Secular changes in cardiovascular risk factors and attack rate of myocardial infarction among men aged 50 in Gothenburg, Sweden. Accurate prediction using risk models


Lars Wilhelmsen, Fregattgatan 16, SE-42674, V.Frölunda, Sweden.
(fax: +46 31 690379; e-mail:


Aims.  Coronary risk factor changes were related to attack rate of acute myocardial infarction (AMI).

Methods and results.  Cross-sectional population samples of 50-year-old men were examined every 10th year from 1963 to 2003. Attack rates of AMI were recorded from 1975 to 2004. Prevalence of smoking decreased from 56% in 1963 to 22% in 2003. Leisure time physical activity decreased (n.s.), while psychological stress remained the same. Diabetes prevalence increased from 3.6% to 6.6%. Body mass index (BMI) increased from 24.8 to 26.4 kg m−2. Blood pressures decreased from 138.2/90.6 to 134.7/84.9 mmHg (P = 0.00001). Serum total cholesterol decreased from 6.42 to 5.50 mmol L−1 (P = 0.0001), but serum triglycerides increased from 1.26 to 1.71 mmol L−1 (P = 0.0001). The multivariable risk according to total cholesterol, blood pressure and smoking for AMI decreased from the set value 1.0 in 1963 to 0.418. From 1975–1979 to 2000–2004 attack rates for AMI for the age groups 35–44, 45–54 and 55–64 declined to 45%, 46% and 45%, respectively. The 28-day case fatality declined from 30%, 38% and 46% to 12%, 16% and 20%.

Conclusion.  The more than 50% decline in attack rate of AMI during 30 years was comparable with the decline in risk factors.


Coronary disease is a leading cause of death and disability in Western countries, but a decline in mortality and in some reports incidence has been reported [1–5]. However, some regions display increasing trends [4, 6]. Declines in coronary risk factor levels have been reported from the Framingham study in the USA and were suggested to contribute to the 60% decline in cardiovascular mortality there [7]. Similarly, a report from Finland estimated that about a half of the decline in coronary mortality was associated with a decline in coronary risk factors [8–10]. Similar findings have been reported from England and Wales [5].

In Gothenburg we have examined random population samples of men aged 50 years every 10th year in 1963, 1973, 1983, 1993 and 2003. A special Acute Myocardial Infarction (AMI) Register was begun in 1970, and reliable data are available from 1975. Since 1987 there has been a national register of nonfatal and fatal AMI based on hospital admissions and mortality. We were able to analyse attack rate of AMI (nonfatal and fatal) by 5-year age groups from 1975 until 2004.

The main objective of this study was to analyse to what extent changes in the attack rate of AMI could be related to changes in coronary risk factors, using multivariable risk functions to estimate risk.

Study populations and methods

Population samples

From 1963 and every 10th year thereafter population samples of 50-year-old men living in the city of Gothenburg were examined. In 1963 all men born on dates divisible by three were invited to participate [11]. The study in 1923 comprised all men born on the 15th and 27th day of each month. For the following investigations random samples of 50% of all men living in the city were invited to participate [12]. Number invited, number examined and participation rates are given in Table 1.

Table 1.   Participation rates in population studies of 50-year-old men examined from 1963 to 2003 in Gothenburg, Sweden
Birth yearExamined yearAge at examinationNumber invitedNumber examinedParticipation rate, %

Risk factor levels

Data on smoking habits, leisure time physical activity and experience of mental stress were collected by postal questionnaires. Smoking habits were coded as (1) never smoker; (2) former smoker of more than 1 month’s duration (ex-smoker); (3) smoking tobacco of 1–14 g day−1; (4) smoking 15–24 g day−1; (5) smoking 25 g day−1 or more. One cigarette was considered to contain 1 g of tobacco, a cigarillo 2 g and a cigar 5 g.

A questionnaire on physical activity during leisure time was first introduced for the study of men born in 1913 at the examination in 1967 and was graded from 1 to 4. Grade 1 was defined as mainly sedentary, grade 2 as moderate activity, grade 3 as regular exercise and grade 4 as athletic training. Grades 3 and 4 were analysed together because a very limited percentage engaged in athletic training at any period [13].

A simple questionnaire on mental stress was also introduced for the 1967 investigation. Stress was defined as feeling tense, irritable or filled with anxiety or having sleeping difficulties as a result of conditions at work or at home. There were six response options: (i) never experienced stress; (ii) one period of stress ever; (iii) some periods during the last 5 years; (iv) several periods during the last 5 years; (v) permanent stress during the last year; (vi) permanent stress during the last 5 years. Items v–vi have been shown to be related to AMI during follow-up [14–16]. Diabetes was noted if the man had been told by a doctor that he had diabetes or blood glucose levels were ≥ 6.0 mmol L−1 at the screening examination.

All examinations were performed in the morning after an overnight fast. Body weight was measured on a lever balance to the nearest 0.1 kg. Height was recorded barefoot to the nearest centimetre and body mass index (BMI) was calculated. Waist circumference was measured at the umbilicus, being first introduced in 1967 when the men were 54 years old. Blood pressure was measured by the study physician except in 1983, when a study nurse did it. A standard cuff and a mercury manometer were used. Diastolic blood pressure, Korotkoff phase 5, was recorded.

Blood lipid measurements were performed according to available methods at the time and were analysed at the same hospital laboratory. Details have been reported previously [12].

Rate of nonfatal myocardial infarction and fatal coronary disease

During the years 1987–1994, the AMI register in Gothenburg ran in parallel with the national register and thus it was possible to check whether the two registers were comparable. The agreement for the relevant age groups was very good with only isolated discrepancies. The attack rates according to the local AMI Register active during the period 1974–1994 and the National AMI Register that has been active since 1987 until the present have been used.

The definition of attack rate is the total number of AMI during the year, whereas incidence denotes the number of new attacks excluding patients with previous attacks. We have previously published the AMI attack rates and incidence results until 1994 [1].

Statistical methods

Means, prevalences and trend tests (according to Cochrane–Armitage) of values over the years were calculated by traditional methods. P-values < 0.05 were regarded as statistically significant.

Calculation of multivariable risk

A multivariable logistic function for risk of nonfatal AMI and fatal coronary disease was developed in 1973 using the first cohort of the present study population [17]. This function, described in detail in the 1973 publication, primarily concerned risk during 9 years and 4 months, was later adapted to give risk for 10 years. This function was used to analyse the predicted risk of the different cohorts of 50-year-old men in the present study. For a sensitivity analysis, we used the Framingham function for AMI [18], the SCORE function, which includes the risk of fatal stroke and coronary disease in the prediction and does not take amount of tobacco smoked into account [19]. Even though the absolute risk of nonfatal events is different from fatal events, the relative changes of risk could be used for the present comparisons. The relative risks are given in relation to the risk estimate of the year 1963. The studies were approved by the Ethics Committee of Göteborg University, and the participants in the population studies had all given their informed consent.


Smoking habits, physical activity during leisure time, prevalence of diabetes and experience of mental stress are shown in Table 2. The prevalence of smoking decreased considerably from 56% in 1963 to 22% in 2003. This was attributable both to more men never having started and to more men having stopped smoking in recent years.

Table 2.   Smoking habits, physical activity during leisure time, experience of mental stress and prevalence of diabetes among 50-year-old men examined at 10-year intervals 1963–2003 in Gothenburg, Sweden
Year of examination (Number of men)1963 (855)1973 (226)1983 (776)1993 (798)2003 (595)P for trend
  1. *Values from the 1967 examination at age 54 years. These values were not included in the trend tests.

Smoking habits     0.0001
 Never smokers, %24.220.936.332.337.5 
 Ex-smokers, %19.726.426.836.741.0 
 Smoking 1–14 g day−1, %34.629.115.614.88.9 
 Smoking 15–24 g day−1, %17.318.215.611.89.4 
 Smoking ≥ 25 g day−1, % 
Physical activity grade 3–4, %32.4 *
Mental stress grade 5–6, %17.0 *
Prevalence of diabetes, %

As stated under Study populations and methods, the questionnaire regarding leisure time physical activity was not introduced until 1967 for the men born in 1913 at which time 32% reported an activity of grade 3–4, which decreased somewhat during the following years to 24% in 2003. The decline was not statistically significant. The percentage reporting moderate activity increased somewhat from 56% in 1973 (men born in 1923) to 59% in 2003 (men born in 1953).

Mental stress was also first measured in 1967, when grade 5–6 was reported by 17%. This prevalence was largely the same over the years and was still 17% in 2003 (n.s.).

The prevalence of diabetes increased significantly from 3.6% during the first year to 6.6% in 2003, and the percentage of known diabetes from 2.5% to 4.0%.

Anthropometric measurements, blood pressures and serum lipid levels are given in Table 3.

Table 3.   Anthropometry, blood pressure and blood lipids among 50-year-old men examined at 10-year intervals 1963–2003 in Gothenburg, Sweden
Year of examination (Number of men)1963 (855)1973 (226)1983 (776)1993 (798)2003 (595)P for trend
  1. *Values from the 1967 examination at age 54 years.

Height, cm, mean (SD)175.0 (6.0)175.8 (6.4)177.4 (6.5)178.4 (6.9)179.4 (7.1)0.00001
Weight, kg, mean (SD)75.9 (11.0)79.3 (11.7)80.7 (11.8)82.8 (12.1)84.8 (13.3)0.00001
BMI, kgm−2, mean (SD)24.8 (3.2)25.6 (3.2)25.6 (3.4)26.0 (3.4)26.4 (3.7)0.00001
Waist circumference, cm, mean (SD)87.0 (9.6) *91.7 (9.8)94.4 (9.1)95.2 (9.5)94.5 (9.9)0.00001
Systolic blood pressure, mmHg, mean (SD)138.2 (20.9)137.4 (22.1)134.0 (16.9)128.7 (17.1)134.7 (17.8)0.00001
Diastolic blood pressure, mmHg, mean (SD)90.6 (12.5)91.6 (14.2)86.2 (10.6)84.4 (10.6)84.9 (11.0)0.00001
Serum cholesterol, mmol L−1, mean (SD)6.42 (1.12)6.46 (1.34)6.12 (1.12)5.88 (1.04)5.50 (1.01)0.0001
Serum triglycerides, mmol L−1, mean (SD)1.26 (0.82)1.30 (0.77)1.46 (1.10)1.67 (1.14)1.71 (1.18)0.0001

Height increased by about 1 cm for each 10-year cohort, while body weight increased 8.9 kg over the 40-year period, which resulted in a significant increase of BMI, by 1.6 units.

Waist circumference also increased significantly during the period. There was a steady decline in blood pressure until 1993, but an upswing in 2003, which, however, did not change the significant trend of declining blood pressures.

There was a significant decline of total cholesterol from 6.42 to 5.50 mmol L−1, but a significant increase of triglycerides from 1.26 to 1.71 mmol L−1 during the period.

For the multivariable risk calculation the three risk factors smoking, systolic blood pressure and total cholesterol were first inserted into the risk function [17].

The changes relative to the first examination are given in Table 4. The relative risk decreased from 1.0 in 1963 to 0.418 in 2003. In a sensitivity analysis we also used the risk functions from the Framingham study for AMI [18] and from the SCORE project [19].

Table 4.   Change of multiple risk prediction based on smoking habits, serum total cholesterol and systolic blood pressure among 50-year-old men examined at 10-year intervals 1963–2003 in Gothenburg, Sweden. Risk was calculated according to Wilhelmsen et al. [17] and in a sensitivity analysis according to Anderson et al. [18] and Conroy et al. [19]. The relative risks are given in relation to the risk estimate of the year 1963
Year of examination (Number of men)1963 (855)1973 (226)1983 (776)1993 (798)2003 (595)
  1. Predicted risk per 1000 person years in 1963 according to Function 1: 3.3 (nonfatal and fatal AMI); Function 2: 7.8 (nonfatal and fatal AMI during 10 years); Function 3: 3.2 (death from stroke and coronary disease).

1. Ref. [17]11.0270.6580.4970.418
2. Ref. [18]10.9800.7240.5750.509
3. Ref. [19]10.9720.7560.6230.593

Attack rates of nonfatal AMI and fatal coronary disease for men aged 35–44, 45–54 and 55–64 during the period 1975 to 2004 in Gothenburg are shown in Table 5 and Fig. 1.

Table 5.   Total population, sum of nonfatal and fatal coronary disease attacks, attack rate per 1000 population per year as well as relative attack in relation to the period 1975–1979 given as 1.00. Data for Gothenburg, Sweden
Men aged 35–44Men aged 45–54Men aged 55–64
PeriodTotal populationNumber of eventsAttack rate per 1000Relative attack rateTotal populationNumber of eventsAttack rate per 1000Relative attack rateTotal populationNumber of eventsAttack rate per 1000Relative attack rate
1975–197925 4151301.021.0025 5566905.401.0027 184178513.131.00
1980–198429 5071450.980.9622 2135725.150.9525 414168013.221.01
1985–198931 8471360.850.8322 8594774.170.7721 861126011.530.88
1990–199431 1061200.770.7527 0794403.250.6019 4008628.890.68
1995–199932 2101050.650.6429 9784112.740.5120 6707497.250.55
2000–200434 931810.460.4529 6613712.500.4624 9107345.890.45
Figure 1.

 Coronary attack rate per 1000 men per year and from different age groups between 1975 and 2004 in Gothenburg Sweden.

For all age groups, attack rate per 1000 population of men declined to half or less during these 30 years.


The main findings of this paper were declining levels of the coronary risk factors total cholesterol, blood pressure and tobacco smoking, whereas diabetes, BMI and waist circumference as well as triglycerides increased significantly among 50-year-old men over the 40 years of observation. The multivariable risk for a coronary event calculated from four different risk functions, and calculated for 10-year periods, decreased to about 50% of the original value. During about the same period (1975–2004) attack rates for coronaries in Gothenburg declined to 50% or lower for men aged 35–44, 45–55 and 55–64 years in spite of more sensitive enzyme markers such as troponin T had come into use. Concomitantly there was a decline in case fatality during the first 28 days of the same magnitude. With the exception of certain infectious diseases such as polio and tuberculosis, for which vaccination and effective drugs have been found, it is difficult to find any other disease for which such impressive improvements have been registered. The decline is in good agreement with the decline in the major coronary risk factors, for which the declines were not related to increased use of pharmaceuticals.

In Sweden as a whole there was an age-standardized 20% decline of incidence among men from 1000/100 000 population to 800/100 000 (P = 0.0001) during the latter part of the period analysed in this study 1987–2001, and among women somewhat less from 420/100 000 to 400/100 000 (P = 0.0001) [20].

In a previous paper [1] we reported declining levels of risk factors and attack rate as well as incidence during the first period 1975–1994, and according to the present findings the positive trend continued. As discussed, we then saw improved hospital mortality attributable to acceptance of results of clinical trials in AMI [21]. We also found that only 20% of those who suffered a first coronary event were aware that they had coronary disease and could have received prophylactic treatment. The increased hospital survival over the last 20 years has resulted in more patients having survived an AMI. These survivors, who are at high risk of recurrence, will therefore build up a risk population that will recruit new AMI events. The increased efforts to improve their fate by secondary prevention may have been of importance for the decreasing incidence, as reported above. We therefore analysed the extent to which the AMI events were first or recurrent events. It was found that 20% were recurrences in the age groups 45–64 years. Thus the incidence of new AMI events was 80% of the attack rates. It is of interest that the percentage of recurrent AMIs had not increased from our previous report [1]. Thus, only a minority of the present AMI events was recruited from recurrent AMIs indicating that most of the decline in attack rate could be explained by primary preventive actions (mostly thanks to increasing knowledge regarding smoking, healthier diet and perhaps physical exercise) in the population.

Pharmaceutical interventions to lower serum cholesterol were very uncommon in the population (<1.0%) in 1963 and had increased to 3.9% in 2003, but this is considered to be of limited importance for the decline of mean cholesterol levels as compared with dietary changes recorded in large parts of the general population.

Treatment of hypertension was also limited in 1963, when only 1.6% of the men in the population were on drug treatment, but it had increased to 8.2% already in 1973, and to 10.2% in 1983 after which there was no increase until 2003, when it was down to 7.1%. The extent to which anti-hypertensive treatment affects the mean blood pressure in the population is questionable. This issue has been addressed by Tunstall-Pedoe et al. [22] in an analysis of blood pressure in the WHO MONICA study. They concluded that blood pressure fell across 38 MONICA populations at all levels of readings, with no differential fall in high readings attributable to better control of hypertension. In spite of the higher BMI in our population, we saw decline of blood pressure, which may be attributable to changes of diet. There is limited data available regarding salt intake during the period in question.

Other variables than the three used have also been associated with risk of AMI and most recently reported in the case–control study INTERHEART [23]. Of variables measured in the present long-term analysis, results regarding diabetes, mental stress, physical activity and waist circumference were available. The only factors that differed significantly among the five Gothenburg cohorts were diabetes and waist circumference.

Inclusion of the increasing waist circumference would have increased the risk, but it is difficult to estimate to what extent, as it is not known how much waist circumference influenced the risk in the present population. It should be mentioned that increased BMI (and probably waist circumference) was not associated with increased risk for AMI among men in the cohort of men born in 1963 and not even during 28 years follow-up of med born in 1915–1925 [15]. Thus, obesity does not seem to have had influenced AMI risk among men in Gothenburg yet.

The increased prevalence of diabetes has been calculated to correspond to a risk increase of 7.8%, and instead of the relative risk in 2003 of 0.418 the relative risk would be 0.450. The increased prevalence of known diabetes may well be due to the increased obesity.

During the same period as analysed in this study, there was a decline of 28-day case fatality, including both in-hospital and out-of-hospital deaths, from 30%, 38% and 46% among the same age groups of men to 12%, 16% and 20%, thus a decline of 50% or more.

During the same period there was a 50% decline in AMI incidence (0.22–0.11 per 1000) among women aged 35–44, and a somewhat smaller decline among women aged 45–54 (1.10–0.65 per 1000) and among women aged 55–64 (3.65–2.04 per 1000).

The present study has certain strengths primarily the internationally uncommon availability of risk factor data for random population samples of 50-year-old men each decade for 40 years, as well as the AMI register, available for more than 30 years.

There are also limitations, one of the most important being the declining participation rates in the population samples from 1963 to 2003. Higher mortality among nonparticipants in population studies in Gothenburg has been reported previously [24]. It is possible, but not sufficiently well studied, that nonparticipants have higher coronary risk profiles than participants. If so, the lower risk in recent years may be attributable to the lower number of participants in the samples of 50-year-old men. The percentage of men born outside Sweden increased from 5% in 1963 to 23% in 2003, and the participation rate among men born outside Sweden is probably lower than for Swedish-born men while risk factor levels may be higher. If so, the decline of risk factor levels in the entire population may have been overestimated to some extent.

In conclusion, this study showed a remarkable decline of attack rate including incidence of AMI as well as of 28-day case fatality, and the rate was close to that expected, based on the decline of multivariable risk concomitantly registered in the community. This finding may be taken as the best indication available for the effects of primary prevention as it is clear that community based trials are difficult to conduct. This is primarily because of difficulties in achieving differences between control and intervention groups living in the same community [25, 26].

Conflict of interest

None of the authors declared any conflict of interest.


We would like to thank Max Köster at the Epidemiological Centre of the Swedish Board of Health and Welfare for providing data from the National AMI Register. The study was supported by the Swedish Heart and Lung Foundation.