Abstract

BACKGROUND: Vegetables & fruit may protect against CVD risk and some types of cancer, but little evidence was found for an association with CVD death and total cancer risk or death (Hoenselaar R). This raises questions about the association with death from all causes.
OBJECTIVES:

• To review all prospective studies which provided data about total fruits or vegetables, their subgroups, and specific types of vegetables and fruits, in relationship with all-cause mortality/survival.
• To define the amount of consumption found to be related with the described effects on death.
• To define possible effect modification by confounders.

DATA SOURCE: The Pubmed database was searched (No start date - July 6, 2009) for relevant articles using the keywords "prospective, cohort, follow-up, longitudinal" combined with "vegetables, fruit", and a fair amount of other keywords. The exact search term is described in the methods.
Prospective studies published in the English language were included. Reference lists were searched for additional articles.
RESULTS: 49 articles were found which provided information about 32 different cohorts. Of these, 0 articles were excluded.
Results are described when any evidence for an association - as defined in the Methods - was found. In addition, data about total vegetables or fruits is described.

• Data about total vegetables was provided by 16 cohorts, including 18,554 cases. Significant protective effects were found in 4 cohorts. Inconclusive evidence was found for an association between total vegetables and all-cause mortality (RR = 0.96).
• Data about total fruit was provided by 20 cohorts, including 27,905 cases. Significant protective effects were found in 7 cohorts, including just over 50% of all cases. Total fruit consumption possibly protects against all-cause mortality (RR = 0.88).
• Data about soy was provided by 3 cohorts, including 73,844 cases. When all soy items were combined, significant protective effects were found in all 3 cohorts which were of moderate to (very) large size. Soy consumption possibly protects against mortality. No level of consumption could be defined for this effect, but the effect size was weak (RR = 0.96). Results were confined to publications about Japanese cohorts.
• Data about (green) leafy vegetables was provided by 3 cohorts, including 18,275 cases. Significant protective effects were found in 2 cohorts which were of moderate-large size, and in part of the third cohort. Green leafy vegetables possibly protect against mortality. The effect size was weak (RR = 0.92).

CONCLUSION: Total fruit consumption possibly protects against all-cause mortality (- 12%), but inconclusive evidence was found for an association with total vegetables. Both green leafy vegetables and soy products possibly protect against all-cause mortality, but the strength of the effect is weak for both foods (- 8%, and - 4%, respectively). Results from the association with soy were confined to publications about Japanese cohorts. Since few significant effects were published in servings/cups/grams over a given time period, it is not possible to define specific amounts of consumption for these effects.
LIMITATIONS: No results were available from any of the cohorts of very large size (except for green-yellow vegetables and soy). It is to be expected that, in the coming decades, within these cohorts hundreds of thousands of subjects will decease. Since the current evidence is based on 18,554 and 27,905 deaths (for total vegetables and total fruits, respectively), it is not hard to understand that current evidence could change in any direction in the near future.
PERSPECTIVE: Evidence was judged about the relation with mortality only, but within the tables in this review results were added from publications about the relation with total major chronic disease. Three publications about 2 cohorts provided information about total major chronic disease in relation with various subgroups of vegetables & fruits, as well as with total vegetables & fruits (18). Inclusion of total major chronic disease in the model of mortality strenghtens the evidence for an association with green leafy vegetables, and attenuates the association with total fruits. But no change of evidence was found for any of the remaining variables.

|References:
Hoenselaar R. Vegetables, fruit, and cardiovascular disease. Canceranddiet.nl Link.
Hoenselaar R. Vegetables, fruit, and total cancer. Canceranddiet.nl Link.|

Total vegetables and all-cause mortality/survival.

25 articles, providing information about 16 different cohorts were found, including a total of 18,554 cases. Survival was the end point in one article (Nube M [2]), and mortality was the end point in all remaining articles.

Use of data from The EPIC Study: 7 publications were found about The EPIC Study (21). Publications included data about the same population with different length of follow-up (Trichopoulou A. 2003 vs Trichopoulou A. 2009), or about a subgroup of- vs all diabetic subjects from the cohort (Trichopoulou A. 2006 vs Nöthlings U. 2008). Obviously, the publications about longer length of follow-up, and about larger parts of the cohort were included.
The EPIC-Elderly study provided data about 74,607 elderly (aged ≥ 60) subjects - with or whithout diabetes - from 9 countries (including Greece & Spain), and including 4,047 deaths (Trichopoulou A. 2005b). But 3 other publications provided information about subgroups of this cohort from which the data partly overlapped with data from The EPIC-Elderly Study:

• Agudo A (2007) 41.358 subjects aged 30-69 from EPIC-Spain, including 562 deaths.
• Nöthlings U (2008) 10.449 subjects aged 35-70 with diabetes from 10 countries, and including 1,346 deaths.
• Trichopoulou A (2009) 23,349 subjects aged 20-86 from EPIC-Greece, including 1,075 deaths.

The latter 3 publications were included in the systematic review, excluding data from The EPIC-Elderly Study. These 3 subcohorts included less deaths than the former subcohort (2,983 vs 4,047). But they all provided RR's as a categorized variable, allowing for analysis about effects at different levels of consumption, and for more precise analysis of the effect size. Whereas The EPIC-Elderly Study provided RR's as a continuous variable.
Also, a 4th publication about this cohort was included: Trichopoulou A (2005a) provided data about subjects from EPIC-Greece with CHD only, whereas he provided data about subjects from EPIC-Greece without CHD only in a later publication (Trichopoulou A. 2009).

Results: Significant protective effects were found in 3 cohorts (Sahyoun NR [7], Nöthlings U [21], and Seccareccia F [24]), and among men in a 4th cohort (Nube M [2]). These 4 cohorts included 3,383 cases (18 % of all cases). No other associations were found of high vs low consumption.
The average RR = 0.96 for all cohorts (excluding Knekt P [9], because no amount of cases was specified, stratified by sex. And excluding Strandhagen E [16], and Chang-Claude J [31] because of incomplete data).
Conclusion: Protective effects were found in 4 cohorts, including 18% of all cases. Inconclusive evidence was found for an association between vegetable consumption and all-cause mortality.

Total fruit and all-cause mortality/survival.

32 articles, providing information about 20 different cohorts were found, including a total of 27,905 cases. Survival was the end point in one article (Nube M [2]), and mortality was the end point in all remaining articles.

Use of data from The EPIC Study: 7 publications were found about The EPIC Study (21). Publications included data about the same population with different length of follow-up (Trichopoulou A. 2003 vs Trichopoulou A. 2009), or about a subgroup of- vs all diabetic subjects from the cohort (Trichopoulou A. 2006 vs Nöthlings U. 2008). Obviously, the publications about longer length of follow-up, and about larger parts of the cohort were included.
The EPIC-Elderly study provided data about 74,607 elderly (aged ≥ 60) subjects - with or whithout diabetes - from 9 countries (including Greece & Spain), and including 4,047 deaths (Trichopoulou A. 2005b). But 3 other publications provided information about subgroups of this cohort from which the data partly overlapped with data from The EPIC-Elderly Study:

• Agudo A (2007) 41.358 subjects aged 30-69 from EPIC-Spain, including 562 deaths.
• Nöthlings U (2008) 10.449 subjects aged 35-70 with diabetes from 10 countries, and including 1,346 deaths.
• Trichopoulou A (2009) 23,349 subjects aged 20-86 from EPIC-Greece, including 1,075 deaths.

The latter 3 publications were included in the systematic review, excluding data from The EPIC-Elderly Study. These 3 subcohorts included less deaths than the former subcohort (2,983 vs 4,047). But they all provided RR's as a categorized variable, allowing for analysis about effects at different levels of consumption, and for more precise analysis of the effect size. Whereas The EPIC-Elderly Study provided RR's as a continuous variable.
Also, a 4th publication about this cohort was included: Trichopoulou A (2005a) provided data about subjects from EPIC-Greece with CHD only, whereas he provided data about subjects from EPIC-Greece without CHD only in a later publication (Trichopoulou A. 2009).

Results: Significant protective effects were found in 6 cohorts (Appleby PN [8], Knoops KT [10], Kouris-Blazos A [14], Barzi F [26], Nagura J [32], González S [33]) and among men in a 7th cohort (Nube M [2]), including 14,040 cases (50% of all cases). In addition, nonsignificant protective effects were found in 2 cohorts, including 952 cases (Strandhagen E [16], Agudo M [21]. Two out of these cohorts were of moderate size (Agudo M [21], Nagura J [32]), the rest was of (very) small size.
When the effect of fresh fruit - instead of total fruit - was used from one cohort (Agudo A [21]), the nonsignificant protective effect became significant (HR = 0.79; 95% CI = 0.62-1.00; P = 0.029).
(Non)significantly increased risks were found in 2 cohorts of very small size (Nube M [2], Lasheras C [15]), but the effect was restricted to women once (Nube M [2]), and to subjects aged ≥ 80 in the other cohort (Lasheras C [15]). These subgroups included 506 cases (2% of all cases). For total fruits all RR's > 1 were restricted to cohorts of very small size.
The average RR = 0.88 for all cohorts (excluding incomplete data from Seccareccia F [24], and Chang-Claude J [31]).
Conclusion: Protective effects were found in 7 cohorts, including just over 50% of all cases. While increased risks were found in 2 subgroups from 2 cohorts of very small size, and including 2% of all cases. Fruit consumption possibly protects against all-cause mortality (- 12%).