Tight glucose control in critically ill adults.

To the Editor: Dr Soylemez Wiener and colleagues reported the results of a meta-analysis that included 29 studies on tight glucose control in critically ill adults. Several methodological issues limit the conclusions to be drawn from this study. First, the included studies have substantial differences, most importantly concerning the type of intervention (fixed vs titrated insulin dose) and the glucose levels actually achieved. The proof-of-concept study on tight glucose control in the intensive care unit (ICU) hypothesized that normalization of blood glucose (80-110 mg/dL) with insulin protects against complications (eg, severe infections, organ failure, death). (To convert glucose values to mmol/L, multiply by 0.0555.) The scientific concept was that any degree of hyperglycemia above normal ( 110 mg/dL) in a condition of ischemia/reperfusion additionally damages those cells that take up glucose passively. The study by Soylemez Wiener et al pooled all published studies on insulin therapy (including glucose-insulin-potassium) and those that targeted glucose of 150 mg/dL or lower. Studies not achieving normoglycemia in the majority of patients should not be expected to provide the same benefit as in the original Leuven study. Methodological quality assessment in the metaanalysis should have addressed this issue of achievement of normoglycemia, as well as the varying blood glucose levels in the different control groups. Overlap of glucose levels in controls and the intervention groups should not exceed 20%. Achieving a glycemic target cannot be assessed using mean glycemia levels; high and low values may cancel out and falsely suggest goal achievement. The number of patients on average within the target would have been a better parameter. Second, the primary outcome measure used in the metaanalysis was inaccurate because hospital mortality is not identical to 30-day mortality. Benefit from tight glucose control in this population can only be expected after at least 30 days. Pooling lacks a scientific basis and predictably negates any possible benefit of tight glucose control. Third, presence of heterogeneity was not handled according to Cochrane guidelines. Studies contributing to statistical heterogeneity were identified and subsequently eliminated. This is methodologically inadequate as heterogeneity may indicate a true difference in effect size among studies differing in certain clinical or methodological characteristics (eg, actual achievement of normoglycemia). Cochrane guidelines advise the use of meta-regression for this purpose rather than calculating a summary effect estimate across a series of very different studies. This was not performed in the meta-analysis. Greet Hermans, MD Department of General Internal Medicine Miet Schetz, MD, PhD Greet van den Berghe, MD, PhD greet.vandenberghe@med.kuleuven.be Department of Intensive Care Medicine Catholic University of Leuven Leuven, Belgium