Abstract

BACKGROUND: Meta-analysis of prospective studies has shown that adherence to the "mediterranean diet" decreases risk of mortality (Sofi F. 2008). The mediterranean diet consists of 8 components on average which are thought to be either protective or detrimental. Dairy consumption is thought to be one of the detrimental components. Still, few systematic reviews were created about the specific components of the mediterranean diet, so labeling these components as protective or detrimental may not be justified.
OBJECTIVES:

• To review all prospective studies which published information about total dairy, milk, and other dairy products in relationship with all-cause mortality/death or survival.
• To define the amount of consumption found to be related with possible effects on mortality/survival.
• To define possible effect modification by confounders.

DATA SOURCE: The Pubmed database was searched (No start date - May 25th, 2010) for relevant articles using the keywords "butter, cheese, cream, dairy, dietary calcium, dietary fat, lactose, margarine, milk, saturated fat, or yoghurt" combined with "prospective, cohort, follow-up, or longitudinal". The exact search term is described Here.
Prospective studies published in the English language were included. Reference lists were searched for additional articles.
RESULTS: 32 articles were found which provided information about 22 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 dairy and milk is described.

• Total dairy. Data about 11 different cohorts was found, including 11,740 deaths. Inconsistent findings were done. No evidence was found for an association between total dairy consumption and all-cause mortality.
• Milk. Data about 10 different cohorts was found, including 93,663 deaths. No evidence was found for an association between high vs low milk consumption and all-cause mortality. The possibility of a protective effect at an intermediate level can not be excluded, but quantification of effects is difficult. No evidence was found for effect modification by milk fat

CONCLUSION: No evidence was found for an association between total dairy, milk, cheese, yoghurt, (ice) cream, butter, or margarine and mortality. In addition, no evidence was found for a modifying effect of dairy-, milk-, and cheese fat on the association between the related items and mortality. And no evidence was found for a difference in effect between butter and margarine.

|Reference: Sofi F. Adherence to Mediterranean diet and health status: meta-analysis. BMJ. 2008 Sep 11;337:a1344. doi: 10.1136/bmj.a1344. Full text|

Total dairy and mortality or survival.

Data about 11 different cohorts was found, including 11,740 deaths.

Use of data: For the overall analysis, results from 3 articles were not included. Data about these 3 articles can be found in the "extended version of the table". About the 3 excluded articles:

• Trichopoulou A (2003). Examined 22.043 subjects aged 20-86 from The Greek part of The EPIC Study, which included 275 deaths.
• Trichopoulou A (2006). Examined 1,013 subjects with diabetes mellitus from The Greek part of The EPIC Study, which included 80 deaths.
• Trichopoulou A (2009). Examined 23,349 subjects aged 20-86 from the Greek part of The EPIC Study, which included 1,075 deaths.

Reason for exclusion in this analysis is as follows:
In 2005 Trichopoulou published results from the EPIC-Elderly Study, examining subjects aged ≥ 60 from 9 European countries, and including data from EPIC-Greece. Most deaths from the first- (2003) and last (2009) publication about The Greek part of The EPIC Study will probably have occured among subjects aged ≥ 60, and therefore inclusion of the data would have led to using the same data twice. In addition, subjects with a history of diabetes mellitus were not excluded from analysis in the EPIC-Elderly Study, so inclusion of the subjects with diabetes mellitus (2006) would also have led to using the same data twice.

Results: (Non)significantly increased risks were found in 3 cohorts (4, 11, 15a), including 242 cases. But (non)significant protective effects were found in 2 other cohorts (7, 21), including 2,852 cases.
Effect modification: Increased risks were restricted to findings from Southern European countries (Greece, Spain), and among cohorts of very small size. But few results were found, and no data was provided about interactions with possible confounders.
Conclusion: Inconsistent findings were done. No evidence was found for an association between total dairy consumption and all-cause mortality.

Milk and mortality or survival.

11 articles, providing information about 10 different cohorts were found providing information about total milk consumption in relation to total mortality, including 93,663 cases. In addition, data about survival was available from 1 additional cohort (Nube M. 1987).

Results: Significant protective effects were found in 3 cohorts (14, 19, 21), including 18,971 cases. Though the trend, but not the RR, was significant in one of these cohorts (14). No other associations were found of high vs low consumption.

Inclusion of intermediate levels of consumption:
Some protective effects were found at different intermediate levels of consumption. No (non) significantly increased risks were found at any level of consumption in any cohort.
Protective effects (RR) of total milk at any level of consumption were as follows:

• Mann JI (1997) Significant at 284 ml/day.
• Ness AR (2001) Significant at 190-757 ml/day.
• Paganini-Hill A (2007) Nonsignificant at < 245 ml/day.
• Iso H (2007) Significant at frequency of consumption ≥ 3 times/week among both men and women (not included in graphics).
• van der Pols JC (2009) Significant at > 199 ml/day.

Effect modification: Interactions with milk fat were as follows:

• Whiteman D (1999) No difference in effect was found between whole milk and semiskimmed or skimmed milk.
• Jamrozik K (2000) No difference in effect was found between full-fat and reduced-fat/skim milk.
• Elwood PC (2005) No difference in effect was found between full fat milk and total milk.
• Andersen LF (2006) A nonsignificantly increased risk (trend) of whole milk was found, while no association was found with skim/low-fat milk.

In only one out of four cohorts (Andersen 2006), a difference in effect was found between whole and (semi)skimmed milk. In addition, whole milk was (almost) the only type of milk consumed in 2 other cohorts (Fawzi W. 1994; van der Pols JC. 2009), and the RR's were < 1 in both cohorts.
The average effect for whole milk from all 4 cohorts providing an RR for the association (Fawzi W. 1994; Elwood PC. 2004; Andersen LF. 2006; van der Pols JC. 2009) gives an RR of 1.02.
No suggestion was found for possible effect modification by other confounders including gender, and continent.
Conclusion: No evidence was found for an association between high vs low milk consumption and all-cause mortality. The possibility of a protective effect at an intermediate level can not be excluded, but quantification of effects is difficult. No evidence was found for effect modification by milk fat or any other confounders.