Meta-analysis or Systematic Review

M - Click here for explanation of classification scheme.

POSITIVE: See Quality Criteria Checklist below.

To assess the relationship between fruit, vegetable and fruit, and vegetable consumption and the risk of stroke.

• Published prospective studies that assessed the relationship between stroke and consumption of fruit and vegetables, separately or combined
• Fruit and vegetable intake was assessed with validated methods
• Inclusion criteria for events were all cerebrovascular events, including ischemic, thrombotic, embolic, hemorrhagic and transient ischemic attack
• Inclusion criteria for exposure variable was fruit and vegetable consumption per se and not one of their respective nutrients (i.e., only food groups named as follows in the articles: All vegetables, total vegetables, all fruit, total fruit or fruit).

Outcomes that combined cerebrovascular and cardiac events.

Recruitment (Search Sstrategy)

• Searches were conducted in MEDLINE and EMBASE from 1970 to September 2004
• References from the extracted articles, reviews, and previous meta-analyses were also used
• Terms (MeSH terms and truncated free text) included cardiovascular disease, cerebrovascular disease, cerebrovascular disorders, stroke, cohort studies, prospective studies, follow-up studies, fruit and vegetables.

Design

Meta-analysis.

Statistical Analysis

• The relative risk (RR) of stroke for an increment in one standard portion per day of fruit and vegetables and related confidence intervals adjusted for the greatest number of potentially confounding variables was extracted form the original report
• The RR for ischemic stroke was used if the results for more than one event were available
• The portion size was 106g. When the RR was expressed in frequency, the exposure level was calculated from the mean value of intake in the validation study
• Heterogeneity among studies was assessed using the Cochran's Q test
• The pooled estimate of the RR was computed assuming a log linear model. 

Dependent Variables

All cerebrovascular events, including ischemic, thrombotic, embolic, hemorrhagic and transient ischemic attack.

Independent Variables

Fruit and vegetable consumption (i.e., food groups named as follows in the articles, including all vegetables, total vegetables, all fruit, total fruit or fruit).

Control Variables

Study results that controlled for the greatest number of potential confounders were used.

• Initial N: 11 studies (four studies and part of a fifth did not meet inclusion criteria)
• Attrition (final N): Seven independent studies comprising 232,049 subjects (90,513 men and 141,536 women), 2,955 events and an average of 10.7 years of follow-up
• Age: 25 to 103 years
• Location: Studies conducted in US, Japan and Europe.

 Key Findings

• The pooled RR (95% CI) of stroke for each increment of one portion of fruit per day was 0.89 (0.85 to 0.93)
• The pooled RR (95% CI) of stroke for each additional portion of vegetable per day was 0.97 (0.92 to 1.02) (NS)
• The pooled RR (95% CI) of stroke for each increment of one portion per day of fruit and vegetable was 0.95 (0.92 to 0.97)
• There was a linear relationship between fruit intake and stroke, but not between vegetable intake and stroke
• There was no evidence of heterogeneity among studies and no evidence of a publication bias.

The meta-analysis shows a linear association between consumption of fruit and decreased occurrence of stroke, but no significant reduction in stroke rates with vegetable consumption.

• Study strengths:
  • Results adjusted for potential confounders were used
  • Only studies that used dietary assessments that were validated were used
  • Tests for heterogeneity among studies and publication bias were used
• Study limitations:
  • The analysis is based on observational studies, which may include residual confounding and measurement error
  • There were many differences among the studies including dietary assessment methods, the variety of fruit and vegetable investigated, the selection of the reference group and the choice of exposure categories
  • The characteristics of events were not always comparable among studies.