Blood pressure, haemorrhagic stroke, and ischaemic stroke: the Korean national prospective occupational cohort study

BMJ 2004; 328 doi: (Published 05 February 2004) Cite this as: BMJ 2004;328:324
  1. Yun-Mi Song, epidemiologist1,
  2. Joohon Sung, epidemiologist2,
  3. Debbie A Lawlor, lecturer in public health4,
  4. George Davey Smith, professor of clinical epidemiology4,
  5. Youngsoo Shin, epidemiologist3,
  6. Shah Ebrahim, professor of epidemiology of ageing (shah.ebrahim{at}
  1. 1Department of Family Medicine, Samsung Medical Center, SungKyunKwan University School of Medicine, 50 Irwon-Dong, Gangnam-Gu, Seoul, Korea
  2. 2Department of Preventive Medicine, Kangwon National University College of Medicine, Hyoja-2-Dong, Chunchon, Kangwon-Do, Korea
  3. 3Department of Health Policy and Management, College of Medicine, Seoul National University, 28 Yeongeon-dong, Jongno-Gu, Seoul
  4. 4Department of Social Medicine, University of Bristol, Bristol BS8 2PR
  1. Correspondence to: S Ebrahim


    Deaths from haemorrhagic stroke declined consistently through the 20th century, but deaths from ischaemic stroke showed a rise and fall, mirroring the coronary heart disease epidemic.1 Blood pressure has also declined,2 and if blood pressure is more strongly associated with haemorrhagic stroke than with ischaemic stroke, this might contribute to the divergent trends. Previous meta-analyses have shown contradictory findings; one showed similar associations for both stroke subtypes,3 and another, of Asian studies, showed a stronger association with haemorrhagic stroke than with ischaemic stroke.4 To resolve this uncertainty, we examined the association of blood pressure with subtype of stroke in a large cohort of Korean civil servants.

    Participants, methods, and results

    The Korean National Health System for public servants and teachers provides medical expenses and biennial multiphasic health examinations at which blood pressure is measured in the seated position by trained staff using a standard mercury sphygmomanometer (fifth Korotkoff sound used for diastolic pressure) or an electronic manometer.5 We grouped mean blood pressures for individuals between 1986 and 1996 according to the joint national committee on prevention, detection, and treatment of high blood pressure categories of normal, stages 1, 2, and 3 (table). We included deaths attributed to ICD-10 (international classification of diseases, 10th revision) codes of I60-I69 for all strokes, I61 for haemorrhagic stroke, and I63 and I67.8 for ischaemic strokes between 1991 and 2000 in these analyses. We categorised non-fatal strokes using data on the use of medical care and found an accuracy of 83.4% and 85.7% for ischaemic stroke and haemorrhagic stroke.

    Mean blood pressure and stroke subtype from the Korean National Health System Study, 1986-2000. Data are adjusted relative risks (95% confidence intervals) unless otherwise indicated*

    View this table:

    In 9.5 million person years of observation of 955 271 people; they had 14 057 strokes, giving crude and age standardised incidences of 1.48 and 2.24 for every 1000 person years. Of these, 10 716 (76%) strokes had complete information on major exposure variables and we included these in our analyses; we classified 2695 strokes as haemorrhagic, 5326 as ischaemic, 1731 as undetermined, and 964 as subarachnoid haemorrhage.

    We calculated fully adjusted relative risks and 95% confidence intervals using logistic regression. The gradient of risk with blood pressure was steeper for fatal than non-fatal stroke, reflecting a relative excess of haemorrhagic strokes among fatal events. Both ischaemic stroke and haemorrhagic stroke had strong gradients with blood pressure, but these were much steeper for haemorrhagic stroke with a stage 3 category relative risks of 9.56 (95% confidence interval 8.46 to 10.80) and 28.83 (24.89 to 33.40) for ischaemic and haemorrhagic strokes. For each higher 20 mm Hg of systolic blood pressure, the relative risk of ischaemic and haemorrhagic stroke increased by 2.23 (2.17 to 2.30) and 3.18 (3.06 to 3.30), z test for difference between odds ratios 11.40, P < 0.00001. Those excluded because of incomplete data had similar distributions of stroke subtypes, and including them in analyses resulted in similar age-sex adjusted blood pressure gradients.


    The gradient of the relationship between blood pressure and haemorrhagic stroke is steeper than that for ischaemic stroke. Falls in blood pressure observed over the 20th century may lead to bigger reductions in the incidence of haemorrhagic stroke compared with ischaemic stroke and thereby provide a partial explanation for the differential trends in stroke subtypes.1 Falls in blood pressure cannot be ascribed solely to antihypertensive drugs as they have been seen at young ages and during times when treatment was not widely used. Factors in early life, rather than treatment of hypertension, may have contributed to population declines in both blood pressure and risk of stroke, particularly haemorrhagic. Our findings also emphasise the importance of controlling blood pressure, particularly in countries with a high risk of haemorrhagic stroke.


    • Contributors Y-MS contributed to design, analysis, interpretation and revision of the manuscript JS contributed to design, interpretation and revision of the manuscript YS contributed to interpretation and revision of the manuscript. DAL contributed to the design, interpretation, and revision of the manuscript. GDS contributed to the design, interpretation, and revision of the manuscript. SE wrote the first draft of the paper and contributed to design, interpretaton, and revision of the manuscript. Y-MS and SE are guarantors.

    • Funding Korean Ministry of Health and Welfare (01-PJ1-PG1-01CH10-0007)

    • Competing interests None declared.

    • Korean Health iNsurance Corporation, who are legally able to use health insurance data for public health purposes.  


    View Abstract