Dietary pattern and 20 year mortality in elderly men in Finland, Italy, and the Netherlands: longitudinal cohort studyBMJ 1997; 315 doi: http://dx.doi.org/10.1136/bmj.315.7099.13 (Published 05 July 1997) Cite this as: BMJ 1997;315:13
- Patricia Huijbregts, research fellowa,
- Edith Feskens, senior epidemiologista,
- Leena Räsänen, assistant professor of nutritionb,
- Flaminio Fidanza, professor of nutritionc,
- Aulikki Nissinen, professor of community health and general practiced,
- Alessandro Menotti, professor of epidemiologye,
- Daan Kromhout, professor of public health researcha
- a Department of Chronic Diseases and Environmental Epidemiology, National Institute of Public Health and the Environment, PO Box 1, NL-3720 BA Bilthoven, Netherlands
- b Division of Nutrition, Department of Applied Chemistry and Microbiology, University of Helsinki, 00014 Helsinki, Finland
- c Institute of Food Sciences and Nutrition, University of Perugia, Perugia, 00610 Italy
- d Department of Community Health and General Practice, University of Kuopio, 70211 Kuopio, Finland
- e Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, MN 55454-1015, USA
- Correspondence to: Ms Huijbregts
- Accepted 15 April 1977
Objective: To investigate the association of dietary pattern and mortality in international data.
Design: Cohort study with 20 years' follow up of mortality.
Setting: Five cohorts in Finland, the Netherlands, and Italy.
Subjects: Population based random sample of 3045 men aged 50-70 years in 1970.
Main outcome measures: Food intake was estimated using a cross check dietary history. In this dietary survey method, the usual food consumption pattern in the 6-12 months is estimated. A healthy diet indicator was calculated for the dietary pattern, using the World Health Organisation's guidelines for the prevention of chronic diseases. Vital status was verified after 20 years of follow up, and death rates were calculated.
Results: Dietary intake varied greatly in 1970 between the three countries. In Finland and the Netherlands the intake of saturated fatty acids and cholesterol was high and the intake of alcohol was low; in Italy the opposite was observed. In total 1796 men (59%) died during 20 years of follow up. The healthy diet indicator was inversely associated with mortality (P for trend <0.05). After adjustment for age, smoking, and alcohol consumption, the relative risk in the group with the healthiest diet indicator compared with the group with the least healthy was 0.87 (95% confidence interval 0.77 to 0.98). Estimated relative risks were essentially similar within each country.
Conclusions: Dietary intake of men aged 50-70 is associated with a 20 year, all cause mortality in different cultures. The healthy diet indicator is useful in evaluating the relation of mortality to dietary patterns.
Studying dietary patterns instead of single nutrients in relation to mortality takes into account the intercorrelation of nutrients in the diet
A healthy diet, as measured by an indicator based on WHO recommendations, is associated with a reduction of 13% after 20 years in all cause mortality for men aged 50-70
The dietary pattern as a whole is more important than specific dietary components with respect to survival among older people
The WHO dietary recommendations for the prevention of chronic diseases seem to be effective
The healthy diet indicator is useful for evaluating the relation of dietary patterns and mortality in a cross cultural setting
Selected nutrients or food groups have often been studied in relation to mortality,1 2 3 4 5 but studying dietary patterns in relation to mortality has several advantages. Using dietary patterns takes into account the high intercorrelation of nutrients within the diet, which is due to the choice of foods in which these nutrients occur or to the consumption of a particular food at the expense of another one.6 Also, studying dietary patterns in relation to mortality provides a practical way to evaluate the health effects of adherence to dietary guidelines by individuals.7 It can help to identify groups with an unhealthy dietary pattern and disclose possibilities for prevention of chronic diseases or disability.
To our knowledge, dietary pattern has been used to study the association with mortality only within individual countries.7 8 9 10 We developed a healthy diet indicator, based on the World Health Organisation's dietary recommendations for the prevention of chronic diseases,11 and investigated its association with all cause mortality during 20 years of follow up in population based samples from three different countries (Finland, Italy, and the Netherlands).
From 1958 to 1964, 16 population samples of men aged 40-59 years from seven countries were enrolled and examined at baseline for the seven countries study.12 Since 1984, Finland, Italy, and the Netherlands have continued the examination of their cohorts, focusing on health and its determinants in elderly people. The population has been described in detail.13 The participating cohorts were eastern and western Finland; Crevalcore and Montegiorgio, Italy; and Zutphen, the Netherlands.
The study was started in 1959 in Finland and in 1960 in Italy and the Netherlands. Baseline dietary information used in this study was gathered in 1969 in the Finnish cohorts and in 1970 in the Dutch cohort and one Italian cohort (Crevalcore). As the 1970 data from Montegiorgio were available only on a non-random subset of men, the dietary data gathered in 1965 from the men who were still alive in 1970 were used as an approximation for dietary intake in 1970. This is justifiable because dietary intake in Finland, Italy, and the Netherlands was relatively stable within the cohorts during the first 10 years of the seven countries study.12
Information on diet around 1970 was available for 612 men in eastern Finland, 694 men in western Finland, 615 men in Zutphen, and 592 men in Crevalcore. We had dietary information in 1965 for 662 men in Montegiorgio. The analyses in this study are based on subjects for whom complete information on diet and confounding variables was available at baseline (1969-70): 606 men from eastern Finland (89%), 683 men in western Finland (91%), 608 men in Zutphen (79%), 591 men in Crevalcore (69%), and 557 men in Montegiorgio (85%)
Collection of information
Food intake around 1970 was estimated with the cross check dietary history method, adapted to the local situation and carried out by experienced dietitians and nutritionists. This method provides information about the usual food consumption pattern in the 6-12 months preceding the interview.14 15 16 17 18 The usual food consumption pattern of the subject during weekdays and weekends was assessed by recording foods eaten at breakfast, lunch, and dinner, and between meals. This information was checked by registering from a comprehensive food list the frequency and amount of foods consumed and, in the case of discrepancies, discussing the recorded food consumption pattern with the participant. This method has satisfactory reproducibility.19 20 In Finland the interviews were carried out in autumn and in the Netherlands and Italy in spring. Nutrient intakes were assessed by using computerised versions of food tables for each country.16 18 21
Information about cigarette smoking was collected by a standardised questionnaire. Subjects were classified as men who had never smoked, men who had stopped smoking, and current smokers. Men who never drank alcohol were classified as abstainers. On the basis of mortality in Italian men, moderate drinkers were categorised as men who consumed up to six glasses of alcohol a day. Those who drank more than six glasses a day were categorised as heavy drinkers.22
The men were followed for mortality for 20 years. None of the men was lost to follow up. The underlying cause of death was coded in a standardised way by one reviewer (AM), using the eighth revision of the international classification of diseases (ICD-8). The cause of death was based on information from the official death certificate, in combination with information from medical and hospital records. In Finland, after the fifth year of follow up only death certificates were available for assigning the cause of death. The coder of the causes of death was blind to the risk factor status of the subject. In the case of multiple causes of death, priority was given to accidents, followed by cancer in advanced stages, coronary heart disease, and stroke. For the present analyses cardiovascular disease was defined as ICD-8 codes 390-450 and A795 (code A795 was a special choice of the study group, identifying sudden death of probable coronary origin, occurring within two hours of the onset of symptoms). Cancer was defined as ICD-8 codes 140-239. Other causes included all other deaths not covered in these rubrics.
The healthy diet indicator was calculated by using the dietary guidelines for the prevention of chronic diseases, defined by the WHO.11 A dichotomous variable was generated for each food group or nutrient that was included in these guidelines (table 1). If a person's intake was within the recommended range this variable was coded as 1; otherwise it was coded as 0. The healthy diet indicator was the sum of all these dichotomous variables.
To avoid overlap, total fat and total carbohydrates were omitted in the calculation of the healthy diet indicator. Salt was not included because only information about the sodium content in foods was available and it was not known how much salt was added during preparation of meals and at the table. We used the variable “monosaccharides and disaccharides” instead of free sugars because the free sugars variable was not comparable between the countries. The use of monosaccharides and disaccharides has overestimated the actual intake of free sugars, especially in Finland, where the intake of milk products and therefore of lactose is high. Alcohol consumption in Italy was higher than in the two other countries, so for all three countries, before the intake of macronutrients was entered in the healthy diet indicator it was calculated as a percentage of energy intake without the energy provided by alcohol.
Both in the pooled populations and in each country separately, participants were divided into thirds (low, medium, and high) according to their healthy diet indicator. In Finland and the Netherlands, and also in the pooled populations, cut off values were <2, 2, and >2. In Italy the cut off values were <3, 3-4, and >4.
Differences in baseline characteristics between cohorts were tested with analysis of variance, using Scheffé's test for multiple comparisons. Frequencies of categorical variables in the different cohorts were compared by the χ2 test. Cox's proportional hazards survival analysis was used to investigate the relation between healthy diet indicator groups and mortality in the total study population after a model including the interaction between healthy diet indicator and country had been tested. Adjustments were made for age at baseline, cigarette smoking, and alcohol consumption. All confounders except age at baseline were entered in the model as dummy variables. Analyses were repeated for each country separately. All P values were based on two sided tests, and a P value of 0.05 was considered to be significant. The sas statistical analysis computer package (version 6.10) was used.
Out of the total study population of 3045 men, 1796 men (59%) died during 20 years of follow up (table 2). Mortality was highest in eastern Finland and lowest in Montegiorgio. Mean age at baseline varied between 58 and 60 years and was significantly lower in eastern Finland than in western Finland, Zutphen, and Crevalcore. The percentage of smokers varied from 42% to 57%, and alcohol intake varied greatly among the cohorts.
Dietary intake around 1970 varied greatly between the cohorts (table 3). Median energy intake ranged from 11 MJ to 15 MJ, and nutrient intake varied greatly: for example, median saturated fatty acid intake was 22% in Finland and 9% in Montegiorgio, Italy. The minimum healthy diet indicator was 0; the maximum healthy diet indicator varied from 5 in Finland to 8 in Italy. The healthy diet indicator was positively associated with alcohol intake (r=0.42, P=0.0001). A significant but small inverse association (r=-0.05, P=0.005) with cigarette smoking was found.
The healthy diet indicator was inversely related to all cause mortality in the pooled population analyses (table 4). The crude analysis showed a 15% lower risk of deaths from all causes in the group with the highest healthy diet indicator. After adjustment for age, cigarette smoking, and alcohol consumption the all cause mortality risk was 13% lower. In an additional analysis the interaction between healthy diet indicator and country was not significant (P>0.20); we therefore considered it appropriate to do a pooled analysis. In separate analyses, the healthy diet indicator and all cause mortality were inversely associated in each country; the association was similar to that in the pooled populations but was not significant.
Although the relative risks for all cause mortality were within the same order of magnitude for the three countries, the absolute mortality was highest in Finland and lowest in Italy (fig 1). Men with the healthiest diet in Finland had higher mortality than the men with an unhealthy diet in Italy, but absolute risks in the Netherlands and Italy were similar.
Overall, after adjustment for confounders, the group with the highest healthy diet indicator had an 18% lower risk of death from cardiovascular disease than the group with the lowest healthy diet indicator (P for trend <0.05). Risk of death from cancer was 15% lower in the highest group than in the lowest group (P for trend=0.13).
This study shows that 20 year mortality is lowest in men with the healthiest diet according to WHO recommendations. After adjustment for age, cigarette smoking, and alcohol consumption, the group with the highest healthy diet indicator had a 13% lower risk of death than the group with the lowest. The healthy diet indicator had an even stronger inverse association with mortality from cardiovascular diseases.
Several other studies have examined dietary patterns or a combination of nutrients instead of single nutrients or dietary components, but these were all done within single countries.7 8 9 10 23 24 Trichopoulou and coworkers recently used an approach similar to ours. They assessed the influence of a specific dietary pattern on overall survival among 182 elderly residents of three rural Greek villages. A diet score was calculated on the basis of eight component characteristics of the traditional Mediterranean region; an increase of one unit was associated with a 17% reduction in overall mortality.8 Nube and coworkers investigated the effect of a dietary score on longevity among 2820 middle aged Dutch civil servants. Using the recall of the frequency of intake of particular foods instead of an extensive dietary history, they found a significant linear trend for 25 year survival from highest (healthiest) to lowest scores.7 Our results confirm these national results.
Since dietary patterns are highly determined by cultural influences (for example, the Mediterranean dietary pattern25 26 27), we did not adjust for country in the pooled population analyses. Country has a strong cultural component which is responsible for (part of) the variation in dietary patterns. Adjustment for this variable would result in an overcorrection and hence an underestimation of the true association between the quality of the diet and mortality. When the countries were analysed separately, the associations between the healthy diet indicator and all cause mortality were essentially the same, although they no longer reached significance. This was due to a low statistical power resulting from the smaller numbers of subjects within a country.
In our analyses we assumed that the diet around 1970 is indicative of diet between 1970 and 1990. The general dietary patterns in the different cultures can still be recognised after 20 years, but the differences between them have become smaller.13 This may have resulted in attenuation of the association.
Healthy diet indicator
To assess whether one of the components of the healthy diet indicator could be responsible for the observed association with mortality, we analysed the same models for each of the components of the healthy diet indicator separately. For most of the components the association was not significant (data not shown), and different components were responsible for the association in different countries. Trichopoulou and coworkers, too, found significant results for only one of the individual components.8 We therefore concluded that the dietary pattern as a whole, as reflected in the healthy diet indicator, was responsible for the observed association.
Though some of the criteria of the healthy diet indicator overlap and some of the variables may be U shaped, we found a significant inverse association with 20 year mortality. The healthy diet indicator will be refined in future studies.
Alcohol intake and energy intake
Alcohol intake varied greatly among the cohorts. In Italy mean alcohol intake was on average 10 times higher than in Finland. Giving macronutrient intakes within total energy intake would be misleading, since the high alcohol intake in the Italian cohorts would dilute macronutrient intake in comparison to that in Finland and the Netherlands. To separate the effect of alcohol from that of the healthy diet indicator, in each of the countries we calculated macronutrient intake relative to energy intake excluding energy from alcohol. We regarded alcohol intake as a possible confounder in the association of the healthy diet indicator with mortality. The high cut off for moderate alcohol consumption was chosen for practical reasons. Although these categories were rather crude, residual confounding would not have resulted in large artefacts since inverse associations between alcohol intake and mortality have been observed for up to six glasses per day.22
To adjust for physical activity in the survival analyses, we calculated energy intake per kilogram of body weight as an approximation.28 The results were essentially the same as without the adjustment (data not shown). Since no other reliable measure of physical activity was available we could not adjust for it in the present study.
Besides all cause mortality, we were also able to investigate mortality from cardiovascular diseases and cancer, the most important causes of premature death in developed industrialised countries.11 There is abundant evidence that diet affects these diseases, and this was recognised in the WHO recommendations for the prevention of chronic diseases. Our results show that these recommendations may be effective. The healthy diet indicator, which was based on the WHO recommendations, was associated not only with reduced all cause mortality but with an even stronger reduced mortality from cardiovascular diseases.
We are indebted to the many people who were involved in this longitudinal study: the men who took part in the surveys; Dr S Giampaoli, who organised the late part of the field work in Italy; Professor M Pekkarinen, who was responsible for the data collection in Finland in 1969; and the fieldwork team in Zutphen, especially Drs EB Bosschieter and B Bloemberg, who helped prepare a common dataset.
Funding: National Institute of Aging, Bethesda, USA; Netherlands Prevention Foundation (Praeventiefonds), The Hague, Netherlands; NATO Research grant (A Menotti).
Conflict of interest: None.