Dietary habits and mortality in 11 000 vegetarians and health conscious people: results of a 17 year follow up

BMJ 1996; 313 doi: http://dx.doi.org/10.1136/bmj.313.7060.775 (Published 28 September 1996)
Cite this as: BMJ 1996;313:775
  1. Timothy J A Key, research scientista,
  2. Margaret Thorogood, senior lecturerb,
  3. Paul N Appleby, scientific officera,
  4. Michael L Burr, senior lecturerc
  1. a Imperial Cancer Research Fund, Cancer Epidemiology Unit, Radcliffe Infirmary, Oxford OX2 6HE
  2. b Department of Public Health and Policy, London School of Hygiene & Tropical Medicine, London WC1E 7HT
  3. c Centre for Applied Public Health Medicine, University of Wales College of Medicine, Cardiff CF1 3NW
  1. Correspondence to: Dr Key.
  • Accepted 11 July 1996

Abstract

Objective: To investigate the association of dietary habits with mortality in a cohort of vegetarians and other health conscious people.

Design: Observational study.

Setting: United Kingdom.

Subjects: 4336 men and 6435 women recruited through health food shops, vegetarian societies, and magazines.

Main outcome measures: Mortality ratios for vegetarianism and for daily versus less than daily consumption of wholemeal bread, bran cereals, nuts or dried fruit, fresh fruit, and raw salad in relation to all cause mortality and mortality from ischaemic heart disease, cerebrovascular disease, all malignant neoplasms, lung cancer, colorectal cancer, and breast cancer.

Results: 2064 (19%) subjects smoked, 4627 (43%) were vegetarian, 6699 (62%) ate wholemeal bread daily, 2948 (27%) ate bran cereals daily, 4091 (38%) ate nuts or dried fruit daily, 8304 (77%) ate fresh fruit daily, and 4105 (38%) ate raw salad daily. After a mean of 16.8 years follow up there were 1343 deaths before age 80. Overall the cohort had a mortality about half that of the general population. Within the cohort, daily consumption of fresh fruit was associated with significantly reduced mortality from ischaemic heart disease (rate ratio adjusted for smoking 0.76 (95% confidence interval 0.60 to 0.97)), cerebrovascular disease (0.68 (0.47 to 0.98)), and for all causes combined (0.79 (0.70 to 0.90)).

Conclusions: In this cohort of health conscious individuals, daily consumption of fresh fruit is associated with a reduced mortality from ischaemic heart disease, cerebrovascular disease, and all causes combined.

Key messages

  • This study examined dietary factors associated with mortality among 11 000 health conscious people followed for an average of 17 years.

  • After smoking was adjusted for, daily consumption of fresh fruit was associated with a 24% reduction in mortality from ischaemic heart disease, a 32% reduction in mortality from cerebrovascular disease, and a 21% reduction in all cause mortality compared with less frequent consumption.

Introduction

In the early 1970s interest grew in the relation between diet and health, and it was suggested that the risk of death from ischaemic heart disease might be reduced by a high intake of dietary fibre1 or by a vegetarian diet. To test these hypotheses a cohort of about 11 000 British men and women was recruited between 1973 and 1979 from among the customers of health food shops and other people with an interest in health foods or vegetarianism. Diet was assessed by a short questionnaire which asked about vegetarianism and intake of a few high fibre foods. Follow up of the cohort until 19802 and 19853 showed that habitual consumption of wholemeal bread was not significantly associated with mortality but that a vegetarian diet was associated with a significant reduction in mortality from ischaemic heart disease. We report here the results of follow up until 1995. The aim of this analysis was to examine the relation of six dietary factors (vegetarian diet and consumption of wholemeal bread, bran cereals, nuts and dried fruit, fresh fruit, and raw salad) with mortality from conditions for which associations with diet have been suggested.

Subjects and methods

RECRUITMENT

Subjects were recruited by distributing a short questionnaire to customers of health food shops and clinics, subscribers to health food magazines and a Seventh Day Adventist publication, and members of vegetarian and health food societies. This questionnaire asked about current smoking, weight (but not height), and dietary habits. Subjects were asked whether they were vegetarian (this was not defined further) and to record their usual frequency of consumption of wholemeal bread, bran cereals, nuts or dried fruit, fresh fruit, and raw vegetable salads. There were three frequency categories: at least daily, less than daily but at least once a week, and less than once a week. Most subjects consumed these foods at least once a week so subjects were grouped into the two categories daily and less than daily. Subjects also provided their name and address, date and place of birth, and NHS number for tracing at the NHS central register. We recruited and successfully traced 10 977 subjects between November 1973 and November 1979.

To assess the stability of dietary patterns we interviewed 289 subjects between one and a half and six years after the recruitment questionnaire was completed. Of these subjects, 66% of those who initially reported that they were vegetarian were still eating meat or fish less than once a month, and 67% of those who initially reported consuming wholemeal bread daily were still doing so.2

FOLLOW UP

Participants were followed up until death (2562, 23.3%), emigration (519, 4.7%), or 31 March 1995 (7896, 71.9%). Copies of all the death certificates were obtained, and the underlying causes of death were coded according to the International Classification of Diseases (8th revision for 1973-78 and 9th revision for 1979-95).4 5

STATISTICAL METHODS

We analysed all deaths occurring up to age 79 in participants aged 16 years or above at recruitment. Subjects with a previous cancer registration (other than ICD 173, non-melanoma skin cancer) were excluded, as were those with missing data on smoking or diet. These exclusions left 10 771 subjects. Subjects were censored on reaching the age of 80.

We estimated the numbers of deaths expected from each cause in each sex by multiplying the number of person years at risk in each of nine age groups (16-39, 40-44, 45-49…75-79) by the corresponding national mortality for England and Wales. Numbers were calculated separately for 1973-74, 1975-79, 1980-84, 1985-89, and 1990-95. We calculated standardised mortality ratios for men and women in the whole cohort as the ratio of observed to expected deaths and 95% confidence intervals assuming that the number of observed deaths had a Poisson distribution.

We examined the associations of smoking and dietary factors with risk of death by Poisson regression using GLIM-4 to calculate mortality ratios and 95% confidence intervals. The ratios were adjusted firstly for age and sex and secondly for age, sex, and current smoking. Smoking was categorised as non-smoking, pipe or cigars only, 1-14 cigarettes a day, >/=15 cigarettes a day. If smokers did not declare how much they smoked they were put in the category 1-14 cigarettes a day (n = 21; 8 men and 13 women).

The dietary variables were dichotomised as vegetarian or non-vegetarian and as daily versus less than daily consumption of wholemeal bread, bran cereals, nuts or dried fruit, fresh fruit, and raw salad. Information on weight was available for 10 223 subjects; the relation of weight with mortality was examined by dividing men and women into thirds of the weight distribution and calculating mortality ratios for the middle and top thirds relative to the lowest. Two sided P values are quoted.

Results

Table 1 shows that men and women were similar in age and in the proportion who were vegetarian and ate wholemeal bread, bran cereals, or nuts or dried fruit daily. Fewer women than men smoked and more women reported eating fresh fruit or raw salad daily.

View this table:
Table 1

Characteristics and intake of certain foods of men and women. Values are numbers (percentages) unless stated otherwise

After a mean of 16.8 years of follow up (maximum 21.3 years) mortality was substantially lower than in the general population; the standardised mortality ratio for all causes of death was 0.56 (95% confidence interval 0.53 to 0.59) for men and women combined. Table 2 shows the standardised mortality ratios for major causes of death in men and women. Among men, standardised mortality ratios were significantly below one for all malignant neoplasms, cancer of the stomach, cancer of the large intestine and rectum (subsequently referred to as colorectal cancer), cancer of the bronchus and lung, diabetes mellitus, diseases of the circulatory system, diseases of the respiratory system, diseases of the digestive system, and diseases of the genitourinary system. Among women, standardised mortality ratios were significantly below one for all malignant neoplasms, cancer of the bronchus and lung, diabetes mellitus, mental disorders, diseases of the circulatory system, diseases of the respiratory system, and diseases of the digestive system.

View this table:
Table 2

Observed and expected numbers of deaths and standardised mortality ratios (95% confidence interval) for major causes of death in men and women

ASSOCIATIONS OF MORTALITY WITH SMOKING AND SIX DIETARY FACTORS

Table 3 shows the mortality ratios associated with smoking and six dietary factors for all causes of death combined and for six cause of death categories. Smoking was associated with a 52% increase in all cause mortality, increasing to 100% in people who smoked >/=15 cigarettes a day. After smoking was adjusted for there were significant protective associations for consumption of wholemeal bread and fresh fruit. The significant reduction in mortality associated with eating fresh fruit daily remained after wholemeal bread was adjusted for (mortality ratio 0.81 (95% confidence interval 0.71 to 0.92)), but the reduction associated with eating wholemeal bread daily was not significant after adjustment for consumption of fresh fruit (0.91 (0.81 to 1.02)).

View this table:
Table 3

Mortality ratios (95% confidence interval) for smoking and six dietary factors after adjustment for age and sex and for age, sex, and smoking

For ischaemic heart disease the mortality ratio associated with smoking was 1.43 (1.11 to 1.84). After smoking was adjusted for, daily consumption of fresh fruit and raw salad were each associated with a significant reduction in mortality from ischaemic heart disease. An adjustment for raw salad increased the mortality ratio associated with daily consumption of fresh fruit to 0.82 (0.64 to 1.05), while adjustment for fresh fruit increased the ratio associated with daily consumption of raw salad to 0.77 (0.61 to 0.97). Smoking was also associated with an increased risk of death from cerebrovascular disease (1.53 (1.02 to 2.29)). After adjustment for smoking, daily consumption of fresh fruit had a significant protective effect (0.68 (0.47 to 0.98)).

Smoking was associated with an increased risk of death from all malignant neoplasms combined (1.45 (1.16 to 1.82)), but the dietary factors were not significant once smoking was adjusted for. Similarly no significant dietary associations were found for lung cancer after adjustment for smoking. Mortality from colorectal cancer was not significantly associated with smoking or dietary factors. Vegetarian diet was associated with an increased mortality from breast cancer in women (1.64 (1.01 to 2.67)), and this estimate was not altered by adjusting for smoking (which was not associated with breast cancer mortality).

ASSOCIATION OF MORTALITY WITH WEIGHT

Table 4 shows that weight was not associated with death rates for all cause mortality, cerebrovascular disease, all malignant neoplasms combined, colorectal cancer, and breast cancer after age, sex, and smoking were adjusted for. The mortality ratio in the heaviest third was raised for ischaemic heart disease (1.28 (0.97 to 1.69)) and was low for lung cancer (0.63 (0.33 to 1.23)), but these apparent trends were not significant.

View this table:
Table 4

Mortality ratios (95% confidence interval) for people in the middle and top thirds of body weight compared with those in the lowest third*

Discussion

The low standardised mortality ratio of the cohort, 0.56 for men and women combined, is similar to that in comparable cohorts: 0.46 in the Oxford Vegetarian Study,6 0.48 in a cohort of German vegetarians and health conscious people,7 and 0.56 in a cohort of Californian Seventh Day Adventists.8 The low overall mortality was mostly due to low death rates for diseases of the circulatory system, diseases of the respiratory system, and cancer of the bronchus and lung compared with the general population. This is probably mainly accounted for by the low proportion of smokers in the cohort (19% smokers overall).

LIMITATIONS

We followed the cohort for a mean of 17 years. The long follow up has the advantages of yielding a large number of deaths and of ensuring that dietary habits were recorded, on average, long before the onset of symptoms of the diseases studied. The disadvantage is that dietary habits will have changed during the follow up. The validation study suggested that subjects' diets did change during the first few years of follow up in respect to vegetarian diet and consumption of wholemeal bread, but these changes are a combination of real change plus imperfect repeatability and therefore overestimate the extent of change. Changes in dietary habits would be expected to result in underestimation of any associations found.

Another limitation is that the questionnaire was short and did not include several important food groups (for example, dairy products, fish, alcoholic drinks), did not allow us to estimate energy intake, and did not include other factors known to be associated with health (exercise, socioeconomic status, past smoking habits). We were therefore unable to explore whether the significant associations observed were partly due to confounding by other dietary or non-dietary variables.

ASSOCIATIONS OF DIETARY FACTORS WITH MORTALITY

This study was initially set up to test the hypotheses that daily consumption of wholemeal bread (as an indicator of a high fibre diet) and vegetarian diet are associated with a reduction in mortality from ischaemic heart disease; the reduction in mortality associated with both of these dietary factors was not significant. We did, however, find that daily consumption of fresh fruit was associated with a significant reduction in mortality from ischaemic heart disease (24%), cerebrovascular disease (32%), and all causes of death combined (21%), and was associated with non-significant reductions in mortality from all the other cause of death examined.

Fruit was included on the questionnaire as a possible indicator of a high fibre diet but was not part of the original hypothesis, so caution should be applied in looking at these results in isolation. Nevertheless, our findings are broadly consistent with the results of several other studies. For all cause mortality, Kahn et al reported odds ratios of 0.89 and 0.72 for frequent versus infrequent consumption of fruit or fruit juice and green salad respectively,9 Pandey et al reported a 31% reduction with a high intake of foods rich in vitamin C and in β carotene,10 and Enstrom et al reported a 23% reduction associated with high vitamin C consumption.11 Fruit is an important source of vitamin C. Fruits and vegetables,12 13 carotene containing fruit and vegetables,14 apples,15 and foods rich in vitamin C and β carotene10 have all been reported to protect against ischaemic heart disease. Fruit and vegetables have also been reported to protect against stroke,16 17 as have potassium18 and vitamin C19—nutrients for which fruit is an important source.

We found that a vegetarian diet was associated with a 15% reduction in mortality from ischaemic heart disease. This was not significant and was less than the roughly 30% reductions reported in earlier analyses of this cohort,2 3 other British vegetarians,6 and Californian vegetarians.20 The 0.61 mmol/l lower average total plasma cholesterol concentration reported among vegetarians than non-vegetarians in a sample of this cohort21 might be expected to result in a 24% lower mortality from coronary heart disease.22 Any protective effect may have been attenuated by crossover between the vegetarian and non-vegetarian groups.

A vegetarian diet was also associated with a significant increase in mortality from breast cancer. However, the confidence interval was wide, and the result might be due to chance, perhaps combined with differences in parity. In the Oxford Vegetarian Study, for example, 37% of vegetarian women aged 40 and above were nulliparous, compared with 28% of meat eating women (unpublished data). Other studies have reported no association between vegetarian diet and risk of breast cancer or mortality.23 24 25

Of the other associations examined, only two were significant. Daily consumption of wholemeal bread was associated with a 12% reduction in all cause mortality, but the protection was much less than that for fresh fruit (21%) and became non-significant after fruit was adjusted for. Daily consumption of raw salad was associated with a 26% reduction in mortality from ischaemic heart disease, slightly greater than that for fresh fruit (24%). After each of these variables was adjusted for the other, salad was more closely associated with mortality, but the reduction in risk associated with fruit remained substantial (18%) and both these foods may have a protective effect.

The numbers of deaths for individual cancer sites were small and the mortality ratios have wide confidence intervals. The 41% reduction in mortality from lung cancer associated with daily consumption of fresh fruit was not significant but is consistent with previous studies.26 Perhaps surprisingly, none of the dietary variables was significantly associated with mortality from colorectal cancer.27

We found no significant association between weight and mortality, perhaps because of the low numbers of obese subjects. Only 3% of men were heavier than 91 kg and 2.5% of women heavier than 78 kg, the weights at which men and women of average height are considered obese.28

We thank the participants in this study, the staff of the NHS registers and the Office of Population Censuses and Surveys for tracing the subjects, and Drs Valerie Beral and Gary Fraser for commenting on the text.

Funding: Medical Research Council and Imperial Cancer Research Fund.

Conflict of interest: TJAK and PNA are members of the Vegetarian Society.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.

THIS WEEK'S POLL