In Reply: Dr Hermans and colleagues raise several concerns, many of which were explicitly addressed by our methods. They cite substantial differences among included trials (specifically fixed vs titrated insulin dose and glucose levels actually achieved). We did not include trials that used a fixed insulin infusion rate without a glucose goal lower than 150 mg/dL in the tight control arm (such as early studies of glucose-insulin-potassium). Because we too hypothesized that degree of glycemic control (“normalization of blood glucose”) might be important, we stratified analyses for all outcomes based on the glucose goal in the tight control arm (very tight, ≤110 mg/dL, vs moderately tight, 111-150 mg/dL). We also addressed the issue of trials not achieving the specified glucose goal by performing a sensitivity analysis in which studies were stratified into very tight and moderately tight glucose control according to actual achieved glucose level. We addressed the issue of overlap between tight glucose control and usual care arms by restricting analysis to trials that achieved at least a 20-mg/dL difference between the 2 groups. Finally, realizing that 30-day mortality and hospital mortality are not identical, we performed a sensitivity analysis (not reported in the article) restricting analysis to studies reporting hospital mortality and found no difference in our results.
Wiener RS, Wiener DC, Larson RJ. Tight Glucose Control in Critically Ill Adults—Reply. JAMA. 2008;300(23):2725-2728. doi:10.1001/jama.2008.819