Vegetables, Fruit, and total Cancer.

Abstract

BACKGROUND: Early articles often found protective effects of vegetables and fruits against various types of cancer, but results were mostly based on data from case-control studies (Riboli E. 2003). Later reviews which could be based on more data from prospective studies found little association with most types of cancer.
The WCRF recently published a systematic review providing information about various dietary items in relation to various types of cancer (World Cancer Research Fund. 2007). A protective effect of vegetables and fruits was found against few cancer types. Since no information was provided about the relation with total cancer, information may still be difficult to translate to recommendations, because vegetables and fruits may negatively impact other - less studied - cancer sites. This theory is strenghtened by a recent article examining a large cohort, and providing information about > 50,000 cancer cases (George SM. 2009). In this cohort, protective effects were found for some cancer sites, but increased risks were found for other types of cancer by high consumption of vegetables or fruits. The balance in these effects resulted in (almost) no association with total cancer risk.
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 total cancer risk, disease progression and mortality/survival.
  • To define the amount of consumption found to be related with the described effects on cancer.
  • To define possible effect modification by confounders.

DATA SOURCE: The Pubmed database was searched (No start date - Sep 11, 2009) for relevant articles using the keywords "cancer, neoplasm, or neoplasms" combined with "vegetables, vegetable, fruit, or fruits" 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: 34 articles were found which provided information about 24 different cohorts.
Of these, 0 articles needed to be excluded because of language restrictions. And no articles were excluded for any other reasons.
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.

  • Total disease risk.
    -For total vegetables, 14 articles providing information about 10 different cohorts were found, including 96,960 incident cancer cases. Among men, weak significant protective effects were found in 2 cohorts of very large size, and no (non)significantly increased risks were found. Total vegetables consumption possibly protects against total cancer risk among men. The effect size is weak (RR = 0.95), and the evidence is restricted to an intermediate level of consumption (235-306 g/d). Among women inconsistent findings were done. Inconclusive evidence was found for an association between total vegetables and total cancer risk among women (RR = 0.97). Stratified analysis included the possibility of effect modification by smoking status, alcohol drinking, and use of dietary supplements, but the evidence is inconclusive.
    -For total fruits, 15 articles providing information about 10 different cohorts were found, including 97,192 incident cancer cases. Some protective effects were found, but these were restricted to few cohorts. And effect sizes showed little evidence for an effect (RR = 0.96 for men, and 0.99 for women, respectively). Stratified analysis showed inconclusive evidence for a modifying effect of alcohol drinking, and use of dietary supplements, but showed a consistent lack of any protective effect among never smokers.
    -A significantly increased cancer risk of citrus fruit juice among multivitamin supplement users was found in a pooled analysis of 2 cohorts. No other associations were found. Citrus fruit juice possibly increases total cancer risk among multivitamin supplement users.
  • Advanced stage/metastatic disease risk or disease progression.
    No article provided information about any of the dietary variables in this review in relation to advanced/metastatic stage disease, or disease progression.
  • Mortality risk.
    -For total vegetables, 6 articles providing information about 6 different cohorts were found. No associations were found in any cohort. In addition, no evidence was found for an association when 5 cohorts examining green-yellow vegetables were added to this model.
    -For total fruits, 12 articles providing information about 11 cohorts were found. Significant protective effects were found in 2 cohorts. Evidence was judged suggestive for a protective effect of total fruits against total cancer mortality. This effect was found at the level of daily vs less frequent consumption.

CONCLUSION: Among men, total vegetables consumption possibly protects against total cancer risk (- 5%). The evidence was restricted to an intermediate level of consumption (235-306 g/d). No evidence was found for an association among women. Suggestive evidence was found for a protective effect of total fruits against total cancer mortality (- 12%). This effect was found for (almost) daily vs less frequent consumption. No evidence was found for an effect of total fruit with total cancer risk, or total vegetables with total cancer mortality. Smoking status, alcohol drinking, and dietary supplement use may modify these effects, but little data is provided and therefore, no firm conclusions can be drawn. Though current evidence excludes the possibility of a protective effect from fruit consumption against total cancer risk among never smokers.
Citrus fruit juice possibly increases cancer risk among multivitamin supplement users. No further evidence was found for an association between any vegetable/fruit subgroup or item and total cancer risk or mortality.
PERSPECTIVE: Little evidence was found for an association. Effect are of small-moderate size and may be confined to specific levels of consumption. Few results were published stratified by possible confounders, so the possibility of any effect among specific subgroups of the population can not be excluded. Very little data is available about the relation with specific vegetable or fruit items.
The author is aware that at least one article was published following the closing date of the literature review (Boffetta P. 2010). This article was included in the current analysis.



|REFERENCES:
-Boffetta P. Fruit and vegetable intake and overall cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC). J Natl Cancer Inst. 2010 Apr 21;102(8):529-37. Abstract
-George SM. Fruit and vegetable intake and risk of cancer: a prospective cohort study. Am J Clin Nutr. 2009 Jan;89(1):347-53. Abstract
-Riboli E. Epidemiologic evidence of the protective effect of fruit and vegetables on cancer risk. Am J Clin Nutr. 2003 Sep;78(3 Suppl):559S-569S. Full text
-World Cancer Research Fund / American Institute for Cancer Research. Food, Nutrition, Physical Activity, and the Prevention of Cancer: a Global Perspective. Washington DC: AICR, 2007. Full text|


Total vegetables and total cancer risk.

Background: Almost all articles presented data stratified by gender. Therefore the choice was made to present results for men and women separately. As a consequence, results from two cohorts will not be included in the following analysis: In "The Finnish Mobile Clinic Health Examination Survey Cohort" (Knekt P [7]), and in "The Boyd Orr Cohort" (Maynard M [2003]. See extended table), including 1,480 cancer cases, no associations were found between total vegetables consumption and total cancer risk: RR = 0.93 (0.74-1.17), and 1.34 (0.93-1.93; P = 0.27), respectively.

Men: Data for men was provided by 7 cohorts, including a total of 50,109 cases.
Weak, significant protective effects were found in 2 cohorts of very large size (Boffetta P [18], George SM [22]), including 44,675 cases (89% of all cases). No associations were found in the remaining cohorts. The average RR = 0.95 (excluding incomplete data from Strandhagen E [9]).
Women: Data for women was provided by 5 cohorts, combining 45,371 cases.
A weak, significant protective effect was found in one cohort of very large size (Boffetta P [18]). But a weak, nonsignificantly increased risk (trend) was found in another cohort of very large size (George SM [22]). No other associations were found. The average RR = 0.97.
Inclusion of intermediate levels of consumption:
Among men, (non)significant effects at any level of consumption are as follows:

  • Boffetta P (18) Though the trend was significant, the RR was significant only at an intermediate level of consumption (209-306 g/d).
  • George SM (22) Significant protective effects were found at 2 different levels of consumption (235-312 g/d, and > 424 g/d)

Among women, (non)significant effects at any level of consumption are as follows:

  • Boffetta P (18) At ≥ 307 g/day.
  • George SM (22) Though the trend showed a nonsignificantly increased risk, no RR differed (non)significantly from 1 at any level of consumption.

These analysis show that among men, signicant protective effects were found at an overlapping level of consumption (235-306 g/day). The effects size is RR = 0.95 (including data from the 3 cohorts which could be provided in this analysis), and therefore is identical to the overall effect size for high vs low consumption.

RRs for the association between total vegetables and total cancer risk among men (g/day):



RRs for the association between total vegetables and total cancer risk among women (g/day):


Effect modification.
Smoking status: Effects stratified by smoking status were provided by 4 cohorts (see extended table):

  • George SM (22) A significant protective effect was found among the male part of the cohort (RR = 0.94; 95% CI = 0.91-0.97), but not among the male nonsmokers (RR = 0.97; 95% CI = 0.91-1.04).
  • Takachi R (20) No effects were found among never smokers (HR = 0.97; 95% CI = 0.83-1.13), or ever smokers (HR = 0.94; 95% CI = 0.80-1.11).
  • Boffetta P (18) Significant protective effects were found among never smokers (HR = 0.97; 95% CI = 0.96-0.99), and current smokers (HR = 0.97; 95% CI = 0.95-0.98), but not among past smokers (HR = 0.98; 95% CI = 0.96-1.00).
  • Hung HC (11) No effects were found among never smokers (RR = 1.02; 95% CI = 0.94-1.10), past smokers (RR = 1.01; 95% CI = 0.94-1.08), or current smokers (RR = 0.94 (0.85-1.05).

Some results are hard to interpret. In one cohort results were only given for male never smokers (George SM [22]). And in another cohort, results were only given for nonsmokers vs ever smokers, categorizing past-, and current smokers into one group (Takachi R [20]). Briefly, a small protective effect among never smokers was found in one out of 4 cohorts, and no associations were found among past smokers in 2 cohorts. Also, a small protective effect among current smokers was found in one out of 2 cohorts, and the RR tended to be lower than among the other groups in a second cohort (Hung HC [11]).
Average RR's are 0.98, 0.99, and 0.96 for never, past, and current smokers, respectively.
Alcohol drinking: The significant protective effect in one cohort of very large size was restricted to subjects with heavy alcohol consumption (Boffetta P [18]. see extended table). No difference in effect was found in another cohort (Takachi R [20]).
Dietary supplement use: Some articles showed stronger evidence for a possible protective effect among non-users of multivitamins. (Pooled results from 2 cohorts; Hung HC [11]) or dietary supplements (Olsen A [2005]). In contrast, the effect was slightly more protective among dietary supplement users in a fourth cohort (Benetou V [18]).

Conclusion: Among men, weak significant protective effects were found in 2 cohorts of very large size, and no other associations were found. Total vegetables consumption possibly protects against total cancer risk among men (- 5%). Though this evidence was restricted to an intermediate level of consumption (235-306 g/d), the effect size was identical to the one for high vs low levels of consumption.
Among women, inconsistent findings were done. Inconclusive evidence was found for an association between total vegetables - at any level of consumption - and total cancer risk.
Stratified analysis included the possibility that any protective effect may be stronger among current smokers. Inconclusive evidence was found for a modifying effect of alcohol consumption and dietary supplement use.

Prospective studies of total vegetables and total cancer risk (men):
AuthorCohort nameCasesRelative Risk (RR)
22) George SM (2008)The NIH-AARP Diet and Health Study35,071RR = 0.94 (0.91-0.97; P = 0.004).
20) Takachi R (2007)The JPHC Study1,925HR = 0.95 (0.82-1.10; P = 0.38).
18) Boffetta P (2010)The EPIC Study9,604HR = 0.95 (0.87-1.03; P = 0.04).
15) Jansen MC (2004)The Zutphen Elderly Study138RR = 0.83 (0.54-1.25; P = 0.36).
11) Hung HC (2004)The Health Professionals' Follow-Up Study2500RR = 0.99.
9) Strandhagen E (2000)The Study Of Men Born In 1913226No significant association (no data shown).
2) Shibata A (1992)The Leisure World Study6451.05 (0.89-1.27).
Total number of cases: 50,109Average RR = 0.95



Prospective studies of total vegetables and total cancer risk (women):
AuthorCohort nameCasesRelative Risk (RR)
22) George SM (2008)The NIH-AARP Diet and Health Study15,792RR = 1.04 (0.98-1.09; P = 0.084).
20) Takachi R (2007)The JPHC Study1,305HR = 0.94 (0.78-1.12; P = 0.19).
18) Boffetta P (2010)The EPIC Study21,000HR = 0.92 (0.87-0.97; P = 0.005).
11) Hung HC (2004)The Nurses' Health Study6584RR = 0.99.
2) Shibata A (1992)The Leisure World Study6900.84 (0.70-1.01).
Total number of cases: 45,371Average RR = 0.97



Prospective studies of total vegetables and total cancer risk.
Stratified by smoking status:
AuthorCohort nameCasesRelative Risk (RR) for never smokersRelative Risk (RR) for past smokersRelative Risk (RR) for current smokers
22) George SM (2008)The NIH-AARP Diet and Health StudyNot definedRR = 0.97 (0.91-1.04; P = 0.47).--
20) Takachi R (2008)The JPHC Study1,631 never smokersHR = 0.97 (0.83-1.13; P = 0.26).--
18) Boffetta P (2010)The EPIC Study13,728 never smokers,

8,832 past smokers, and

7,388 current smokers
HR = 0.97 (0.96-0.99).HR = 0.98 (0.96-1.00).HR = 0.97 (0.95-0.98).
11) Hung HC (2004)The Nurses' Health Study

&

The Health Professionals' Follow-up Study
3,577 never smokers,

3,945 past smokers, and

1,694 current smokers
RR = 1.02 (0.94-1.10).RR = 1.01 (0.94-1.08).RR = 0.94 (0.84-1.05).
Average RR = 0.98Average RR = 0.99Average RR = 0.96


Click here for an extended version of these tables.

Total vegetables and total cancer mortality.

6 articles, providing information about 6 different cohorts were found. No associations were found in any cohort. The average RR was 1.07 (excluding incomplete data from Strandhagen E [9], and Hung HC [11]).

Conclusion: No evidence was found for an association between total vegetables and total cancer mortality.

Prospective studies of total vegetables and total cancer mortality:
AuthorCohort nameCasesSexRelative Risk (RR)
18) Nöthlings U (2008)The EPIC Study319Men & WomenRR = 1.09 (0.87-1.39).
13) Maynard M (2003)The Boyd Orr Cohort158Men & WomenOR = 1.14 (0.75-1.72).
11) Hung HC (2004)The Nurses' Health Study &

The Health Professionals' Follow-up Study
Not defined.Men & WomenNo significant association (no data shown).
9) Strandhagen E (2000)The Study Of Men Born In 1913121MenNo significant association (no data shown).
5) Sahyoun NR (1996)No cohort name57Men & WomenRR = 0.80 (0.36-1.76; P = 0.82).
Total number of cases: 655 + XAverage RR = 1.07


Click here for an extended version of this table.

Total fruit and total cancer risk.

Background: Almost all articles presented data stratified by gender. Therefore the choice was made to present results for men and women separately. As a consequence, results from two cohorts will not be included in the following analysis: "The Finnish Mobile Clinic Health Examination Survey Cohort" (Knekt P [7]), and "The Boyd Orr Cohort" (Maynard M [2003]. See extended table), including 1,480 cancer cases. A significant protective effect was found in the cohort of very small size (OR = 0.62; 95% CI = 0.43-0.90; P = 0.02 for Maynard M [13]), but no association was found in the other cohort (RR = 0.90; 95% CI = 0.71-1.14 for Knekt P [7]).

Men: Data for men was provided by 8 cohort, including a total of 50,210 cases.
Significant protective effects were found in 2 cohorts (Jansen MC [15], Boffetta P [18]), including 9.742 cases (19% of all cases). No other associations were found, but RRs were ≤ 1. The average RR = 0.96 (excluding incomplete data from Strandhagen E [9]).
Women: Data for women was provided by 6 cohorts, including a total of 45,502 cases.
(Non)significant protective effects were found in two cohorts of (very) small size (Shibata A [2], Cox BD [10]), including 821 cases. And a nonsignicant protective effect was found in another cohort, though the trend but not the RR was nonsignificant (George SM [22]). The average RR = 0.99.

Inclusion of intermediate levels of consumption:
Among men, (non)significant effects at any level of consumption are as follows:

  • Boffetta P (18) Nonsignificant protective at 163-246 g/day. Significant protective at ≥ 247 g/day.
  • Jansen MC (15) Significant protective at > 200 g/day.

Among women, (non)significant effects at any level of consumption are as follows:

  • George SM (22) Though the trend showed a nonsignificant protective effect, no RR differed (non)significantly from 1 at any level of consumption.
  • Boffetta P (18) Significant protective at 163-246 g/day.
  • Cox BD (10) Though risk decreased per frequency category of consumption, the cagegorized models showed no evidence for an effect at any level of consumption.
  • Shibata A (2) Significant protective at ≥ 192 g/day.

These analysis show no consistent effects at any intermediate level of consumption.


RRs for the association between total fruits and total cancer risk among men (g/day):



RRs for the association between total fruits and total cancer risk among women (g/day):


Effect moficiation.
Smoking status: Effects stratified by smoking status were provided by 3 cohorts (see extended table):

  • Takachi R (20) No effects were found among never smokers (HR = 1.12; 95% CI = 0.95-1.33; P = 0.83), or ever smokers (HR = 0.95; 95% CI = 0.79-1.14; P = 0.79).
  • Boffetta P (18) Any possibly protective effect seemed to be restricted to current smokers (HR = 0.98; 95% CI = 0.97-1.00), though the effect size was very weak. No associations were found among never smokers (HR = 1.00; 95% CI = 0.99-1.01), and former smokers (HR = 0.99; 95% CI = 0.98-1.01).
  • Hung HC (11) No effects were found among never smokers (RR = 1.03; 95% CI = 0.94-1.12), past smokers (RR = 0.97; 95% CI = 0.89-1.06), or current smokers (RR = 1.07 (0.94-1.22).

In one cohort cohort, results were only given for nonsmokers vs ever smokers, categorizing past-, and current smokers into one group (Takachi R [20]). No significant effects were found in any cohort for any subgroup, stratified by smoking status. RR's were ≥ 1 among never smokers in all 3 articles, examining 4 cohorts.
Average RRs are 1.02, 0.98, and 1.00 for never, past, and current smokers, respectively.
Alcohol drinking: Results from one cohort of very large size were presented, stratified by drinking status (Boffetta P [18]). Though no significant effects were found in all 3 subgroups of drinkers, the effect size suggests any possible protective effect may be restricted to heavy drinkers (HR = 0.96; 95% CI = 0.93-1.00). No associations were found among weak drinkers (HR = 1.00; 95% CI = 0.99-1.01), or moderate drinkers (HR = 0.99; 95% CI = 0.97-1.00).
Dietary supplement use: One article showed stronger evidence for a possible protective effect among non-users of multivitamins. (Pooled results from 2 cohorts; Hung HC [11]). In contrast, the effect was slightly more protective among dietary supplement users in a third cohort (Benetou V [18]). No difference in effect of dietary supplements was found in a fourth cohort (Olsen A [2005]).

Conclusion: Some protective effects were found, but these were restricted to few cohorts, inluding a small minority of the total amounts of cases. Also, the average RR's (adjusted for the amounts of cases/cohort) showed little evidence for an effect. Stratified analysis showed inconclusive evidence for a modifying effect of alcohol drinking, or use of dietary supplements. Analysis stratified by smoking status showed that any protective effect among never smokers seems very unlikely since all 3 RR's were ≥ 1.

Prospective studies of total fruit and total cancer risk (men):
AuthorCohort nameCasesRelative Risk (RR)
22) George SM (2008)The NIH-AARP Diet and Health Study35,071RR = 0.98 (0.95-1.02; P = 0.17).
20) Takachi R (2007)The JPHC Study1,925HR = 0.90 (0.77-1.07; P = 0.29).
18) Boffetta P (2010)The EPIC Study9,604HR = 0.89 (0.82-0.97; P = < 0.001).
15) Jansen MC (2004)The Zutphen Elderly Study138RR = 0.62 (0.40-0.96; P = 0.04).
11) Hung HC (2004)The Health Professionals' Follow-up Study2,500RR = 0.98.
10) Cox BD (2000)The British HALS101Fruit in winter: OR = 0.91 (0.78-1.06; P = 0.23).
Fruit in summer: OR = 1.00 (0.85-1.17; P = 0.97).
9) Strandhagen E (2000)The Study Of Men Born In 1913226No significant association (no data shown).
2) Shibata A (1992)The Leisure World Study645RR = 0.94 (0.78-1.14).
Total number of cases: 50,210Average RR = 0.96


Prospective studies of total fruit and total cancer risk (women):
AuthorCohort nameCasesRelative Risk (RR)
22) George SM (2008)The NIH-AARP Diet and Health Study15,792RR = 0.99 (0.94-1.05; P = 0.059).
20) Takachi R (2007)The JPHC Study1,305HR = 1.14 (0.93-1.39; P = 0.29).
18) Boffetta P (2010)The EPIC Study21,000HR = 0.97 (0.92-1.02; P = 0.5).
11) Hung HC (2004)The Nurses' Health Study6,584RR = 1.02.
10) Cox BD (2000)The British HALS131Fruit in winter: OR = 0.87 (0.76-1.00; P = 0.052).
Fruit in summer: OR = 0.88 (0.75-1.02; P = 0.097).
2) Shibata A (1992)The Leisure World Study690RR = 0.76 (0.63-0.91; P = < 0.05).
Total number of cases: 45,502Average RR = 0.99


Prospective studies of total fruit and total cancer risk.
Stratified by smoking status:
AuthorCohort nameCasesRelative Risk (RR) for never smokersRelative Risk (RR) for past smokersRelative Risk (RR) for current smokers
20) Takachi R (2008)The JPHC Study1,631 never smokersHR = 1.12 (0.95-1.33; P = 0.83).--
18) Boffetta P (2010)The EPIC Study13,728 never smokers,

8,832 past smokers, and

7,388 current smokers
HR = 1.00 (0.99-1.01).HR = 0.99 (0.98-1.01).HR = 0.98 (0.97-1.00).
11) Hung HC (2004)The Nurses' Health Study

&

The Health Professionals' Follow-up Study
3,577 never smokers,

3,945 past smokers, and

1,694 current smokers
RR = 1.03 (0.94-1.12).RR = 0.97 (0.89-1.06).RR = 1.07 (0.94-1.22).
Average RR = 1.02Average RR = 0.98Average RR = 1.00


Click here for an extended version of these tables.

Total fruit and total cancer mortality

12 articles, providing information about 11 different cohorts were found including a total of > 5,052 cases (no data was available from 2 cohorts: Hung HC; 2004).
Significant protective effects were found in 2 cohorts, including 3,816 cases (4, 14), though in one cohort the association was significant for men only (4). In addition, a nonsignificant protective effect was found in a third cohort examining men only, and including 114 cases (6). No other associations were found. The effect was of moderate size (excluding incomplete data from Strandhagen E [9], and Hung HC [11]: Average RR = 0.88).

Conclusion: Significant protective effects were found in 2 cohorts. Within one cohort, an effect was found for daily vs less frequent consumption (4), and in the other one for (almost) daily consumption vs < 2 servings/wk (14), indicating that a protective effect may only be found compared with very low or infrequent consumption. Since no data about the amount of cancer cases was available from 2 cohorts, evidence was judged suggestive for a protective effect of total fruits against total cancer mortality for daily vs less frequent consumption (- 11%).

Prospective studies of total fruit and total cancer mortality:
AuthorCohort nameCasesSexRelative Risk (RR)
18) Nöthlings U (2008)The EPIC Study.319Men & WomenRR = 1.08 (0.98-1.19).
16) Khan MM (2004)No cohort name.155 men, and

89 women
Men & WomenMen: RR = 1.00 (0.6-1.6).

Women: RR = 1.1 (0.4-2.9).
14) Sauvaget C (2003)The Hiroshima/Nagasaki Life Span Study.3,136Men & WomenRR = 0.88 (0.80-0.96; P = 0.0044).
13) Maynard M (2003)The Boyd Orr Cohort.158Men & WomenOR = 0.73 (0.47-1.11; P = 0.17).
11) Hung HC (2004)The Nurses' Health Study & The Health Professionals' Follow-up Study.Not defined.Men & WomenNo significant association (no data shown).
9) Strandhagen E (2000)The Study Of Men Born In 1913121MenNo significant association (no data shown).
8) Whiteman D (1999)The OXCHECK Study.223Men & WomenRR = 0.91 (0.63-1.32).
6) Hertog MG (1996)The Caerphilly Study114MenRR = 0.5 (0.3-1.0).
5) Sahyoun NR (1996)No cohort name.57Men & WomenRR = 1.26 (0.53-2.99; P = 0.96).
4) Appleby PN (2002)The Health Food Shoppers Study.680Men & WomenMen: RR = 0.72 (0.56-0.93; P = < 0.05).

Women: 0.88 (0.68-1.13).
Total number of cases: 5,052 + XAverage RR = 0.88


Click here for an extended version of this table.

Total fruit & Vegetables and total cancer.

Primary variables of interest were either total fruits, or total vegetables consumption.
Two articles did not stratify the association by consumption variables, but only provided information about the combined effect of total vegetables & total fruits. Both articles examined the relation with total cancer mortality, and both showed a nonsignificant protective effect (trend).

Both cohorts were of (very) small size. Within the first cohort (Hertog 1996), the highest level of consumption was defined as > 258 g/day (> 3.3 servings/day), and within the second cohort (Genkinger 2004), The strength of the association differed little between the 2nd and 5th quintile of consumption (Table).

Conclusion: Both articles do not provide strong evidence for a protective effect of high consumption, but show a possible increased risk of low levels of consumption.