Authors
Tunstall-Pedoe H. Woodward M. Tavendale R. A'Brook R. McCluskey MK.
Institution
Cardiovascular Epidemiology Unit, Ninewells Hospital, and Medical School,
Dundee. h.tunstallpedoe:dundee.ac.uk
Title
Comparison of the prediction by 27 different factors of coronary heart
disease and death in men and women of the Scottish Heart Health Study: cohort
study [published erratum appears in BMJ 1998 Jun 20;316(7148):1881].
Source
BMJ. 315(7110):722-9, 1997 Sep 20.
Abstract
OBJECTIVE: To compare prediction by 27 different factors in men and women of
coronary heart disease events, coronary deaths, and deaths from all causes.
DESIGN: Cohort study. SETTING: Scottish population study. SUBJECTS: In 1984-7
random sampling of residents aged 40-59 produced 11,629 men and women who
generated survey clinic questionnaires, examination findings, and blood and
urine specimens. MAIN OUTCOME MEASURES: Subsequent death, coronary artery
surgery, and myocardial infarction. Risks were calculated for each category
of factor or fifth of continuous variables. 27 factors were ranked by
descending age adjusted hazard ratio of the top to bottom class in each
factor, by sex and end point. RESULTS: Follow up averaged 7.6 years, during
which the 5754 men had 404 coronary events, 159 coronary deaths, and 383
deaths and the 5875 women 177, 47, and 208 respectively. The rankings for
factors for the three end points were mainly similar in men and women,
although hazard ratios were often higher in women. Classical risk factors
ranked better for predicting coronary risk than newer ones. Yet strong
prediction of coronary risk was no guarantee of significant prediction of all
cause mortality. Findings included an anomalous coronary
protective role for type A behaviour in women; raised plasma fibrinogen as a
strong predictor of all end points; and an unexpectedly powerful protective
relation of dietary potassium to all cause mortality.
CONCLUSIONS: These initial unifactorial rankings and comparisons must be
interpreted with caution until potential interaction, confounding, and
problems of measurement and causation are further explored.