Fruit and vegetable consumption and risk of stroke.

Epidemiologic studies of the association between fruit and vegetable intake and coronary heart disease (CHD) have shown that greater fruit and vegetable intake is associated with decreased CHD. A recent pooled analysis of cohort studies noted that consumption of fiber from cereals and fruit is associated with a decreased risk of CHD. These observations support the recommendation to eat fruit and vegetables to prevent CHD.

CHD and stroke share risk factors. The association between fruit and vegetable consumption and the risk of stroke has been examined in cohort studies. In general, these studies reported favorable associations although sometimes inconsistent. In addition, the magnitude of this association remains unclear due to differences in methodologic approaches, analytical techniques, and event definitions. Therefore, we sought to assess the nature and the intensity of the relationship between fruit, vegetable, and fruit and vegetable consumption and the risk of stroke. Only cohort studies that used validated dietary assessment tools were used for this meta-analysis.

Discussion

The results of the present meta-analysis show a consistent association between consumption of fruit and decreased occurrence of stroke. The reduction in risk is 11% for each additional portion per day of fruit intake and 5% (3 to 8%) for fruit and vegetable intake. These associations are linear, suggesting a dose-response relationship. In contrast, we found no significant evidence for a significant reduction in stroke rates with vegetable consumption.

The protective effect of fruit consumption against stroke is supported by clinical and experimental evidence. First, the relation is biologically plausible, with numerous clinical and laboratory data showing that micro- and macro-constituents of fruit favorably affect important risk factors of stroke, such as hypertension, dyslipidemia, and diabetes. Second, the association is linear, supporting a dose effect, and is homogeneous in cohorts from North America, Europe, and Asia despite contrasted lifestyles and cardiovascular incidence rate. Finally, the association was observed after adjustment on classic risk factors, suggesting a specific effect of fruit.

Altogether, these observations support the concept that fruit consumption might protect men and women from stroke. In contrast, other elements are not in favor of a causal relationship. For example, in observational studies fruit consumption correlates with healthy behaviors and lifestyles, which may explain the lower stroke rates among fruit consumers. Generally, usual fruit consumers smoke less, exercise more, and are better educated than non-consumers.

Although most studies adjust for these lifestyle factors with statistical procedures, a possibility remains that confounders not measured in these studies contribute to the association between fruit intake and stroke. Consumption of fruit is also associated with changes of other food intakes during the meal. Thus, the apparent benefit of fruit could be the consequence of the substitution of fruit and vegetables for unhealthy foods. Finally, the hypothetical mechanisms involved in the protective effect of fruit have not always been confirmed experimentally. For example, in randomized clinical trials, supplementation with antioxidant vitamins has not been associated with significant reduction in stroke incidence. Thus, it appears that the association between fruit intake and decreased risk of stroke observed in cohort studies could as well be explained by confounding factors or healthy behaviors in fruit consumers. Therefore, the results of the present study support the concept that regular fruit consumption and related healthy behaviors are associated with lower rates of stroke in population studies.

The relation between vegetable intake and stroke remained statistically non significant. This finding is consistent with the results of pooled analysis of cohort studies that reported lower CHD risk with increasing intake of dietary fibers from fruit, but not from vegetables. One possible explanation for the lack of association between vegetable intake and stroke may be that assessment of vegetable intake is less accurate than fruit, resulting in a dilution bias and thus reduced power to find an association. Another hypothesis is that vegetables included a wide variety of items, some of which may present unfavorable properties. For example, starchy vegetables increase dietary glycemic load and may increase the risk of diabetes and CHD. Also, cooking vegetables could alter the properties of some nutrients such as thermo-labile vitamins. Another possible explanation is that consumption of vegetables may not be as strongly associated with favorable behaviors as fruit, which may result in non detectable associations.

Finally, the sample size may still be too limited to detect a mild protective association. Altogether, the results of the meta-analysis do not support a major protective effect of vegetable intake toward stroke.

This study has several limitations. First, the meta-analysis is based on observational studies, which leaves the possibility of residual confounding or bias, including measurement errors, to affect the relationship between fruit and vegetable intake and stroke. Second, there were many differences among studies including dietary assessment methods, the variety of fruit or vegetable investigated, the selection of reference group, and the choice of exposure categories. These differences might affect the estimation of the true relationship. For example, fruit and vegetable intake was measured once at entry in most studies, which precludes the possibility of controlling for daily variability or changes in food intake with time. Third, the characteristics of events were not always comparable among studies. However, heterogeneity analyses do not support any major effect of these factors. Fourth, like other meta-analyses, a publication bias cannot be excluded. Four out of 11 relevant studies were excluded because the presentation of the results did not allow a meta-analysis. However, funnel analyses do not support a publication bias. Finally, due to the limited number of studies, subgroup analyses to test the possible impact of sex, geographic area, and other factors on these associations were not done.

Controlled trials provide the strongest evidence regarding the evaluation of a nutritional intervention. However, the assessment of the health benefit of fruit and vegetable consumption is hardly feasible in a randomized blinded controlled study. Alternatively, this meta-analysis provides evidence of a beneficial effect of fruit and fruit and vegetable consumption on risk of stroke from observational studies in North America, Europe, and Japan. These data support the recommendation to eat fruit and vegetables to lower the risk of stroke. Finally, this meta-analysis also points out the limited number of prospective studies that analyzed the effects of fruit and vegetable intake on stroke, in Southern Europe and especially in Asia, where the incidence of stroke is elevated.