Body Weight, Cardiovascular Risk Factors, and Coronary Mortality : 15-Year Follow-up of Middle-aged Men and Women in Eastern Finland

Risk factor surveys were conducted in 1972 and 1977 in the provinces of North Karelia and Kuopio in eastern Finland. A random sample of 6.6% of the population aged 30 to 59 years was drawn in both areas. In the present analyses, the two cohorts were pooled together. The total sample size was 19 894. The participation rate was 90% among men and 93% among women. Those subjects who participated in both surveys were included only in the first survey cohort. Of 17 682 participants, 683 were excluded because of reported history of myocardial infarction, stroke, or diabetes before the survey. Another 886 participants were excluded because of incomplete data on either BMI or at least one of the risk factors (smoking, serum cholesterol, or BP). Thus, 7740 men and 8373 women were included in the present analyses.

The survey methods have been described earlier in detail.³¹ A self-administered questionnaire was sent to the participants in advance, and it included questions about medical history and health behavior. At the study site, specially trained nurses measured height, weight, and BP using a standardized protocol. Weight was measured with subjects wearing light clothing and height was measured without shoes. BMI (kg/m²) was used as a measure of relative body weight. The BP measurement was taken from the right arm of the subject, who had been asked to sit for 5 minutes before the measurement. After the BP measurement, a venous blood specimen was taken. Serum cholesterol was determined from frozen samples by the Lieberman-Burchard method. All samples were analyzed in the same laboratory.

Smoking was assessed in the surveys by a set of standardized questions in a self-administered questionnaire. On the basis of the responses, the participants were classified into three categories: (1) current smokers, or persons who had smoked regularly for at least 1 year more than once a day on average and who had smoked during the preceding month; (2) ex-smokers; and (3) lifelong nonsmokers, or those who had never smoked. In the present study, those ex-smokers who had not smoked during the past 6 months were considered nonsmokers, and those ex-smokers who had quit smoking <6 months earlier were considered smokers.

Mortality data were obtained from the Central Statistical Office of Finland and linked to the risk factor data by use of the identification numbers assigned to every resident of Finland. The rate of deaths ascertained in the study cohorts was therefore complete. The eighth revision of the ICD was used in Finland from 1969 to 1986 and the ninth revision was adopted at the beginning of 1987. ICD codes 410 through 414 were classified as CHD deaths. The follow-up time of each subject considered in our present analyses was 15 years. The number of CHD deaths during the follow-up was 480 among men and 103 among women.

The association between BMI and serum cholesterol and between BMI and BP at baseline was analyzed by use of a multiple regression model. The association between BMI and smoking prevalence was analyzed by use of a logistic regression model. In both analyses, BMI was used as a continuous variable and the analyses were adjusted for age. To calculate CIs for mortality rates, deaths during the follow-up were assumed to follow a Poisson distribution. Multivariate analyses were performed by use of a Cox proportional hazards model.³² The estimates of relative risks and their 95% CIs were based on this model. Furthermore, to assess the extent to which the risk of CHD mortality associated with BMI may be modified by or mediated through the known cardiovascular risk factors, smoking, serum cholesterol, and systolic BP were included in the models. Interactions between BMI and other risk factors were analyzed by adding all first-level interactions between BMI, smoking, serum cholesterol, and systolic BP, one at a time, into the model together with all main variables. A stratified analysis by smoking status, dichotomized values of serum cholesterol and BP, and different levels of BMI was also performed. Because of the relatively small number of deaths due to CHD among women, the stratified analysis by smoking status was performed only for men. Statistical analyses were done with use of the SAS statistical programs.³³

Serum cholesterol and both diastolic and systolic BP had a positive association with BMI (Tables 1 and 2). Smoking, on the other hand, had an inverse association with BMI, except that the most obese men smoked more than the moderately overweight men. Among men, the inverse association of smoking with BMI was explained by smoking cessation. The proportion of ex-smokers increased as BMI increased, but the proportion of lifelong nonsmokers did not have a significant association with BMI. Among women, in contrast to men, the proportion of lifelong nonsmokers increased as BMI increased, but the proportion of ex-smokers did not associate with BMI. There was a strong positive association between age and BMI in women but not in men. However, in both sexes, the positive association of BMI with serum cholesterol and BP and the negative association with smoking were still statistically significant after adjustment for age.

Age-adjusted coronary mortality increased from 39 per 10 000 person-years among men with a BMI between 20 and 22.5 kg/m² to 78 per 10 000 person-years among men with a BMI ≥32.5 kg/m² (Fig 1). Among the lean women with a BMI <22.5 kg/m², age-adjusted CHD mortality was very low, <2 per 10 000 person-years (Fig 2). It increased to about 10 per 10 000 person-years among women with a BMI between 22.5 and 32.5 kg/m² and to 13 per 10 000 person-years among the most obese women. In both sexes, the number of subjects with a BMI <20 kg/m² was small, and therefore CIs for mortality rates in the leanest group are wide.

Among men, the BMI-associated RR of CHD mortality, adjusted for age and study year, was 1.04 per kg/m² (P<.001) (Table 3). Inclusion of smoking in the model increased the RR, whereas inclusion of serum cholesterol decreased it slightly. Inclusion of systolic BP in the model decreased the RR to 1.02 per kg/m² (P=.092). In the model that included all three major cardiovascular risk factors, the BMI-associated RR was 1.03 per kg/m² (P=.027). Among women, the respective RRs were 1.05 per kg/m² (P=.023), 1.02 per kg/m² (P=.417), and 1.03 per kg/m² (P=.151).

In men aged 30 to 49 years, age- and study year–adjusted BMI-associated RR of CHD mortality was 1.02 per kg/m² (P=.299) (Table 4). After including smoking in the model, the RR was 1.04 per kg/m² (P=.042). After further inclusion of serum cholesterol and systolic BP in the model, the association between BMI and CHD mortality disappeared (P=.837). In men aged 50 to 59 years, age- and study year–adjusted RR was 1.05 per kg/m² (P<.001) and the multivariate-adjusted RR was 1.04 per kg/m² (P=.006).

Unlike in men, the BMI-associated RR of CHD mortality was higher among younger compared with older women. In women aged 30 to 49 years, the BMI-associated RR of CHD mortality, adjusted for age and study year, was 1.09 per kg/m² (P=.020). Inclusion of smoking in the model increased the RR to 1.10 per kg/m² (P=.012). The further inclusion of serum cholesterol and systolic BP decreased the RR to 1.05 per kg/m² (P=.266). In women aged 50 to 59 years, the age- and study year–adjusted RR was 1.03 per kg/m² (P=.185) and the multivariate adjusted RR was 1.02 per kg/m² (P=.374).

The first-level interaction between BMI and systolic BP as continuous variables was statistically significant among men (P=.033) but not among women. In both sexes, interactions between BMI and smoking and between BMI and serum cholesterol were not significant statistically.

Among men, as stratified by smoking status and BMI, there was a trend toward an increase in the risk of CHD mortality as BMI increased among lifelong nonsmokers, ex-smokers, and smokers (Table 5). This trend diminished after adjustment for serum cholesterol and systolic BP. Among lean and moderately overweight men, ie, BMI <30 kg/m², the risk of CHD mortality was similar among lifelong nonsmokers and ex-smokers, but among the most obese men, the RR was higher among ex-smokers than lifelong nonsmokers.

When stratified by serum cholesterol and BMI, the risk of CHD mortality increased as BMI increased both among men with serum cholesterol <6.5 mmol/L and among men with serum cholesterol >6.5 mmol/L (Table 6). Among women, there was also an increasing tendency, but the trend was not as regular as among men. In both sexes and at all BMI levels, the RR was higher in the high-cholesterol group compared with the low-cholesterol group.

In both sexes as stratified by BP and BMI, the risk of CHD mortality increased as BMI increased only among normotensive subjects (diastolic BP <95 mm Hg and systolic BP <160 mm Hg) (Table 7). Among hypertensive subjects (diastolic BP ≥95 mm Hg or systolic BP ≥160 mm Hg), the risk of CHD mortality did not vary by BMI level. At low and moderate BMI levels (BMI <30 kg/m²), the RR was higher among hypertensive than among normotensive subjects, but among the most obese subjects, the RR was similar in both low and high BP groups.

Obesity was an independent risk factor for CHD mortality among men and also contributed to the risk of CHD death among women. Starting at a BMI of ≈22 kg/m², an increase in body weight equivalent to 1 BMI unit (kg/m²) was related to a 4% to 5% increase in CHD mortality. In other words, an increase in body weight of ≈1 kg increased the risk of CHD mortality by 1% to 1.5%.

BMI was closely related to the other major cardiovascular risk factors (smoking, serum cholesterol, and BP). Several studies showed that smokers have, on average, a lower BMI than nonsmokers and that smoking cessation is often associated with weight gain.³⁴ This may be due to metabolic effects of smoking that increase energy consumption or to psychological factors that affect eating habits. Thus, smoking has a rather complicated association with CHD. Smoking is, without a doubt, one of the most important risk factors for CHD mortality, but at the same time, smoking seems to protect against another risk factor, obesity.^{35 36 37} Weight control should be an integral part of CHD prevention among both nonsmokers and smokers, but among smokers, smoking cessation is obviously the most important single preventive measure. Particular emphasis is also needed to prevent weight gain after quitting smoking. Risk of coronary death among lean and moderately overweight ex-smokers was similar to that of lifelong nonsmokers, but among the most obese ex-smokers, the risk seemed to be higher than among lifelong nonsmokers at the same BMI level. One possible explanation for this could be that the obese ex-smokers may have started to smoke again to control weight. Also, the concurrent presence of severe obesity and smoking in an individual may be an indicator of a particular unhealthy lifestyle in general.

Serum cholesterol level correlated positively with BMI at baseline. In the longitudinal analyses, part of the CHD mortality risk associated with overweight was mediated through serum cholesterol. Overweight was associated with increased CHD mortality both in high- and low-cholesterol groups. Even though we did not assess the effect of weight change on serum cholesterol level, it can be assumed that weight reduction also decreases serum cholesterol level.^{14 38} Therefore, weight control probably can prevent CHD mortality more than is estimated from analysis of BMI alone.

Our results are in agreement with the well-documented association between body weight and BP.^{9 14 17 18 19 20 21 22 23 24} Several studies showed that hypertension is more common in obese than in lean persons. Body weight also correlates with BP within the so-called normotensive range of BP. In longitudinal studies, weight gain is independently related to increased BP.^{18 19 20} Among people with slightly increased BP (borderline hypertension), long-term intervention with weight loss has been shown to be effective in the prevention of hypertension.²¹ Since obesity is the strongest determinant of hypertension, weight control could be the most effective way to prevent hypertension in a population and to reduce BP among overweight hypertensive subjects.

In the present study, a substantial part of the CHD mortality risk associated with BMI was mediated through the association between body weight and BP, as the BMI-associated risk of CHD mortality decreased when systolic BP was added to the model. Nevertheless, this finding does not diminish the practical value of weight control as an integral part of cardiovascular disease prevention. To the contrary, it is a very strong argument for nonpharmacological control and prevention of high BP.

Some studies^{39 40} suggested that lean hypertensive subjects may have a higher risk of cardiovascular disease than obese hypertensive subjects. In the current study, the risk of CHD death did not increase much as BMI increased among hypertensive subjects. It is likely that there are several etiologic subtypes of hypertension. A fraction of hypertensive subjects may have a genetic predisposition to hypertension that is independent of body weight. On the other hand, it has been observed that at least part of the excess mortality among lean hypertensive subjects is due to deleterious lifestyles, such as smoking and excessive alcohol intake.⁴¹ Among the majority of hypertensive patients who are also overweight, overweight is probably the central factor in the origin of hypertension. Estimating the effect of weight control on CHD mortality among hypertensive subjects should not be based on BMI level alone but also on its simultaneous effect on BP.

Diabetes is the third possible mediator between obesity and the risk of CHD. Obesity increases the risk of diabetes, which is a known cardiovascular risk factor.^{9 14 25 26 27} Even though those subjects who had diabetes before the baseline measurements were excluded from the present study, diabetes developing during the follow-up may still play a role in the risk estimates of obesity-associated CHD mortality in our study. Findings that obesity, lipid abnormalities, and diabetes often coexist in hypertensive subjects suggest that hyperinsulinemia may be the common link between the four phenomena.^{25 26 27} Several studies⁴² also showed that subjects with an increased level of these risk factors have a highly increased risk of cardiovascular disease. In the interrelation between hyperinsulinemia, obesity, diabetes, hypertension, and lipid abnormalities, the most natural target of primary intervention is obesity, which further stresses the importance of weight control in cardiovascular disease prevention.

From the public health point of view, the question of whether obesity is an independent risk factor for CHD among subjects with elevated BP, high serum cholesterol, or diabetes is not very relevant. The two components of the risk associated with obesity, operating either independently or through other risk factors, cannot be separated in healthcare practice. Most likely, the best prediction of the practical value of weight control for the prevention of cardiovascular disease can be obtained by using the models presented in the current study without adjustment for other risk factors or by adjustment for smoking only. Furthermore, by replacing and supporting the current drug therapy in the treatment of hypertension, lipid abnormalities, and non–insulin-dependent diabetes with nonpharmacological treatment methods, substantial savings in costs can be obtained. Simultaneously, an optimal health effect may be reached without the risk of the possible side effects of antihypertensive, lipid-lowering, and antidiabetic drugs.^{43 44}

Among young men, the association between BMI and CHD mortality was weaker than among older men, and among younger men, this risk was mainly mediated through serum cholesterol and BP. Because body weight in young adulthood usually correlates with body weight in older age, we can assume that in the younger age group, the risk may also increase during a longer follow-up. Some studies^{3 45} showed that the risk associated with body weight appears only after a relatively long follow-up. In contrast to men, the association between BMI and CHD mortality among young women was stronger than that among older women. Among women, body weight increases with age much more than among men. Young women are leaner than young men but, because of hormonal and other factors, women gain weight later. After 50 years of age, women are more obese than men of the same age. It is possible that among women, the weight gain at an older age is not as dangerous as if the weight gain occurred at a younger age. Also, body fat distribution differs between sexes. Men more often have central adiposity, which may associate more strongly with the risk of CHD mortality than peripheral adiposity, which is more common among women.⁴⁶

The current study did not include weight history and the effect of weight change on CHD mortality. Therefore, we cannot conclude from our data how the optimal weight should be achieved. Some reports^{7 11 47 48} showed a negative health effect associated with weight fluctuation, but the data are still limited in this regard. Nevertheless, obesity ideally should be prevented in young adulthood when the risk of weight gain is at its highest, particularly among men. The safest way to reach the optimal weight by weight reduction is probably to lose weight over a relatively long time period. There are no data on the long-term health effects of using very low caloric diets or other methods that cause rapid weight reduction.

BMI is the most commonly used indicator of obesity in population studies, although it is not a perfect one. It does not take into account body fat patterning as waist-hip ratio and skin-fold measurements do. It seems that increased central or visceral fat, independent of relative body weight, is associated with a variety of metabolic disorders and increased cardiovascular mortality.^{49 50 51} Furthermore, weight is usually positively related to increased morbidity and mortality, whereas height is often associated with good health. Therefore, among obese subjects, BMI can reflect the negative effects of both fatness and shortness. The risks of fatness and shortness are most likely mediated via different mechanisms. However, BMI also has several advantages compared with other methods of measuring obesity. BMI measurement is simple, inexpensive, and reliable. It is widely used, and the results of different studies are therefore easily compared. Results are also easily transferred for use in practical health care and disease prevention.

We conclude that weight control should be an integral part of the prevention of cardiovascular disease. The question of whether obesity is an independent risk factor for cardiovascular diseases or whether its effect is mediated via BP, lipid abnormalities, impaired glucose metabolism, or other mechanisms is not very important in health practice because these components cannot be separated. Similarly, although fat distribution plays an important role in the research of the pathophysiological mechanism of obesity and its relation to other diseases, in practical prevention it can be used only in individual counseling. In community-based prevention programs, identification and use of different subtypes of obesity are difficult. By preventing overweight in early adulthood, it is likely that a substantial amount of CHD mortality can be prevented.