Adherence to Mediterranean diet and health status: meta-analysis

BMJ 2008; 337 doi: (Published 11 September 2008) Cite this as: BMJ 2008;337:a1344

Are the olives just as green on the other side of the Mediterranean? Meta-analysis of a concept rather than an intervention.

Considering the empirical and public health attention to the
Mediterranean diet, a meta-analysis of studies examining the association
between adherence to this diet and associated mortality and morbidity
outcomes might further our understanding of how nutritional patterns in
accordance with this diet may be of both individual and public health
importance. Sofi et al. [1] reported some potentially important meta-
analytic findings. However, unclarity as to what is understood by the
Mediterranean diet makes these findings suggestive at best. This is
compounded by additional methodological concerns.

Foremost, Sofi et al. do not specify what constitutes the
Mediterranean diet – as if they assume that it is generally known or that
it is a narrowly defined and operationalized diet. Instead, they provide
a definition of the overall composition (7-9 elements), rather flatly so;
without prioritization and proportioning as provided, for instance, in the
Mediterranean diet pyramid [2]; ignoring it is based on dietary habits
from the 1960s; and skirting the call for redefinition [3]. Instead, the
authors focus on adherence to various elements of not much more than a
concept (the Mediterranean diet), operationalizing this adherence (and not
the concept) as the summed extent to which people consume above-median
amounts of each element – a statistical boundary without much empirical
substantiation and devoid of statistical context.

The lack of clarity is compounded by the omission of nutrition-
impacting variation within the Mediterranean region - in terms of topo-
and hydrography, agri- and pesciculture, cultures and religions, and
economic variations. Should one assume that the nutritional pattern
referred to as the Mediterranean diet show a certain uniformity across
this large an area? And how far should one leave the 46000km of shoreline
before dietary patterns begin to diverge significantly? Furthermore, how
does the (ill-defined) Mediterranean diet differentiate itself from other
regional diets that are comparable in part or in all nutritional
components - e.g., diets along the Atlantic coast of Morocco, Portugal,
Spain and France? In the end, what is the Mediterranean diet, where is it
found, how is it operationalized, and how much does it vary across
geography and populations? In the absence of answers to these questions,
the Mediterranean diet is at best a concept, certainly not a definable

Other concerns compound this primary problem of not meta-analyzing an
intervention but rather a concept. Might there be a study cohort effect,
as the Sofi et al. meta-analysis is limited to studies in which adherence
to the diet, not the diet itself, could be quantified? Granted, and using
randomized clinical trial terminology, by studying adherence to the
Mediterranean diet one asserts a degree of “adherence to the protocol”.
Yet this occurs at the risk of a collusion between adherence to a concept
and adherence to a replicable (or at least imitable) intervention
protocol. Might there be a “lead author” effect as five of the nine
studies included in Figure 2 are by two lead authors, including one
covering one-third of the studies? As Sofi et al. argue correctly,
interactions between the components of a dietary pattern must be
considered. However, what this interaction is and how it is influenced by
nonnutritional factors is not addressed: one does not eat in isolation and
instead one is thought familially and socially certain nutritional
patterns – based on (seasonal) availability. Thus, Sofi et al.’s decision
to combine studies with subjects indigenous to the Mediterranean region
and studies involving subjects from, for instance, the US merits a
convincing argument.

Notwithstanding the high position of meta-analysis in the hierarchy
of evidence, sometimes individual and public health might be served better
by an integrative review of a large corpus of studies that addresses the
full range of scientific issues from conceptualization,
operationalization, study design, statistical analysis, to interpretation.
The Sofi et al. findings may provide some direction, however the
destination remains elusive.


[1] Sofi F, Cesari F, Abbate R, Gensini GF, Casini A. Adherence to
Mediterranean diet and health status: a meta-analysis. BMJ 2008;337:1344.

[2] Willett WS, Sacks F, Trichopoulou A, Drescher G, Ferro-Luzzi A,
Helsing E, et al. Mediterranean diet pyramid: a cultural model for
healthy eating. Am J Clin Nutr 1995;61:1402-6.

[3]Serra-Majem L, Trichopoulou A, de la Cruz JN, Cervera P, Alvarez
AG, La Vecchia A, et al. Does the definition of Mediterranean diet need
to be updated ? Publ Health Nutr 2004;7:927-9.

Competing interests:
None declared

Competing interests: No competing interests

06 October 2008
Ivo Abraham
Chief Scientist, Matrix45; Professor, University of Arizona Colleges of Nursing and Pharmacy
Earlysville, VA 22936, USA (Matrix45) and Tucson, AZ 85721 (University of Arizona)