Twenty years ago, Finland had the highest coronary heart disease-related mortality rate in the world. Within Finland, the highest mortality rate was seen in the eastern part of the country, particularly in the North Karelia province [1, 3]. These observations led to the planning and launching of the North Karelia Project in 1972 [12]. The original goal of the Project (the first of its kind) was to test the feasibility and effects of a community-based program for preventing cardiovascular diseases by changing general lifestyle and risk factors in North Karelia. The main intermediate objective was to reduce the prevalence of smoking, high cholesterol levels, and high blood pressure in the entire population, with special emphasis on middle-aged men. The activities initially used to prevent cardiovascular diseases in North Karelia were gradually implemented in other parts of Finland. In the past 10 years, the North Karelia Project has been a national model for new innovations in chronic disease prevention and health promotion [13].

The effects of the North Karelia Project have been evaluated every 5 years, beginning with the baseline survey in 1972, using large and representative cross-sectional population samples [12, 14]. In 1982, the evaluation protocol was adapted so that it would comply with the risk factor survey protocol of the multinational MONICA (monitoring trends and determinants in cardiovascular disease) program initiated by the World Health Organization [15]. In 1992, after the fifth consecutive risk factor survey had been completed, we had a unique database on cardiovascular risk factors and health behavior in two provinces in eastern Finland: North Karelia and the neighboring Kuopio. The main results of these surveys, including comparisons of risk factor trends between the two provinces, have been published [14]. Cholesterol levels decreased markedly in both provinces between 1972 and 1992. The decrease was steeper in North Karelia between 1972 and 1977, but since then the extent of the decrease has been similar in both areas.

In this analysis, we combine the data on the two provinces. We analyze the dynamics in serum cholesterol levels in relation to age, sex, birth cohort, time period, and mortality rate and to the change in dietary intake of saturated fat among the middle-aged population in eastern Finland.

Methods

Top Methods Results Discussion Author & Article Info References

To assess the levels of risk factors for cardiovascular disease in North Karelia and Kuopio, an independent cross-sectional population survey was done in each of the following years: 1972, 1977, 1982, 1987, and 1992 [12, 14]. In 1972 and 1977, a sample of 6.6% of the population born between 1913 and 1947 was randomly selected from both provinces. In 1977, an additional sample of 6.6% of the population born between 1948 and 1952 was randomly drawn from North Karelia only. In 1982, 1987, and 1992, the sample included persons 25 to 64 years of age; the samples were stratified according to the MONICA protocol so that at least 250 persons of each sex and in each 10-year age group were chosen from each province [15]. In the 1970s, the population of 25- to 64-year-old persons in eastern Finland was equally distributed to each 10-year age group. Thus, the effect of this change in the sampling procedure on the results is minimal. The study was conducted with five new independent, randomly selected cross-sectional samples of participants. The total sample size of the five surveys was 41 478. The participation rate was more than 90% in the first survey and was satisfactory in the later surveys (Table 1). Our present analysis includes the persons in the 25- to 59-year-old age group in the 1972 survey and the persons in the 25- to 64-year-old age group in the later surveys. The 1977 estimates for the 25- to 29-year-old age group are based on data from North Karelia only. We excluded 793 men and 837 women from our analysis because data on serum cholesterol level or intake of saturated fat were missing. Persons in the first birth cohort were born in 1913, and persons in the last birth cohort were born in 1967. Thus, the analyses include 11 five-year birth cohorts (Table 2). Persons in four 5-year cohorts (who were born between 1928 and 1947) were followed throughout the entire 20-year period.

Our survey methods followed those used in the MONICA protocol in 1982, 1987, and 1992 and are similar to the methods used in 1972 and 1977 [14, 15]. The surveys included a self-administered questionnaire about health behavior and other related factors. The questionnaires were returned to the survey site, where nurses checked them to ensure that they were complete, asked additional questions, measured blood pressure and other variables, and obtained a venous blood specimen for measurement of cholesterol levels.

In 1972 and 1977, serum cholesterol levels in frozen samples were measured using the Lieberman-Burchard reaction. In 1982, 1987, and 1992, an enzymatic assay method was used (cholesterol oxidase-phenol aminophenazone peroxidase, Boehringer Mannheim, Mannheim, Germany). In our laboratory, values were 2.4% lower with the enzymatic assay method than with the Lieberman-Burchard method. Cholesterol values from 1972 and 1977 were therefore corrected by 2.4%. Cholesterol levels were measured in the same central laboratory, and the laboratory data were standardized against samples from international reference laboratories.

Change in dietary intake of saturated fat—including change in intake of two major sources of saturated fat among the Finnish population: liquid dairy products and spreadable fat (for example, butter)—was assessed by using an index. The index was based on questions about the type and amount of liquid dairy products consumed daily and questions about the type of fat used on bread. The amount of fat spread on bread was determined by showing study participants models of bread slices on which different amounts of fat were spread. Participants were asked which slice represented the amount of fat they usually used on bread and how many slices of bread were usually consumed daily. Data on the saturated fat content of liquid dairy products and spreadable fat available in Finland were obtained from nutritional information listed on product labels. Daily intake of saturated fat from these two sources was then calculated on the basis of this information.

To assess the potential effect of selective mortality, cholesterol levels and trends in persons who died during follow-up were compared with those in persons who survived for the entire follow-up period. Data on the participants' mortality in the study cohorts were obtained from the national mortality register of the Central Statistical Office of Finland. Each participant was followed until the last risk factor survey in his or her birth cohort was administered. Four hundred fifty women and 1326 men died during follow-up.

The dynamics of cholesterol levels in the population are presented graphically using sex, age, birth cohort, and time period as explanatory variables. We used linear regression analysis to test the statistical significance of the association between explanatory variables and the levels of and trends in serum cholesterol levels and saturated fat intake. Variable estimates are presented for each 5-year increase in age, 5-year period, and 5-year difference in birth cohort. Estimates are based on models that included only the main effects. First-level interactions between main effects were included in the model together with the main effects; if the interaction was statistically significant at a 95% level, a stratified analysis was done. Regression models, variable estimates, and the significance levels of these estimates are given in the footnotes of the tables and in the Figure legends. We used SAS statistical software (SAS Institute, Cary, North Carolina) for all statistical analyses [16].

Results

Top Methods Results Discussion Author & Article Info References

In both sexes, mean cholesterol levels decreased between 1972 and 1982. This trend leveled off between 1982 and 1987, but levels began to decrease again from 1987 to 1992 (Figure 1). Levels decreased faster in women than in men. In every survey, cholesterol values in the youngest age group (25 to 29 years of age) were lower in women than in men (Figure 2). The relation of age to cholesterol level differed between men and women. In the cross-sectional analyses, cholesterol levels in men increased steadily from the youngest age group to the 45- to 49-year-old age group. In the latter group, the increase leveled off. In women, cholesterol levels continued to increase with age; the increase was steepest between 40 and 54 years of age. When women were approximately 50 years of age, their cholesterol values were as high as those of men.

View larger version (19K): [in this window] [in a new window]   Figure 1. Serum cholesterol levels in men and women according to age group and study year. The following are the variable estimates of the linear regression analysis. Model 1 (sex, age, study year, interaction between sex and study year): sex, –0.13(P < 0.001); age, 0.22 (P < 0.001); study year, –0.25(P < 0.001); interaction between sex and study year, P < 0.001. Model 2 (age, study year, interaction between age and study year): in men: age, 0.15 (P < 0.001); study year, –0.22(P < 0.001); interaction between age and study year, P = 0.08; in women: age, 0.28 (P < 0.001); study year, –0.28(P < 0.001); interaction between age and study year, P > 0.2.

View larger version (17K): [in this window] [in a new window]   Figure 2. Serum cholesterol levels in men and women according to age group and study year. The following are the variable estimates of the linear regression analysis. Model 1 (sex, age, study year, interaction between sex and age): sex, –0.13(P < 0.001); age, 0.22 (P < 0.01); study year, –0.25(P < 0.001); interaction between sex and age, P < 0.001. Model 2 (age, study year, interaction between age groups): in men: age, 0.15 (P < 0.001); study year, –0.22(P < 0.001); interaction between age groups, P < 0.001; in women: age, 0.28 (P < 0.001); study year, –0.28(P < 0.001); interaction between ages, P < 0.001. Model 3 (age, study year): in men by age group—25 to 44 years of age: age, 0.26 (P < 0.001); study year, –0.22(P < 0.001); 45 to 64 years of age: age, –0.01(P > 0.2); study year, –0.21(P < 0.001); in women by age group—25 to 44 years of age: age, 0.19 (P < 0.001); study year, –0.26(P < 0.001); 45 to 64 years of age: age, 0.29 (P < 0.001); study year, –0.29(P < 0.001).

In men, the analysis by birth cohorts showed that cholesterol levels decreased with age in the cohorts whose members were born between 1913 and 1937 (Figure 3). Cholesterol levels did not change in the cohort whose members were born between 1938 and 1942. In younger cohorts (comprising members born between 1943 and 1962), cholesterol levels increased slightly with age during follow-up; most of this increase occurred during the first 5 years of follow-up. In men, cholesterol levels changed little between 30 and 45 years of age. In all birth cohorts that were followed long enough, cholesterol levels began to decrease between the ages of 45 and 49 years. In the youngest birth cohorts, which were entering the surveys for the first time (that is, the cohort of persons 25 to 29 years of age), cholesterol levels were markedly lower than those in the same age group of the previous surveys.

View larger version (18K): [in this window] [in a new window]   Figure 3. Serum cholesterol levels in men and women according to age group and birth cohort. The following are the variable estimates of the linear regression analysis. Model 1 (sex, age, birth cohort, interaction between sex and birth cohort): sex, –0.13(P < 0.001); age, –0.04(P < 0.001); birth cohort, –0.25(P < 0.001); interaction between sex and birth cohort, P < 0.001. Model 2 (age, birth cohort, interaction between age and birth cohort): in men: age, –0.08(P < 0.001); birth cohort, –0.22(P < 0.001); interaction between age and birth cohort, P < 0.001; in women: age, 0.00 (P > 0.2); birth cohort, –0.28(P < 0.001); interaction between age and birth cohort, P < 0.001. Model 3 (age): in men by birth cohorts: 1913 to 1917, –0.24(P < 0.01); 1918 to 1922, –0.29(P < 0.001); 1923 to 1927, –0.20(P < 0.001); 1928 to 1932, –0.13(P < 0.001); 1933 to 1937, –0.10(P < 0.001); 1938 to 1942, –0.03(P > 0.2); 1943 to 1947, 0.04 (P < 0.05); 1948 to 1952, 0.11 (P < 0.001); 1953 to 1957, 0.13 (P < 0.01); 1958 to 1962, 0.02 (P > 0.2); in women by birth cohorts: 1913 to 1917, –0.16(P < 0.05); 1918 to 1922, –0.10(P < 0.01); 1923 to 1927, 0.06 (P < 0.05); 1928 to 1932, 0.03 (P = 0.11); 1933 to 1937, 0.03 (P = 0.11); 1938 to 1942, –0.02(P > 0.2); 1943 to 1947, –0.07(P < 0.001); 1948 to 1952, –0.05(P = 0.08); 1953 to 1957, 0.01 (P > 0.2); 1958 to 1962, –0.05(P > 0.2).

In women, cholesterol levels tended to decrease in both the oldest and the youngest birth cohorts and to increase in the cohorts whose members were born between 1923 and 1937. In most birth cohorts, however, the overall change was not statistically significant. Among women, cholesterol levels in most cohorts decreased with age between 25 and 40 to 45 years of age. In different birth cohorts at different ages, this decrease changed to an increase between 1982 and 1987. As was seen with men, the youngest birth cohorts of women in each survey year also had markedly lower cholesterol levels than did the same age group in the previous surveys. In every survey year, the new birth cohorts of women had lower cholesterol levels than did the corresponding birth cohorts of men.

Mean cholesterol levels were higher in persons who died during follow-up than in persons who survived for the entire follow-up period (Table 3). The difference was most prominent in middle-aged men who belonged to a birth cohort that was followed for 10 to 15 years. The difference in mean cholesterol levels between survivors and nonsurvivors was smaller, or even inverse, among the oldest and the youngest persons who were followed for only 5 years, particularly women.

In both sexes, daily intake of saturated fat from the use of liquid dairy products and spreadable fat decreased markedly between 1972 and 1992: from 45 g/d to 16 g/d in men and from 25 g/d to 8 g/d in women (Table 4). Intake of saturated fat from these sources was slightly associated with age but was strongly related to the time period. According to the cross-sectional data, the daily intake of saturated fat slightly increased with age, but the association with age was weaker than the association with time period. In each birth cohort, intake of saturated fat from liquid dairy products and spreadable fat decreased markedly with age.

Discussion

Top Methods Results Discussion Author & Article Info References

Our study represents the first database sufficient for the analysis of changes in serum cholesterol levels associated with both time period and birth cohort. Our cross-sectional data showed that with time, cholesterol levels decreased in a fairly continuous manner in all age groups and in both sexes. However, cholesterol levels also increased with age. A birth cohort analysis provides useful information that can increase understanding of these dynamics.

Contrary to the cross-sectional data, cholesterol levels in any one birth cohort did not increase with age. In birth cohorts of men, cholesterol levels changed little between 30 and 45 years of age, after which they started to decrease. In most birth cohorts of women, cholesterol levels were almost the same at the beginning and end of follow-up. We found that the cholesterol levels of the youngest birth cohorts entering each survey were lower than levels in the youngest cohorts in the previous surveys. The observed decrease in cholesterol levels among middle-aged men in eastern Finland in the past 20 years can largely be explained by two factors: 1) a smaller increase in cholesterol levels with age before age 30 years and 2) the real decrease in cholesterol levels seen after age 45 years. In middle-aged women, the cohort effect alone explained the observed decrease in cholesterol levels. Differences in cholesterol levels between men and women (that is, higher mean values in men) were already determined before 30 years of age.

Our knowledge of the natural dynamics of cholesterol during the human life span is primarily based on cross-sectional analyses or repeated measurements in the same persons. Many cross-sectional surveys have reported an association among cholesterol level, age, and sex that is similar to our results [17]. It has been debated whether this often-seen increase in serum cholesterol levels with age is a physiologic phenomenon or an adverse effect of western lifestyle and diet [18]. Cross-sectional assessments of trend have shown that cholesterol levels have decreased in the past 20 years in many western European countries, the United States, and Canada but may have increased in Asian and developing countries [15, 19, 20]. These different trends are usually explained by the new health-oriented lifestyle of well-off populations in the wealthiest countries and the urbanization and westernization of lifestyles and diet in many populations outside of Europe and the United States. As was seen in other studies, our cross-sectional analysis indicated that the relation of age to cholesterol level differed between men and women. In men, the increase in cholesterol levels started earlier and was steeper in early adulthood than it was in women; the increase then leveled off around age 50 years. In women, the increase in cholesterol levels is usually steepest during and after menopause [21].

In our study, cholesterol levels in persons from eastern Finland who were 25 years of age in 1992 were about 1 mmol/L lower than levels in the same age group in 1972. We can assume that cholesterol levels during childhood and adolescence have also decreased similarly in the youngest birth cohorts and the older ones. A 6-year follow-up of Finns who were 3 to 18 years of age showed that cholesterol levels decreased by about 1% per year and that the greatest decrease was seen in eastern Finland [22]. It has been estimated that the total decrease of cholesterol levels in Finnish adolescents has been about 1 mmol/L since the early 1970s; this finding is in agreement with our results.

In the past 20 years, the rate of death from coronary heart disease in eastern Finland has decreased to less than half its previous level [23]. About half of that observed decrease can be explained by the decrease in cholesterol levels in the population. Atherosclerosis is a cumulative process that starts at a fairly young age [24]; thus, the contribution of decrease in serum cholesterol levels to the decrease in the rate of death from coronary heart disease may be even larger than was estimated on the basis of the measurement of cholesterol levels at middle age.

Contrary to our results, investigators for the Seven Countries Study [25] found that cholesterol levels increased in most of the 16 study cohorts (including the 2 Finnish cohorts), which had been selected from different populations in a 10-year follow-up of men who were 40 to 59 years of age at study entry. Cholesterol levels decreased only in two Japanese cohorts. Repeated measurements of cholesterol levels in 65- to 95-year-old persons participating in the Florida Geriatric Research Program [26] showed that cholesterol levels started to decrease only after age 75 years. Another result contrary to ours was the finding that cholesterol levels were higher in younger birth cohorts than in older cohorts.

Differences between our results and the results of other surveys might be explained by period-related environmental factors such as lifestyle and diet: These factors have changed more in the past 20 years in eastern Finland than in most other countries. Thus, although metabolism of cholesterol may differ among populations, our data are consistent with the hypothesis that much of the natural course of serum cholesterol levels during the human life span can be determined and therefore modified by diet and dietary interventions.

High cholesterol levels were associated with total mortality, particularly among men. In women, this association was not as constant. Because the age of the participants at the end of follow-up was 30 to 64 years, relatively few women died. In addition, few participants died of coronary heart disease. In women, death from coronary heart disease occurs about 10 years later than in men, and the possible contribution of high cholesterol level to death is also expected to be seen later in women than in men. The association was most prominent in our middle-aged cohorts, which were followed for a relatively long time. In the youngest cohorts and in the cohorts that were followed for only 5 years, the difference was smaller or even inverse. This latter finding can be explained by the decrease of serum cholesterol levels due to the clinical or preclinical disease that preceded death [27]. In the youngest cohorts, the number of deaths was also too small to allow any definitive conclusions to be drawn. In the older cohorts with a short follow-up period, selective mortality in persons who had high cholesterol levels before the baseline survey may have enriched the respective study cohort in which participants had relatively low cholesterol values. In our analysis, we compared survivors with persons who died of any cause and showed the association between serum cholesterol level and total mortality. If we had used cardiovascular-related mortality instead of total mortality, the difference in mean cholesterol levels between survivors and nonsurvivors would have been more marked. Even though death during follow-up was selective in relation to serum cholesterol levels, it did not have a remarkable effect on the observed dynamics of cholesterol levels in the population because the relative contribution of nonsurvivors to the overall analysis was small.

In Finland, intake of saturated fats used to be very high and that of polyunsaturated fats was very low. In the late 1960s, the total intake of saturated fat among a sample of the population in central Finland was about 77 g/d in men and 55 g/d in women [28]. Despite the lack of data in eastern Finland, the intake in this part of the country may have been even higher. Of the total intake of saturated fat, approximately 58% in men and 44% in women came from liquid dairy products and spreadable fat. In 1982, the total intake of saturated fat in eastern Finland had decreased to 62 g/d in men and 44 g/d in women; in 1992, the total daily intakes were 46 g/d and 32 g/d, respectively [29, 30]. According to our data and available reference data, intake of saturated fat from sources other than liquid dairy products and spreadable fat in men was approximately 32 g/d in the late 1960s and 30 g/d in 1992; in women, intakes were 21 g/d and 16 g/d, respectively. In the same period, the total intake of polyunsaturated fat increased from 11 g/d to 14 g/d in men and from 8 g/d to 11 g/d in women. In men, 22% of total energy intake in the late 1960s came from saturated fats and 3.2% came from polyunsaturated fats; in 1992, these percentages were 16% and 5.0%, respectively. In women, these percentages were 21% and 3.2% in the late 1960s and 15% and 5.1% in 1992. Dietary intake of cholesterol decreased during the same period—from 610 mg/d to 400 mg/d in men and from 450 mg/d to 310 mg/d in women.

In our study, intake of saturated fat from liquid dairy products and spreadable fat decreased by two thirds between 1972 and 1992. Decrease in intake of saturated fat from these two major sources can explain almost all of the decrease in total intake of saturated fat in men and about 75% of the decrease in women. The decreasing trend was similar in both sexes even though the absolute level was higher in men than in women. The decrease was primarily correlated with time period. The association between time period and intake of saturated fat corresponds well to changes in Finnish nutrition policy and eating habits. Four major changes during the last 20 years can explain the observed decrease in intake of saturated fat: 1) introduction and promotion of low-fat milk, 2) introduction and promotion of oil-based soft margarine in the 1970s, 3) introduction of fat-free milk in the early 1980s, and 4) introduction of low-fat oil-based and water-based margarine in the late 1980s. The trends in intake of liquid dairy products and spreadable fat seen in the middle-aged population probably also occurred in younger persons. This would explain the marked decrease seen in cholesterol levels in the youngest birth cohorts entering the survey for the first time compared with levels in the same age group in previous surveys.

Twenty years ago, most saturated fat was consumed in the form of liquid dairy products and spreadable fat. As noted above, the use of these sources of saturated fats decreased, but the amount of saturated fat that men consumed from other sources changed little; in women, the decrease was also relatively small. The other main sources of saturated fat include meat and meat products, cheese, yogurt, ice cream, cooking fat, and fat used in bakery products and chocolate. We have no detailed time-trend data on the use of these products; however, on the basis of sales records, we can assume that the use of saturated cooking fat has decreased somewhat in the past 20 years and that intake of saturated fat from cheese and yogurt may have increased.

Why did intake of saturated fat from liquid dairy products and spreadable fat decrease so dramatically in the past 20 years in eastern Finland but intake from other sources changed only marginally? Although health education messages emphasized the importance of reducing saturated fat intake from all sources, the sources of food available to consumers may have limited their ability to respond to the messages. When low-fat milk and oil-based margarine became available, consumers could easily select the healthier product without making an economic, social, or gastronomic sacrifice. In most other food groups, the availability of products low in saturated fat has been limited. The change in the form of spreadable fat consumed (from butter to oil-based margarine) has been particularly important because it led not only to a decreased intake of saturated fat and dietary cholesterol but also to an increased intake of polyunsaturated fat. The recent observation of the cholesterol-lowering effect of oil-based margarine fortified with oil-soluble natural sterol also provides a new way to control cholesterol [31].

The observed decrease in cholesterol levels in eastern Finland has been somewhat greater than the decrease that was estimated from the changes in intake of saturated and polyunsaturated fat and dietary intake of cholesterol only [7]. In both sexes, the estimated decrease in serum cholesterol levels (determined using the Keys equation) was approximately 0.70 mmol/L, whereas the observed decrease was about 1 mmol/L. Changes in the type of coffee consumed, a decreased prevalence of obesity, and the possible effect of cholesterol-lowering medication may also have contributed to the decrease in cholesterol levels. Finland consumes more coffee than any other country in the world. Twenty years ago, most coffee was boiled; in 1992, filtered coffee was predominantly used. This change may explain most of the difference between estimated and observed decrease in serum cholesterol levels [32]. Obesity, on the other hand, is now more common in men than it was 20 years ago and is only slightly less common in women [33]. Cholesterol-lowering medication has been infrequently used in eastern Finland. In 1992, about 2.5% of the population was using these drugs; during the earlier surveys, almost no one used them.

The dynamics of saturated fat intake according to birth cohort differed from those of serum cholesterol levels in the same cohorts. Saturated fat intake was fairly independent of age, and the change in intake was similar in both sexes even though the absolute level of intake was higher in men. In both men and women, the level of intake and the change in saturated fat intake were determined primarily by time period. The change seen in cholesterol levels differed between sexes and across age groups; an interaction between age and sex was also seen. These differences are attributable to the fact that diet is an important but not the only determinant of cholesterol level; other factors are age and sex. Cholesterol levels tend to increase with age, even when intake of saturated fat remains constant. This increase may be due to a decrease in the metabolic use of cholesterol as persons age, but this phenomenon is not fully understood. These data suggest that intake of saturated fat should decrease with age so that serum cholesterol levels do not increase.

The difference seen between sexes and the interaction between age and sex are probably attributable to dietary and hormonal differences between the sexes, particularly menopause in women [21]. The higher cholesterol levels in men during early adulthood can be at least partly explained by the higher intake of saturated fat in men. At older ages, the situation of women is similar to that of men, probably because of hormonal factors. However, cholesterol levels in postmenopausal women may be substantially influenced by dietary factors, suggesting that cholesterol levels can be modified by dietary interventions in this group as well as in older men.

Our study has some limitations. First, because sample sizes were smaller and participation rates were lower in the later population surveys than in the earlier ones, we may have slightly overestimated the decrease in cholesterol levels that occurred with time, particularly in men. Second, collecting data on nutrition is always difficult, especially when comparable data are collected at different points during a long period. Our data on the consumption of liquid dairy products and spreadable fat are probably reliable because the products are easy to define and because we used the same methods in all five surveys. More problematic are the reference data on total saturated fat intake, dietary cholesterol intake, and total energy intake: The studies from which these data were obtained used different methods, and the selection of available food products has changed much during the 20-year period.

Third, our estimates of saturated fat intake in the late 1960s are based on a sample of the population in central Finland, where the intake of saturated fat may have been lower than that in eastern Finland. Therefore, the real decrease in saturated fat intake in eastern Finland and its contribution to the decrease in cholesterol levels may have been larger than that used in our calculations.

Fourth, we compared the change in saturated fat intake with the change in serum cholesterol levels at the population level rather than at the individual level. The effect of change in diet on change in cholesterol level cannot be proven in epidemiologic analyses or cross-sectional studies; only clinical trials can establish this effect. On the basis of the data from clinical trials, equations have been developed for estimating the effect of dietary change on cholesterol levels [7]. The Keys Equation is probably the best known of several equations. A well-controlled trial in eastern Finland has shown that the Keys Equation functions relatively well in this population [34]. In our study, we applied this Equation at the population level, but our results concur with the results of trials in which analyses were done at the individual level. We also analyzed the cross-sectional association between saturated fat intake and serum cholesterol at the individual level (data not shown). The association was positive and statistically significant but was weaker than results obtained from prospective intervention trials.

The usefulness of data obtained from clinical trials is also limited. For example, the duration of dietary trials is short, usually lasting only weeks or, at best, months. Therefore, we do not know whether the changes in diet introduced in controlled trials can be maintained during the long run in free-living populations. The strength of our study is that it has been able to address these questions at a population level: Large dietary changes were introduced in large populations, and these changes were maintained over a long period.

In conclusion, our data suggest that a community-based strategy for preventing cardiovascular diseases may be able to affect a large part of a population. In our study, cholesterol levels markedly decreased across the population, including persons 30 years of age or younger. The latter group, however, was not the primary target group for most of the preventive strategies. An interesting finding was that targeting an intervention to a population rather than limiting the approach to individual persons at high risk had effects that were more extensive than expected. Our results also indicate that a large part of the increase in cholesterol levels that occurs with age is not an inevitable physiologic phenomenon but is closely associated with diet; the increase appears to be largely preventable through dietary changes.