The traditional dietary pattern of Mediterranean populations offers a palatable choice of healthy eating. This diet is generally characterized, despite regional disparities, by high consumption of plant foods, moderate consumption of alcohol and low consumption of meat. There is consistent evidence for the health-promoting properties of the Mediterranean diet(Reference Serra-Majem, Roman and Estruch1). Individual components of this diet can partially explain the health-protecting effects(Reference Schroder2).
Analysis of quality of life is a valid measurement for self-perceived mental and physical health status, which is closely associated with CVD and all-cause mortality(Reference Kaplan, Goldberg and Everson3, Reference Wannamethee and Shaper4). It has been argued that the social and cultural aspects related to the Mediterranean diet may additionally provide health benefits(Reference Trichopoulou and Lagiou5). A favourable environment and social support may contribute to having a better self-perceived health-related quality of life (HRQL)(Reference Molarius, Berglund and Eriksson6). Furthermore, there is evidence for the association of self-perceived health status and diet. Poor diet quality, characterized through low consumption of particular healthy foods or nutrients, has been associated with low mental and physical health(Reference Silvers and Scott7–Reference Blank, Grimsley and Goyder10). Furthermore, evidence from dietary intervention studies indicates an improvement of quality of life through healthy diets(Reference Hislop, Bajdik and Balneaves11, Reference Plaisted, Lin and Ard12). However, to our knowledge it is unknown if healthy diet, such as the Mediterranean diet, affects self-perception of HRQL in the general population. The aim of the present study was to analyse the association between adherence to the Mediterranean diet and self-perceived mental and physical health, controlled for confounding effects of age, smoking, BMI, alcohol consumption, educational level, leisure-time physical activity and the presence of chronic conditions in a representative population of Spanish men and women.
Methods
Study participants
Data were obtained from two population-based cross-sectional surveys conducted in Gerona (Spain) in 2000 and 2005. The first survey included 3054 randomly selected free-living men and women between 25 and 74 years of age. The second survey included 6322 men and women between 35 and 88 years of age. Response rates for the two surveys were 71·0 and 71·5 %, respectively. Participants from both surveys aged 35–74 years (n 8195) were included in the present analysis. Participants with unrealistic dietary energy intake (n 420), defined as energy intake less than 3347 kJ/d (800 kcal/d) or more than 18 828 kJ/d (4500 kcal/d), were excluded from analysis. Furthermore, data on self-reported mental and physical health function were missing for 650 participants. Finally, 7145 men and women (87·2 %) were included in the analysis. The protocol was approved by an Ethics Committee (CEIC-IMAS, Barcelona, Spain) and the results of the examination were sent to all participants.
Quality of life
We chose one of the most widely used generic tests in health outcomes research to analyse self-reported HRQL. The SF-12 Health Survey was developed to provide a shorter alternative to the MOS SF-36 Health Survey(Reference Ware, Kosinski and Keller13) to measure overall physical and mental health outcomes. The SF-12 is a valid measurement tool for the exploration of self-perceived vitality, pain, mental health, physical, social, emotional and mental functioning, and general health; it has been validated for use in the Spanish population(Reference Gandek, Ware and Aaronson14). This test consists of twelve questions addressing eight items that are summarized in a physical health score (items: physical functioning, role – physical, bodily pain, general health), and a mental health score (items: vitality, social functioning, role – emotional, mental health). The questionnaire uses Likert scales to analyse the intensity or frequency of each response. The outcomes are standardized according to the population-based values, considering 50 (sd 10) as the mean of the population. In Spain, mean values of physical and mental health scores are 49·9 (sd 9·0) and 51·0 (sd 9·0), respectively(Reference Gandek, Ware and Aaronson14).
Dietary assessment
Food consumption and nutrient intake were measured by a validated FFQ(Reference Schroder, Covas and Marrugat15) administered by a trained interviewer. An optical readable FFQ asked for the usual intake over the past year. In a 165-item food list, including alcoholic and non-alcoholic beverages, the participants indicated their usual consumption and chose one of ten frequency categories, from never or less than once per month to six or more times per day. Instead of standard questions about portion size, this FFQ uses medium servings defined by natural (e.g. one orange, one standard slice) or households units (e.g. one cup, one spoon).
A Mediterranean diet score (MDS) based on the intake of cereals, vegetables, fruits, legumes, nuts, fish, dairy products, olive oil, meat and red wine was computed to calculate overall diet quality. Multiple food components were operationalized as a single dietary exposure variable, similar to the approach used by other authors(Reference Trichopoulou, Kouris-Blazos and Wahlqvist16–Reference Martinez-Gonzalez, Fernandez-Jarne and Serrano-Martinez18). The characteristics of the scoring system are based on the traditional food consumption of the Mediterranean region(Reference Trichopoulou, Kouris-Blazos and Wahlqvist16). With the exception of red wine, the MDS was calculated according to the tertile distribution of energy-adjusted(Reference Willett19) consumption of cereals, vegetables, fruits, legumes, nuts, fish, dairy products, olive oil, meat and red wine. The lowest tertile was coded as 1, medium as 2 and highest as 3 for cereals, fruits, vegetables, legumes, fish, olive oil and nuts. The highest tertile was coded as 1, medium as 2 and lowest as 3 for meat and dairy products. Moderate red wine consumption was included as a favourable component in the MDS. For this purpose we coded up to 20 g alcohol intake proceeding from red wine as 3. Exceeding this upper limit or reporting no red wine consumption was coded with 0. The distribution values of all dietary components were calculated. The resulting MDS ranged from 10 to 30.
Measurement of non-dietary variables
Information on demographic and socio-economic variables, medical history, diet and lifestyle factors including tobacco smoking and alcohol consumption was obtained through structured standard questionnaires administered by trained personnel.
Leisure-time physical activity was measured by the Minnesota leisure-time physical activity questionnaire, also administered by a trained interviewer. This questionnaire has been previously validated for Spanish men and women(Reference Elosua, Garcia and Aguilar20, Reference Elosua, Marrugat and Molina21).
Anthropometric data
Measurements were performed by a team of trained nurses and interviewers who used the same standard methods in the two surveys. A precision scale of easy calibration was used for weight measurement with participants in underwear. Body weight was rounded up to 200 g and height was measured to the nearest 0·5 cm. BMI was calculated as weight/height squared (kg/m2).
Statistical analyses
Differences in continuous variables were compared using the Student t test. The Mann–Whitney U test was used for non-parametric variables. Categorical variables were tested using the χ2 goodness of fit test. The general linear modelling procedures (PROC GLM version 8.0; SAS Institute, Cary, NC, USA) were used to estimate lifestyle, anthropometric, medical history and socioeconomic variables according to the tertile distribution of the SF-12 mental component score (MCS) and physical component score (PCS). Linear trend was tested through including the categorized variable (tertile distribution of the SF-12 MCS and PCS) as continuous in this model. The polynomial contrast was used to determine P for linear trend for continuous variables. Age-adjusted logistic regression was performed to calculate P for linear trend of dichotomous variables (educational level, current smokers, dyslipidaemia, diabetes and hypertension) according to tertile distribution of physical and mental health. Dichotomous variables were included in these models as the dependent variable and tertiles of PCS and MCS as continuous explanatory variables (PROC LOGISTIC procedure of SAS version 8.0; SAS Institute). Multiple linear regression models (PROREG version 8.0; SAS Institute) were fitted to determine the confounder-controlled association of the MDS with self-reported mental and physical health function. The data of these models followed a normal distribution. Differences were considered significant if P < 0·05.
Results
No significant differences were found for age (P = 0·163), leisure-time physical activity (0·460) and smoking status (P = 0·422) between subjects who furnished complete and valid information on the study variables and those who failed to do so. Participants who were excluded from analysis drank less alcohol (not selected participants 9·5 (sd 19·3) g/d v. 10·0 (sd 15·0) g/d selected participants, P = 0·001) and had lower education levels (not selected participants 35·5 % chose ‘more than primary school’ v. 47·3 % of selected participants, P < 0·001).
Men spent more time in leisure-time physical activity, had more education, consumed more alcohol, had higher BMI and showed higher scoring for self-reported physical and mental health than women (Table 1). Furthermore, the prevalence of smoking and medical history and/or drug treatment of hypertension, diabetes and dyslipidaemia was higher in men than women (Table 1).
MCS, mental component score; MET, metabolic equivalents; PCS, physical component score; SF-12, SF-12 Health Survey.
* Categorical variables are presented as relative frequencies.
† Higher than primary school.
The proportion of smokers decreased with higher scoring of mental health in both sexes (Table 2). Men and women showed an opposite linear trend of educational level according to scoring of self-reported mental health. The proportion of higher-educated men decreased across tertile distribution of self-reported mental health and increased in women. Alcohol consumption in men and leisure-time physical activity in women increased across tertile distribution of self-reported mental health. Women with higher scoring of self-reported mental health had a lower BMI and a lower prevalence of medical history and/or drug treatment of dyslipidaemia (Table 2).
LTPA, leisure-time physical activity; MET, metabolic equivalents.
* For details of subjects and procedures, see Methods.
† Higher than primary school.
Age, BMI and the prevalence of medical history and/or drug treatment of diabetes and hypertension decreased across tertile distribution of self-reported physical health in both sexes (Table 3). A direct association was observed between physical health and both leisure-time physical activity and the proportion of higher educational level in men and women. Higher scoring of self-reported physical health was associated with a lower prevalence of smoking and medical history in men but with higher alcohol consumption in women (Table 3).
LTPA, leisure-time physical activity; MET, metabolic equivalents.
* For details of subjects and procedures, see Methods.
† Higher than primary school.
Multiple linear regression models were fitted to analyse the association of the MDS with self-reported mental and physical health in men and women (Table 4). The age-adjusted models revealed a significant direct association of the MDS with self-reported mental and physical health in both sexes. An increase of 5 units of the MDS was directly associated with changes of 0·74 and 1·15 units in men and women, respectively, in the MCS after controlling for potential confounders. The age-adjusted direct association of the MDS with self-reported scoring of physical health remained stable after adjusting for several confounders in men but was attenuated in women. There were no significant interactions between MDS and BMI, leisure-time physical activity or smoking.
* Model 1 adjusted for age.
† Model 2 adjusted for age, leisure-time physical activity, BMI, educational level, smoking, alcohol consumption (excluded alcohol proceeding from red wine), and drug treatment and/or medical history of hypertension, diabetes and dyslipidaemia.
‡ Model 2 was additionally adjusted for menopause in women.
Discussion
To our knowledge this is the first study to take in a holistic approach in analysing the relationship between self-perception of HRQL and diet. In the present population, self-perceived mental and physical health was directly associated with adherence to the Mediterranean diet in men and women. This association held, with the exception of physical health function in women, after controlling for confounders.
Intervention studies lowering dietary fat have reported improvement in mood, hostility and psychological well-being and a decrease in depression and anxiety(Reference Halyburton, Brinkworth and Wilson22–Reference Weidner, Connor and Hollis26). Additionally, intervention with a high-carbohydrate diet and a Mediterranean diet positively affect mental health-related variables(Reference Halyburton, Brinkworth and Wilson22, Reference Wells, Read and Laugharne23). Long-term effects of a low-fat, high-carbohydrate diet on quality of life were studied in the Canadian Diet and Breast Cancer Prevention Trial(Reference Hislop, Bajdik and Balneaves11). Participants showed a high score for self-reported mental and physical health after a median follow-up of 6·8 years. However, a similarly high score was observed in the control group, which reporting approximately 10 % higher fat consumption. This result indicates that it may not be a single nutrient but rather the entire food pattern that affects quality of life. The effect of three different types of diets on quality of life were analysed in a sub-study of the Dietary Approaches to Stop Hypertension Trial(Reference Plaisted, Lin and Ard12). The control diet of this trial represented typical food intakes of the American population. The fruit and vegetable diet was rich in fruit and vegetables but otherwise similar to the control diet. The third diet, called the combination diet, resembled a Mediterranean diet in several ways. Although participants' perception of quality of life improved with all dietary interventions, the greatest improvement was reported with the combination diet. Earlier epidemiological studies focused on the association of particular foods, food groups or nutrients with self-reported mental and physical health(Reference Silvers and Scott7–Reference Blank, Grimsley and Goyder10). Higher fish, vegetable, fruit and dietary fibre consumption has been associated with better self-perceived health, particularly mental health(Reference Silvers and Scott7–Reference Blank, Grimsley and Goyder10). It is worth noting that high consumption of these dietary components is part of the characteristics of the Mediterranean dietary pattern. The analysis of the impact of a single food or nutrient on health outcomes has its limitations(Reference Willett19). Therefore, dietary patterns have gained considerable attention in the past two decades. Indeed, people do not consume nutrients or single foods but combinations of foods. Furthermore, dietary components may interact, making the search for associations between single dietary factors and health outcome more difficult.
The reported health effects of the Mediterranean diet have a solid biological foundation(Reference Serra-Majem, Roman and Estruch1, Reference Schroder2). In addition, this diet is embedded in a social network that may contribute to increased personal well-being. Although speculative, it is reasonable to assume that a higher adherence to the Mediterranean diet also reflects a higher adherence to the Mediterranean lifestyle, including eating at home, expending time cooking, sharing lunchtime with other people and going to the market to buy foods. Indeed, using lunchtime for social communication and regeneration within family structures may improve self-perceived quality of life. It is known that social and family support can promote health by providing persons with positive experiences, socially rewarding roles or improved ability to cope with stressful events(Reference Seeman27). On the other hand, a recent report from the Centers for Disease Control and Prevention suggested an association between perceived social support and HRQL, at least in older people(28).
Although HRQL is a measure of perceived health rather than a biological measure, self-related health status has been shown to be a powerful predictor of mortality at long term(Reference Wannamethee and Shaper4).
Previous explanations argued that perceived health reflects an individual's awareness of symptoms, diagnoses and performance decrements that are associated with mortality risk(Reference Kaplan, Goldberg and Everson3). In the same way, the British Heart Regional Study showed that men who reported poor health were more likely to report medication use, sickness absence, major illnesses, angina pectoris, breathlessness and other characteristics indicative of poorer health(Reference Wannamethee and Shaper4).
It is important to note that, with the exception of self-perceived physical function in women, the associations between HRQL and adherence to the Mediterranean diet held after adjusting for cardiovascular risk factors, leisure-time physical activity and educational status. The present findings are in line with previous studies showing a worse cardiovascular risk profile and low levels of physical activity among men and women reporting low HRQL scores(Reference Spertus, Jones and McDonell29, Reference Bize, Johnson and Plotnikoff30). It is of interest to note that people affected by dyslipidaemia, hypertension or diabetes did not score negatively in perceived mental health.
Levels of self-reported physical functioning tended to be higher in both men and women with more than basic education; mental health was better in women with a higher educational level. The effect of socioeconomic class on HRQL may be a consequence of job insecurity and psychosocial and physical working conditions in men and material well-being at home and amount of household labour in women(Reference Borrell, Muntaner and Benach31). On the other hand, greater health inequalities described previously in women in Spain could support the differences found in mental health outcomes in this group in the present study(Reference Regidor, Gutierrez-Fisach and Dominguez32).
The cross-sectional nature of the study design precludes drawing causal relationships. Furthermore, all dietary instruments such as FFQ measuring past food intake are vulnerable to random and systematic measurement errors. Finally, misreporting is an acknowledged source of measurement error in prospective or retrospective methods of dietary assessment using self-reported food intake records.
Since nutritional exposure has a paramount effect on health, it is not surprising that diet is a key public health concern. Although policy-makers have generated many public health efforts promoting various healthy diets for the general population, the impact of a healthy diet on self-perceived mental and physical health has not been analysed. The present study outcomes provide new evidence for the health benefits of the Mediterranean diet and should encourage public health policymakers to promote this dietary pattern.
In conclusion, self-perceived mental and physical health function was directly associated with adherence to the Mediterranean diet in both sexes. With the exception of physical health function in women, the magnitude and direction of these associations did not change, after adjusting for several confounders.
Acknowledgements
We appreciate the English revision made by Dr Elaine M. Lilly (Writers First Aid). The CIBEROBN and CIBERSP are initiatives from the Instituto de Salud Carlos III, Madrid, Spain. This research was supported by grant 2FD097-0297-CO2-01 from Fondo Europeo de Desarrollo Regional (FEDER) and by parts of grants from Spain's Ministerio de Sanidad y Consumo, Instituto de Salud Carlos III, Red HERACLES RD06/0009, and Fondo de Investigación Sanitaria ISCIII CP 03/00115. M.-A. M. led data analysis and writing of the manuscript. H. S. and M. F. participated in the discussion and interpretation of results, and writing of the manuscript. J. M. and M.-I. C. provided their expertise in data analysis and in interpretation and discussion of results. They made substantial suggestions on the manuscript. The authors declare no conflict of interest.