Diabetes Linked to Improved Survival in Critically Ill Patients

The analysis covered more than 1.5 million subjects in a retrospective cohort and more than 36,000 subjects in a prospective cohort.

Senior author Dr. Ivor S. Douglas wrote, "Our observations robustly correlate with those of other investigators in the field.... Mortality was lower in diabetes for all ... diagnoses except acute myocardial infarction -- consistent with smaller studies."

In the January 10th issue of Critical Care Medicine, Dr. Douglas of the University of Colorado, Denver, and colleagues report that in the larger retrospective cohort, patients with a history of diabetes had a lower risk of in-hospital mortality compared to nondiabetics. Before adjustment, the odds ratio was 0.90; after adjustment, it was 0.75.

In the prospective cohort, after adjustment for age and severity of illness, the odds ratio for mortality was 0.88 in the diabetes group.

In the retrospective dataset, diabetic patients in all severity-of-illness categories had lower hospital mortality compared to nondiabetics. In the prospective dataset, diabetic patients with extreme severity of illness had decreased age-adjusted hospital mortality.

"The surprising finding highlights that despite decades of investigation we continue to have a limited understanding of the pathophysiological modifying effects of diabetes in critical illness," Dr. Douglas said. "Our data suggest that there is a potential protective effect in diabetics of hyperglycemia but relative intolerance of hypoglycemia."

He continued, "Importantly, there was a particularly detrimental effect of acute hyperglycemia in nondiabetic patients which may account for the significant difference in mortality we found between diabetic and nondiabetic patients in critical illness."

Dr. Douglas and his coauthors acknowledge that their findings "may seem counterintuitive." But they also point out that "preexisting diabetes is not a chronic health parameter in the calculation of APACHE III or IV," because it did not independently predict mortality when those scores were derived.

Now, the researchers conclude, it's time "to evaluate potential mechanisms for this association that might be therapeutically exploitable."