Mortalitätsrate diabetischer und nicht diabetischer Intensivstationspatienten – Analyse relevanter Risikofaktoren mit Hilfe eines Cox-Regressionsmodells

Introduction: The global prevalence of diabetes mellitus in the world’s adult population is estimated to be 8.3 % and still continues to increase dramatically. Approximately one fourth of all diabetic patients in the United States are not diagnosed. Because of its multifactorial complications diabetes mellitus implicates a lot of challenges in patient care. The HbA1c-value is measured routinely in all critically ill patients to differentiate between stress hyperglycemia and exacerbation of diabetes mellitus. Hyperglycemia in critical illness has been associated with increased mortality under a variety of clinical conditions, most notably myocardial infarction, stroke and after major surgery. Methods: This prospective observational study included 1005 intensive care patients of the McGill University Health Center in Montreal between June 2011 and June 2012. The study was designed to investigate the association between abnormal HbA1c-values on admission to Intensive Care Unit with poor outcomes as well as the influence of other possible variables on mortality. The data was presented using Cox regression, Kaplan-Meier survival estimates and T-test. As secondary outcomes of this study the prevalences of known and unknown diabetes mellitus have been determined. In the context of subgroup analysis the mortality and mean survival time of cardiac surgery, other surgical and medical intensive care patients as well as diabetic and non-diabetic patients have been compared with each other. Moreover, variables with significant influence on the mortality of these groups have been investigated. Results: The prevalence of diabetes mellitus in this study was 26 %. 6,3 % of the intensive care patients showed an undiagnosed diabetes mellitus. Within the scope of Kaplan-Meier estimates and T-test it was demonstrated that diabetic patients had higher mortality rates and significantly lower mean survival times compared with non-diabetic patients. At the same time the Cox regression model showed that the HbA1c-value itself had no significant influence on the mortality rate of all intensive care patients as well as in every subgroup analysis. However, the following variables showed significant influence on mortality of the whole intensive care patient group: Age, preexisting or new-onset renal failure, active cancer, high minimal glucose levels, hypoglycemic events as well as belonging to the cardiac surgery patient group, whereas the last-mentioned variable had a protective influence. Age and preexisting renal failure are the only variables that showed significant influence on the mortality rate in every subgroup analysis. Conclusion: Hypoglycemic events as well as high minimal glucose levels showed significant influence on the mortality rate in the multivariate analysis. Opposite to the current international data this effect could not be proven for the glycemic variability. As the glucose parameters are the only modifiable variables with significant influence on the mortality rate within the scope of this study it is recommendable to implement a continuous recording of the plasma glucose levels in future studies to investigate whether other variables (e.g. amplitude or abruptness of glucose changes, frequency of fluctuation, length of hypoglycemic episodes) show relevant effects on the mortality rate. A continuous real-time plasma glucose measuring of the interstitial glucose levels with simultaneous use of computer-assisted insulin-application systems could improve glycemic control and therefore contribute to a lower mortality rate.