Introduction Lipoprotein(a) (Lp(a)) plays an important role in atherosclerosis. Some observational studies report Lp(a) as positively associated with myocardial infarction, but the relationship between Lp(a) and stroke is unclear. We examined the relationship between Lp(a) and the incidence of stroke in the general population.

Methods A multi-center population-based cohort study was conducted. A total of 10 444 men and women were eligible. Data were obtained from April 1992 to July 1995 in 12 rural districts in Japan. Subjects were divided into tertiles of Lp(a) levels. We analysed the risks of all stroke and of stroke subtypes in each sex using Cox's proportional hazard models.

Results Mean follow-up duration was 10.7 years. Cut-off levels of Lp(a) for tertiles were 10 mg/dl and 23 mg/dl. Risks for all stroke were 1.34 (95% CI 1.03 to 1.74) and 1.00 (95% CI 0.77 to 1.31) in the lower and the higher Lp(a) group, respectively, with reference to the middle group after adjustment for age, smoking status, drinking status, systolic blood pressure, and body mass index. Risks for cerebral haemorrhage (lower tertile 2.25, 95% CI 1.28 to 3.94 and higher tertile 0.93, 95% CI 0.49 to 1.77), were similar to all stroke and no significant relationships were seen between Lp(a) and cerebral infarction (lower tertile 1.15, 95% CI 0.83 to 1.60 and higher tertile 1.02, 95% CI 0.74 to 1.41),or subarachnoid haemorrhage (lower tertile 1.04, 95% CI 0.52 to 2.09 and higher tertile 0.96, 95% CI 0.48 to 1.90).

Conclusion Lower Lp(a) was an independent risk factor for stroke, especially, for cerebral haemorrhage in the general population.