently associated with increased risk of mortality, but among patients with mean BG of 110–140 mg/dL, diabetes was independently associated with decreased risk of mortality, leading the authors to speculate that those with antecedent diabetes may have developed a tolerance for hyperglycemia that was lacking among those experiencing stress hyperglycemia.9
Of course, such an interpretation that diabetes confers a “protective effect” implies that hyperglycemia as a stress response does itself contribute to adverse outcomes. Another speculation, more relevant on general wards, would be that the beneficial actions of insulin are more likely to be experienced by patients with diabetes, because of greater readiness to introduce insulin therapy.38,57 Difficulty of interpretation results from the observational design of most studies. There is uncertainty on diagnosis of diabetes in some reports; uncertainty whether stress hyperglycemia in general is simply a marker of severity of illness or other downstream consequence of the illness or its treatment, possibly even beneficial58; and, conversely, uncertainty whether stress hyperglycemia at some given level of severity becomes maladaptive, itself contributing to adverse outcomes.
It has been suggested that stress hyperglycemia may be no worse a prognostic indicator than lactate levels. The incremental overall increase of lactate observed in hospitalized populations with stress hyperglycemia is so small that measurements would be incapable of discriminating between normal or stressed physiology in most individual cases. Still, lactate levels have been shown to correlate with outcomes, in a manner similar to stress hyperglycemia.59–61 An emergency room follow-up study of patients with sepsis found that mortality risk did not increase with hyperglycemia unless associated with simultaneous hyperlactatemia.60 In one retrospective ICU study, no independent association between hyperglycemia and mortality was identified, once lactate levels were considered.62
Optimal Control of Hospital Hyperglycemia in the Presence of Diabetes
It is unknown what the optimal targets for glycemic control should be for subgroups of hospitalized patients with the diagnosis of diabetes along with another admitting diagnosis. When pooled data of critically ill medical and surgical patients from Leuven, Belgium, were studied, no mortality reduction from intensive control was demonstrable in the subgroup with diabetes.63 Examining mortality in relation to bands of glycemia, a recent large multicenter observational study of critically ill patients found that among patients without diabetes, the lowest mortality rate was seen with a mean BG 80–140 mg/dL, whereas the lowest mortality among those with diabetes was seen at mean BG 110–180 mg/dL.9
It has been speculated but not proven that rapid reduction to low targets may be harmful for those patients with diabetes who have become acclimatized to chronic hyperglycemia. The speculation is based partly on a two-center ICU observational study of 415 ICU patient with diabetes, showing that if time-weighted hospital glycemia was in a higher rather than lower range, then patients with diabetes with higher HbA1c (>7%) had a lower mortality rate than those with diabetes with lower HbA1c (≤7%) (see Figure 3.2).64 For those with HbA1c >7%, among the nonsurvivors, there was a lower time-weighted average BG than among the survivors. At lower HbA1c, the mortality difference according to time-weighted average BG was not apparent. The authors speculated that rapid reduction of hyperglycemia, among patients habituated to chronic hyperglycemia, might be acutely harmful. By not capturing the glucose on first admission to compare against the HbA1c, however, the effects of medical stress could have been underestimated.64