**3. Results**

In Table 1 the component loadings and the eigenvalues from the factor analysis are presented. The size of eigenvalues strongly suggests the formation of three distinct dietary patterns, explaining 51.78% of the total variance of the information. The first dietary pattern could be characterized as the "lacto-fish-vegetarian", the second as the "meat-eaters" and the third as the "saturated fat and added sugars".


**Table 1.** Factor loadings of the food components included in the extracted dietary patterns.

Abbreviations: LFV, lacto-fish-vegetarian; and, SFAS, saturated fat and added sugars. In each factor only the components with values > 0.30 are included in the table.

The sociodemographic, lifestyle, and clinical characteristics of participants are presented by anxiety status in Table 2. Compared to the participants without anxiety, those with anxiety were more likely to be sedentary (*p* = 0.001), smokers (*p* = 0.027), and su ffered depressive symptoms (*p* < 0.001). The rest of measured variables (i.e., age, gender, education, income, family status, diet, energy intake, BMI, hypertension, diabetes, hypercholesterolemia, and cognitive distortion) were not significantly associated with the level of anxiety.


Abbreviations: MedDiet, mediterranean diet; BMI, body mass index; LFV, lacto-fish-vegetarian; SFAS, saturated fat and added Sugars; diet anti-inflammatory index (D-AII).

As a first step, results of multi-adjusted analysis assessing the energy intake by tertiles on anxiety levels are presented in Table 3, without adjusting for dietary patterns. Initially the association between energy intake (kcals/day) as a continuous variable and anxiety levels was tested. Based on this analysis, it was observed that the higher levels of energy intake were positively related with higher anxiety levels (non-standardized b (95% CI): 0.01 (0.003 to 0.2)), after various adjustments (i.e., sex, smoking habits, physical activity, etc.) (data shown only in text). When the analysis was applied by energy intake tertiles, it was shown that the 1st energy intake tertile had an independent inverse association with anxiety levels as compared with the highest one (3rd tertile) (non-standardized b (95% CI): −11.65 (−22.83 to −0.48), *p* = 0.04). The 2nd energy intake tertile as compared with the 3rd tertile, was no related with anxiety levels (non-standardized b (95% CI): −7.61 (−18.55 to 3.34), *p* = 0.16). In addition, female gender (non-standardized b (95% CI): 11.96 (0.53 to 23.38), *p* = 0.04), family status (non-standardized b (95% CI): 18.17 (4.56 to 31.77), *p* = 0.012), and depression (non-standardized b (95% CI): 1.07 (0.27 to 1.87), *p* = 0.01) were positively related to higher anxiety levels. Cognitive distortion (non-standardized b (95% CI): 0.20 (0.09 to 0.32), *p* = 0.001) was also observed to be associated with the presence of anxiety symptoms, while physical activity was inversely related with anxiety levels (non-standardized b (95% CI): −8.58 (−15.90 to −1.25), *p* = 0.02). Finally, the results remained similar when basic metabolic rate was inserted as a confounding variable in the model.


**Table 3.** Correlates of anxiety among adults aged ≥ 50 years estimated by multivariable linear regression, in the ATTICA study, *n* = 758.

Data are presented as non-standardized coefficients (b) and their 95% confidence intervals (CIs). \* *p* < 0.05, \*\* *p* < 0.01. Abbreviations: BMI, body mass index. Model is adjusted for all the covariates in the table.

As a next step, Table 4 illustrates the results from multiple linear regression analysis that evaluated the association between dietary patterns, energy intake and anxiety (models I and II). After adjusting for sociodemographic, lifestyle and clinical characteristics, as in Table 3, plus for confounding due to obesity, central obesity (waist circumference), and energy intake, the dietary pattern characterized by the consumption of saturated fats and added sugars (SFAS dietary pattern) was consistently associated with higher anxiety levels (non-standardized b (95% CI): 5.82 (0.03 to 11.61), *p* = 0.04) (Model II). Moreover, anxiety was positively associated with female gender (non-standardized b (95% CI): 21.06 (3.19 to 38.94), *p* = 0.02), and family status (non-standardized b (95% CI): 19.83 (2.47 to 37.19), *p* = 0.02). In addition, depressive symptomatology was related with higher level of anxiety (non-standardized b (95% CI): 1.88 (0.48 to 3.28), *p* = 0.01). Finally, LFV dietary pattern as well as meat-eaters and SFAS dietary patterns were replaced with the D-AII, a dietary index that is picturing pro- and antiinflammatory dietary habits. In this additional analysis, there was not association between the D-AII and the anxiety levels (*p* = 0.94) (data shown only in text).

**Table 4.** Correlates of anxiety among adults aged ≥ 50 years estimated by additive multivariable linear regression (Model I and Model II), in the ATTICA study, *n* = 758.



**Table 4.** *Cont.*

Data are presented as non-standardized coe fficients (b) and their 95% confidence intervals (CIs). Abbreviations: LFV, lacto-fish-vegetarian; SFAS, saturated fat and added sugars. Model I and Model II are adjusted for all the respective covariates in the table.
