Food Labels Use Is Associated with Higher Adherence to Mediterranean Diet: Results from the Moli-Sani Study

Mediterranean diet (MD) has been associated with lower risk of ischemic cerebro- and cardio-vascular disease, neurological degenerative disease, and breast and colonrectal cancers. Nevertheless, adherence to this pattern has decreased. Food labels are a potentially valid means to encourage towards healthier dietary behavior. This study, conducted on a subsample of 883 subjects enrolled in the Moli-sani Project, evaluated whether food labels reading (LR) is associated with MD adherence. Participants completed a questionnaire on nutrition knowledge, information, and attitudes, with a specific question on food labels reading. Biometric measurements, socio-economic status, education, physical activity, and smoking habits were collected. The European Prospective Investigation into Cancer and Nutrition (EPIC) food frequency questionnaire was used to collect dietary habits, and subsequently evaluated by both the Mediterranean diet score (MDS) and Italian Mediterranean index (IMI), a priori dietary patterns. Food consumption patterns were generated by Principal Components Analysis (PCA), an a posteriori approach. Multivariable odds ratios were calculated to quantify the association of LR categories with dietary habits. LR was significantly associated with greater adherence to both MDS (p = 0.0004) and IMI (p = 0.0019) in a multivariable model. LR participants had 74% (MDS) or 68% (IMI) higher probability to be in the highest level of adherence to Mediterranean diet-like patterns. Moreover, they showed greater adherence to Mediterranean-like food consumption patterns (0.1 vs. −0.2, p < 0.0001) and lower adherence to two Western-like patterns (0.01 vs. 0.2, p = 0.009 and 0.1 vs. 0.2, p = 0.02). These findings support an association between food label use and consuming a Mediterranean-type diet.


Introduction
Mediterranean diet (MD) has been associated with lower incidence of chronic degenerative diseases such as cardiovascular disease, stroke, neurodegenerative diseases, and some forms of cancer [1][2][3][4]. Moreover, evidence of a protective effect of MD on cardiovascular and neurovascular diseases has been provided by recent intervention studies [5]. In the wake of these data, the Mediterranean diet has been proposed as an optimal strategy of public health intervention [6].
However, in spite of the growing scientific and cultural [7] support, MD is being abandoned in the areas from which it originated, such as in Southern European and Northern African countries [8,9]. The adherence to this dietary pattern is, particularly, decreasing among young people in countries with strong Mediterranean traditions [10,11], foreseeing a deteriorating situation in future adults, probably also favored by economic factors. Indeed, low income is associated with lower adherence to MD-like eating patterns [12]. In this context, examination of approaches to improve MD adherence, especially among those population historically linked to Mediterranean tradition, is warranted.
Nutrition information on food labels, that have been standardized in the USA by the 1990 Nutrition Labeling and Education Act [13], and made mandatory by the European Commission in December 2011 [14], is regarded as a potentially valid mean to encourage people towards generally healthier dietary behaviors [15]. Research show that label use is effectively associated with the consumption of healthier nutrients [16,17] and, in particular, it can have a positive effect on choosing beneficial nutrient components and a negative effect on harmful ones [18].
To our knowledge, no study addressed the impact that food labels could have on the adherence to complex dietary patterns. Therefore, the aim of the present study was to evaluate the association between labels reading and adherence to the Mediterranean diet in an adult population living in a central-southern region of Italy, with strong MD traditions.

Study Population
The study includes a subsample of subjects enrolled in the Moli-sani Project, a population-based cohort study on citizens resident in the Molise region, placed between Central and Southern Italy. The Moli-sani study started in 2005 and randomly recruited men and women aged ≥35 years from subjects included in the city-hall registries of Molise [19]. Exclusion criteria were pregnancy, disturbances in understanding/willing processes, ongoing poly-traumas or coma, refusal to sign the informed consent; 30% of subjects refused to participate; these were generally older and had a higher prevalence of cardiovascular disease. At the end of recruitment, in April 2010, 24,325 subjects were enrolled. In 2009, we elaborated an additional self-administered questionnaire on nutrition knowledge and mass media exposure. This questionnaire was proposed to participants, recruited from May 2009 to April 2010 (n = 1571). The rate of participation to the nutrition knowledge and mass media questionnaire was 61%. After exclusion of 75 questionnaires with missing data for relevant questions, the final sample size was 883 subjects.
The Moli-sani study was approved by the Ethics Committee of the Catholic University of Rome. All participants signed the informed consent before taking part in the study.

Dietary Information
The Italian validated European Prospective Investigation into Cancer and Nutrition (EPIC) food frequency questionnaire (FFQ) was administered in-person to determine food intake [20,21]. The questionnaire, computerized with tailor-made software, allowed to interview participants in an interactive way, including illustrations of sample dishes of definite sizes or by reference to standard portion sizes. To simplify interpretation of data and to minimize within-person variations in intakes of individual foods, 188 food items were classified into 45 predefined food groups on the basis of similar nutrient characteristics or culinary usage.

Dietary Analysis
To obtain a broader picture of healthy habits, we evaluated the adherence to a Mediterranean diet using two scores. The first one was the Mediterranean diet score (MDS), elaborated by Trichopoulou et al. [22], whereas the Italian Mediterranean index (IMI) [23], was specifically elaborated to better capture healthy eating in Italy, including foods, such as pasta, more typically consumed in this country.
MDS scoring was based on the intake of the following nine items: vegetables, legumes, fruits and nuts, dairy products, cereals, meat and meat products, fish, alcohol, and the ratio of monounsaturated-saturated fats. For most items, consumption above the study median received 1 point; all other intakes received 0 points. For dairy products and meat, of which consumption is considered unhealthy, an intake below the median received 1 point. Medians are gender specific. For ethanol, men who consumed 10-50 g/day and women who consumed 5-25 g/day received 1 point; otherwise, the score was 0. The possible scores ranged between 0 and 9, the latter value reflecting the maximal adherence.
IMI scoring was based on the intake of 11 items: a high intake of six typical Mediterranean foods (pasta; typical Mediterranean vegetables such as raw tomatoes, leafy vegetables, onions, and garlic, salads, and fruiting vegetables; fruit; legumes; olive oil; and fish); a low intake of four non-Mediterranean foods (soft drinks, butter, red meat, and potatoes); and alcohol consumption. If the consumption of typical Mediterranean foods was in the third tertile of the distribution, the participant received 1 point; all other intakes received 0 points. If the consumption of non-Mediterranean foods was in the first tertile of the distribution, the participant received 1 point. Ethanol received 1 point for intake up to 12 g/day; abstainers and participants who consumed >12 g/day received 0 points. Possible scores ranged from 0 to 11. High adherence to MD, as stated by MDS, was defined when the score was >4 points, whereas low adherence when the score was <4 points. For IMI, respectively, high and low adherence were at >3 and <3 points. Subjects with intermediate values (MDS = 4 or IMI = 3) were excluded from this analysis to focus on the two extreme categories of adherence to MD.
The food consumption pattern was obtained using principal component analysis (PCA). The purpose of PCA is to derive a small number of components that can account for the variability found in a relatively large number of measures. This procedure, called data reduction, is typically performed when a researcher does not want to include all of the original measures in analyses, but still wants to work with the information that they contain. PCA was conducted on the correlation matrix of 45 food groups [24]. In determining the number of factors to retain in PCA we used the criterions of an eigenvalue >1.0, the scree test and the interpretability of the final solution [25]. Three main factors emerged with PCA, in agreement with previous findings in the same population [24]. The first pattern, named "Olive Oil and Vegetables", was characterized by high positive loadings of olive oil, vegetables, legumes, soups, fruits, and fish. The second pattern, named "Pasta and Meat", was characterized by high positive loadings of pasta, cooked tomatoes, red meat, animal fats, and alcoholic beverages, and negative loadings of breakfast cereals, and yogurt. The "Eggs and Sweets" pattern was characterized by high positive loadings of eggs, margarines, processed meat, sugar, and sweets. Food grouping used in the dietary pattern analyses are shown in Table 1.

Data Collection
Body mass index (BMI) was calculated as kg/m 2 . Lifestyles were assessed by a structured questionnaire including smoking habit and physical activity. Regarding the latter, 24 questions were asked on working time, leisure time, and sport participation. Physical activity was expressed as daily energy expenditure in metabolic equivalent task-hours (MET/day) [26].
Education level was considered as low (≤8 years) or high (>8 years). Socioeconomic status (SES) was expressed as a score based on five variables: dwelling ownership and ratio between the number of rooms and number of living-in family members (rooms per person), both currently and during childhood, and availability of hot water at home during childhood. The five components were dichotomized according to the median value, and a score of one was attributed to the category supposed to be marker of higher social status in comparison with the opposite category: thus we assigned a score of 1 to people living in a house with living-in family members/room density >0.6 or dwelling ownership or with availability of hot water, and a score 0 to people with living-in family members/room density ≤0.6, no dwelling ownership or with unavailability of hot water. We recorded missing values for SES for 5.7% of the subjects.

Food Labels Reading
The instrument for measuring food labels use was included in a larger questionnaire aimed at studying exposure to media information, trust in various health information sources, health, and dietary knowledge and attitude toward healthy behavior (the "Moli-info" questionnaire, developed by some of us (AB, MB, LI) and available upon request) [27,28]. A specific question was devoted to food labels reading ( Table 2). To establish validity, test-retest method was used by administering twice, with a two-weeks separation, the information, knowledge, and attitude questionnaire to a group (n = 28) of researchers at the Fondazione di Ricerca e Cura "Giovanni Paolo II" of Campobasso. Consecutive subjects (n = 1132) recruited in the Moli-sani study from May 2009 to April 2010 were asked to fill the self-administered questionnaire; participants reporting missing values were excluded.
People declaring to read any of the food labels' items were classified as readers while those answering "usually I do not read food labels" or "do not know" constituted the non-reader group.

Validation
Taken as a whole, the Moli-info questionnaire have been validated for a Nutrition knowledge score and a Mass media exposure score. Scores between test and retest groups were highly correlated (Sperman correlation coefficient r = 0.71, P < 0.0001 for knowledge and r = 0.84, P < 0.0001 for Mass media exposure). The consistency with which participants answered the single question regarding label reading was assessed using Cohen's Kappa, showing substantial consistency (Kappa = 0.63).

Statistical Analysis
Values for continuous variables are means ± Standard Deviation. In validation, scores between test and retest groups were evaluated by using Spearman correlation. Consistency with which participants answered the single question regarding label reading was assessed using Cohen's Kappa.
Multivariable analysis of variance for continuous or categorical variables was used for testing the associations of general characteristics or adherence to Mediterranean diet scores or other dietary pattern (considered as the dependent variables) with LR categories. The potential predictors tested for association with labels reading included socio-demographic variables (age, sex, income, smoking, educational level, and socioeconomic status), BMI, total caloric intake, smoking habit, and physical activity (expressed as MET/day). Basic model included age and sex, whereas fully adjusted model included age, sex, total caloric intake, socio-economic status, income, BMI, smoking, physical activity, and educational level. By using multivariable logistic regression analysis, odds ratio (ORs) with corresponding 95% confidence intervals (95% CI) were calculated to quantify the association of LR categories with food consumption patterns.
A specific category "non-respondent" was created for income and SES in order to consider the missing values in the multivariable analyses.
The data analysis was performed using SAS/STAT software [29].

Study Sample
The final study sample was 883 subjects. Subjects who were excluded from the analysis because of missing relevant data (n = 75) were comparable for sex distribution, but were older (52.5 ± 9.5 vs. 58.5 ± 11.3; p value < 0.0001), and had lower social status (6.40 ± 2.34 vs. 7.42 ± 2.35; p value < 0.0001) in comparison with subjects included. It is likely that the whole questionnaire was too articulated to be completed by older participants and those having lower social status because they might find it, largely addressing modern technologies, too far away from their daily experience, and refrained from completing it properly. Yet the 883 analyzed subjects were comparable with the whole Moli-sani Project population in terms of sex distribution, cardiovascular disease, and metabolic syndrome prevalence, BMI, cardiovascular risk and dietary patterns (all p > 0.05), whereas mean age was lower (52.5 ± 10 vs. 55.6 ± 11.9, p < 0.0001). Table 3 shows characteristics of the sub-sample of the Moli-sani project included in this study. 56.2% of participants reported to consult information in food labels. Both in univariate and in multivariable (adjusted for age and sex) analysis, the label reading attitude was significantly more diffused among women than men. People using food labels information were slightly younger, an observation significant in univariate analysis and borderline significant in multivariable. No significant associations were found for smoking habits, BMI, total caloric intake, physical activity, social status score, education level, and income.

Association between Label Reading and Dietary Patterns
People reading food labels information showed a higher adherence to Mediterranean diet according both to MDS and IMI (Table 4). This association was observed in the basic model (p = 0.0005 MDS, p = 0.002 IMI), as well as in the fully adjusted model (p = 0.0004 MDS, p = 0.0019 IMI). Regarding food consumption patterns, the label reading group showed a higher adherence to the pattern "Olive oil and vegetables" than non-readers, both in the sex/age adjusted model (p < 0.0001) and in the fully adjusted model (p < 0.0001). Readers group had a lower adherence to both patterns "Pasta and meat" and "Eggs and sweets" in multivariable model (respectively p = 0.009 and p = 0.02).
Such findings remains unchanged when we considered specific components of the food label. In particular, we did not find difference when only subjects declaring to pay attention to calories (n = 194) or fats (n = 227) or other items (n = 75) were compared to non-readers (data not shown).
MDS was then divided into two categories to get both a minimum (score from 0 to 3) and a maximum adherence (score > 4). The same was done with IMI score, obtaining a minimum from 0 to 2 and a maximum > 3.
In multivariable logistic regression model controlling for age, sex, total caloric intake, socio-economic status, income, BMI, smoking, physical activity, and educational level, odds of eating Mediterranean were higher in the label reader group than in non-readers. In fact, participants reading food labels had 73% higher probability to have greater adherence to MDS (OR 1.73, 95% CI: 1.24-2.42) ( Table 5), and 67% to IMI (OR 1.67, 95% CI: 1.2-2.33) ( Table 6).

Discussion
In our study, 56.2% of participants reported to consult information on food labels at least for one item. This percentage is similar to the one found in a USA study (61.5%) [16], albeit ample variability has been found by previous researches. In the UK, food label reading was found to be 27% [30], while in other studies conducted in Europe it ranged from 18% to 63% [31]. This discrepancy could be accounted by cultural and socio-economic difference among populations included in the studies. According to the prevalence of reading observed in our study, food labels may represent one of the means by which a typical Mediterranean population receive information that, arguably, can be part of the decision process about food consumption in the same way it has been observed in other geographical areas [15][16][17][18].
In this context, we found that food labels consulting is associated with greater adherence to Mediterranean diet according to two different scores, MDS [22] and IMI [23]. These findings contribute to provide evidence that food labels are associated not merely to single healthy-oriented item choices, as stated by other studies [15][16][17][18], but also to a whole, complex dietary pattern. This is corroborated by odds-ratio analysis showing higher probability for label readers to be in the highest level of MD adherence, again both for MDS and for IMI.
To further evaluate the association between label reading and dietary patterns, beside an "a priori" scales like MDS and IMI, we used PCA, which is able to empirically derive dietary patterns not set a priori. This analysis showed that in our population food labels reading is positively associated with a healthy food pattern, close to the Mediterranean diet ("Olive oil and vegetables") and negatively with other, less healthy, patterns ("pasta and meat", "eggs and sugar").
These observations suggest that people paying attention to food labels have a nutrition behavior, not only healthier, but closer to the Mediterranean model, albeit with an observed limited difference of just 0.3 points both for MDS and IMI. A limited amount of variance in nutrients intake explained by food labels reading have been reported also in other studies [16,32], but the association could still have sufficient magnitude to have potential impact [16].
In our findings the only socio-demographic characteristics having effect on food label reading are gender and, marginally, age, unlike other studies in which also education level and income do have an effect [16,18]. Social and cultural difference between Italian and other populations included in the studies could be a possible explanation for the discrepancy.

Limitations of the Study
The first limitation in this study is its cross-sectional nature. Nevertheless, this type of investigation is useful to set new hypotheses to be tested in future prospective studies. Secondly, caution is needed in extending the results presented here to larger contexts as data were collected in a region located between Central and Southern Italy, Mediterranean by tradition and culture [19]. However, the main characteristics of our sample are comparable to those of the Italian Cardiovascular Epidemiological Observatory [33], a large survey including random samples of the general population all over Italy; therefore our sample could be considered representative, at least, of the Italian population.
Another weak point is the possibility of over-reporting of food label reading, due to social desirability bias, as suggested by previous findings [34]. We think we partially overcame this issue with the specific question asking not if participants read labels in general, but inquiring about which item they consult more often.
Finally, we must consider that the observed association between labels reading and Mediterranean diet adherence does not prove that labels use by consumers has impact on dietary habits. This much-debated issue was not addressed by the present study. Changing decisions in buying or not a specific product after reading labels have been previously reported for one-third of consumers [35], but it must be considered that a recent study observed no modification toward healthy choices in supermarket sales after introduction of "traffic light" labels [36].

Conclusions
This study highlights, to our best knowledge for the first time, the association between food labeling and healthy nutrition choices, not just in terms of single food or single nutrients, but also in the wider range of a complete dietary pattern. We cannot exclude that this association may only reflect a healthy lifestyle in general, and the potential role of food labeling in encouraging healthy dietary habits cannot be concluded without further research. More evidence in this direction could lead to a new concept of labeling, aimed to present foods according to whether they belong to the Mediterranean diet, and to what extent. This could be used in the framework of already used, new informative messages (Front-of-Package systems), such as the "traffic light" system developed by the United Kingdom Food Standards Agency [37]. Our findings should address specific public health campaign to implement food label use and reading to maintain Mediterranean dietary habits and the linked health advantages.
LI critically reviewed the manuscript; they were also responsible of the whole Moli-sani project and obtained financial support for it. All authors were involved in the subsequent edits of the manuscript, and have read and approved the final version. The Moli-sani Project was partially supported by research Grants from Pfizer Foundation (Rome, Italy) and the Italian Ministry of University and Research (MIUR, Rome, Italy)-Programma Triennale di Ricerca, Decreto no. 1588.
Moli-sani research group would like to thank Vittorio Krogh and Sabina Sieri from Istituto Nazionale dei Tumori, Milan, Italy for their precious contribution to dietary questionnaire analysis and interpretation.