Results

There were 1027 participants with a main discharge diagnosis of stroke of which 239 (23.3%) were classified as ischaemic and 90 (8.8%) as haemorrhagic. The remaining 698 strokes (68%) were ill defined. We found an inverse association between height and all stroke and ill defined stroke (table 1) while for haemorrhagic strokes there was a stronger inverse relation, each 10 cm increment in height resulting in an almost 30% decline in risk. The association between height and ischaemic stroke was weak and did not reach statistical significance. The difference between the relative rates for haemorrhagic and ischaemic stroke adjusted for age and sex was not statistically significant (p=0.29). Adjustment for confounding variables had little effect on the relative rates and did not account for the different associations with ischaemic and haemorrhagic stroke.

Discussion

To our knowledge only one other population-based study has examined the association between height and stroke subtype in men and women.2 The results were similar to those reported here but are limited because in that study it was not possible to control for socioeconomic position, an important confounding factor.

In the current study 86% of the cohort were linked to either a record of hospital admission or a death record.5 In a study from a more affluent population around a quarter of deaths occurred in people who had spent little or no time in hospital.6 It is therefore probable that in the current cohort, many of the 14% of participants who were not computer linked may never have been admitted to hospital, indicating that record linkage is close to complete. Misclassification bias related to differential hospital attendance among socioeconomic groups is also unlikely as the majority (74%) of people with a stroke were admitted to one hospital, with only small proportions admitted to several other hospitals.

The major weakness of this study is the low proportion of strokes that were accurately classified. However, as haemorrhagic strokes usually account for 10%–15%7 of all strokes it is likely, given that almost 10% of strokes were already so defined and that haemorrhagic strokes are more likely to be accurately diagnosed, that the majority of the ill defined strokes are ischaemic. The relative rates for ill defined stroke, intermediate between those for haemorrhagic and ischaemic stroke, are in agreement with this.

Our findings add to the evidence that early life factors may determine risk of stroke. In a recent Finnish study haemorrhagic stroke was more closely associated with birth weight adjusted for head circumference than ischaemic stroke.8 The authors hypothesised that reduced fetal growth, in which the brain is spared through differential shunting of blood at the expense of the trunk, may permanently change the cerebral arterial structure resulting in reduced deposition of elastin and increased risk of hypertension and haemorrhagic stroke. Poor fetal growth was also associated with increased risk of ischaemic stroke, but less than that of haemorrhagic stroke. Persons of tall stature are likely to have had optimal fetal and childhood growth reducing the overall risk of stroke, in particular haemorrhagic stroke. Stroke has been declining throughout the past century and the substantial decline in the incidence of haemorrhagic stroke in the 1960s and 1970s traditionally attributed to use of antihypertensives may also be related to contemporaneous increases in height.