Progress in Nutrition 2018; Vol. 20, N.

3: 344-360 DOI: 10.23751/pn.v20i3.5544 © Mattioli 1885

Original article

“Mediterranean Diet ‘reflections’ ”. Estimating adherence to

the Mediterranean diet through secondary data
Corrado Finardi1, Luca Bucchini2, Aida Turrini3
1
Confederazione Nazionale Coldiretti - E-mail: corrado.finardi@hotmail.it; 2Hylobates Consulting; 3CRA NUT Food Con-
sumptions and DB.

Summary. Purpose: to compare several countries against many Mediterranean adherence indices, calculated
by looking at 19 European Member States. The value of a population-level Mediterranean Diet Index, the
Mediterranean Adequacy Index (P-MAI) is at the core of the analysis. Design/methodology/approach: the EFSA’s
Concise European Food Consumption Database (mean g/day/per capita) and the FAO-FBS dataset (grams
and calories/day/per capita values) were used as the unique sources currently available, in order to derive a
simple yet harmonised secondary-data framework, which could serve for policy analysis and policy making
therein of. Findings: The adherence to a Mediterranean-like dietary pattern outlines a general rank correla-
tion among countries, and a broader north-south divide within Europe. Scores remain relatively stable across
time. Although there has been a decrease in Mediterranean adherence in southern Europe, some central and
northern European regions have seen gains. Research limitations/implications: Several data gaps do not allow a
full comparison across all the indices used (i.e., lack of foodstuff detail of key-foods of the Med diet). A further
problem of Med-adherence indices is that it does not consider the overall caloric intake. Practical implications:
The relatively low discriminatory power of the emerging clusters of countries, reflecting the national diets- limits
their usefulness in terms of policy-making recommendations. Furthermore the indices used were originally built
on first-hand data (i.e., cohort studies relying on real persons), and not on aggregated mean-median values at
population level (secondary data). Social implications In a period in which the interest for the health outcomes
of the Mediterranean diet is on the rise in terms of preventive medicine, the P-MAI is an interesting indicator
due to its user-friendliness, which allows the classification of European countries’ diets using food intake data.
Originality/value: Mediterranean adherence indices may be useful as synthetic indicators for monitoring the
evolution of diets and for identifying sub-regions with similar dietary patterns or changes. The P-MAI index in
particular, due to its simplicity, may help to monitor the overall healthiness of national diets, and could help to
inform subsequent nutrition policies, including emerging labelling provisions both at National and European
level, in order to achieve public health targets (i.e., reduction of NCDs).

Keywords: Mediterranean Diet, National Food Patterns, monitoring

Introduction The Mediterranean Diet is globally recognised

as one of the healthiest dietary patterns, and one of
The “Mediterranean Diet” is a dietary pattern the most studied as well. In spite of existing only one
relying for the most part on fresh fruits&vegetables model, it is apparent that there are different regimes
(including seeds, nuts and legumes), whole grains which may fit to the “Mediterranean Diet” definition,
and olive oil, and on relatively low consumption (2-3 also outside the key-countries of the Mediterranean
times per week) of animal (by)products such as meat basin, and which proved to be healthful as well (3). The
and dairy products (but with high intake of fish), and health and social benefits of the Mediterranean diet
a sparing ethanol intake, generally during meals (1,2). have been extensively documented (4-8) and although
“Mediterranean Diet ‘reflections’” Estimating adherence to the Mediterranean diet through secondary data 345

a causal link between Mediterranean Adequacy and EU- harmonized data. The increasing attention paid
mortality (prevention) has been suggested (9), it has to comparable, compatible and reliable individual food
not been definitely demonstrated (10). Since 2010 consumption data –with in mind a wide EU- risk as-
UNESCO awarded the Med Diet with the Intangible sessment standardization- led EFSA to launch the
Cultural inheritage status. EU-MENU programme to support Members State in
However, is still unclear if Mediterranean countries collecting harmonized data (17).
maintained along time adherence to the Med diet, or -due This seems particularly relevant considering pre-
to increasing trade in processed foods, long food chains liminary cues for the worsening of national dietary
and globalisation- transited to other dietary patterns. patterns in recent decades, at least with respect to a
Another concern regards the real “existence” of some- number of EU countries. According to a set of 88
thing like the Mediterranean diet outside abstractions. health indicators collected in the European Commu-
Criticisms have been raised about the delimitation and nity Health Indicators Monitoring (ECHIM, www.
definition of “what is” eventually a Mediterranean di- echim.org/) database, fruit and vegetable consump-
etary pattern. tion ranks 49th and 50th respectively (www.echim.org/
This paper tries to answer to both questions. Even indicators.html) reflecting a minor importance to the
if there are several metrics and underneath biological health-status - in front of other, more prominent in-
rationales endorsing Mediterranean-style diets (link- dicators.
ing food items to health outcomes), there is a core of Furthermore, the number of EU countries that
food which eventally cannot be missed inside a balanced failed to comply with WHO guidelines for sugar con-
Med Diet. Also, as authors are going to demonstrate, sumption has increased from 1961 to 2003 from 8 to 10
the resilience of dietary patterns-including the Med diet and in the last 40 years, sugar energy shares converged
in southern EU countries is apparent, in spite of a gen- with all countries at around 11% level (18, 19). Al-
eral loss of adherence to the original model, and gains though- according to identical sources during the same
from Northern countries due to specific policy making period-, countries with an adequate intake of fruits and
interventions along the last decades. vegetables more than doubled. The Food Balance Sheets
(FBS) published by the Food and Agriculture Organisa-
tion (FAO) of the United Nations1 showed an increase
The challenge of comprehensive database to moni- in the consumption of animal protein and saturated fat
tor food patterns in the same last 40 years timeframe, particularly within
Mediterranean countries, such as Greece, Italy, Spain,
National dietary assessment and its consequent and Portugal (20). However, there remain remarkable
monitoring represent a key aspect in public health differences across countries, especially in terms of satu-
management. While cross-country comparisons are rated fat intake (Fig. 1), in spite of an apparent conver-
complex to study, they have received support from gence phenomenon: countries with high consumption
the European Commission (EC) (11,12). Further- levels of saturated fatty acids (15%) like Finland or Ire-
more, complicated, resource-intensive, nationwide land reduced them to close to the recommended maxi-
food consumption surveys, which aim to estimate di- mum, (18,19). Given the difficulties associated with
etary patterns, are not currently carried out on an an- analysing individual dietary patterns (16,21), and hence
nual basis. This was examined in the European Food in inferring how dietary patterns converge or diverge
Consumption Survey Methods project (EFCOSUM with the Mediterranean diet across countries and across
project) (13) and subsequently taken on by the Euro- time, it would be useful to examine existing aggregated
pean Food Safety Authority (EFSA), which then went population-level food consumption datasets. This is an
on to establish the Network of Food Consumption “a priori approach” of food consumption patterns – as
Data (former Expert Group on Food Consumption
Data) and set up the Concise Database (14,15) and 1
FAO – Food and Agriculture Organisation of the United Nations-
Food Balance Sheet in FAOSTAT (http://faostat3.fao.org/faostat-
the Comprehensive Database (16), in order to collect gateway/go/to/home/E).
346 C. Finardi, L. Bucchini, A. Turrini

defined by Efsa- and “is based on prevailing knowledge the adequacy of national diets against the Mediterra-
concerning favourable or adverse effects of various di- nean diet. Both MAI and P-MAI are calculated as the
etary constituents. Diets are assessed for the presence or ratio between the summed weight (or the summed en-
absence of certain food or nutrient characteristics, and ergy value) of food items from the core Mediterranean
the resulting score is then operationalised as a dietary Diet (vegetables, fruit, cereals, red wine, vegetable oils,
exposure variable” (22)- but without an empirical, a- potatoes and fish) and of non-core foods (meat, dairy
posteriori assessment of the health outcomes as integral products, animal fat, eggs and sugar) (9 ,25).
part of the research. The Mediterranean Adequacy In- The P-MAI was computed using the average in-
dex (MAI) is one of the most predictive indicators of a take of 15 food groups and 21 sub-groups from the
Mediterranean diet (9, 23, 24). The MAI inversely cor- EFSA CONCISE European Food Consumption Da-
relates with 25 years of figures for deaths from coronary tabase (g/day/per capita intakes), hereafter referred to
heart disease (6). as the CONCISE database/data, and from the FAO
This study aims to analyse the possibility of build- Food Balance Sheets (FBS database/data) (g or kcal/
ing a Mediterranean diet adequacy index-, the Popula- day/per capita intakes 1961-2007). The CONCISE
tion level Mediterranean Adequacy Index, P-MAI- to database comprises mean food consumption data for
allow for monitoring dietary trends using figures from adults (aged 16-64 years) departing from different
national Food Balance Sheets (FBS), which are pub- food categories. FBS data were calculated by dividing
lished annually on the FAO website (2011).While the the total amount of food available for consumption by
index per se is not new (MAI), it has never been applied the aggregate population of a given country2.
before to aggregated data at population-level (P-MAI). It is important to point out that some foods were
At the same time, more precise food-intake data col- not included in the CONCISE database, for exam-
lected at EU Member State level (EFSA Concise Food ple, vegetable oils and red wine. Red wine was also not
Consumption Database, hereinafter “CONCISE”) are available in the FBS database. Instead, ‘wine’ (FBS
available, but under a more limited timeframe (i.e., dif- database) and ‘wine and substitutes’ (CONCISE da-
ferently from FAO FBS, only in selected years of surveys, tabase) were used. Furthermore, as vegetable oils were
not harmonised at the EU level and with each Member not recorded separately from animal fats, the broader
State having different surveys years). category ‘fats’ was excluded when using the CON-
These two sources of data –even if different by CISE database.
nature-enable us to draw some kind of comparison on Countries were selected based on national sur-
specific periods of time (i.e. the years during which the veys availability inside the CONCISE and conse-
dietary surveys were carried out in the Member States quently, more available FBS data were in turn includ-
based on individual national dietary surveys (14). ed for comparison. This means that all FBS data were
Therefore, the index proposed in this study is the aligned to those in the CONCISE database according
P-MAI (Population-level Mediterranean Adequacy to survey year. However, in a separate analysis rely-
Index), which provides geographical and temporal in- ing on longitudinal FBS data only -Spain and Greece
sights into food consumption patterns across the EU were added, as examples of Mediterranean countries.
Member States. In this way, it creates a user-friendly tool Whilst data for Estonia in 1961 was not available in
for public health policy-making, at a time when there is the FBS database. In fact, in order to assess histori-
increasing focus on food-related diseases and costs. cal trends, P-MAI scores were estimated in both 2007
and 1961 using FBS data (the only dataset allowing for
this diachronic assessment).
Methods
2
The total amount is obtained by examining production and import
figures, less export and re-use figures (supply fed to livestock or used
The P-MAI, as previously stressed, is an extension for seed, and losses during storage and transportation). divided by
of the original (and well-established) concept (MAI) the national population level for the given year. FBS are inherently
advantageous as they take into account both domestic and non-domestic
proposed by Alberti and Fidanza (2004) to measure food consumption (catering, restaurants, etc.).
“Mediterranean Diet ‘reflections’” Estimating adherence to the Mediterranean diet through secondary data 347

Energy intake was derived from the FBS (grams tained, the better the diet.
instead were used for CONCISE). b) The Global Nutrition Index (GNI), (28), account-
All data was entered into Excel spreadsheets ver- ing for three indicators of nutritional status: defi-
sions (2010 and 2013) and Scott’s choice analysis, was cits, excess, and food security.
used to identify the number of classes with internal and c) The Mediterranean Score (29). This score seizes
external consistency (26). The Scott’s choice test is a sim- the adherence to the Med-diet and relies on spe-
ple rule for describing an optimal grouping for the iden- cific cut-off points for healthy vs unhealthy foods
tification of clusters of countries (see Figure 1 below). (i.e. 1 point for healthy foods such as cereals, fruit,
Concordance between the different elaborations of vegetables and legumes, fish and moderate ethanol
P-MAIs (CONCISE vs. FBS, FBS grams and FBS cal- amounts; 0 points for ‘unhealthy’, non-Mediterra-
ories, and FBS time series) was measured using Spear- nean foods such as meat and dairy products). Un-
man’s rank (r) and the Kendall’s Tau (t) correlational fortunately, the ratio between monounsaturated/
analysis. Other correlations were included once added saturated fats, as originally indicated in Trichopou-
other diet-focused indices, in order to compare the re- lou (2003)(29) could not be provided. Therefore,
sulting classification with that determined by P-MAI. olive oil consumption versus animal fat consump-
The indices are: tion was used instead as a proxy for the monoun-
a) A simplified form of the Diet Quality Index for the saturated fats-saturated fats ratio. Nor was it possi-
Mediterranean Region (Med-DQI) developed by ble to analytically separate legumes from vegetables
Gerber (2006)(27). The Med-DQI is a screening using the CONCISE database.
tool which gives scores (from 0 to 2) for the intake Other indices were also initially considered with-
of the following food items, meats, olives, fish, ce- in this analysis, namely, the Mediterranean-Style Di-
reals, fruit and vegetables, as based on Table 1. etary Pattern Score (MSDPS) proposed by Sanchez-
With regard to the analysis of the Med-DQI, nei- Villegas et al. (2002) (30); the Mediterranean Dietary
ther cholesterol nor SFAs were included –in spite Pattern (MDP) by Rumawas et al. (2009) (31); and the
of being present in the original DQI- due to a lack Mediterranean diet score by Panagiotakos et al. (2006)
of European population-level data. Instead, to (32). However, these were subsequently excluded either
make directly comparable the DQI with the other due to a lack of available data (i.e. on trans fatty acids),
Indexes, the complement to 10 of the Gerber’s In- or due to other classification difficulties.
dex was calculated, such that the higher value ob- A one-way analysis of variance (ANOVA) was
performed to measure the impact of the aforemen-
tioned data sources (CONCISE or FBS) as a major
contributor to changes in the P-MAI value, in order to
depurate results from dataset effects. The rank correla-
tion coefficients (the Spearman’s Rho r and the Ken-
dall Tau t) were used in the analysis to measure cor-
relations among GNI and Med-DQI with the other
indices.
Figure 1. The Scott’s choice test for the optimal number of
classes/members of a class

Table 1. Scoring system derived as simplification of the Diet Quality Index for Mediterranean Region
Scores Meats (g) Olive oil (ml) Fish (g) Cereals (g) Vegetables+fruit (g)
0 <25 >15 >60 >300 >700
1 25-125 15-mag 60-30 300-100 700-400
2 >125 <5 <30 <100 <400
348 C. Finardi, L. Bucchini, A. Turrini

Results and Discussion From the results obtained, a geographical gradient

can be seen in Table 3, for example, Italy was among
Results showed that, the estimated Population- the highest in terms of P-MAI scores according to all
level Mediterranean Adequacy Index (P-MAI) scores three calculations (P-MAI, MDQI, and MSC). In
in European countries when estimated from average general, a North-South trend can be observed with
food intake from the CONCISE (Table 2) varied from northern countries in the cluster of lower P-MAI
0.86 to 2.34, whereby higher scores indicated increased scores (i.e. lower adherence to the Mediterranean diet).
adherence to a Med-Dietary pattern. At the lowest levels, Scandinavian countries maintain
FBS data were aligned to those in the CONCISE their ranking in the first two (lower adherence) clus-
database according to survey year3. ters, despite differences between the CONCISE da-
tabase and FBS database computations. The Nether-
3
Where figures in the CONCISE dataset referred to multiple years, lands, Iceland and Finland remain in the first cluster,
the mean of the corresponding years in the FBS was used (e.g. if
CONCISE(country i)1986-87 was the reference period of the survey, whereas Norway moves from the first cluster to the
then a mean of FBS(i)1986 and FBS(i)1987 was calculated for country i).

Table 2. P-MAI (Population-level Mediterranean Adequacy Index) scores for the aggregated average national diets, estimated by
food weight (g) from the FAO Food Balance Sheets (“FBS”) and the EFSA CONCISE database (“CONCISE”). For comparabil-
ity reasons
FBS g* CONCISE g* Delta Delta Years of
(mean values) (% difference (absolute g) reference
between (EFSA)
the 2 values)
AUT 1.18 2.20 86.42 1.02 Average 2005-2006
BEL 1.14 2.12 85.94 0.98 2004
BGR 1.35 1.88 39.35 0.53 2004
CZE 1.13 1.59 40.30 0.46 Average 2003-2004
DEU 1.05 1.67 59.29 0.62 1988
DNK 1.09 1.31 20.63 0.22 Average 2000-2001-2002
EST 1.39 1.49 7.23 0.10 1997
FIN 0.81 0.86 6.34 0.05 2002
FRA 1.1 1.70 54.25 0.60 1999
GBR 1.31 1.53 16.97 0.22 Average 2000-2001
HUN 1.33 1.48 11.01 0.15 Average 2003-2004
IRL 1 1.67 66.80 0.67 Average 1997-1998-1999
ISL 0.98 1.01 3.48 0.03 2002
ITA 1.5 2.34 56.30 0.84 Average 1994-1995-1996
NLD 0.79 1.17 48.20 0.38 Average 1997-1998
NOR 1.07 0.95 11.11 0.12 Average 1993-1997
POL 1.47 2.30 56.43 0.83 2000
SVK 1.42 2.26 59.26 0.84 2006
SWE 0.78 1.28 63.78 0.50 Average 1997-1998-1999
*EFSA, European Food Safety Authority; FBS, FAO-Food Balance Sheets. P-MAI calculated as the ration between Med foods (vegetables,
fruit, cereals, red wine, vegetable oils, potatoes and fish) and of non Med foods (meat, dairy products, animal fat, eggs and sugar).
Austria (AUT); Belgium (BEL); Bulgaria (BGR); Czech Republic (CZE); Germany (DEU); Denmark (DNK); Estonia (EST); Finland
(FIN); France (FRA); United Kingdom (GBR); Hungary (HUN); Ireland (IRL); Iceland (ISL); Italy (ITA); the Netherlands (NLD);
Norway (NOR); Poland (POL); Slovakia (SVK).
“Mediterranean Diet ‘reflections’” Estimating adherence to the Mediterranean diet through secondary data 349

second from the CONCISE database to the FBS da- section later on). We should however take into account
tabase. Sweden moves from the second to the first and here the fact that the FBS P-MAI based on grams is
Denmark remains in the second cluster in both the slightly lower than he FBS P-MAI based on kilocalo-
CONCISE and FBS calculations. Interestingly, when ries (1.47 and 1.49). This could mean that allegedly
adopting the synth-MDQI, Norway scores in the first calorie-dense, healthy foods play some minimal role in
cluster of Mediterranean Diet adherence. However, meliorating the score (again, potatoes, or alcohol).
this may be biased, as cholesterol and SFAs were not As for Germany, according to the CONCISE da-
included in this synth MDQI, and historically Nor- tabase, vegetable consumption is relatively high (252g/
dic countries have a high intake of these, as observable day, and 125g/day for potatoes) as is fruit consumption
when considering food matrices of departure. In both (190g/day), while dairy product consumption stands
the CONCISE and FBS databases, Poland, Slovakia at 313g/day. Beer consumption, which covers 184g/
and Italy showed/had the best Mediterranean adher- day of the 231g of alcoholic beverages consumed on a
ence scores. daily basis, is not taken into account in this computa-
Italy shows relatively good fruit and vegetable con- tion because of the potential bias of the indicator used
sumption (respectively 203g/day and 249g/day, about (excessive consumption, which is unhealthy, equally
4 to 5 portions, against a virtual recommendation of enhances the score a higher value since - to determine
at least overall 4 portions-or 400 g/day- from WHO the “right amount” of alcohol to be consumed in order
in 1991) (33), low meat intake (137g/day- against the to have health benefits- no thresholds are in place for
standard Med Diet advice of a moderate consumption the traditional P-MAI- as on the contrary, other indi-
of 2-3 servings per week- no WHO recommendations cators do, such as the MDS).
here) and low sugar intake (19g/day.- well below the In fact, FBS-based computations for Germany
10% of total energy intake as suggested by WHO). In show lower P-MAI values (0.97 based on kilocalo-
general, Italy has a more homogeneous ranking along ries and 1.05 based on grams), most likely due to the
the different datasets used with only minor variations FAO’s more detailed food categories (in particular, the
in the ranking in response to the use of different Med vegetable oil/animal fats ratio and wine).
Diet indicators or datasets used. Despite this, there is a noticeable divergence ver-
The high P-MAI value (i.e. adherence to the sus the same indicator (P-MAI) when relying on the
Mediterranean diet) for Italy is not surprising. How- CONCISE dataset (with a value of 1.67) as illustrated
ever, other figures require insight into the data in or- via a comparison of the CONCISE database vs. FBS
der to be explained further. For example, taking into (Z-scores) (see Figure 2).
account the CONCISE database result for Austria, Results from the historical trend observing the P-
the relatively high P‑MAI (2.20) was due to its rela- MAI score both in 2007 and 1961 using FBS dataset,
tively high fruit and vegetable consumption (202g/day showed that during this time period, P-MAI scores
and 211g/day respectively, and 59g/day of potatoes), decreased in most countries. Across all countries (i.e.
as well as its low intake of dairy products (171g/day) Mediterranean and northern European countries), av-
and sugar (23g/day). As regards Poland, which has a erage P-MAI decreased from 1.83 to 1.37 (from 1961
P-MAI value of 2.30 in the CONCISE database, we to 2007) and the standard deviation (SD) decreased
note a high consumption of vegetables (292g/day), po- from 1.27 to 0.34 (Table 3). This may be interpreted
tatoes (304g/day) and fruit (282g/day). Although con- on the one hand as result of more globalised lifestyles
sumption levels for meat (259g/day) and dairy prod- and dietary patterns; on the other one, as public health
ucts (181g/day) are high, the overall P-MAI remains policies in charge to Nordic countries governments
relatively good and potatoes play a key role here, as to meliorate dietary behaviours since the ‘70es. In the
they are considered as ‘vegetables’ inside the traditional same period, all core-Mediterranean countries expe-
P-MAI score (even if this is questionable from a pub- rienced decreases in P-MAI (Italy: -1.32; Portugal:
lic health perspective: in the UK potatoes are not valid -2.03; Spain: -2.25; and Greece: -2.59). On a rela-
for the “5 a day” F&V purposes- see the Discussion tive scale, southern European countries (Greece, Italy,
Table 3. Clusters determined by Scott’s choice applied to -P-MAI calculated from FBS - calories and grams and CONCISE database (g). by CONCISE database coun-
tries (*) and according FBS years. The MDQI and MSC scores and rankings (“Cluster”) are added for comparison. 350
P-MAI Cluster Country P-MAI Clusters Country P-MAI Clusters Country MDQI Clusters Country MDQI Clusters Country MDS Clusters Country MDS Clusters
Country
CONCISE FAO FAO CONCISE FAO CONCISE FAO
(g) FBS FBS (g) FBS (g) FBS
(g) (cal) (g) (g)

FIN 0.86 1 SWE 0.78 1 ISL 0.68 1 CZE 3 1 SVK 2 1 NLD 1 1 ISL 2 1

NOR 0.95 NLD 0.79 DNK 0.82 DNK 3 AUT 3 2 NOR 2 2 DEU 3 2

ISL 1.01 FIN 0.81 NLD 0.93 EST 3 POL 3 GBR 2 NOR 3

NLD 1.17 ISL 0.98 2 DEU 0.97 2 IRL 3 DEU 3 IRL 2 NLD 3

SWE 1.28 2 IRL 1 FIN 1.03 HUN 3 EST 3 ISL 2 FIN 3

DNK 1.31 DEU 1.05 SWE 1.04 NLD 3 IRL 3 FIN 2 CZE 4 3

HUN 1.48 NOR 1.07 FRA 1.05 BEL 4 2 HUN 3 EST 2 SWE 4

EST 1.49 DNK 1.09 NOR 1.15 BGR 4 CZE 3 BEL 3 3 DNK 5 4

GBR 1.53 FRA 1.1 BEL 1.16 DEU 4 GBR 3 SWE 3 AUT 5

CZE 1.59 3 CZE 1.13 HUN 1.18 SVK 4 NLD 3 CZE 3 BGR 5

IRL 1.67 BEL 1.14 IRL 1.22 3 SWE 4 FRA 4 3 DNK 3 EST 5

DEU 1.67 AUT 1.18 3 AUT 1.23 GBR 4 BGR 4 HUN 3 POL 5

FRA 1.70 GBR 1.31 CZE 1.32 ISL 5 3 DNK 4 SVK 3 GBR 5

BGR 1.88 HUN 1.33 4 GBR 1.35 AUT 5 FIN 4 BGR 3 FRA 5

BEL 2.12 4 BGR 1.35 SVK 1.47 4 FIN 5 ISL 4 POL 4 4 IRL 5

AUT 2.20 EST 1.39 POL 1.49 FRA 5 SWE 4 FRA 4 ITA 5

SVK 2.26 SVK 1.42 EST 1.52 ITA 5 BEL 5 4 AUT 4 BEL 6 5

POL 2.30 POL 1.47 BGR 1.89 5 POL 5 NOR 5 DEU 5 5 HUN 6

ITA 2.34 5 ITA 1.5 ITA 1.95 NOR 5 ITA 7 5 ITA 6 6 SVK 6

(*) Country indicated by the three digits standard

P-MAI CONCISE (g): Population-level Mediterranean Adequacy Index calculated on EFSA’s Concise database (grams-based calculation); P-MAI FAO FBS (g): Population-level Mediterranean Adequacy Index calculated on FAO
Food Balance Sheets database (grams-based calculation); P-MAI FAO FBS (cal): Population-level Mediterranean Adequacy Index calculated on FAO Food Balance Sheets database (calories-based calculation); MDQI CONCISE (g):
Mediterranean Diet Quality Index (synth version based on available aggregated data) calculated on EFSA’s Concise database (grams-based calculation); MDQI FAO FBS (g): Mediterranean Diet Quality Index (synth version based on
available aggregated data) calculated on FAO Food Balance Sheets database (grams-based calculation); MDS CONCISE (g): Mediterranean Diet Score calculated on EFSA’s Concise database (grams-based calculation); MDS FAO FBS
(g): Mediterranean Diet Score calculated on FAO Food Balance Sheets Database (grams-based calculation).
Austria (AUT); Belgium (BEL); Bulgaria (BGR); Czech Republic (CZE); Germany (DEU); Denmark (DNK); Estonia (EST); Finland (FIN); France (FRA); United Kingdom (GBR); Hungary (HUN); Ireland (IRL); Iceland
(ISL); Italy (ITA); the Netherlands (NLD); Norway (NOR); Poland (POL); Slovakia (SVK).
C. Finardi, L. Bucchini, A. Turrini
“Mediterranean Diet ‘reflections’” Estimating adherence to the Mediterranean diet through secondary data 351

Figure 2. Biplot of P-MAI (Population-level Mediterranean Figure 3. P -MAI (Population-level Mediterranean Adequa-
Adequacy Index) Z- scores for the Efsa mean national diet, cy Index) scores for the aggregated average national diets from
estimated from the EFSA concise database, by food weight FAO FBS in 2007 – X axis- and 1961 Y axis-, and change
versus the FAO FBS data. Scores are standardized and com- in selected EU Member States (calories-based computation)
pared with the average of the variable. Data collection period (Pearson correlation: 0,84 p-value not significant 1.58).
as per Table 1. Norway (NOR); United Kingdom (GBR);
Belgium (BEL); Ireland (IRL); Iceland (ISL); Finland (FIN);
Sweden (SWE); Czech Republic (CZE); Denmark (DNK); CONCISE datatabase, FBS database (g), or FBS da-
Netherlands (NLD); Poland (POL); Germany (DEU); Hun- tabase (kcal) (see Tables 2 and 3).
gary (HUN); France (FRA); Austria (AUT); Slovakia (SVK); Given that one of the goals of this study was to
Italy (ITA); Bulgaria (BGR); Estonia (EST).
assess the P-MAI’s discriminatory power, it was tested
against rank correlation both internally (CONCISE
Spain and Portugal) halved their P-MAI, whereas P-MAI and FBS P-MAI) and externally, taking into
northern European countries (Norway and the UK, consideration other food quality indices capable of
but also Finland and Sweden) increased their adher- providing further insight. In particular, the simplified
ence to the Mediterranean Diet. form of the Diet Quality Index for the Mediterra-
For these FBS-based historical trends, as expect- nean Region (synth Med-DQI) developed by Gerber
ed, g-based values and kcal-based values were positive- (2006) (27) was added for comparative purposes (lim-
ly correlated (Pearson = 0.77), as was the FBS database ited to food items with and without cholesterol and
correlated with the CONCISE database (0.71). saturated fatty acids), as the Global Nutrition Index
However, a substantial stability of patterns for (GNI) of Rosenbloom et al. (2008) (28).
the interested countries emerges (Figure 3)-, P -MAI Spearman’s rank (rs) and Kendall’s tau (t) correla-
scores for the aggregated average national diets from tions were calculated to compare the list of countries
FAO FBS in 2007 – X axis- and 1961 Y axis-, in se- as classified by the P-MAI ranks (derived from the
lected EU Member States (calories-based computa- CONCISE and FBS(g)), obtaining r = 0.72 (p-val-
tion) showed a Pearson correlation of 0,84 and a R- ue 0.002) and t = 0.54 (p-value 0.001). When com-
squared of 0,617. paring P-MAI ranks derived from the CONCISE
Finally, P-MAI values did not appear to be asso- and FBS(kcal), r =0.67 (p-value 0.004) and t = 0.50
ciated with total energy intake, regardless of the data- (0.003) were found. All coefficients showed a high
set used for the P-MAI estimate (CONCISE, FBS(g), level of concordance (value³ 0.5).
or FBS(kcal)), with r varying from ‑0.16 to 0.25 and An overview on how such indicators performed
0.07, respectively, in 2007 in the FBS data. in the clustering analysis is shown in Table 3, whereas
For the sake of comparison we provide detailed a correlation matrix across the indicators used can be
P-MAI values for the indicator used, whether it be the seen in Table 4.
352 C. Finardi, L. Bucchini, A. Turrini

Comparison with the Rosenbloom Global Nutri- foods), but also by the cut-off based system (assigning
tional Index (GNI), which sums up both food dep- values ranging from 0 to 2 according to consumption
rivation and overweight burden (the index combines thresholds). This implies an underestimation of specif-
deficits, excesses and food security), failed to show any ic food consumption not reaching the threshold level
correlation with P-MAI. Examining the CONCISE of the specific nutrient.
P‑MAI ranks versus the GNI ranks for the same
countries results in no correlation found for any of
the rank-correlation indices and for any of the data- Discussion
sets used. The Diet Quality Index for Mediterranean
Region (Med-DQI) developed by Gerber (2006) (27), The health status of persons and even more of
provided clustering with some differences from the population is linked to a number of factors, only a part
P-MAI. This may be explained by the different items of which refer to a healthy diet (Waxman, 2005/34). A
taken into account (i.e. mostly nutrients instead of real healthy diet is in fact only part of broader healthy life-

Table 4. A comparison across the indicators, either considering absolute values or ranks
Spearman rank nCLUS-PMAI- nCLUS-PMAI- nCLUS-PMAI- Rank MDQI- Rank MD Rank MDS Rank
correlation CONCISE-mean FBS-calories FBS-grams CONCISE QI-FBS -CONCISE MDS-FBS

nCLUS-PMAI-
CONCISE-mean
1 1 1 0 0 1 0

nCLUS-PMAI-
FBS-calories 1 1 0 0 0 1

nCLUS-PMAI-
FBS-grams 1 0 0 0 1

Rank MDQI-
CONCISE 1 0 0 0

Rank-MDQI-FBS 1 0 0

Rank MDS
CONCISE 1 0

Rank MDS-FBS 1

nCLUS-PMAI-CONCISE-MEAN sign + sign + sign + sign - sign +

sign +

nCLUS-PMAI-FBS-CALORIES sign + sign + sign + sign +

sign +

nCLUS-PMAI-FBS-GRAMS sign + sign - sign +

sign +

Rank-MDQI-CONCISE sign + sign +

sign -

Rank-MDQI-FBS sign + sign -

Rank-MDS-CONCISE
sign +

Rank-MDS-FBS
“Mediterranean Diet ‘reflections’” Estimating adherence to the Mediterranean diet through secondary data 353

style. However, to measure the adherence to a healthy When ranking and clustering was carried out us-
diet is of the upmost interest since it is one of the most ing the Scott’s choice test (Tab. 3) for the FBS(g) P-
controllable variable either individually or inside pub- MAI, most countries maintained their position or at
lic policy initiatives. Furthermore, the focus on diets least maintained their original cluster from the CON-
instead than on food items is increasingly prevalent, as CISE (Tab.6). The only major changes were seen with
a full risk- assessment cycle requires also exposure data Estonia (from 0.71 to 1.39 and from cluster 2 to 4),
(consider the Trans Fatty Acids-TFAs debate at the Poland (from 0.76 to 1.47 and from cluster 2 to 4) and
EU level, if provide for labelling on food or not, due to the UK (from 0.72 to 1.31 and from cluster 2 to 3). It-
the unquestionable risk characterization, but also with aly and Slovakia remained at the top of the ranking of
a low exposure of the population) (35EC, 2005)) and countries for Mediterranean Adequacy in both cases,
also a risk-benefit assessment -(i.e., for eating oily fish- other countries worsened passing onto calories-based
with benefits from omega 3 fatty acids and risks from computation and others ameliorated. However, non-
PCBs, heavy metals and dioxins- (EFSA,2015/36)). Med foods (meat, dairies, fats) are generally denser
Clear trade-offs hence refer more and more to the bal- from a caloric perspective than Med- foods (cereals,
anced consumption of foods more than on the mere F&V, legumes..,), i.e., a lower intake of Non-Med foods
intake (or avoidance) of a food matrix.. gives better scores than a proportionally increased intake of
Once cleared this aspect, it can be disclosed that Med foods- even if hardly there is a general rule (olive oil
to a certain extent, P-MAI appears to discriminate is more caloric than butter; cereals have the same 4 calories
between different diets across Europe and to detect per g than meat). Many of the Nordic countries main-
changes over time. Across databases, eastern Euro- tained both their position and clusters. Despite their
pean countries seem to have a higher adherence to a relatively high potato consumption, scoring positively-
Mediterranean-like dietary pattern than northern and adherence to the Mediterranean diet was low and no
central European countries. This is perhaps due to a significant effect was seen by including starchy foods
lower meat consumption, which can most probably be in the list of Mediterranean foods.
attributed to reduced access to meat and to higher po- When examining calories instead of grams, the
tato consumption levels. The apparent phenomenon of effects of ranking variations were subtle, yet present-
converging diets in Europe is interesting and has been the SD –reasonably- increased. The UK and Ireland
investigated elsewhere (18,37). each gained a rank in the clusters partition (from 1.02
As potatoes and starchy foods are a controversial to 1.35 and 0.94 to 1.22 respectively), as does Estonia
category as regards their place in the Mediterranean (from 1.08 to 1.52).
diet, they have alternatively been included either in the It is possible to draw similar conclusions from the
list of Mediterranean foods or on the contrary, in the EFSA database. Here, it is important to bear in mind
list of non-Mediterranean foods (Tables 5 and 6). that starchy foods only play a minor role in the diet,
For this reason, P-MAI scores were higher when regardless of whether calories or grams are used as in-
potato values were calculated in grams rather than in dicator units.
calories. This is due to the relatively low energy den- It is possible to advance similar considerations to
sity of potatoes compared to other food categories. those pinpointed for potatoes about other food items
On average, the difference between the P-MAI with that do not yet have a clear nutritional status, such as
potatoes and the P-MAI without potatoes equals a beer or fruit juices with low fruit content4. When in-
value of 47.5% for FBS g/day computations, but “only” cluding them in the P-MAI (CONCISE g)-by food
20.9% for calories/day computations (Table 5). The weight-, with the assumption of beer at the numerator
inclusion of potatoes in Med Foods seems to increase (ethanol was positively considered within the MDS)
discrimination between MS dietary patterns -resulting and soft drinks at the denominator, the indicator
in higher standard deviation values-, but the scientific
rationale of such inclusion is questionable from a pub- 4
Soft drinks are considered drinks with a fruit content lower than
‘nectar’, as defined by the European Commission (EC) Directive
lic health perspective. 2001/112, typically containing 25-50% fruit, but with added sugars.
354 C. Finardi, L. Bucchini, A. Turrini

Table 5. PMAI on the datasets FBS (grams and calories) and CONCISE (mean and median values), with potatoes/starchy foods
alternatively included in Med foods (NUMERATOR) or in non Med foods (DENOMINATOR)
FBS Grams FBS Calories CONCISE MEAN
Potatoes in potatoes in delta(%) potatoes in potatoes in delta(%) potatoes in potatoes in % Delta
“non-med foods” “med-foods” “non-med” foods “med” foods “non med foods” “med foods”
AUT 0,9 1,18 31,1 1,08 1,23 13,9 1,78 2,20 23,9
BEL 0,81 1,14 40,2 0,99 1,16 17,2 1,48 2,12 43,6
BGR 1,07 1,35 26,2 1,69 1,89 11,8 1,32 1,88 42,5
CZE 0,75 1,13 50,7 1,1 1,32 20 1,09 1,59 45,8
DEU 0,68 1,05 54,4 0,78 0,97 24,4 1,18 1,67 42,1
DNK 0,77 1,09 41,6 0,68 0,82 20,6 0,94 1,31 39,8
EST 0,71 1,39 95,8 1,08 1,52 40,7 0,79 1,49 88,6
FIN 0,58 0,81 39,7 0,85 1,03 21,2 0,61 0,86 40,5
FRA 0,82 1,1 34,1 0,91 1,05 15,4 1,39 1,70 22,2
GBR 0,72 1,31 81,8 1,02 1,35 32,7 1,04 1,53 48
HUN 0,94 1,33 41,5 1,01 1,18 16,8 1,04 1,48 41,6
IRL 0,55 1 81,8 0,94 1,22 29,8 0,78 1,67 114,2
ISL 0,75 0,98 30,7 0,6 0,68 13,3 0,85 1,01 19,8
ITA 1,26 1,5 19 1,78 1,95 9,6 1,97 2,34 18,7
NLD 0,54 0,79 46,3 0,77 0,93 20,8 0,79 1,17 48,6
NOR 0,74 1,07 44,2 0,96 1,15 19,5 0,64 0,95 48,9
POL 0,76 1,47 93,4 1,1 1,49 35,5 1,09 2,30 110,1
SVK 1,2 1,42 18,3 1,23 1,47 19,5 1,52 2,26 48,3
SWE 0,59 0,78 32,2 0,91 1,04 14,3 0,84 1,28 52,2
Mean 0,8 1,15 47,53 1,03 1,23 20,89 1,11 1,62 50,96
SD 0,2 0,22 0,29 0,32 0,37 0,45
(*) Country indicated by the three digits standard; Austria (AUT). Belgium (BEL); Bulgaria (BGR); Czech Republic (CZE); Germany
(DEU); Denmark (DNK); Estonia (EST); Finland (FIN); France (FRA); United Kingdom (GBR); Hungary (HUN); Ireland (IRL);
Iceland (ISL); Italy (ITA); the Netherlands (NLD); Norway (NOR); Poland (POL); Slovakia (SVK).Mediterranean Foods as by original
Mediterranean Adequacy Index: vegetables, fruit, cereals, red wine, vegetable oils, potatoes and fish. Non Mediterranean Foods as by original
Mediterranean Adequacy Index: meat, dairy products, animal fat, eggs and sugar

shows substantial variations as compared to the base- with added sugar and a low fruit content require also
line model. The mean P-MAI value without including proper examination from a public health perspective.
beer and soft drinks is 1.37 (SD 0.43), but changes to In fact, there are also quite high consumption levels for
1.28 (SD 0.40) with both foods included in the above- these fruit juices across the EU (Norway 330g, Iceland
mentioned positions. 339g, Belgium 275.2g, and the UK 219g).
Although the inclusion of beer inside Med-foods With regard to scores other than the P-MAI, af-
is questionable, as several countries have a high average ter taking into account the basic Mediterranean Score
beer consumption (i.e. the Czech Republic 373g/day, index (MDS), the scope was enlarged to include non-
Ireland 299g/day, and the UK 257g/day), fruit juices Mediterranean foods, as well as potatoes and eggs,
Table 6. PMAI on the datasets FBS (grams and calories) and CONCISE (mean and median values), with potatoes/starchy foods (“POT”) alternatively included in Med
foods or in non-Med foods. Clusters provided depending on the rankings.
Pot Pot Pot Pot Pot Pot
“Non-Med” “Med” “Non-Med” “Med” “Non-Med” “Med”

FBS g Cluster FBS g Cluster FBS cal Cluster FBS cal Cluster CONCISE Cluster CONCISE Cluster
mean mean

NLD 0,54 1 SWE 0,78 1 ISL 0,6 1 ISL 0,68 1 FIN 0,61 1 FIN 0,86 1

IRL 0,55 NLD 0,79 DNK 0,68 DNK 0,82 NOR 0,64 NOR 0,95

FIN 0,58 FIN 0,81 NLD 0,77 NLD 0,93 IRL 0,78 ISL 1,01

SWE 0,59 ISL 0,98 2 DEU 0,78 DEU 0,97 2 NLD 0,79 NLD 1,17

DEU 0,68 IRL 1 FIN 0,85 2 FIN 1,03 EST 0,79 SWE 1,28 2

EST 0,71 2 DEU 1,05 FRA 0,91 SWE 1,04 SWE 0,84 DNK 1,31

GBR 0,72 NOR 1,07 SWE 0,91 FRA 1,05 ISL 0,85 HUN 1,48

NOR 0,74 DNK 1,09 IRL 0,94 NOR 1,15 DNK 0,94 2 EST 1,49

CZE 0,75 FRA 1,1 NOR 0,96 BEL 1,16 GBR 1,04 GBR 1,53

ISL 0,75 CZE 1,13 BEL 0,99 HUN 1,18 HUN 1,04 CZE 1,59 3

POL 0,76 BEL 1,14 HUN 1,01 IRL 1,22 3 CZE 1,09 IRL 1,67

DNK 0,77 AUT 1,18 3 GBR 1,02 AUT 1,23 POL 1,09 DEU 1,67

BEL 0,81 GBR 1,31 AUT 1,08 CZE 1,32 DEU 1,18 FRA 1,70
“Mediterranean Diet ‘reflections’” Estimating adherence to the Mediterranean diet through secondary data

FRA 0,82 HUN 1,33 4 EST 1,08 GBR 1,35 BGR 1,32 3 BGR 1,88

AUT 0,9 3 BGR 1,35 CZE 1,1 3 SVK 1,47 4 FRA 1,39 BEL 2,12 4

HUN 0,94 EST 1,39 POL 1,1 POL 1,49 BEL 1,48 AUT 2,20

BGR 1,07 4 SVK 1,42 SVK 1,23 EST 1,52 SVK 1,52 4 SVK 2,26

SVK 1,2 5 POL 1,47 BGR 1,69 4 BGR 1,89 5 AUT 1,78 POL 2,30

ITA 1,26 ITA 1,5 ITA 1,78 ITA 1,95 ITA 1,97 5 ITA 2,34 5

(*) Country indicated by the three digits standard. Austria (AUT); Belgium (BEL); Bulgaria (BGR); Czech Republic (CZE); Germany (DEU); Denmark (DNK); Estonia (EST); Finland (FIN); France
(FRA); United Kingdom (GBR); Hungary (HUN); Ireland (IRL); Iceland (ISL); Italy (ITA); the Netherlands (NLD); Norway (NOR); Poland (POL); Slovakia (SVK). Mediterranean Foods as by
original Mediterranean Adequacy Index: vegetables, fruit, cereals, red wine, vegetable oils, potatoes and fish. Non Mediterranean Foods as by original Mediterranean Adequacy Index: meat, dairy products,
animal fat, eggs and sugar.
355
356 C. Finardi, L. Bucchini, A. Turrini

which within Trichopoulou’s highly relevant results Table 7. The MDS with potatoes and eggs regarded as Mediter-
ranean entries or not (original model of Trichopoulou)- EFSA
showed an increase in the mortality rate of 1.07 (0.95-
mean data considered, grams.
1.21 and 0.98-1.17 respectively). For comparative pur-
MDS (pot + eggs MDS
poses, in this study there was an increased hazard of as MED) Original
1.05 for saturated fats, 1.06 for meat and 1.11 for dairy
GBR 2 NLD 1
products. Results suggest that when eggs and potatoes
NOR 2 NOR 2
were considered part of the Mediterranean diet entries,
as expected they gave rise to a higher apparent adher- NLD 2 GBR 2

ence to the Mediterranean diet, with a number of coun- IRL 2 IRL 2

tries benefitting from same (including Italy, Austria and EST 3 ISL 2
Estonia). This loophole, due to the inability of the tra- HUN 3 FIN 2
ditional Med Diet to reflect on either new foods nor on
FIN 3 EST 2
foods traditionally outside the Med pattern, requires for
BEL 3 BEL 3
sure additional research and modelling (Table 7).
Also, some considerations can be drawn about DNK 4 SWE 3

the different data sources used. Previous international SWE 4 CZE 3

comparisons (38, 395-“Dafne”) relied exclusively on ISL 4 DNK 3
other datasets, such as the FAO Food Balance Sheets, BGR 5 HUN 3
a food supply database, or Dafne, a household food
SVK 5 SVK 3
availability database. Both of these datasets provided
FRA 5 BGR 3
data obtained via the food balance method (FBS) or
from household food purchases (Dafne) whereas on DEU 5 POL 4

the contrary the EFSA dataset collected data from POL 5 FRA 4
national food consumption surveys. While the known CZE 5 AUT 4
limitations of the EFSA dataset include its lack of har- AUT 6 DEU 5
monisation in collection and survey methods, the dif- ITA 7 ITA 6
ferent timeframes of the national surveys and under-
(*) Country indicated by the three digits standard. Austria (AUT); Belgium
reporting, it may still provide more accurate informa- (BEL); Bulgaria (BGR); Czech Republic (CZE); Germany (DEU); Den-
tion than previously used datasets. As for the P-MAI, mark (DNK); Estonia (EST); Finland (FIN); France (FRA); United King-
the apparent discrepancies between the CONCISE dom (GBR); Hungary (HUN); Ireland (IRL); Iceland (ISL); Italy (ITA);
the Netherlands (NLD); Norway (NOR); Poland (POL); Slovakia (SVK).
and FBS average values could be due to waste along
the food consumption chain (from distribution to con-
sumption). In fact, FBS are corrected for food reused
for other production purposes, but not for retail/kitch- to damage health. Moreover, starchy foods (refined,
en waste or table leftovers. as opposed to whole foods) contribute to a higher P-
For future research, it would be worth exploring MAI. However, if they are not consumed as a sub-
whether Mediterranean diet adherence depends on stitute animal protein, they do not lead to a healthier
socio-economic or cultural factors, i.e. the degree of diet, which can lead to an increased BMI (9, 41).
inequality in income distribution inside a given coun- In this study, we estimated the P-MAI with weight
try (using the Gini Index; (40)). We also used wine and or energy content, depending on the available data.
wine substitutes as well as beer as a proxy of moderate Where data was missing, it was recommended to use
alcohol consumption (an element of the Mediterrane- the number of grams consumed daily (25, 9). In coun-
an diet). This may be misleading, as alcohol is known tries where EFSA data was available, the P-MAI (es-
timated in 1961 and 2007) demonstrated a decline in
all southern European countries surveyed, confirming
5
Dafne stands for DAta Food NEtworking, and aims at the creation
of a pan-European food data bank changes in the consumption of the Mediterranean diet.
“Mediterranean Diet ‘reflections’” Estimating adherence to the Mediterranean diet through secondary data 357

According to FBS data, the reason for this lies Table 8. P-MAI (Population-level Mediterranean Adequacy
Index) scores for the aggregated average national diets from
in increased meat consumption, rather than in the
FAO FBS in 2007 and 1961, and change in selected EU Mem-
reduced consumption of fruit and vegetables or veg- ber States (calories-based computation).
etable oils. Reductions have been observed in the 1961 2007 Change Change %
consumption of beans and wine only. Furthermore,
Norway 0.7 1.36 0.66 94.7
southern European diets have also increased their total
Great Britain 0.78 1.36 0.58 74.3
food and energy intake, which is not reflected in the
P-MAI. Several countries in northern and central Eu- Belgium 1.56 1.37 -0.19 -12.2

rope had higher levels of Mediterranean diets in 2007 Ireland 0.89 1.43 0.54 60.4
than in 1961, as measured by the P‑MAI. However, Iceland 0.64 0.82 0.18 27.5
this change coincided with a general increase in energy Finland 0.88 1.04 0.16 18.3
intake (kcal). More generally, most national diets tend
Sweden 0.84 1.07 0.23 27.6
to display some sort of ‘inertia’ in terms of the P-MAI,
Czech Republic 1.29 1.48 0.19 14.4
which can easily be attributed to dietary cultures and
recipes. Denmark 0.87 0.77 -0.10 -11.8

When measured with the FBS, more food in terms The Netherlands 1.17 1.11 -0.06 -5.3
of both weight and energy was consumed in 2007 than Poland 1.66 1.45 -0.21 -12.5
in 1961 in all countries. National diets, which were be- Germany 1.13 1.11 -0.02 -1.3
low average P-MAI in 1961, were generally still below
Hungary 1.39 1.20 -0.19 -13.8
average in 2007. The 2007 P-MAI scores are corre-
France 1.41 1.25 -0.16 -11.1
lated with those in 1961 (Pearson’s r = 0.84), Table 8.
For comparative purposes for each food catego- Austria 1.43 1.26 -0.17 -12.1

ry in the EFSA database, Confidence Intervals (CI) Slovakia 2.07 1.44 -0.63 -30.2
have been derived in order to assess whether related Italy 3.28 1.96 -1.32 -40.3
FAO data fall inside them. The results obtained were Portugal 3.62 1.59 -2.03 -56.1
informative and lead to the following consideration. Bulgaria 4.08 1.77 -2.31 -56.5
Categories that were too broad in scope did not allow
Spain 4.04 1.79 -2.25 -55.7
values to overlap in the 2 datasets (i.e. values inside the
Greece 4.63 2.04 -2.59 -55.9
confidence intervals), even if there appeared to be a
strong correlation (i.e. countries with a high consump- Mean 1.83 1.37
tion of dairy products were the same across databases, St.Dev 1.27 0.34
although the intake varied significantly depending on Pearson correlation 0.75
the source used, which in turn may depend upon a dif- Pearson correlation 0.07
ferent level of aggregation of single food items). (calories/PMAI 2007)
Other categories, even if they were more restrict-
ed, showed that despite the same direction being seen record at home, could easily be explained by the fact
in the CONCISE and FBS (expressed by the correla- that wine is not processed at home meaning that waste
tion); the magnitude was quite different to the well- can be reduced. Hence, wine intake and availability are
known variation expected between FBS and real con- similar, unlike other food categories for which waste
sumption (fish and starchy products). Hence, it gener- within the home environment is to be expected.
ally did not allow FBS values to fall inside the confi- Such comparisons may also drive reflections on
dence intervals of the CONCISE DB. As for wine, the impact that adult consumption levels have on the
there was a degree of correlation and, in some national overall population. Although, as confidence intervals
cases, an overlapping between FBS and survey data is from the CONCISE data in most cases did not in-
apparent. This similarity between data from the CON- clude corresponding FBS values, it could be deduced
CISE and FBS, which on the contrary food availability that this is not in fact the case.
358 C. Finardi, L. Bucchini, A. Turrini

Table 9. Correlation and overlapping between FAO and EFSA DB (grams) for specific food categories
Food Category Pearson’s correlation Countries presenting overlapping
r CONCISE - FBS (g) Confidence Intervals (i.e., comparable values)
Fish 0.59 Austria (P <0.01)
Cereals 0.25 France (P <0.01 and P< 0.05), Germany (P< 0.01)
Fruit -0.3 none
Vegetables 0.34 none
Dairy products 0.72 none
Meat 0.08 Slovakia (P <0.01 and P< 0.05)
Sugars and sweeteners 0.41 none
Starchy products 0.62 Austria (P <0.01 and P< 0.05)
Wine products 0.81 Austria (P <0.01 and P <0.05), Czech Rep. (P <0.01 and P <0.05),
Slovakia (P <0.01 and P <0.05), Sweden (P <0.01).
Eggs 0.22 Poland P < (0.01)

However, we believe that the P-MAI, for all its during diverse temporal windows- as well as the dif-
simplicity, may be worth exploring as an initial sum- ferent survey methodologies (i.e., 24 hours recall, 48
mary population-level measure of the level. Such an hours recall, etc)
index is currently available in several forms (42, 9, 41). Also, with regard to the FAO FBS, they are not
Another outcome of this study was the compari- deflated for domestic food waste (i.e., apparent con-
son between different databases, which can provide sumptions may be exaggerated).
problematic results from a public health perspective. In addition, the use of dietary indices, which, in
When political and administrative resources are lim- this case, rely on average or median data do not account
ited and there is a need to address emerging issues for the variability in the population’s dietary habits.
(such as diet) at population level, the use of different Furthermore, many indices used relying on popula-
indicators and databases may result in different policy tion-level data discount the lack of all the information:
indications. Although we were already aware (43) that i.e, for the Med Diet Score the ratio between monoun-
variations in percentiles of national populations exist, saturated/saturated fats, as originally indicated in Tri-
whereby there are higher variations between countries. chopoulou (2003) (28) was absent, nor was it possible to
This contribution once again underlines the difficul- analytically separate legumes from vegetables departing
ties encountered when managing the need for coun- from the CONCISE database; neither the Med- DQI
try-specific policies and the need for specific relief for was able to capture TFAs or cholesterol.
specified target groups (percentiles) of the populations, Another major limitation of this study is that the
at the time of adopting aggregated data. indices used were originally built on first-hand data (i.e.,
cohort studies relying on real persons), and not intended
for aggregated mean-median values analysis at popula-
Limitations tion level (secondary data). Eventually, a further problem
of Med-adherence indices is that it does not consider the
A first limitation is that some foods were not in- overall caloric intake and also, specific nutrients or other
cluded in the CONCISE database- for example, veg- indicators of nutritional status such as BMI. Further con-
etable oils and red wine-distorting the overall value of siderations reflect the uncertain status inside the Medi-
the indices. terranean diet of food items, i.e. soft drinks, potatoes or
Another limitation depends on the different years beer/ethanol. This will likely also present a challenge in
of surveys in the CONCISE, which make inherently terms of the scientific background and overall conceptual
difficult to compare Mediterranean Diet adherence framework.
“Mediterranean Diet ‘reflections’” Estimating adherence to the Mediterranean diet through secondary data 359

Nonetheless such indicators fitted the clustering pur- 2.  Willett W, C Sacks, F, Trichopoulou A, Drescher G, Ferro-
poses only and obviously, the paper did not intend to infer Luzzi A, Helsing E, Trichopoulos D. Mediterranean diet pyra-
mid: a cultural model for healthy eating. Am J Clin Nutr 1995;
health-based outcomes, such as hazard ratios presented in 61:1402–1406.
the original works. 3.  Tognon G, Moreno L A, Mouratidou, T, Veidebaum T, Mol-
In conclusion, it is apparent that there is something nár D, Russo P, Siani A, Akhandaf Y, Krogh V, Tornaritis M,
like a real “Mediterranean Diet” and Mediterranean di- Börnhorst C, Hebestreit A, Pigeot I and Lissner L. Adherence
to a Mediterranean-like dietary pattern in children from eight
etary lifestyle. However, internationalisation of food con-
European countries. The IDEFICS study. International Journal
sumption and trade, global lifestyles and attention to pre- of Obesity 2014; 38:108–114.
ventive medicine-(starting from healthier diets)- result in 4.  Alexandratos N. The Mediterranean Diet in a World Context.
an increase in Med diet patterns in Nordic EU countries Public Health Nutrition 2006; 9(1A): 111–117.
and a relative loss in traditional Southern ones. 5.  Estruch R, Ros E, Salas-Salvadó J, Covas M I, Corella D,
Arós F, Gómez-Gracia E, Ruiz-Gutiérrez V, Fiol M, Lapetra
In the end, different dietary patterns seem to con- J, Lamuela-Raventos RM, Serra-Majem L, Pintó X, Basora
tinue to exist across Europe, which is consistent with the J, Muñoz M A, Sorlí J V, Martínez, J A, Martínez-González
findings of other studies. Future refinements require a bet- M A. Primary Prevention of Cardiovascular Disease with a
ter definition of the Mediterranean diet (42). Moreover, Mediterranean Diet. N Engl J Med 2013; 368:1279-1290.
6.  Martinez-Gonzalez M A, Bes-Rastrollo M, Serra-Majem
proper consideration should be paid to food categories
L, Lairon D, Estruch R, Trihopoulou A. Mediterranean food
with still uncertain Mediterranean-taxonomy (i.e. pota- pattern and the primary prevention of chronic disease: recent
toes, beer or fruit juices with a low fruit content). Equally, developments. Nutrition Rev. 2009; 67 -1: S111-6.
a future refinement could intend to model FAO FBS 7.  Fidanza F, Alberti A, Lanti M, Menotti A. Mediterranean
Adequacy Index: correlation with 25-year mortality from coro-
data in order to mimic real survey data, should there be
nary heart disease in the Seven Countries Study. Nutrition,
a lack of such data. Consequently, correction factors for Metabolism & Cardiovascular Diseases 2004; 14(5):254-258.
Mediterranean countries could be extrapolated (i.e., food 8.  Trichopoulou A, Bamia C, Trichopoulos D. Anatomy of health
waste- in order to better describe real at home intake of F&V, effects of Mediterranean diet: Greek EPIC prospective cohort
for which the difference between intake surveys data and FBS study. BMJ 2009; 338:2337.
9.  Alberti A, Fruttini D, Fidanza F. Mediterranean Adequacy
seems relevant). Index: Further Confirming Results of Validity. Nutrition, Me-
The P-MAI is an interesting indicator due to its tabolism and Cardiovascular Diseases 2009; 19(1): 61-66.
user-friendliness, lack of a-aprioristic assumptions or 10.  Knoops K T, Groot de L C, Fidanza F , Alberti-Fidanza A
modelling with a minimal recourse to hypothesis - and , Kromhout D, van Staveren W A. Comparison of three dif-
ferent dietary scores in relation to 10-year mortality in elderly
allows the classification of European countries’ diets us-
European subjects: the HALE project. Eur J Clin Nutr. 2006;
ing food intake data. In fact, exploring variations with a 60(6):746-55.
one-way ANOVA and considering the 3 comparable data 11.  Elmadfa I, Weichselbaum E. European Nutrition and Health
sources used (FAO FBS calories, FAO FBS grams, and Report 2004. Annals of Nutrition and Metabolis 2004; 48(2).
EFSA mean values in grams), we cannot reject the null 12.  Elmadfa I, Meyer A L, Nowak V. European Nutrition and
Health Report 2009”. Annals of Nutrition and Metabolism
hypothesis that mean values are not the same. Therefore, 2209;55(2)- available at: https://www.univie.ac.at/enhr/.
using different data can lead to not completely different 13.  Brussaard J H, Johansson L & Kearney J. Rationale and Meth-
results (F 4.73, F critical 3,16). The residual variance/over- ods of the EFCOSUM Project. European Journal of Clinical
all variance ratio, is in fact 0.88 (values closer to 1 imply Nutrition 2002; 56(2): S4-S7.
14.  EFSA. Food Consumption Concise Database. 2008, available
that -regardless the dataset used- the results can still be at: http://www.efsa.europa.eu/EFSA/ScientificPanels/DA-
compared, suggesting that different dietary patterns count TEX/efsa_locale-1178620753812_ConciseEuropeanCon-
more than diverging collections methods/dataset). sumptionDatabase.htm
15.  EFSA. Guidance Document for the Use of the Concise Euro-
pean Food Consumption Database in Exposure Assessment.
2008
References 16. 
EFSA. Use of the EFSA Comprehensive European Food
Consumption Database in Exposure Assessment. EFSA Jour-
1.  Keys A. Mediterranean diet and public health: personal reflec- nal 2011; 9(3):2097, available at: http://www.efsa.europa.eu/
tions. Am J Clin Nutr 1995; 61(6): 1321-1323. en/efsajournal/doc/2097.pdf.
360 C. Finardi, L. Bucchini, A. Turrini

17.  EFSA. General Principles for the Collection of National Food 33.  WHO. Food based dietary guidelines in the WHO European
Consumption Data in the View of a Pan-European Dietary Region”. Document E7983 2003. Copenhagen, WHO Re-
Survey., EFSA Journal 2009; 7(12):1435, available at: http:// gional Office for Europe.
www.efsa.europa.eu/en/efsajournal/pub/1435.htm. 34.  Waxman A. Why a global strategy on diet, physical activity and
18.  Schmidhuber J, Traill W B. The changing structure of diets in health? In Nutrition and Fitness: Mental Health, Aging, and
the European Union in relation to healthy eating guidelines. the Implementation of a Healthy Diet and Physical Activity
Public Health Nutrition 2006; 9(5):584–595. Lifestyle 2005; 95:162-166. Karger Publishers.
19.  Schmidhuber J. The EU Diet – Evolution, Evaluation and Im- 35.  European Commission, COM (2015) 619 final-Report from
pacts of the CAP”. Montreal Conference acts. 2008 the Commission to the European Parliament and the Council
20.  De Boer J. Protein Consumption and Sustainability: Diet Di- regarding trans fats in foods and in the overall diet of the Union
versity in EU-15. Ecological Economics 2006; 59(3):267-274. population.
21.  EFSA. Guidance on the EU Menu methodology. EFSA Jour- 36.  EFSA Scientific Committee. Statement on the benefits of fish/
nal 2014; 12(12):3944, available at: http://www.efsa.europa.eu/ seafood consumption compared to the risks of methylmercury
de/efsajournal/doc/3944.pdf in fish/seafood. EFSA Journal 2015; 13(1):3982.
22.  EFSA. Scientific Opinion on establishing Food-Based Dietary 37.  Finardi C, Arfini F, Turrini A. Dietary changes over time in
Guidelines. EFSA Journal 2010; 8(3):1460, available at: http:// Europe, between Cyclops and Phaeacians. The role of supply
www.efsa.europa.eu/it/search/doc/1460.pdf factors in driving (changes in) food demand: a first, empirical
23.  De Lorenzo A, Di Renzo L. Working Paper Nutrire per pre- insight, presented at the EAAE 115 Seminar “The Econom-
venire: quali nuovi indicatori di rischio nutrizionale? INEA, ics of Food, Food Choice and Health”, 15th-17th September
Rome 2007. 2010, ACTA, available at: http://ageconsearch.umn.edu/han-
24.  Sofi F, Cesari F, Abbate R, Gensini G F, Casini A. Adherence dle/116443.
to Mediterranean diet and health status: meta-analysis” BMJ 38. 
WHO. GEMS/Food Consumption Cluster Diets, 2003.
2008; 337:a1344. Available at: http://www.who.int/foodsafety/publications/
25.  Alberti, A. and Fidanza F. Mediterranean Adequacy Index of chem/regional_diets/en/
Italian diets. Public Health Nutrition 2004; 7(7): 937-41. 39.  Trichopoulou A, DAFNE contributors. The DAFNE data-
26. 
Scott D W. On optimal and data-based histograms”. Bi- bank as a simple tool for nutrition policy. DAta Food Network-
ometrika 1979; 66(3): 605–610- available at: doi:10.1093/bi- ing”. Public Health Nutr. 2001;4(5B):1187-9.
omet/66.3.605. 40.  Gini C. Measurement of Inequality and Incomes. The Eco-
27.  Gerber M. Qualitative methods to evaluate Mediterranean diet nomic Journal 1921; 31:124-126.
in adults. Public Health Nutrition 2006; 9 (1A):147-51. 41.  Silva R, Bach-Faig A, Raidó Quintana B , Buckland G, Vaz de
28.  Rosenbloom J I, Kaluski D N, Berry E M. A global nutri- Almeida M D, Serra-Majem L. Worldwide variation of adher-
tional index. Food Nutr Bull. 208; 29(4):266-77. ence to the Mediterranean diet, in 1961-1965 and 2000-2003”,
29.  Trichopoulou A, Costacou T, Bamia C, Trichopoulos D. Ad- PHN 2009;12(9A):1676-1684.
herence to a Mediterranean Diet and Survival in a Greek Popu- 42.  Bach A, Serra-Majem L, Carrasco J, Roman B, Ngo J, Ber-
lation. The New England Journal of Medicine 2003;.348:2599- tomeu I, Obrador B. The use of indices evaluating the adher-
608. ence to the Mediterranean diet in epidemiological studies: a
30.  Sanchez-Villegas A, Martinez J A, De Irala J, Martinez-Gon- review. Public Health Nutrition 2006;9(1):132-146.
zalez M A.. Determinants of the adherence to an “a priori” de- 43.  Scali J, Richard A, Gerber M. Diet profiles in a population
fined Mediterranean dietary pattern. Eur J Nutr 2002; 41:249– sample from Mediterranean southern France. Public Health
57. Nutr. 2001;4:173–82.
31.  Rumawas M E, Dwyer J T, McKeown N M, Meigs J B, Rogers
G, Jacques P F. The development of the Mediterranean-style
dietary pattern score and its application to the American diet in
the Framingham Offspring Cohort. J Nutr. 2009; 139:1150–6.
32.  Panagiotakos D B, Pitsavos C, Stefanadis C. Dietary patterns: Correspondence:
a Mediterranean diet score and its relation to clinical and bio- Dr. Corrado Finardi
logical markers of cardiovascular disease risk. Nutr Mtab Car- Confederazione Nazionale Coldiretti
diovasc Dis 2006; 16:559–68. E-mail: corrado.finardi@hotmail.it