Other models find that hyperglycaemia is not independent predictor

Editor—Christopher J Weir and colleagues conclude from their study of a cohort of 750 non-diabetic patients with stroke that hyperglycaemia (plasma glucose concentration >8 mmol/l) during the acute phase has an adverse influence on outcome and that this is independent of severity of stroke.1 Stroke severity was assessed in a limited way using only the Oxfordshire community stroke project classification and time to resolution of symptoms (≤72 hours or >72 hours), both of which are relatively inaccurate measures. When two variables are closely correlated—for example, stroke severity and glucose concentration—the one that is most accurately measured (glucose concentration) will always emerge as the strongest explanatory variable in multiple regression even if it is, in fact, less important.2

We have produced a series of validated models to predict the probability of survival and disability using the 530 patients from the Oxfordshire community stroke project who were seen within 30 days of their stroke.3 A history of diabetes mellitus and the presence of acute hyperglycaemia (glucose concentration >12 mmol/l) were two of about 20 variables that were entered into these models, in addition to several measures related to stroke severity (eye, motor, and verbal components of the Glasgow coma score; arm power; and ability to walk). Although diabetes mellitus was an adverse and independent predictor of death (relative hazard 2.01; 95% confidence interval 1.36 to 2.99), hyperglycaemia was not (1.66; 0.93 to 2.97). Neither of these variables was an independent predictor of death or disability (modified Rankin score >2) at six months.

We repeated our analyses using only the 249 patients seen within 72 hours of onset with no known history of diabetes; we redefined hyperglycaemia as glucose concentration >8mmol/l to allow direct comparison with the results of Weir …