Customize your JAMA Network experience by selecting one or more topics from the list below.
Joshipura KJ, Ascherio A, Manson JE, et al. Fruit and Vegetable Intake in Relation to Risk of Ischemic Stroke. JAMA. 1999;282(13):1233–1239. doi:10.1001/jama.282.13.1233
Author Affiliations: Departments of Epidemiology (Drs Joshipura, Ascherio, Manson, Stampfer, Rimm, Spiegelman, and Willett), Nutrition (Drs Ascherio, Stampfer, Rimm, and Willett), Environmental Health (Dr Speizer), and Biostatistics (Dr Spiegelman), Harvard School of Public Health; Department of Oral Health Policy and Epidemiology (Dr Joshipura), Harvard School of Dental Medicine; Channing Laboratory (Drs Stampfer, Rimm, Speizer, and Willett) and the Division of Preventive Medicine (Dr Manson), Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Mass. Dr Hennekens now resides in Boca Raton, Fla.
Context Few studies have evaluated the relationship between fruit and vegetable
intake and cardiovascular disease.
Objective To examine the associations between fruit and vegetable intake and ischemic
Design, Setting, and Subjects Prospective cohort studies, including 75,596 women aged 34 to 59 years
in the Nurses' Health Study with 14 years of follow-up (1980-1994), and 38,683
men aged 40 to 75 years in the Health Professionals' Follow-up Study with
8 years of follow-up (1986-1994). All individuals were free of cardiovascular
disease, cancer, and diabetes at baseline.
Main Outcome Measure Incidence of ischemic stroke by quintile of fruit and vegetable intake.
Results A total of 366 women and 204 men had an ischemic stroke. After controlling
for standard cardiovascular risk factors, persons in the highest quintile
of fruit and vegetable intake (median of 5.1 servings per day among men and
5.8 servings per day among women) had a relative risk (RR) of 0.69 (95% confidence
interval [CI], 0.52-0.92) compared with those in the lowest quintile. An increment
of 1 serving per day of fruits or vegetables was associated with a 6% lower
risk of ischemic stroke (RR, 0.94; 95% CI, 0.90-0.99; P=.01, test for trend). Cruciferous vegetables (RR, 0.68 for an increment
of 1 serving per day; 95% CI, 0.49-0.94), green leafy vegetables (RR, 0.79;
95% CI, 0.62-0.99), citrus fruit including juice (RR, 0.81; 95% CI, 0.68-0.96),
and citrus fruit juice (RR, 0.75; 95% CI, 0.61-0.93) contributed most to the
apparent protective effect of total fruits and vegetables. Legumes or potatoes
were not associated with lower ischemic stroke risk. The multivariate pooled
RR for total stroke was 0.96 (95% CI, 0.93-1.00) for each increment of 2 servings
Conclusions These data support a protective relationship between consumption of
fruit and vegetables—particularly cruciferous and green leafy vegetables
and citrus fruit and juice—and ischemic stroke risk.
In several articles, nutrients in fruit and vegetables such as dietary
fiber, potassium, and antioxidants have been associated with reduced risk
of cardiovascular disease (CVD).1-5
However, the data directly relating fruit and vegetable intake and CVD are
sparse, as summarized in a recent review,6
and the results of studies that specifically examined fruit and vegetables
as a group were inconsistent.6 Furthermore,
many studies were small with few end points, dietary assessments were often
crude and available only at baseline, and the distinction was often not made
between hemorrhagic and ischemic strokes, which have different causes. To
understand better the relationship between diet and stroke, we examined intakes
of specific fruits and vegetables, as well as overall fruit and vegetable
consumption, in relation to the incidence of ischemic stroke in 2 large cohorts
of women and men.
Data for this analysis were derived from 2 ongoing cohorts, the Nurses'
Health Study (NHS)7 and the Health Professionals'
Follow-up Study (HPFS).1 The 2 studies are
similar in design; participants in both cohorts have been completing mailed
questionnaires every 2 years to provide information on medical history, health
behaviors, and the occurrence of cardiovascular and other outcomes. The NHS
enrolled 121,700 female registered nurses aged 30 to 55 years in 1976.
The HPFS study population is composed of 51,529 male health professionals
including dentists, veterinarians, pharmacists, optometrists, osteopathic
physicians, and podiatrists who were 40 to 75 years of age in 1986.
After 4 mailings, 98,462 women returned the 1980 NHS dietary questionnaire.
We excluded those with 10 or more items blank, those with implausible scores
for total food intake (<2092 kJ/d or >14,644 kJ/d) (n=5994), and those
with previously diagnosed cancer, diabetes, or CVDs (n=7254). Among the men,
we excluded 6442 participants who reported CVD, cancer, or diabetes prior
to the follow-up period. We further excluded 1349 men who reported daily energy
intake outside the plausible range of 3360 to 17,640 kJ/d or who left 70 or
more of the 131 dietary questions blank. More than 90% of the baseline population,
on average, responded to at least 1 follow-up questionnaire, and about 80%
completed each of the repeated dietary questionnaires during follow-up in
these cohorts. The total number of subjects included in this analysis were
75,596 women and 38,683 men (Table 1).
Participants (or next of kin for the deceased) reporting an incident
stroke on a follow-up questionnaire were asked permission to have their medical
records reviewed. Using these records, including imaging results recommended
by the National Survey of Stroke,8 we subclassified
the strokes into ischemic (embolic or thrombotic), hemorrhagic (subarachnoid
or intracerebral), or unknown type. Cases attributed to infections or malignant
processes or cases of indeterminate age discovered on computed tomographic
or magnetic resonancing imaging scans without acute symptoms were not included.
Physicians reviewing the medical records were unaware of the participants'
dietary status. Deaths were reported by next of kin or coworkers or were obtained
from postal authorities or from the National Death Index. Only strokes confirmed
by medical records were included in this analysis. We considered nonresponders
who were not listed on the National Death Index as noncases.
In the NHS, diet was assessed in 1980, 1984, 1986, and 1990. A 61-item
semiquantitative food frequency questionnaire (FFQ) including 6 fruit and
11 vegetable items, and potatoes was included in 1980. An expanded 116-item
FFQ including 15 fruit items, 28 vegetable items, and potatoes were included
in 1984; similar questionnaires were repeated in 1986 and 1990. In the HPFS,
diet was assessed in 1986 and 1990 using an expanded semiquantitative FFQ
similar to that used in the NHS. Further details about the NHS and the HPFS
have been described elsewhere.1,7
On dietary questionnaires, participants reported their average intake
of a specified portion size for each food over the past year. For each food
item on the questionnaire, 9 responses were possible, ranging from never or
less than once per month to 6 or more times per day. Detailed descriptions
of the reproducibility and validity of the FFQ for both men and women have
been published previously.9-11
Validity of the dietary data has been documented by comparisons with multiple
weighted dietary records correcting for within-person weekly variation in
diet. The correlation coefficients for responses to the 61-item questionnaire
in the NHS for fruits and vegetables compared with four 1-week diet records
averaged 0.57 (ranging from 0.27-0.93). In the HPFS, Pearson correlations
between intake from the questionnaire and the diet records, corrected for
within-person weekly variation in diet record, were on average approximately
0.7 for intake of individual fruits and 0.5 for intake of individual vegetables.
Responses to the individual food items were converted to average daily
intake of each fruit and vegetable item for each participant. The average
daily intakes of individual food items were combined to compute total fruit
and vegetable intake, as well as intakes of composite fruit and vegetable
groups. The definitions of composite groups were based on a report by Steinmetz
et al12; the groups were modified to relate
to our questionnaires and hypothesis (available on request from the authors).
Vitamin C–rich fruits and vegetables included items containing more
than 30 mg of vitamin C per serving. For total vegetable intake, we did not
include potatoes, tofu and soybeans, dried beans and lentils, or items with
small portion sizes such as chili sauce and garlic. Missing individual components
were counted as zero intake when aggregating items to compute the composite
We recorded 670 total strokes among women: 366 new cases of ischemic
stroke, 198 new cases of hemorrhagic stroke, and 106 of unknown type. There
were 317 total strokes among men: 204 ischemic, 64 hemorrhagic, and 49 of
unknown type. Due to the relatively small number of cases of hemorrhagic stroke
to date and the differences in cause between ischemic and hemorrhagic stroke,
this article focuses only on ischemic stroke.
Person-time for each participant was calculated from the date of return
of the 1980 questionnaire (in the NHS) or the 1986 questionnaire (in the HPFS)
to the earlier of the date of first stroke event, death, or the cutoff date
(June 1, 1994, for the NHS, and January 31, 1994, for the HPFS). Participants
who reported CVD or cancer or diabetes prior to completion of the baseline
dietary questionnaires were excluded. Each participant thus contributed only
1 end point, and the cohort at risk for each follow-up period included only
those who remained free from stroke at the beginning of each 2-year follow-up
Due to the difference in sex, follow-up time, and the questionnaires
in the 2 cohorts, the analyses were performed separately in each cohort and
combined to achieve better control of confounding. The risk for ischemic stroke
was compared by fruit/vegetable intake using pooled logistic regression methods
proposed by Cupples et al13 and D'Agostino
et al,14 which summarize the relative risk
(RR) across all 2-year follow-up periods. Analyses were adjusted for age (5-year
categories); smoking (never, former, current: 1-14, 15-24, and ≥25 cigarettes
per day); alcohol consumption (5 categories in women, 7 categories in men);
family history of myocardial infarction (<65 years of age in women, <60
years of age in men); body mass index (quintiles); multivitamin supplement
use; vitamin E use, aspirin use, physical activity (2 categories in women,
quintiles in men); and reported hypertension and hypercholesterolemia and
time period. Additionally, among the women, we controlled for postmenopausal
hormone use. We also controlled for total energy intake.15,16
Because of the long follow-up period, exposures and confounders were
updated to better represent long-term dietary patterns.17
In the NHS we used the 1980, 1984, 1986, and 1990 questionnaires, and in the
HPFS we used the 1986 and 1990 questionnaires. We computed fruit and vegetable
intake for each follow-up period as a cumulative average of intake from all
available FFQs up to the start of the 2-year follow-up period in which events
were reported. Since intermediate end points may influence diet, in cases
in which the participant experienced angina, coronary artery bypass grafting
or angioplasty, hypercholesterolemia, hypertension, or diabetes, we stopped
updating diet at the beginning of the interval in which they experienced the
The average daily intake for each composite item was coded into quintiles
of intake within each cohort. Each quintile of intake was compared with the
lowest quintile of intake. Trends in stroke risk across individual fruits
and vegetables and composite items were assessed in logistic models by using
a variable for the average number of servings consumed per day. To assess
the trend for composite items, we used the median values of intake for each
quintile of the composite item to minimize the influence of outliers. The
RR for the continuous measures, for both the individual and composite items,
indicates the increase or decrease in risk associated with an average increment
of 1 serving per day of the standard portion size defined in the questionnaires.
Because a protective effect of fruits and vegetables could be explained
by their vitamin content, we also evaluated associations between fruit and
vegetables and ischemic stroke separately among multivitamin supplement users
and nonusers. Nonusers of multivitamins who took other vitamin supplements
at baseline were excluded from these subgroup analyses. Nonusers from the
HPFS who took mineral supplements at baseline were also excluded from the
subgroup analyses; data on mineral supplements were not available in the NHS
in 1980. We also performed the analyses separately for current smokers and
for never or past smokers to determine if smoking influenced the association
between intake of fruits and vegetables and ischemic stroke. For the primary
analyses, the RR from the (pooled) logistic regression pooling across 2-year
time periods from the 2 cohorts were further combined using the pooling methods
proposed by DerSimonian and Laird18 that weighted
the RR in each cohort by the inverse of the SEs. Finally, we examined the
adjusted RR for total stroke, using methods similar to those used for ischemic
Table 1 shows the distribution
of the baseline standard cardiovascular risk factors by quintiles of baseline
fruit and vegetable intake from the NHS and the HPFS (quintile 5 being the
highest intake group). Persons who consumed more fruit and vegetables were
older and generally had healthier lifestyles as indicated by lower rates of
smoking and higher levels of physical activity. Similar trends were seen with
other behavioral variables. Smoking and exercise were more strongly related
to fruit intake compared with vegetables. Intakes of protein, cereal fiber, transfatty acids, saturated fat, polyunsaturated fat, and
cholesterol increased with increasing fruit and vegetable intake.
Median intakes of total fruits and vegetables in 1986 were 5.8 servings
per day for women (2.9 and 10.2 for quintiles 1 and 5, respectively) and 5.1
servings per day for men (2.6 and 9.2 for quintiles 1 and 5, respectively)
(Table 2). Results from both studies
individually, as well as pooled analyses for composite fruit and vegetable
items, are presented in Table 2.
Overall fruit and vegetable intake was inversely related to risk of ischemic
stroke after adjusting for potential confounders. Those in the top quintile
of intake had an RR of 0.69 (95% CI, 0.52-0.92) compared with those in the
lowest quintile; the corresponding RR was 0.74 (95% CI, 0.52-1.05) among women
and 0.61 (95% CI, 0.37-1.00) among men. Assessed as a continuous trend, an
increment of 1 serving per day was associated with a 7% lower risk among women
and a 4% lower risk among men; for the combined population there was a 6%
lower risk of ischemic stroke (P=.01, test for trend).
Similar results were seen for total fruit and for total vegetables separately.
Additional control for intakes of transfatty acids,
saturated fat, polyunsaturated fat, or cholesterol did not appreciably alter
the results. All the composite items except legumes and potatoes were inversely
associated with risk of ischemic stroke.
The lowest risks were observed for high consumption of cruciferous vegetables
(eg, broccoli, cabbage, cauliflower, brussel sprouts), green leafy vegetables,
citrus fruits, and vitamin C–rich fruit and vegetables. The results
were similar across both cohorts, especially the measures of trend. Again,
additional analyses incorporating nutrients that may be potential confounders
including protein, cereal fiber, saturated fat, transfatty
acids, polyunsaturated fat and cholesterol intake, or meat intake did not
change the results. For example, when meat intake was added to the model,
the RR for a 1 serving per day increment of total fruits and vegetables changed
from 0.94 (95% CI, 0.90-0.99) to 0.95 (95% CI, 0.90-0.99); addition of transfatty acid quintiles resulted in an RR of 0.94 (95%
CI, 0.90-1.0). When we examined total stroke as the outcome, an increment
of 1 serving per day of total fruits and vegetables was associated with a
3% lower risk among women and a 5% lower risk among men; the multivariate
pooled RR was 0.96 (95% CI, 0.93-1.00). Results for specific fruits and vegetables
were in the same direction as for ischemic stroke or null.
Figure 1 shows the multivariate
RRs and 95% CIs for ischemic stroke computed for increasing servings per day
of intake of fruit and vegetables (pooled for the NHS and the HPFS), suggesting
a decreasing risk of ischemic stroke with increasing intake. There was no
apparent further reduction in risk beyond 6 servings per day.
Supplement users consumed more fruit and vegetables than nonusers (6.08
servings per day vs 5.64 servings per day among women and 5.40 servings per
day vs 4.94 servings per day among men). Similarly, nonsmokers consumed more
fruit and vegetables than current smokers (6.00 servings per day vs 5.10 servings
per day among women and 5.21 servings per day vs 4.21 servings per day among
men). Pooled RRs for an increment of 1 serving per day (linear trend) by multivitamin
supplement use or nonuse and by smoking status are presented in Table 3. The inverse associations between consumption of fruits,
citrus fruit, and cruciferous vegetables and risk of ischemic stroke were
somewhat stronger in the nonsupplement group compared with the supplement
user group. Although the associations were somewhat weaker and not statistically
significant among multiple vitamin users, the number of cases was smaller
and the CIs included substantial reductions in risk. No striking differences
in associations were seen by smoking status, although the apparent protective
effects appeared to be slightly stronger among current smokers than among
past and never smokers. The CIs indicate that interactions between intake
of fruit and vegetables and smoking (or supplement use) or ischemic stroke
are not significant.
Analyses of individual fruit and vegetable items that constituted the
composite items did not show any single fruit or vegetable that was strikingly
more protective than others. Most items showed null or inverse associations.
In general, the findings were consistent with what may be expected from the
composite analyses (data not shown).
Overall, these findings support a protective effect between consumption
of fruits and vegetables against risk of ischemic stroke in both women and
men. Fruit and vegetables—in particular cruciferous and green leafy
vegetables, citrus fruits, and citrus fruit juices—were inversely associated
with risk of ischemic stroke in both cohorts. Fruit and vegetable intake was
associated with behavioral risk factors such as smoking and exercise. However,
adjustment for behavioral and other risk factors, including nutrients that
were suspected to be confounders, did not explain the apparent benefits of
higher fruit and vegetable consumption, although adjustments for these factors
did attenuate the observed associations.
These cohorts of health professionals tended to have healthier lifestyles
than the general population; for example, the median intake of fruit and vegetables
in these cohorts was over the recommended 5 servings per day.19,20
The existence of a dose-response relationship within these groups, which are
relatively homogenous in occupation and education, is particularly notable,
as confounding by variables related to socioeconomic status is minimized.
We found that intake beyond 6 servings per day provided little further reduction
in risk of ischemic stroke compared with intake of 5 to 6 servings per day.
These findings lend support to the recommendation of eating 5 servings per
day of fruit and vegetables. For specific fruits and vegetables, the RRs for
an increment of 1 serving per day should be interpreted in light of the consumption
level. For example, the intake of legumes was very low, with only 0.5 servings
per day as the median intake in the highest quintile (Table 2 and Table 3);
hence, our ability to study the effect of higher consumption was limited.
Our results are similar to a study of 832 men from the Framingham cohort.
The RR of 0.94 in our study for an increment of 1 serving per day translates
to a decremental risk of 0.83 for 3 servings per day, which is reasonably
comparable with the RR of 0.75 for ischemic stroke (95% CI, 0.55-1.03) for
an increment of 3 servings per day of fruit and vegetables (including potatoes)21 in the Framingham study. No association between green
and yellow vegetables and total stroke was observed among a 16-year follow-up
of 265,118 Japanese people,22 but the study
reported on total strokes and the majority of strokes in Japan at that time
were likely to be hemorrhagic. In a 5-year follow-up23
of 1299 Massachusetts nursing home residents, the RR for stroke was 0.40 for
carrots (95% CI, 0.17-0.98) and 0.49 for salads (95% CI, 0.31-0.71), comparing
1 or more vs less than 1 serving per day, which is stronger than comparable
associations in our study.
The constituents of fruits and vegetables potentially responsible for
their apparent beneficial effects on risk of ischemic stroke include potassium,
folate, and fiber; the associations between ischemic stroke and these nutrients
have been evaluated in other reports.23-27
Dietary flavonoids (which are present in fruits and vegetables) have also
been recently related to decreased risk of ischemic stroke.28
There is little evidence relating dietary fiber to stroke. The evidence for
potassium is limited and most studies were based on a 24-hour recall, which
may not be able to adjust for other foods and nutrients that could explain
the inverse associations reported.4,24
However, in a recent report based on the same cohort of men included in these
analyses,25 a significant inverse association
was seen (multivariate RR, 0.62; 95% CI, 0.43-0.88 comparing the top and bottom
quintile of potassium intake). The evidence for folate is largely indirect
through its effect on homocysteine,26 although
in the Framingham Heart Study, an association was observed between low serum
concentrations and low intakes of folate and increased risk of extracranial
carotid artery stenosis.27 Hence, the evidence
for a protective role of any single constituent of fruits and vegetables is
suggestive but inconclusive, and it is important to consider the role of fruits
and vegetables foods in themselves. Slightly stronger protective associations
were seen in the nonusers of supplements, suggesting that constituents of
multi-vitamins may contribute to the apparent protective effect of fruits
and vegetables. There was no clear modification of the relation of fruit and
vegetable consumption to risk of ischemic stroke by smoking, but current smokers
may gain a slightly greater benefit from eating more fruits and vegetables
than never or past smokers.
These data support a protective effect of higher consumption of fruits
and vegetables—in particular cruciferous and green leafy vegetables,
citrus fruits, and citrus juice—against risk of ischemic stroke. Our
results provide further support for the recommendation to consume at least
5 servings of fruits and vegetables a day.