We investigated the effects of hyperglycaemia on infarct size of 82 acute ischaemic stroke patients with angiographically diagnosed intracranial occlusion in middle cerebral artery territory. There were 9 diabetics, 40 non-diabetic hyperglycaemics and 33 non-diabetic normoglycaemics (mean age 67 +/- 8 SD years, male/female ratio 1:1). For each patient the infarct at CT was compared to that predicted from the location of the arterial occlusion. The extent of the infarct was then classified as equal to, smaller than and larger than estimated, taking a standard anatomical template of arterial territories as reference. The results were analysed separately according to the presence or absence of a collateral blood supply (CBS) at angiography. The clinical outcome at 30 days was also evaluated. The 35 patients lacking CBS had a high frequency of equal to estimated lesions (75%), without substantial differences among the three subgroups (72% of hyperglycaemics, 82% of normoglycaemics and 67% of diabetics; Fisher's exact test not significant for any of the pairwise comparisons). On the contrary, the 47 patients with CBS exhibited an overall predominance of smaller than estimated lesions (66%) but with a very uneven distribution among hyperglycaemics, normoglycaemics and diabetics (82%, 64% and 0%, respectively; p < 0.05 at Fisher's exact test for diabetics vs hyperglycaemics). Finally, the clinical outcome was bad (death and neurological impairment) in 89% of diabetics, 72% of hyperglycaemics and 54% of normoglycaemics (p < 0.05). These results suggest that in patients with intracranial arterial occlusion associated with CBS the effects of hyperglycaemia might be beneficial in non-diabetics and harmful in diabetics.(ABSTRACT TRUNCATED AT 250 WORDS)