点击显示 收起
1 From the Department of Clinical Epidemiology, Aarhus University Hospital and Aalborg Hospital, Aarhus, Denmark (SPJ and HTS); the Department of Epidemiology and Social Medicine, University of Aarhus, Aarhus, Denmark (SPJ and KO); the Danish Cancer Society, Institute of Cancer Epidemiology, Copenhagen (CS and AT); and the Department of Cardiology and Internal Medicine, Aarhus County Hospital and Aarhus University Hospital, Aarhus, Denmark (SEH).
2 Supported by The Danish Cancer Society, Europe Against Cancer, The Danish Heart Foundation (grants 99-1-4-77-22703 and 99-2-4-95-22771), Hjernesagen, and The Danish Medical Research Council through Vestdansk Sundhedsvidenskabeligt Forskningsforum. 3 Reprints not available. Address correspondence to SP Johnsen, Department of Clinical Epidemiology, University of Aarhus, Vennelyst Boulevard 6, DK-8000 Aarhus C, Denmark. E-mail: spj{at}soci.au.dk.
ABSTRACT
Background: Previous studies have suggested that a high dietary intake of fruit and vegetables is associated with a reduced risk of ischemic stroke. The magnitude of the effect is uncertain, and only one study reported data on the intake of specific fruit and vegetables and the risk of stroke.
Objective: We examined whether the intake of fruit and vegetables is associated with a reduced risk of ischemic stroke, with particular attention paid to specific fruit and vegetables and subtypes of ischemic stroke.
Design: In a prospective cohort study of 54 506 men and women who were included in the Danish Diet, Cancer, and Health study from 1993 to 1997, estimated total intakes of fruit and vegetables (in g/d) were extracted from a semiquantitative food-frequency questionnaire completed at baseline. Data about subjects hospitalized with ischemic stroke were obtained from the Danish National Registry of Patients and were verified later by record reviews. The follow-up for ischemic stroke ended on the date of a first hospital admission for stroke or transient ischemic attack, the date of death or emigration, or the end of the study, whichever came first.
Results: We identified 266 cases of ischemic stroke involving hospitalization during 168 388 person-years of follow-up (median follow-up: 3.09 y; range: 0.02–5.10 y). After adjustment for potential confounders, persons in the top quintile of fruit and vegetable intake (median: 673 g/d) had a risk ratio of ischemic stroke of 0.72 (95% CI: 0.47, 1.12) relative to persons in the bottom quintile of intake (median: 147 g/d) (P for trend = 0.04). When comparing the top quintile with the bottom quintile, an inverse association was most evident for fruit intake (risk ratio: 0.60; 95% CI: 0.38, 0.95; P for trend = 0.02). Similar risk estimates were seen for most types of fruit and vegetables, although the risks were significant only for citrus fruit.
Conclusion: An increased intake of fruit may reduce the risk of ischemic stroke.
Key Words: Nutrition • diet • fruit • vegetables • ischemic stroke • prospective study • epidemiology • cerebrovascular disorders • Diet • Cancer • and Health study
INTRODUCTION
Stroke is a leading cause of death and disability, with major global public health implications. Observational studies have provided strong evidence that lifestyle factors, possibly including diet, may play an important role in the etiology of stroke (1–3).
The relation between diet and the risk of stroke has been extensively examined. However, most studies focused on specific dietary components rather than on foods (4–13). Intakes of antioxidants (ie, vitamin C, ß-carotene, and flavonoids), potassium, and fiber have been associated with a reduced risk of stroke (4–8). These findings have generated an increased interest in foods that are rich in these substances, such as fruit, vegetables, and other plant foods (12–15). This interest is further stimulated by previous experiences, which have taught us that, when trying to unravel the complexity of nutrient interactions in the relation between foods and health, focusing on single nutrients may be too simple an approach (16).
The few prospective studies on the intake of fruit and vegetables and the risk of stroke (6, 12, 13, 17, 18) have shown either inverse associations or no associations. Thus, there is still uncertainty regarding the magnitude of any preventive effect. Furthermore, most of the studies had several limitations because they were based on crude assessments of diet and limited adjustment for confounding factors and because they included only a few outcomes, with no distinction between subtypes of stroke. At present, only one study has reported on the association between the intake of specific fruit and vegetables and the risk of stroke (13). The reduced risk of stroke in that study was most evident for the intakes of cruciferous vegetables; green, leafy vegetables; citrus fruit including juice; and citrus fruit juice alone (13). On the basis of data from a prospective study on diet and health, we examined whether the intake of fruit and vegetables is associated with a reduced risk of ischemic stroke, with special reference to specific fruit and vegetables and different subtypes of ischemic stroke.
SUBJECTS AND METHODS
Design and study population
The Diet, Cancer, and Health study is a prospective cohort study with the primary aim of studying the etiological role of diet in cancer risk. The study design is described in detail elsewhere (19). From December 1993 through May 1997, 80 996 men and 79 729 women aged 50–64 y were invited to participate in the study; 27 177 men and 29 876 women accepted the invitation. Eligible cohort members were born in Denmark, were living in the Copenhagen and Aarhus areas, and had no previous cancer diagnosis in the Danish Cancer Registry. The baseline data were linked to the Danish Cancer Registry and other nationwide population-based registries, including the National Registry of Patients and the Civil Registration System, by using the civil registry number, which is the unique number given to every Danish citizen at birth. Information on death or emigration was obtained through record linkage with the Civil Registration System, which has kept electronic records of all changes in vital status, including change in address, date of emigration, and date of death, for the entire Danish population since 1968.
The Danish National Registry of Patients, established in 1977, records 99.4% of all discharges from nonpsychiatric hospitals in Denmark (20). The data include the civil registry number, dates of admission and discharge, surgical procedures performed, and up to 20 discharge diagnoses, which were classified until 1993 according to the Danish version of the International Classification of Diseases, 8th revision (ICD-8), and subsequently according to the corresponding national version of ICD-10 (20). All discharge diagnoses are assigned exclusively by the physician who discharges the patient. To study incident cases of stroke, we excluded participants who had been hospitalized before enrollment with cardiovascular disease, ie, stroke, transient ischemic attack, ischemic heart disease, or peripheral arteriosclerosis, as recorded in the National Registry of Patients (ICD-8 codes: 410–414, 430–438, and 440, and ICD-10 codes G45, I20–25, and I60–70). Both the Diet, Cancer, and Health study and the present substudy were approved by the Regional Ethics Committees in Copenhagen and Aarhus and by The Danish Data Protection Agency.
Baseline data
All cohort members completed a detailed 192-item semiquantitative food-frequency questionnaire (FFQ). Descriptions of the development and validation of the questionnaire were published previously (21, 22). All the participants also completed a questionnaire about other lifestyle factors, such as smoking habits, alcohol intake, physical activity, health, education, and, for women, use of hormone replacement therapy. Height, weight, waist and hip circumferences, blood pressure, and total serum cholesterol concentrations were also measured at baseline.
Intake of fruit and vegetables
We estimated intakes of fruit and vegetables from the FFQ, in which the participants reported their average intake of different food items over the previous year within 1 of 12 possible intake categories, which ranged from never to ≥ 8 times/d. Daily intakes of specific foods and nutrients, including 42 different types of fruit and vegetables, were calculated from the FFQ for each participant by using the software program Food Calc (23). Standard recipes and sex-specific portion sizes were applied to calculate intake in grams per day by using data from different sources, ie, the 1995 Danish National Dietary Survey (24), 24-h diet recall interviews from 3818 of the study participants (25), and various cookery books.
Fruit and vegetables were grouped into the following 10 categories according to the classifications of the American Institute for Cancer Research (26): leafy vegetables (eg, lettuce and spinach), fruiting vegetables (ie, avocados, cucumbers, aubergines, peppers, peas, pumpkins, zucchini, green beans, and tomatoes), root vegetables other than potatoes (eg, carrots), cruciferous vegetables (eg, broccoli, Brussels sprouts, cauliflower, and cabbage), mushrooms, onion and garlic, stalk vegetables (eg, leeks and asparagus), citrus fruit (eg, oranges and grapefruit), other fruit (eg, apples and pears), and fruit and vegetable juices. Thus, fruit and vegetables were grouped by using a combination of botanical and culinary criteria. Potatoes were not included in the analyses.
Stroke data
A detailed description of the assessment of cases of stroke in the cohort was published previously (26). We identified probable cases of ischemic stroke within the cohort from The Danish National Registry of Patients; these were cases with ICD-10 discharge diagnoses of I60–69.8 and G45 through 1998 for persons living in the Copenhagen area and through 1999 for persons living in the Aarhus area. Medical records and hospital discharge letters were retrieved and reviewed by a single reviewer using a detailed standardized form (26).
We used the World Health Organization’s definition of stroke, ie, an acute disturbance of focal or global cerebral function with symptoms lasting more than 24 h or leading to death of presumed vascular origin (27). A computed tomography or magnetic resonance scan, a spinal fluid examination, or an autopsy or operation description was necessary to distinguish between ischemic infarction, intracerebral hemorrhage, and subarachnoid hemorrhage. All cases of ischemic stroke included in this study were verified by a computed tomography or magnetic resonance scan.
We subclassified all cases of ischemic stroke on the basis of the presumed etiology according to the following Trial of Org 10172 (a low-molecular-weight heparinoid) in Acute Stroke Treatment (TOAST) classifications: large-artery atherosclerosis, cardioembolism, small-vessel occlusion, stroke of other determined etiology, and stroke of undetermined etiology (28). These classifications are based on clinical features, ie, cortical or cerebellar dysfunction and lacunar syndrome, and on data (eg, location and size of infarct) collected by tests such as brain imaging, cardiac imaging, duplex imaging of extracranial arteries, arteriography, and laboratory assessments for a prothrombotic state.
Data analysis
Follow-up for ischemic stroke began on the date on which the subjects visited one of the study centers and ended on the date of their first hospital admission for ischemic stroke, the date of a censoring event (ie, hospitalization for intracerebral hemorrhage, subarachnoid hemorrhage, unspecified stroke, or transient ischemic attack; death; or emigration), or 31 December 1998 for participants living in the Copenhagen area or 31 December 1999 for participants living in the Aarhus area, whichever came first. We categorized the intake of fruit and vegetables (in g/d) into quintiles, and the risk of ischemic stroke was compared between the quintiles by using the lowest quintile as the reference. We used the nonparametric test developed by Cuzick (29) to test for trends in the distribution of the potentially confounding variables across the quintiles. A chi-square test was used for categorical variables.
We used Cox proportional hazards regression to obtain risk ratios (RRs) adjusted for possible confounders, ie, sex, total energy intake, smoking status, blood pressure, serum cholesterol, diabetes mellitus, body mass index, alcohol intake, intake of red meat and n-3 polyunsaturated fatty acids, physical activity, and education. We used 2 approaches for fitting the multivariate models to first identify the most important confounding factors and then to adjust for all available possible confounders. First, to adjust for confounding, we used the change-in-estimate method, in which variables were selected on the basis of changes in the risk estimates of interest (adjusted model) (30). Besides diet, the variables included in these models were sex, total energy intake, and smoking. Second, we extended this adjusted model with the additional variables (full model), ie, systolic and diastolic blood pressure, serum cholesterol at baseline, diabetes mellitus, body mass index, alcohol intake, intake of red meat and n-3 polyunsaturated fatty acids, physical activity, and education. Age was used as the time axis, ie, for subjects who entered the study in December 1993, age at the time of inclusion in the study and age at the time of censoring were equal to chronological age at those time points. However, for subjects who entered the study later, we adjusted for their delayed entry at the time of enrollment because the age at which the participants became at risk (ie, the length of time from birth to the start of the study) was not identical with the age at which they came under observation (ie, the length of time from birth to inclusion in the study). We thus adjusted for delayed entry by defining the age at which the participants became at risk and the age at which they came under observation as separate time points, ie, time from birth to the start of the study and time from birth to inclusion in the study, respectively. The assumption of proportional hazards in the Cox models was evaluated by using graphical assessment and was found to be appropriate in all models.
We adjusted for energy intake by using the residual method (31), ie, residuals were computed from a linear regression model with fruit and vegetable intake as the dependent variable and total energy intake as the independent variable. The predicted fruit and vegetable intake for a person having an energy intake equal to the mean value for the study population was added as a constant. The total energy intake was also included in the models.
Present tobacco consumption (in g/d) was calculated by equating a cigarette with 1 g tobacco, a cheroot or pipe with 3 g, and a cigar with 4.5 g. Smoking status was included in the models at 5 levels: never smoker, former smoker, and current smoker of 1–14, 15–24, or ≥ 25 g tobacco/d. Systolic and diastolic blood pressure at baseline, body mass index at baseline, intake of red meat and n-3 polyunsaturated fatty acids (in g/d), and recreational physical activity (in h/wk) were included as continuous variables, whereas total serum cholesterol concentration was included as a dichotomized variable (≤ 6 compared with > 6 mmol/L), and length of education after elementary school was included with 4 categories (0, < 3, 3–4, and > 4 y).
We also used separate Cox models to carry out analyses stratified by sex and smoking status and analyses that included product terms to assess possible effect modification by these variables. Finally, the analyses were repeated according to different subtypes of ischemic stroke, ie, large-artery atherosclerosis, cardioembolism, small-vessel occlusion, and stroke of undetermined etiology; official Danish recommendations (daily intake of fruit and vegetables ≥ 600 g); and length of follow-up (< 1 compared with ≥ 1 y).
We estimated P for trends in stroke risk according to intake of fruit and vegetables on the basis of the Wald test, after inclusion of the individual exposure variables as continuous measures. We calculated 95% CIs throughout the analyses. All analyses were performed with the use of STATA statistical software (release 7.0; Stata Corporation, College Station, TX).
RESULTS
The cohort included 57 053 persons at baseline. We excluded 2500 persons (4.4%) who were previously registered with cardiovascular disease in the Danish National Registry of Patients and 47 persons (0.08%; including 2 persons with stroke) who left ≥ 10 items blank on the FFQ or who had ≥ 7 items with implausible values. The study population thus included 54 506 persons, who provided a total of 168 388 person-years of risk (median: 3.09 y; range: 0.02–5.10 y). Information on all variables was available for 53 035 persons (97.3% of the persons included in the present study).
Characteristics of the study participants by quintile of total fruit and vegetable intake (in g/d) are shown in Table 1. The median intake of fruit and vegetables ranged from 147 g/d in the lowest quintile to 673 g/d in the highest quintile. Total intake of fruit and vegetables was positively associated with the proportion of women (ie, 39.7% of the subjects in the lowest quintile were women, whereas 67.7% of the subjects in the highest quintile were women), with the proportions of subjects having a higher education and diabetes mellitus, and with total energy intake. By contrast, total intake of fruit and vegetables was negatively associated with the proportion of current smokers (ie, 53.2% of the subjects in the lowest quintile were current smokers, whereas 25.1% of the subjects in the highest quintile were current smokers) and with the proportions of subjects having hypertension, high cholesterol, and high alcohol intake.
View this table:
TABLE 1 . Characteristics of study participants by quintile of total fruit and vegetable intake (in g/d)1
We verified 266 instances of first-time acute ischemic stroke among the study participants during follow-up. These events included 27 (10.2%) instances of stroke due to large-artery atherosclerosis, 24 (9.0%) instances of stroke due to cardioembolism, 115 (43.2%) instances of stroke due to small-vessel occlusion, 2 (0.8%) instances of stroke due to other determined etiology, and 98 (36.8%) instances of stroke of undetermined etiology.
RRs of acute ischemic stroke by quintile of fruit and vegetable intake are shown in Table 2. The total combined intake of fruit and vegetables, as well as the total separate intakes of fruit and of vegetables, was associated with a reduced risk of acute ischemic stroke in the unadjusted models; relative to the lowest quintile, the RRs for the highest quintile were between 0.40 and 0.63. Adjustment for sex, total energy intake, and smoking status weakened the association (RRs: 0.52–0.85); however, the directions of the associations remained unchanged. Whether energy intake was adjusted for by residuals of the dietary exposure variable alone or by residuals and total energy intake in combination had no influence on the estimates. The RRs were further weakened after additional adjustment for systolic and diastolic blood pressure, serum cholesterol at baseline, diabetes mellitus, body mass index, alcohol intake, intake of red meat and n-3 polyunsaturated fatty acids, physical activity, and education; however, the effect of adjustment for these factors was modest in most of the models. The lowest RR in the full model was seen in the highest quintile of fruit intake (RR = 0.60; 95% CI: 0.38, 0.95; P = 0.03).
View this table:
TABLE 2 . Risk ratios and 95% CIs of ischemic stroke by quintile (Q) of fruit and vegetable intake1
Regarding specific types of fruit and vegetables, the results indicated a reduced risk of acute ischemic stroke for most of the items except mushrooms, onion and garlic, and stalk vegetables: relative to the lowest quintiles, crude RRs for the highest quintiles were between 0.45 and 0.69 (Table 3). Adjustment for the most influential confounding factors weakened these associations, making most of them nonsignificant. Additional adjustment for potentially confounding factors had a minor effect on the risk estimates. We found the lowest RRs in the full model in the highest quintiles of intake of citrus fruit (RR = 0.63; 95% CI: 0.41, 0.96; P = 0.03) and other fruit (RR = 0.67; 95% CI: 0.43, 1.04; P = 0.08).
View this table:
TABLE 3 . Risk ratios and 95% CIs of ischemic stroke by quintile (Q) of intake of specific groups of fruit and vegetables1
Some variation in the risk estimates, eg, when comparing the highest and the lowest quintiles of total fruit and vegetable intake, was seen when we stratified by smoking status (never or former smokers compared with current smokers [the RRs (and 95% CIs) for the nonsmokers and the smokers were 0.93 (0.47, 1.82) and 0.65 (0.35, 1.20), respectively (P = 0.37)]) and sex [the RRs (and 95% CIs) for the men and the women were 0.88 (0.49, 1.60) and 0.68 (0.34, 1.40), respectively (P = 0.62)]. However, none of these differences was significant.
The associations between fruit and vegetable intake and ischemic stroke were also evaluated separately for different subtypes of ischemic stroke because the pathophysiologic mechanisms involved in the different subtypes are likely to be different. No differences were found; however, some of the risk estimates were imprecise because of the relatively low number of outcomes, particularly for large-artery atherosclerosis and cardioembolism (data not shown). The lowest RRs were found in the highest quintile of fruit intake for strokes due to small-vessel occlusion [RR = 0.67 (95% CI: 0.34, 1.31)] and for strokes of undetermined etiology [RR = 0.50 (95% CI: 0.23, 1.08)]. However, inverse, nonsignificant associations were found for all types of ischemic stroke.
Relative to the participants with a daily intake of < 100 g fruit and vegetables, the participants with a daily intake of ≥ 600 g, which is in accordance with official Danish recommendations, had an RR of acute ischemic stroke of 0.63 (95% CI: 0.33, 1.22) (full model). We searched for any preclinical diseases at baseline that might have affected diet or other lifestyle factors; however, when instances involving potential preclinical diseases were excluded from the first year of follow-up, the risk estimates did not change (data not shown).
DISCUSSION
In this large cohort study, which included persons aged 50–64 y at baseline, the intake of fruit, in particular citrus and other fruit, was associated with a reduced risk of acute ischemic stroke. We found no clear association between the intake of vegetables and the risk of acute ischemic stroke. The reduced risks were most evident for strokes due to small-vessel occlusion and strokes of undetermined etiology. The main strengths of our study are the large number of instances; the validated and comprehensive assessment of dietary intake; the detailed information on a wide range of potentially confounding factors; the complete follow-up through nationwide, population-based registries, which limit the risk of selection and surveillance bias; and the standardized assessment of all registered outcome events.
For the identification of potential instances of stroke, we relied on the coding of cerebrovascular diagnoses by hospital physicians at the time the patients were discharged. Thus, we were able to include only strokes that led to hospitalization. However, given the age profile of our study cohort, it is likely that most patients with clinical symptoms of acute stroke were referred to a hospital for further evaluation and that any loss to follow-up was likely to be nondifferential and modest because of the short follow-up period.
Although we adjusted for several confounding factors in the analyses, the possibility of residual confounding still remains. A high intake of fruit and vegetables, and particularly a high intake of fruit, might also be an indicator of an overall more healthy lifestyle (32) that, besides including measurable behavioral elements such as being a nonsmoker and having a moderate alcohol intake, also includes a wide range of more subtle behavioral and psychosocial elements. Furthermore, knowledge about diet and attitudes toward health and food are likely to be associated with fruit and vegetable intake (33). In contrast, it may also be argued that adjustment for potential intermediate factors in the causal pathway between fruit and vegetable intake and the risk of ischemic stroke, eg, blood pressure and serum cholesterol, is troublesome and could result in risk estimates that are too conservative.
Finally, both the exposure and the confounder data may have been subject to misclassification, which, because of the prospective nature of our study design, was probably independent of the assessment of stroke, ie, nondifferential. Nondifferential misclassification of exposure data, eg, because of the imprecision of the dietary questionnaire at baseline or because of changes in dietary practices among the participants during follow-up, would lessen our ability to detect any association between fruit and vegetable intake and the risk of acute ischemic stroke, whereas nondifferential misclassification of the confounder variables may lead to residual confounding and could result in both attenuation and inflation of the risk estimates.
Our findings are consistent with those of the few previous prospective studies (6, 12, 13, 17, 18). On the basis of the Nurses’ Health Study and the Health Professionals’ Follow-up Study, Joshipura et al (13) reported an RR of 0.69 (95% CI: 0.52, 0.91) for the highest quintile of fruit intake (median servings/d: men, 4.54; women, 4.33) relative to the bottom quintile (median servings/d: men, 0.86; women, 0.72), which is similar to the RR of 0.60 found in our study if servings are assumed to be 100 g each. They also found a weaker inverse association between the intake of vegetables alone and the risk of ischemic stroke (RR = 0.90) when they compared the highest quintiles of intake (median servings/d: men, 6.21; women, 5.37) with the lowest quintiles (median servings/d: men, 1.60; women, 1.36). By contrast, in a study based on the Framingham cohort, Gillman et al (12) did not find that the protective effect of fruit was stronger than that of vegetables. However, their study included only 97 cases of stroke or transient ischemic attack. Nonetheless, each increment of 3 daily servings of fruit or vegetables was associated with an RR for stroke or transient ischemic attack of 0.81 (95% CI: 0.56, 1.19) and 0.74 (95% CI: 0.54, 1.02), respectively.
With regard to specific types of fruit and vegetables, note that a reduced risk of ischemic stroke was most evident for items that are usually eaten raw, ie, leafy vegetables, citrus fruit, and other fruit. The potentially protective effect of citrus fruit in our study is also in accordance with previous findings on fruit and vegetables that are rich in vitamin C (13) and may indicate a role for vitamin C. In contrast, whereas other studies indicated that garlic may have beneficial health effects in relation to cardiovascular disease and cancer (34), we found no reduced risk among the participants with the highest intake of onion and garlic. However, the intake of garlic was very low within our study population, which made us unable to study this issue in further detail.
Several mechanisms may be involved in the apparent protective effect of fruit and vegetables on the risk of ischemic stroke. Several fruit and vegetable constituents, including micronutrients, antioxidants, phytochemicals, and fiber, have been related to a decreased risk of stroke and other cardiovascular diseases in animal models and observational epidemiologic studies (4–8). However, the evidence on the role of the individual constituents has so far been inconclusive, and the results from testing several of these potentially bioactive compounds alone or in combination in randomized trials on cardiovascular disease and cancer have been disappointing, especially in persons consuming a Western diet (15, 35–39). There may be various explanations for these findings, eg, inexpedient dosages and short study duration. Furthermore, the role of single nutrients may not be easily isolated from the complex biochemical content of plant foods. In addition, attempts at making causal inferences in nutritional epidemiology between specific constituents of fruit and vegetables and later risk of disease are hampered by the varying content of biologically active constituents, eg, vitamin C, ß-carotene, and flavonoids, in fruit and vegetables, which depends on geographical origin, season of the year, methods of storing and cooking, etc. These issues make it difficult to study the specific nutrients, and they are also troublesome for attempts to classify fruit and vegetables by amounts of specific constituents and micronutrients.
In conclusion, previous studies have suggested a protective effect of a high intake of fruit and vegetables on the risk of ischemic stroke. Our results from this large Danish cohort study support the hypothesis that a high dietary intake of fruit is independently associated with a reduced risk of ischemic stroke.
ACKNOWLEDGMENTS
SPJ, KO, and HTS designed the study. SPJ, KO, CS, AT, and SEH collected the data. SPJ, KO, and HTS analyzed the data. SPJ prepared the manuscript, which was reviewed by all of the other authors. The authors were not aware of any conflicts of interest.
REFERENCES