点击显示 收起
1 From the Department of Preventive Cardiology (IE, IH, EMH, and ST) and the Center of Clinical Research (MA), Ullev°al University Hospital, Oslo.
2 Address reprint requests to I Ellingsen, Department of Preventive Cardiology, Ullev°al University Hospital, N-0047 Oslo, Norway. E mail: ingrid.ellingsen{at}ulleval.no.
ABSTRACT
Background: In the Oslo Diet and Antismoking Trial, 1232 high-risk men aged 40-49 y were randomly assigned to either a lifestyle
intervention group or a control group for 5 y. The study showed a
significant reduction in ischemic heart disease (IHD) events in the
intervention group.
Objective: Our objective was to examine this cohort 23 y after the start of the trial.
Design: We examined the effect of group assignment on IHD mortality in subjects with a normal (below the median; range: 0.69-2.00 mmol/L; n = 615) or a high (at or above the median; range: 2.01-13.80 mmol/L; n = 617) fasting triacylglycerol concentration in 1972-1973 (at inclusion into the study). We recorded vital status on 31 December 1996 and ascertained causes of death by linkage to Statistics Norway.
Results: In the men with a high triacylglycerol concentration, IHD death occurred in 25 (8.13%) subjects in the intervention group and in 44 (14.2%) subjects in the control group (relative risk: 0.57; 95% CI: 0.36, 0.91; P = 0.02). An adjusted Cox proportional hazards model yielded a hazard ratio of 0.56 (95% CI: 0.34, 0.93; P = 0.027). In the men with a normal triacylglycerol concentration, the intervention had no detectable effect on IHD mortality (adjusted hazard ratio: 1.10; 95% CI: 0.66, 1.83; P = 0.7).
Conclusions: These data suggest that advice to change diet and smoking habits reduced the relative risk of IHD mortality after 23 y in men with high triacylglycerol concentrations. Men with normal triacylglycerol concentrations did not appear to achieve this long-term benefit of lifestyle intervention.
Key Words: Oslo Diet and Antismoking Trial • ischemic heart disease mortality • triacylglycerol • middle-aged men
INTRODUCTION
Several studies suggest that men with a high triacylglycerol
concentration and a high total or LDL-cholesterol concentration may have a particularly high risk of subsequent ischemic
heart disease (IHD) (1-3). High triacylglycerol concentrations
tend to cluster with other features of the insulin resistance or
metabolic syndrome, including abdominal obesity; low HDL-cholesterol concentrations; the presence of small, dense LDL
particles; high blood pressure; and type 2 diabetes (4). Although dietary change, weight loss, and increased physical
activity have been shown to improve several aspects of the
metabolic syndrome (5), the effect of changes in lifestyle on
mortality has not been thoroughly documented.
The Oslo Diet and Antismoking Trial was a 5-y primary prevention trial that showed the protective effect of dietary and antismoking advice against IHD events in high-risk, middle-aged men. The effects of the intervention were described previously at the end of the study (6) and in a further analysis 102 mo after the start of the study (7).
In the present study, we present the results of a 23-y follow-up of the men in the Oslo Diet and Antismoking Trial. We compared the IHD mortality of the men who had a normal triacylglycerol concentration at screening with that of the men who had a high triacylglycerol concentration at screening. Our rationale for doing so included the following considerations. First, the potential of risk reduction in terms of absolute risk may be greater in subjects with a high baseline risk of IHD than in subjects with a low baseline risk (8). Second, subjects with hypertriglyceridemia are often overweight, and dietary changes may reduce body weight in this subgroup. Finally, cigarette smoking worsens insulin sensitivity and may increase the risk of type 2 diabetes (9, 10). Thus, smoking cessation may be especially effective in reducing IHD risk in subjects with the metabolic syndrome.
SUBJECTS AND METHODS
Study population
The Oslo Diet and Antismoking Trial was initiated in 1972 as
a primary prevention trial that sought to establish whether reduction in serum lipid concentrations and cessation of smoking reduced the incidence of IHD (6). The study was extensively described in previous publications (6). Briefly, 1232 healthy men
aged 40-49 y who had an elevated serum total cholesterol concentration or coronary risk score were included in the study from
a pool of 16 202 men who were screened (65% of all men aged
40-49 y in Oslo). All subjects had a normal electrocardiogram at
rest and were free of chest pain during exercise testing; they were
also free of cardiovascular disease, hypertension, diabetes mellitus, cancer, disabling diseases, psychopathologic diseases, and
alcoholism. Men on lipid-lowering diets were excluded. The subjects in the intervention group (n = 604) visited the clinic every 6
mo and were seen by the doctor and dietitian at each session,
whereas the subjects in the control group (n = 628) were seen at
12-mo intervals. Body weight was measured at screening and at
subsequent clinic visits. The study was performed in accord with
the Helsinki Declaration of 1975 as revised in 1983.
Diet and lifestyle history and interventions
A short diet history of 8 food items was evaluated at screening.
The subjects were asked to quantify their consumption of hard
margarine or butter, soybean oil, whole milk, cream, cheese, eggs,
cakes and pastries, and lean or fatty meats. The scores for all 8
dietary items were summed to obtain a total score, with a low total
score indicating a high-risk diet. The maximum possible score was
50 points. Thus, for example, consumption of > 250 g margarine
or butter/wk, consumption of > 1 L whole milk/wk, daily consumption of cream or cakes and pastries, consumption of > 60 g
cheese/d, consumption of > 3 eggs/wk, consumption of 0-50 g
soybean oil/wk, and avoidance of lean meat gave a total score of
0 points (0 points for each item), whereas consumption of 0-50 g
margarine or butter/wk, 0-0.25 L whole milk/wk, or > 250 g
soybean oil/wk gave a score of 8 points for each item. Consumption of wine, beer, and liquor was recorded as none, occasional,
1-2 d/wk, or daily.
At the time the dietary questionnaire was administered, the subjects were asked about marital status, current cigarette smoking (exsmokers and never smokers were counted as nonsmokers), and physical activity at work and during leisure time. Smoking status was reevaluated at 2 and 5 y.
A trained dietitian and the lead investigator (IH) gave dietary advice individually and during group sessions (6). Subjects with elevated total cholesterol concentrations were advised to reduce their saturated fat intake and to slightly increase their polyunsaturated fat intake. Subjects with elevated triacylglycerol concentrations were asked to reduce their total energy intake (mainly by reducing their intakes of sugar, sweet drinks, chocolate, alcohol, and fat). For breakfast and evening meals, all subjects were asked to consume fiber-rich bread with no or only a thin layer of highly polyunsaturated, low-fat margarine. Preferred bread fillings included fish, vegetables, fruit, low sugar jam, and low-fat meats and cheeses. Skim milk and 1 egg/wk were recommended. For main meals, the use of lean meat, fish, and low-fat fish products, together with potatoes and vegetables, was advised. Cigarette smokers were urged to stop smoking. Each subject was monitored individually and given nutritional advice in groups. The results of the blood tests were also used during the nutrition sessions as a monitoring tool.
Laboratory methods
Blood samples were collected by venipuncture at an initial
screening (nonfasting), at a second screening (fasting), at half-yearly intervals, and after 3 and 5 y of follow-up. Concentrations obtained at the second screening were used in determining baseline lipid concentrations for study eligibility and were
collected at the same time as questionnaire and other clinical
data. Thus, the fasting values obtained in the second screening
were used in the present study. These samples were taken after
a 12-h fast. Samples were assayed for serum total cholesterol,
triacylglycerol, and glucose concentrations but not for HDL-cholesterol concentrations (OP Foss, unpublished observations,
1975). The same methods of analysis were used for samples
taken at all visits. For missing values at 5 y, the last available
value was used in the present analysis.
Endpoints
The inclusion date was the date of screening in 1972-1973.
We set a census date of 31 December 1996. Outcomes were
identified by linkage to Statistics Norway. Linkage was based
on the unique 11-digit identification number given to all Norwegian citizens and was approved by the Norwegian data
inspectorate. Outcome status was available for every participant included in the original study. The main outcome measure
was IHD mortality. Codes were defined according to the International Classification of Diseases (ICD). For IHD mortality, the ICD 8th revision (codes 410-414 and 795), the ICD 9th
revision (codes 410-414 and 798.1), and the ICD 10th revision
(codes I21.0-I21.9, I25.0-I25.9, and R96.0) were used. Death
from causes other than IHD was censored (11). Other causes of
death included cerebrovascular disease, other vascular disease,
infections, accidents, alcoholism, asthma, and cancer.
Statistical methods
The subjects were classified according to the median triacylglycerol concentration, and risk factors were compared between
the normal and high triacylglycerol groups according to intervention or control status. For continuous variables, a two-factor analysis of variance with terms for interaction between triacylglycerol
concentration and treatment group was applied. For categorical
variables, the Cochran and Mantel-Haenszel statistics test was
applied by using the Wolf test for interactions.
The subjects' age on 1 January 1972 was used in all analyses. The inclusion period lasted 2 y; thus, the age range of subjects was 40-51 y. Person-time for each subject was calculated from the date of screening to the date of confirmation of death due to IHD or until 31 December 1996, whichever came first. The relative risk (RR) of IHD mortality for assignment to the intervention or the control group was estimated for men whose triacylglycerol concentration was below the median (0.69-2.00 mmol/L) or equal to or greater than the median (2.01-13.80 mmol/L). To test our main hypothesis that the RRs of men with high triacylglycerol would differ from those of men with normal triacylglycerol, we used Cox regression analysis with treatment group (intervention or control), triacylglycerol concentration (below the median or equal to or greater than the median), and their product as confounders (12). A Wald test was applied to test whether the product (interaction term) was statistically significant (13). The test was done by linear regression analysis using triaclyglycerol concentration, treatment group, and triacylglycerol concentration x treatment group. Further analyses were based on 615 men with triacylglycerol concentrations below the median and 617 men with triacylglycerol concentrations equal to or greater than the median. Comparison of univariate survival curves between the treatment groups within each of the triacylglycerol subgroups was performed by using log rank statistics (14).
The following variables were specified for inclusion in the multivariate model because of their potential relation with IHD mortality: age, body mass index (BMI; in kg/m2), cigarette smoking, total cholesterol and triacylglycerol concentrations, glucose concentration, systolic and diastolic blood pressure, total dietary score (or score for each dietary item), alcohol intake, and activity at work or during leisure time. The relations between assignment to the intervention or the control group and IHD mortality were then analyzed with the multivariate Cox proportional hazards model, with initial adjustment for age followed by adjustment for potential confounders. Because of intercollinearity between systolic and diastolic blood pressure, only diastolic blood pressure was entered into the multivariate model. We did not account for the effect of the intervention on IHD risk factors at the end of the intervention because the intention of this analysis was to examine the effect of the intervention on total IHD deaths including deaths duringthe intervention period. All calculations were done with SPSS (version 10; SPSS Inc, Chicago).
RESULTS
Between 1972 and 1996, we accumulated 26 734 person-years of follow up, and 322 deaths, including 136 IHD deaths
(5.1/1000 person-years), occurred. The mean length of follow
up was 22.8 y (maximum: 24.7 y). In the entire cohort, IHD
death occurred in 77 of the 628 men in the control group and
in 59 of the 604 men in the intervention group (P = 0.14). The
Wald interaction term between group assignment and triacylglycerol concentration (ie, below the median or equal to or
above the median) was marginally significant (P = 0.06).
As shown in Table 1, the subgroups with normal or high triacylglycerol concentrations were well balanced with regard to baseline characteristics: there were only minor differences between the 2 subgroups with regard to BMI and egg consumption. In the subgroup with high triacylglycerol concentrations, IHD death occurred in 44 of 310 (14.2%) subjects in the control group and in 25 of 307 (8.1%) subjects in the intervention group, which gave a risk ratio of 0.58 (95% CI: 0.35, 0.94; P = 0.02) (Table 2). Survival curves show that, relative to the control group, the intervention group had an early and persistent decrease in cumulative IHD mortality (log rank 2 = 5.9, P = 0.015) (Figure 1A).
View this table:
TABLE 1. . Baseline characteristics and dietary intakes according to median fasting triacylglycerol concentrations and assignment to the intervention or the control group1
View this table:
TABLE 2. . Relative risks of mortality from ischemic heart disease (IHD) in 1232 men in the Oslo Diet and Antismoking Trial according to median fasting
triacylglycerol concentrations and assignment to the intervention or the control group
FIGURE 1.. Kaplan-Meier survival curves showing the relation between survival from ischemic heart disease and assignment to the intervention
or the control group in men with high (n = 617; log rank 2 = 5.9, P = 0.015) or normal (n = 615; log rank 2 = 0.13, P = 0.7) triacylglycerol
concentrations.
In the subgroup with normal triacylglycerol concentrations,
IHD death occurred in 33 of 318 (10.4%) subjects in the
control group and in 34 of 297 (11.4%) subjects in the intervention group, which resulted in a risk ratio of 1.10 (95% CI:
0.68, 1.78; P = 0.7) (Table 2). Survival curves show that the
cumulative incidence of IHD deaths was not significantly different between the intervention group and the control group
(log rank 2 = 0.13, P = 0.7) (Figure 1B).
The independent multivariate relation obtained with the Cox proportional hazards model between assignment to the intervention or the control group and IHD mortality in the men with high or normal triacylglycerol concentrations is shown in Table 2. Relative to the incidence of IHD deaths in the control group, the incidence of IHD deaths in the intervention group decreased after adjustment for age in the subgroup with high triacylglycerol concentrations but not in the subgroup with normal triacylglycerol concentrations. Adjustment for age, BMI, cigarette smoking, total cholesterol and triacylglycerol concentrations, glucose concentration, diastolic blood pressure, dietary score, alcohol intake, and activity at work and during leisure time reduced the RR minimally, from 0.58 to an adjusted hazard ratio of 0.56 (95% CI: 0.34, 0.93; P = 0.03). In the subgroup with normal triacylglycerol concentrations, there was no effect of assignment to the intervention or the control group on IHD mortality after adjust-ment for the same variables (RR: 1.10; 95% CI: 0.66, 1.83; P = 0.7). Results of the Cox proportional hazards model were not different when separate dietary items (eggs, cream, and cakes or pastries) were entered in the model instead of the total dietary score. The RRs were then 0.50 (95% CI: 0.30, 0.83; P = 0.007) and 1.14 (95% CI: 0.69, 1.90; P = 0.6) in the groups with high and normal triacylglycerol concentrations, respectively.
Total cholesterol and triacylglycerol concentrations were lower in the intervention group than in the control group after 5 y, regardless of baseline triacylglycerol concentrations (Table 3). Smoking prevalence was lower in the intervention group than in the control group.
View this table:
TABLE 3. . Values for ischemic heart disease risk factors at the end of the intervention (after 5 y or last available values) according to median fasting
triacylglycerol concentrations and assignment to the intervention or the control group
DISCUSSION
These data show that advice given to hypertriglyceridemic
men to change dietary habits and stop smoking may be associated with a > 40% reduction in IHD mortality relative to
IHD mortality in the control group after 23 y of follow-up. We
found no reduction in IHD mortality in the men with normal
triacylglycerol concentrations who were given the same advice,
despite apparently similar changes in total cholesterol concentrations in both groups. These findings suggest that the subgroup of high-risk, middle-aged men with hypertriglyceridemia
may benefit more from changes in lifestyle than do their
counterparts without hypertriglyceridemia.
Our dietary history included only 8 food groups, did not estimate energy intake, and was not repeated at the end of the study. Thus, our ability to provide a detailed analysis of the relation between the diet at baseline or changes in the diet and the primary outcome of the study was limited.
There are several likely mechanisms for the effects of the intervention on IHD risk in the hypertriglyceridemic men; these mechanisms include changes in serum lipid concentrations, body weight, and smoking habits. We were unable to attribute the reduction in IHD mortality to any one of these factors, because the changes achieved did not appear to differ between the subjects with high triacylglycerol concentrations and those with normal triacylglycerol concentrations. Thus, the reasons for these results are unclear but may reflect differences in risk, differences in the responses of biological risk factors to lifestyle changes, or chance.
As hypothesized, IHD mortality in the control group was greater in the hypertriglyceridemic men than in the men with normal triacylglycerol concentrations (14.2% compared with 10.4%). These results are in line with previous data showing that the incidence of IHD in men screened for this study who have high total cholesterol and triacylglycerol concentrations is almost triple that of screened subjects who have high cholesterol but low triacylglycerol concentrations (15). This risk in the latter group of subjects (ie, 0.6%/y) is somewhat lower than the usual risk in the population (16). This observation may be explained by the effects of participation in a trial and the yearly follow-up offered to the control group. The risk reduction occurred in the group at higher risk, which mirrored the positive agreement between absolute risk (though not RR) and risk reduction that was reported in a meta-analysis of diet and drug treatments (8). Previous analysis of the results obtained during the Oslo Diet and Antismoking Trial suggested that most of the benefit accrued at the end of the trial was due to the net difference of 10% in serum cholesterol concentrations between the groups (17). Our study may have lacked power to show the effect of this modest reduction in serum cholesterol concentration on long-term IHD mortality among men who carried a lower risk of IHD than did the group of hypertriglyceridemic men.
As expected, the men with hypertriglyceridemia in the present study had a higher BMI than did men with low triacylglycerol concentrations, which reflects a clustering of abnormalities consistent with the metabolic syndrome. Several studies have suggested that this subgroup may benefit more from drug treatments for hyperlipidemia or from lifestyle changes than would subjects without features of the metabolic syndrome. A post hoc analysis of the Helsinki Heart Study suggested that the effects of gemfibrozil on IHD events were substantially greater in the subgroup with a high ratio of LDL to HDL cholesterol and high triacylglycerol concentrations than in other subgroups (18). Similarly, in the Simvastatin Survival Study, patients with elevated LDL-cholesterol, low HDL-cholesterol, and elevated triacylglycerol concentrations received greater benefit with simvastatin therapy than did patients with isolated elevation in LDL-cholesterol concentration (19). Persons with hypertriglyceridemia tend to have a predominance of small, dense LDL particles (subclass pattern B). Studies have shown that LDL subclass is an independent predictor of IHD risk (20). Compared with subclass pattern A, subclass pattern B is associated with a greater reduction in LDL-cholesterol concentration and particle number in response to low-fat, high-carbohydrate diets similar to the diet recommended in our study (21). Subjects with hypercholesterolemia and high triacylglycerol concentrations may not experience the adverse effects on HDL-cholesterol and triacylglycerol concentrations noted with dietary fat restriction in subjects with hypercholesterolemia alone, who in addition might experience greater benefits from reducing egg consumption (22, 23). These findings suggest differential effects of the diet on serum lipid concentrations between normotriglyceridemic and hypertriglyceridemic subjects. Moreover, smoking cessation may be particularly advantageous for subjects with hypertriglyceridemia. VLDLs activate the plasminogen activator inhibitor-1 promoter, which increases thrombogenic capacity (24), and may potentiate the hypercoagulable state induced by smoking (25). However, we found no difference in smoking habits between the high and normal triacylglycerol groups after 5 y, which suggests that the difference in outcome was probably due mainly to other lifestyle factors.
Finally, chance is a possible explanation of these findings. Our study had major strengths, including its ability to trace outcomes among all the original study participants and the length of follow-up in a population given lifestyle advice. However, because of its post hoc design, it remains a hypothesis-generating study that only further investigations can confirm or dismiss. These studies are unlikely to be done, primarily because of ethical reasons.
Two clinical issues emerge. Recommendations to lower the intake of fat and to reduce body weight are widely recognized modalities to prevent death from IHD. In practice, it may be very difficult for patients to follow such advice (26). Moreover, the effects of usual dietary advice on lipid concentrations in large trials involving free-living populations are small (27). These findings suggest that resources to effectuate dietary change could be focused first on men with hypertriglyceridemia, who appear to experience salutary long-term effects. Because cigarette smoking increases morbidity and mortality from a wide range of diseases, advice to quit smoking is applicable to all persons.
ACKNOWLEDGMENTS
We thank Ingar Holme for helpful comments and Lise Lund H°aheim for
assistance with the coding of causes of death.
IE provided the idea for the study, performed the analysis, wrote most of the Subjects and Methods and Results sections, and contributed to the writing of the Introduction and Discussion sections. IH was responsible for the original study and commented on the present analyses. MA was the statistical consultant. EMH assisted in providing the mortality data from Statistics Norway. ST assisted IE with all stages of the work and wrote most of the Introduction and Discussion sections. There were no conflicts of interest.
REFERENCES