点击显示 收起
1 From the Department of Medicine, the Division of Cardiology, the Central Hospital in Rogaland, Stavanger, Norway; the Department of Public Health and Primary Health Care, Section for Medical Statistics, the University of Bergen, Bergen, Norway; and the Department of Pharmacotherapeutics, the University of Oslo.
2 Supported by Pharmacia-Upjohn A/S, Oslo, and by Pronova A/S, Oslo. 3 Address reprint requests to DWT Nilsen, Division of Cardiology, Central Hospital in Rogaland, 4011 Stavanger, Norway. E-mail: dnilsen1{at}chello.no.
ABSTRACT
Background: Results of epidemiologic studies and clinical trials indicate that moderate doses of n-3 fatty acids reduce the risk of cardiovascular disease and may improve prognosis.
Objective: The objective was to evaluate the effect of a high-dose ethylester concentrate of n-3 fatty acids administered early after an acute myocardial infarction (MI) on subsequent cardiac events and serum lipids.
Design: Three hundred patients with acute MI were randomly assigned to a daily dose of either 4 g highly concentrated n-3 fatty acids or corn oil, administered in a double-blind manner over 12–24 mo. Median follow-up time was 1.5 y. Clinical follow-up, including the drawing of blood samples, was performed after 6 wk of treatment and later at 0.5-year intervals.
Results: Forty-two (28%) patients in the n-3 group and 36 (24%) in the corn oil group experienced at least one cardiac event (cardiac death, resuscitation, recurrent MI, or unstable angina). No significant difference in prognosis was observed between groups for single or combined cardiac events. Total cholesterol concentrations decreased in both groups, with no significant intergroup differences. On average, the monthly increase in HDL cholesterol was 1.11% in the n-3 group and 0.55% in the corn oil group (P = 0.0016). Triacylglycerol concentrations decreased by 1.30%/mo in the n-3 group, whereas they increased by 0.35%/mo in the corn oil group (P < 0.0001).
Conclusion: No clinical benefit of a high-dose concentrate of n-3 fatty acids compared with corn oil was found despite a favorable effect on serum lipids.
Key Words: n-3 Fatty acids • randomized double-blind clinical study • myocardial infarction • serum triacylglycerol • serum triglyceride • HDL cholesterol
INTRODUCTION
n-3 Fatty acids have several beneficial cardiovascular properties (1–5), including antiatherothrombogenic, antiarrhythmic, and antihypertensive effects, and are especially noted for their triacylglycerol-reducing ability (5–12). These effects may translate into an improved clinical outcome in patients with coronary artery disease (CAD) (12, 13).
Early studies of Greenland Eskimos who consumed a diet rich in n-3 fatty acids emphasized that population's lower coronary mortality as compared with Danish control subjects (14). Moreover, several prospective epidemiologic studies reported significant associations between fish intake and a lower risk of CAD (15–20), although some studies observed no association (21, 22). The Health Professionals Follow-up Study reported beneficial effects of fish compared with no fish intake, but no dose-response relation (23). Similar results were obtained in the US Physicians Health Study (24). In an extension of that study, no association was found between plasma concentrations of n-3 fatty acids and CAD (25). Siscovick et al (26), however, found a reduced risk of primary cardiac arrest with increasing dietary fish intake and with increasing red blood cell membrane concentrations of n-3 polyunsaturated fatty acids. The strongest effect of fish intake was shown in the randomized open Diet and Reinfarction Trial (DART) (27), in which men instructed to eat fish after myocardial infarction (MI) had a 29% decline in all-cause mortality compared with men in the control group. That study was performed before the standard use of aspirin in patients with acute MI, a drug that may partly mask the beneficial effects of n-3 fatty acids.
In the recent GISSI Prevention Study (28), a daily intake of 1 g n-3 fatty acids introduced within 3 mo (median: 16 d) of MI and continuing for 3.5 y resulted in a 15% lower risk (P = 0.023) of the primary combined cardiac endpoint than that in patients with no additional treatment. However, all the benefits were found in the reduced risk of cardiovascular death. Although inconsistent with the lack of a dose-response relation found in epidemiologic studies (23, 24), the pronounced effects obtained in the GISSI Prevention Study (28) by a supplementation dose of n-3 fatty acids may suggest an even stronger effect of pharmacologic doses.
In this randomized double-blind study we investigated the effect of high-dose polyunsaturated n-3 fatty acids introduced early after an acute MI on serum lipid concentrations and clinical prognosis. The comparison group consisted of patients receiving similar capsules containing corn oil.
SUBJECTS AND METHODS
Subjects and intervention
Three hundred patients with an acute MI were recruited at one hospital center (Central Hospital in Rogaland, Stavanger, Norway) from September 1995 until December 1996. Eligibility was based on 1) verified MI by World Health Organization criteria (29), 2) age >18 y, 3) discontinuation of a regular supplementation of other fish-oil products, and 4) signed informed consent. Exclusion criteria consisted of 1) assumed noncompliance to protocol; 2) expected survival <2 y because of severe heart failure (New York Heart Association class IV), malignancy, or other reasons; 3) ongoing gastrointestinal bleeding or verified stomach ulcer; 4) thrombocytopenia or blood platelets <100 x 109/L; 5) liver insufficiency; 6) participation in any other study; and 7) residence outside the recruitment area of this study. All patients were included between the fourth and the eighth day after an acute MI. All patients were randomly assigned to receive 2 gelatin capsules of Omacor-R (Pronova AS, Oslo) or corn oil twice a day. Each capsule contained either 850–882 mg eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) as ethylesters in the average ratio of EPA to DHA of 1:2 or the same amount of corn oil. -Tocopherol (4 mg) was added to all capsules. A detailed patient history was recorded and a clinical examination was performed. The electrocardiogram was scrutinized for infarct location. The recent injury was classified as an anterior or inferior Q-wave infarction or any non-Q-wave infarction. Treatment was initiated immediately after inclusion and collection of baseline blood samples. The study was approved by the Regional Ethics Committee for Western Norway.
Follow-up
Clinical follow-up, including blood tests and electrocardiogram recordings, was repeated at 6 wk, 6 mo, 1 y, 18 mo, and, for some patients, 2 y after the start of treatment. A deviation of 2 wk at the first follow-up and 1 mo at later follow-ups was allowed. All cardiac events and ongoing medication were recorded. Cardiac events were defined as cardiac death, resuscitation, recurrent MI, and unstable angina. Revascularization and death from other causes were also recorded. The study was closed at the end of 1997. The median follow-up time (ignoring events except deaths) was 1.5 y (range: 1 d to 2 y).
Statistical analyses
Potential differences in demographic and clinical characteristics between the 2 treatment groups were examined by chi-square and two-sample t tests. Univariate time-to-event analyses (Kaplan-Meier method) were performed and the Mantel-Cox (log-rank) test was used to examine differences in prognosis between the treatment groups. Hazard ratios (HRs with 95% CIs) were calculated in a Cox proportional hazards regression model. The assumption of homogeneity in effects over time (proportionality) was visually checked in a log minus log plot. The different predefined endpoints were analyzed both combined and separately, according to intention to treat. In analyses of single events, the occurrence of other cardiac events was ignored, but a patient was not considered for further follow-up in the case of death. Additional analyses in which events during the first month were ignored, and in which the analyses were restricted to patients with complete compliance at 6 wk, were also performed.
Demographic and clinical characteristics were further adjusted for in a multiple Cox proportional hazards regression model with the first cardiac event as the endpoint. The likelihood ratio test was used in the statistical testing for differences between groups. Dietary intake of n-3 fatty acids, represented by the average number of fish meals per week during follow-up (categorized as <3, 3–4, and >4) was also adjusted for. Adjustment for drug use was made by including an indicator variable for a particular drug one at a time. Both initial drug use (first 24 h after MI) and use of drugs at follow-up (time-dependent covariates generated on the basis of time of first use) were considered. Additional analyses with adjustment for revascularization (as the time-dependent covariate) were also performed.
Relative changes in serum lipid concentrations (total cholesterol, HDL cholesterol, and triacylglycerol; log transformed) the first year after MI were analyzed in an analysis of covariance model with repeated measures, applying an unstructured covariance matrix (30). Both linear and nonlinear terms were considered in the modeling of time pattern. On the basis of a model with linear terms only, including a time-by-treatment interaction, the average change in serum lipid concentration per time unit for each group was calculated. The Wald's test was used to test for heterogeneity in linear time trend between the groups. Results were adjusted for sex, age at MI (cubic expression), body mass index (linear trend), and diabetes. To adjust for the slightly different spacing between follow-up recordings, a variable representing the difference from a common predefined time point (defined as the median value for all patients) was calculated and included as the covariate. Descriptive statistics for each single time point were also given. The analyses were restricted to patients with complete information at all time points during the first year (before inclusion and after 6 wk, 6 mo, and 12 mo of treatment). To examine for heterogeneity in treatment effect according to age and sex, and some other covariates, two-way interaction terms were included in the statistical models one at a time.
RESULTS
Demographic and clinical characteristics, ongoing medication, and compliance
No significant group differences in demographic or clinical characteristics at inclusion were seen (Table 1). A somewhat larger proportion of the patients in the n-3 group than in the corn oil group (30.2% compared with 24.7%) had previously taken fish-oil supplements, but the difference was not statistically significant.
View this table:
TABLE 1.. Demographic and clinical information at inclusion by treatment group1
The mean values of maximum creatine kinase muscle-brain and troponin T did not differ significantly between the 2 groups (Table 2). No significant difference in MI location was observed, although the proportion of anterior MIs was larger in the corn oil group than in the n-3 group. No significant differences were seen in drug use between the 2 groups (Table 2). ß-Blockers and aspirin were the most commonly used drugs, both initially and in the follow-up period. Statins were also commonly used but were introduced mainly during follow-up. Of the 150 patients in each group, 123 (82%) of those receiving n-3 fatty acids and 129 (86%) of those receiving corn oil fulfilled the criteria for complete compliance after 6 wk.
View this table:
TABLE 2.. Clinical characteristics at inclusion and medication by treatment group1
Cardiac events and revascularizations
A total of 42 (28%) of the patients in the n-3 group and 36 (24%) of the patients in the corn oil group experienced at least one cardiac event during the follow-up period (Table 3). The number of patients experiencing more than one event (of the same or of a different type) was low in both groups. No significant differences were found between the groups in number, type, or severity of event. Unstable angina was the most common type of the predefined cardiac events.
View this table:
TABLE 3.. Distribution of events by treatment group1
The number of revascularizations was slightly higher in the corn oil group, but no significant difference was noted between groups (Table 4). Most of the revascularized patients (30 and 33 patients in the n-3 and corn oil groups, respectively) did not experience any other cardiac event. Two patients in the n-3 group had an episode of unstable angina and one patient in the corn oil group died of a cardiac cause after revascularization.
View this table:
TABLE 4.. Hazard ratio (HR) of experiencing event by treatment group1
In general, the number of cardiac events was highest in the n-3 group, but this pattern was more pronounced for recurrent MIs than for other predefined cardiac events. However, no significant overall differences in prognosis between the groups for single or combined cardiac events were found (Table 4). Kaplan-Meier curves for the cumulative probability of not experiencing a cardiac event (cardiac death, resuscitation, recurrent MI, or unstable angina; first occurrence) according to treatment group are shown in Figure 1. The overall HR was 1.19 times higher in the n-3 group than in the corn oil group. No significant interaction between treatment and sex or age was found, although the difference between the treatment groups was more pronounced for women.
FIGURE 1. . Event-free survival curves (Kaplan-Meier method) by treatment group (150 patients in each group).
The HR for experiencing a cardiac event (any type; first occurrence) was not notably changed when early events were ignored or when the analyses were restricted to patients with complete compliance at 6 wk, but was slightly higher (HR = 1.25; NS) in analyses that both ignored early events and were restricted to patients with complete compliance. Adjustment for sex, age group, smoking status at inclusion, body mass index, fish-oil supplementation before inclusion, average number of fish meals during follow-up, and diabetes did not notably affect the HR for the combined cardiac event (first occurrence, results not shown). Adjustment for hypertension resulted in an HR somewhat closer to unity. Revascularization and medication generally had no notable influence, but adjustment for use of diuretics resulted in an HR almost equal to 1.
Total cholesterol
A significant overall decrease in total cholesterol concentration was observed, but with no significant intergroup differences (Table 5). In additional analyses, a significant nonlinear term indicated a less-pronounced decrease in total cholesterol concentration in the last part of the follow-up period. A significant heterogeneity across age groups (<55, 55–64, 65–74, and >75 y) was found. This age difference was due to a less-pronounced decrease in total cholesterol concentration in the n-3 group in addition to a greater increase in total cholesterol concentration in the corn oil group among patients >75 y (n = 76) than among patients in the younger age groups.
View this table:
TABLE 5.. Serum lipids before inclusion and average change over time by treatment group
In view of additional significant interaction terms with statin medication, separate analyses among patients with and without change in statin medication were performed (Table 5). The subgroup referred to as the "change statins" group consisted of previously untreated patients who received a statin during hospitalization. The "no change" group included mainly patients who received no lipid-lowering treatment before or during hospitalization, but it also included some patients who took statins on a regular basis before hospitalization. In this group a transient increase in total cholesterol concentration followed by a moderate decrease was found. The initial increase in total cholesterol concentration was more pronounced in the corn oil group. In both groups, however, an overall decrease in total cholesterol was still seen. In the change statins group, total cholesterol concentrations decreased considerably during the first 6 wk, somewhat more in the n-3 group, but then flattened off in both groups. No significant heterogeneity in linear time trend between the 2 treatment groups was seen in either of these subanalyses.
HDL cholesterol
A significant overall increase in HDL-cholesterol concentration was observed, significantly more pronounced in the n-3 group (Table 5). In separate analyses of each group, a significant trend was observed in both groups. Allowing for nonlinear terms, however, no significant time pattern was observed in the corn oil group because of a transient decrease followed by a moderate increase. In the n-3 fatty acid group, the increase was less pronounced in the last part of the observation period.
Triacylglycerol
The change in triacylglycerol concentration differed significantly between the groups (Table 5). In the n-3 group concentrations decreased significantly, whereas no significant trend was observed in the corn oil group. When including nonlinear terms in the models, however, a significant but different time pattern was observed in both groups. In the n-3 group, the decrease in triacylglycerol concentration was less pronounced in the last part of the follow-up period. In contrast, an increase followed by a slight decline and flattening off was observed in the corn oil group.
A significant interaction with sex, as well as with statin medication was found (linear time trends considered only). Additional subgroup analyses were therefore carried out (Table 5). The magnitude of change in triacylglycerol concentration differed somewhat in these analyses, although the trend estimates generally had the same direction as in the main analyses. Among men, triacylglycerol concentrations decreased during the total follow-up period, whereas among women, an initial decrease followed by a slight increase was seen. However, the mean values remained below that observed at inclusion in the study. Both for men and women, the favorable time pattern in the n-3 group was more pronounced among patients in whom statins had been introduced before discharge from the hospital. The increase in the triacylglycerol concentration in the corn oil group was more pronounced among women than men, but the introduction of a statin reduced the magnitude of increase considerably between both sexes.
DISCUSSION
In this randomized, double-blind study, we observed no beneficial effect on the hazard of different cardiac events after an acute MI of a high-dose concentrate of n-3 fatty acids compared with corn oil, despite a favorable effect on serum triacylglycerol and HDL cholesterol. In previous studies, beneficial effects of n-3 fatty acids were reported mainly in relation to the risk of cardiovascular disease (15–20, 23, 24). In randomized studies of patients with MI, intervention with n-3 fatty acids resulted in a reduction in total mortality (27, 28, 31), which was probably attributable to a decrease in the risk of sudden death (28, 31). In our study, the low death rate in both treatment groups may be explained by patient selection before inclusion. A short follow-up time (median: 1.5 y) is less likely to explain the lack of effect in our study. In DART (27), which reported a reduction in all-cause mortality, the difference in prognosis in favor of fish advice appeared as early as 3–6 mo after the dietary changes were introduced. Additionally, an Indian study on prognosis after acute MI observed beneficial effects of n-3 fatty acids during the first year of treatment (32).
Dietary habits in our study population together with a possible lack of dose-response relation may explain our findings. Participants resided in a coastal area and all patients received a diet rich in fish products. In the US Physicians Health Study (24), the reduced risk of sudden death was not lowered further at levels of consumption exceeding one fish meal per week. Some other studies reported a lack of dose-response relation between fish intake and cardiovascular disease (23). Dietary intake of n-3 fatty acids in both of our treatment groups was probably comparable with the recommended intake of fish oil in the fish advice group of patients in DART (27) but was far lower than that of the n-3 fatty acid supplementation groups in the GISSI Prevention Study (28). If no dose-response relation exists, a diet rich in fish products in both patient groups may explain the lack of a beneficial effect of the intervention with n-3 fatty acids in our study.
Although we did not observe any beneficial effect on prognosis of a high-dose concentrate of n-3 fatty acids compared with corn oil, we did find an improvement in the lipid profile, including a significant increase in HDL-cholesterol concentration. The serum triacylglycerol-reducing effect of n-3 fatty acids observed in this study is consistent with results from several previous studies (5, 7–11). This effect was enhanced by the introduction of a statin, which also modified the increase in triacylglycerol concentrations observed in the corn oil group.
There is growing epidemiologic evidence that elevated fasting serum triacylglycerol may be an independent risk factor for CAD (12, 13, 33). In the Helsinki Heart Study (34), the reduction in cardiac events was assumed to be related to the triacylglycerol-reducing effect of gemfibrozil. Such a biological explanation model was also anticipated when designing our randomized, double-blind clinical trial. However, it was not clearly shown that the reported beneficial effects of n-3 fatty acids are related to their triacylglycerol-reducing effects, and results of other studies indicate that the main clinical benefits may be due to the antiarrhythmic properties of n-3 fatty acids (2, 4–6, 26, 31, 35–38). In experimental studies, it was also shown that n-6 fatty acids, including corn oil, exert antiarrhythmic properties similar to n-3 fatty acids (35–38). Corn oil has a rather neutral effect on serum lipid concentrations (5). If antiarrhythmic properties are of essential importance, the apparent lack of a beneficial effect of n-3 fatty acids in this study may be related to similar protective effects of corn oil.
It is also possible that the high doses of concentrated n-3 fatty acids applied in this study exceeded some optimal threshold level, outweighing the beneficial effect or even leading to an apparent adverse effect. We chose a daily dose of 4 g n-3 fatty acids, >10 times the dose given in DART and 4 times the dose given in the GISSI Prevention Study. A possible adverse effect of high doses of n-3 fatty acids on cardiac events was hypothesized in previous studies using high-dose ethylester compounds of EPA and DHA (39, 40) on the basis of observations of a reduction in vitamin E and an increase in thiobarbituric acid–reactive substances. A proinflammatory response induced by peroxidation may serve as a biological mechanism for an adverse effect. The nonsignificant increase of recurrent MIs in the n-3 group may be consistent with a dose-optimum hypothesis.
Finally, competing interventions including aspirin may have masked the potential to demonstrate an effect of n-3 fatty acids in our study. DART (27) was carried out in the era before aspirin and thrombolytic agents, whereas >60% of patients in the Mediterranean study (31) were taking antiplatelet agents. Patients in the GISSI Prevention Study (28) were treated with the same pharmacologic background therapy as in this study, but a smaller proportion of patients in the GISSI Study were given statins during follow-up. The proportion of additional drugs did not differ significantly between our 2 patient groups, but the use of diuretics tended to be greater in the n-3 group. It is, however, difficult to evaluate the direction of a potential bias, because some drugs were used mainly for preventive purposes, whereas other drugs were possibly introduced for symptom alleviation, reflecting the general condition of the patient.
In conclusion, we found no clinical benefit of a high-dose concentrate of n-3 fatty acids administered early after an acute MI compared with corn oil, despite an improvement in serum lipid concentrations. The lack of beneficial effect and a tendency toward an adverse effect may be related to a dose optimum below the chosen dose in this trial. It is also possible that corn oil exerted a protective effect of similar or greater strength than n-3 fatty acids or that competing interventions masked a potential difference between our treatment groups.
ACKNOWLEDGMENTS
thank the steering committee: Dennis WT Nilsen, Knud Landmark, Leik Woie, and Michael Abdelnoor.
REFERENCES