Letters Section Editor: Stephen J. Lurie,
MD, PhD, Senior Editor.
To the Editor: We agree with Dr Finney and
colleagues1 that current evidence suggests
that higher levels of exogenous insulin do not, in themselves, reduce mortality.
However, their conclusion that glucose regulation to levels below 145 mg/dL
(8.0 mmol/L) was responsible for the beneficial effect reported was too strong.
In the authors' regression analysis, glucose regulation was stratified
as tertiles of time spent in 6 predefined glucose bands. When compared with
patients whose majority of glucose levels were above the upper limit of the
111- to 144-mg/dL (6.1-8.0 mmol/L) band, patients in and below this band were
more likely to survive. Glucose level, however, is actually a continuous variable.
Stratification may entail a loss of information that inflates P values and also may arbitrarily group patients together. By contrast,
the authors could have used their data to discover the exact threshold level
in a certain population, perhaps by calculating the coefficient of variance
of the mean glucose level or a time-averaged glucose level.
van der Horst ICC, Gans ROB, Zijlstra F, Ligtenberg JJM. Exogenous Insulin and Hypoglycemia as Prognostic Factors in Critically Ill Patients. JAMA. 2004;291(5):558. doi:10.1001/jama.291.5.558-b