Cohort study of coffee intake and death from coronary heart disease over 12 years

BMJ 1996; 312 doi: (Published 02 March 1996) Cite this as: BMJ 1996;312:544
  1. Inger Stensvold, research fellow Age Tverdal, research directora,
  2. Aage Tverdal,
  3. Bjarne K Jacobsen, professorb
  1. a National Health Screening Service, PO Box 8155, 0033 Oslo, Norway
  2. b Institute of Community Medicine, University of Tromso, Tromso, Norway
  1. Correspondence to: Dr Stensvold.
  • Accepted 1 November 1995

In our previous study we found that coffee intake was related to death from coronary heart disease and that coffee had an effect that was additional to increasing cholesterol concentrations.1 Recent meta-analyses of the effect of coffee on myocardial infarction and death from coronary heart disease showed great heterogeneity among the cohort studies.2 3

We report the results of a further six years of follow up in our cohort of 38500 men and women who were aged 35-56 at screening, with a reanalysis according to the two periods of follow up.

Subjects, methods, and results

The study population is described in our previous study.1 Participants reported that they had no history of cardiovascular disease, diabetes, stroke, hypertension, angina, or intermittent claudication. Follow up started at screening in 1977-82, when data on coffee drinking were obtained from a self administered questionnaire, and ended on 31 December 1992 (mean follow up 12.1 years). The outcome was coronary death according to the ninth revision of the International Classification of Diseases (ICD-9) (codes 410-413, 414.0, 414.1, 414.3, and 414.9), including sudden death of unknown cause (codes 798.1-798.2). The response rate was 83%. Data were stratified by sex and analysed by Cox's proportional hazards regression.

The coefficient for the risk of coronary death with coffee drinking during the first six years of follow up was twice that in the second six years (table 1). After six years an increased risk was found only in subjects who drank nine cups or more a day. When total cholesterol concentration was included as a covariate the coefficient for coffee drinking became 0.158 for the first six years of follow up, 0.051 for the second six years, and 0.083 for the whole follow up.

Table 1

Relative risks of death from coronary heart disease in cohort according to coffee consumption and time of follow up

View this table:

Analysis of follow up for 11 years in each county showed the strongest relation in Oppland, as in our previous study.1 However, when the first six years of follow up was excluded the relation was much weaker in all counties.

About 8% of the study population with the highest coronary risk (3080 subjects) was referred to a doctor. When this group was excluded the coefficient for coffee drinking was 0.143 for the whole follow up period and 0.118 after six years of follow up.

In 61% of the study population (23485 subjects) we had data from an identical study conducted five to 10 years after this study. The coefficient for coffee drinking became slightly higher when we used the risk factors, including coffee consumption, from the latest study for this group in the analysis.


The fairly strong association between coffee consumption and mortality from coronary heart disease in Norway has been distinctly weakened by six more years of follow up. The association was completely absent when the first six years of observation were excluded and cholesterol concentration was adjusted for.

Neither the high risk strategy nor changes in coffee consumption is likely to explain the weaker relations during the last part of the study. During the follow up a nationwide cholesterol campaign was also initiated, but this campaign is unlikely to have substantially influenced the relation between coffee consumption and coronary death.

Our finding may be explained by a change in the type of coffee consumed. A lipid rich fraction from boiled coffee seems to increase serum cholesterol concentration.4 When the boiled coffee is filtered the lipid rich factor is retained in the filter paper5 and the effect on cholesterol is reduced substantially. From unpublished data in Norway we know that the proportion consuming boiled coffee has been declining from 1985, the first year the type of coffee drunk was recorded. In this study we had no information on type of coffee, but in a few years we will have sufficient data for Norway to assess both dose and type of coffee in relation to coronary death.


  • Funding None.

  • Conflict of interest None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract