Association of Western and Traditional Diets With Depression and Anxiety in Women

The Geelong Osteoporosis Study is a large epidemiological study involving women selected from compulsory Australian Commonwealth electoral rolls for the Barwon Statistical Division. An age-stratified, randomly selected population-based sample of 1,494 women (ages 20–94 years) was recruited between 1994 and 1997, with participation of 77.1%. These women have continued to return for biennial follow-up assessments. Between 2004 and 2008, 881 of the original participants returned for a 10-year follow-up appointment, and the response rate among eligible women was 82.1%. An additional sample of women ages 20–29 years was recruited between 2004 and 2008, with a response rate of 70.9%. This allowed for continuing investigation in the full adult age range. Of the 1,127 women who participated in the 10-year follow-up, participants for whom psychiatric or dietary data were not available at the time of the study (N=81) were excluded from the analyses, resulting in a sample of 1,046 women ages 20–93 years available for analysis. The Barwon Health Human Research Ethics Committee approved the study, and written informed consent was obtained from all participants.

Habitual diet was assessed by using the Cancer Council Victoria dietary questionnaire (14). This is a comprehensive food frequency questionnaire, validated for assessing habitual dietary intake in the Australian population 15). This questionnaire reports the participant's usual consumption of 74 foods and six alcoholic beverages over the preceding 12 months on a 10-point frequency scale. Additional questions are asked about the type of breads, dairy products, and fat spreads used. By using the estimated portion sizes and frequencies, the intake of each food was converted to daily equivalents for statistical analyses. Nutrient intakes, including energy, were computed from the dietary data by means of the nutrient composition tables in the NUTTAB95 database (Food Standards Australia New Zealand, Canberra, 1995).

The Structured Clinical Interview for DSM-IV-TR Research Version, Non-Patient Edition (SCID-I/NP) (16), was the primary diagnostic instrument for mental disorders. Diagnoses of current major depressive disorder, dysthymia, and anxiety disorders were our outcomes of interest, as they have the highest prevalences. Researchers trained in psychology and administration of the SCID-I/NP (F.N.J., L.J.W.) conducted these interviews. Psychological symptoms were measured with the 12-item version of the General Health Questionnaire (GHQ-12) (17). This is a well-established screening instrument designed to detect nonpsychotic psychiatric disorders in people in community and medical settings by means of a 12-item self-report questionnaire about the respondent's assessment of his or her present state.

Socioeconomic status was determined by using scores on the Socioeconomic Index for Areas (18) based on the 2006 Australian Bureau of Statistics census data. It was decided a priori to use one of the included measures, the Index of Relative Socioeconomic Advantage and Disadvantage, which accounts for high and low income and the type of occupation, from unskilled employment to professional positions. A low score on this index identifies the most disadvantaged (quintile 1), and a high score identifies the most advantaged (quintile 5). Information regarding the highest educational level attained was derived from baseline assessments and comprised four categories, from primary schooling only (score=1) to university or other tertiary qualification (score=4). Physical activity was assessed by self-report questionnaire and ranged from confined to chair or bed or having limited activity throughout the home (score=1) to very active (score=4). Alcohol consumption in grams per day was ascertained from the dietary questionnaire and was subsequently categorized according to 2001 recommendations for women (19) as the following: zero consumption (score=1), 1–14 standard drinks per week (score=2), or 15 or more standard drinks per week (score=3). The presence or absence of current cigarette smoking was self-reported. Height was measured to the nearest 0.1 cm, and body weight was measured to the nearest 0.1 kg. Body mass index was calculated from these measurements as weight divided by height squared (kg/m²).

Differences in characteristics between subjects with and without categorical diagnoses of depressive and/or anxiety disorders according to the SCID-I/NP were tested by using the Mann-Whitney U test or the chi-square test. Diet quality and dietary factor scores were normally distributed, but GHQ-12 scores were positively skewed and were normalized by using a natural log transformation. All exposure variables and GHQ-12 scores were subsequently standardized by using z-scores and were expressed as standard deviation units.

Multivariate linear regression analyses were used to assess the relationship of the exposure variables of interest and GHQ-12 scores. Logistic regression models were developed to estimate odds ratios with 95% confidence intervals for the outcomes of current major depressive disorder/dysthymia and anxiety disorders. Diet quality scores and dietary factor scores (entered simultaneously) were the exposure variables of interest. For each analysis, the following covariates were entered sequentially in order to test their contribution to the mental health outcomes: age, socioeconomic factors (Index of Relative Socioeconomic Advantage and Disadvantage score and education), health behaviors (physical activity, alcohol consumption, and smoking), and energy intake (for dietary factor scores only). Body mass index was also tested in the models as an explanatory variable. These models took into account potential nonlinear associations and interactions.

There were no missing data for the dietary variables. There were missing data for GHQ-12 scores (N=23), and listwise exclusions were used when GHQ-12 score was the outcome variable, resulting in a sample of 1,023 women for these analyses. There were 24 participants with missing data for body mass index. There were no missing data for any other variable.

Results from analyses, with all three dietary patterns entered simultaneously as exposure variables, demonstrated that a western dietary pattern was associated with higher GHQ-12 scores, both before and after adjustments for potential confounders. After all adjustments, each standard deviation increase in the factor score for western diet was associated with a 0.17 SD increase in mean GHQ-12 score. There were no associations between the traditional or modern dietary pattern and the GHQ-12 score (Table 2).

In analyses conducted with categorical depressive and anxiety disorders as outcomes, there were inverse relationships between the score for traditional diet and the odds for both depressive and anxiety disorders, which were strengthened by adjustments for socioeconomic and lifestyle factors. After adjustments for age, socioeconomic status, education, physical activity, alcohol consumption, smoking status, and energy intake, a 1.00-SD increase in the score for traditional dietary pattern was associated with a 35% reduced odds for major depression or dysthymia and a 32% reduced odds for anxiety disorders (Table 3). The western dietary pattern was associated with increased odds for major depression or dysthymia in the unadjusted analyses. After adjustment for energy intake, the positive association between a western dietary pattern and depressive disorders was reduced but remained evident (Table 3). There was no observed association between a modern dietary pattern and depressive or anxiety disorders before adjustments, but there was a nonsignificant positive association between a modern dietary pattern and major depression or dysthymia after final adjustments for energy intake (Table 3). Further adjustment for body mass index did not affect the reported associations.

Diet quality scores were normally distributed and ranged from 4.0 to 56.0 (mean=30.8, SD=8.6). Multivariate regression analyses demonstrated that higher diet quality scores were associated with lower mean GHQ-12 scores. After all adjustments, there was a 0.08-SD decrease in log-transformed GHQ-12 score for every 1.00-SD increase in diet quality score (Table 2). The relationship between diet quality scores and GHQ-12 scores followed a dose-response pattern (Figure 1). A higher diet quality score was also associated with lower odds for depressive disorders in the unadjusted model, but this association was attenuated by adjustments for socioeconomic and lifestyle factors. Once again, further adjustment for body mass index did not affect the reported associations.

The main limitation of this study is the cross-sectional design, which prevents conclusions being reached regarding the direction of the relationship between diet and mental health. Appetite changes are a common feature of depressive illness, and a poor-quality diet may be a result of mental health symptoms, rather than a causative factor. Thus, without a prospective study design, it is not possible to preclude reverse causality as an explanation for our findings. However, in our study there was little evidence that the energy intakes of individuals with psychopathology were markedly different from those of the other subjects.

Diet is an indicator of overall lifestyle patterns, and health-related behaviors are influenced by mental illness, as well as influencing the risk for depressive illness (27). Smoking, alcohol consumption, and physical activity levels are all indicators of self-care (28) and were adjusted for in our analyses. Nevertheless, the possibility exists that we did not account for these adequately or that other unmeasured confounders were a factor in the results. Moreover, in studies examining diet as a risk factor for medical outcomes, confounding by socioeconomic status must be considered (29). In this study, socioeconomic status and education had little impact on the reported associations, but it is not possible to rule out residual confounding or reporting bias by socioeconomic status as a factor in these results.

The strengths of this study include the use of gold-standard assessment tools for the diagnosis of depressive and anxiety disorders, the inclusion of a wide range of likely confounding variables, the testing of body habitus as a possible confounding or mediating factor, and comprehensive dietary assessments with a tool appropriate for the population under study. Diet quality was also operationalized by using both a priori measures (diet quality score) and naturalistic observations (factor analysis). Moreover, the study cohort was randomly selected, population based, and representative of the wider female Australian population (30).

Depressive illness is influenced by genetic, hormonal, immunological, biochemical, and neurodegenerative factors. Diet modulates each of these factors and, as a result, has a plausible impact on the development and course of this illness (4).

Inflammatory processes are thought to play an etiological role in the onset and maintenance of depressive disorders (31), as well as being of central importance in the high-prevalence noncommunicable diseases, such as cardiovascular disease, diabetes (32, 33), and cancer (34). Inflammation may, in part, explain the association between these medical illnesses, depression, and mortality, as well as the associations with diet. Adherence to a Mediterranean diet, high in vegetables, fruits, legumes, whole grains, fish, olive oil, and low-fat dairy products, correlates with lower levels of inflammatory markers (35), whereas western-type diets and diets high in refined carbohydrates are associated with higher levels of C-reactive protein, a marker of low-grade inflammation (5).

Nutritional factors also exert a direct and potent influence on neural physiology (7). In experimental studies a western-style diet lowered brain-derived neurotrophic factor (BDNF) levels within a short time, an effect that was independent of obesity or nutritional deficits (6). BDNF protects neurons from oxidative stress and promotes neurogenesis (36) and is believed to play a central role in depressive illness (37). Thus, by modulating the expression of BDNF, diet may influence psychiatric status.

Finally, diet influences oxidative processes, which may also be implicated in the pathophysiology of depressive illnesses (38). For example, antioxidant-rich diets appear to be effective in slowing or preventing age-associated pathophysiological and cognitive changes (39) and reducing the risk of age-associated diseases (9). In this way, dietary factors may modulate the risk for neurostructural and cognitive changes in the brain that affect psychological symptoms across the life course (38, 40).

The potential impact of diet on pathogenesis of depression may compound the impact of depressive illness on appetite and self-care. Thus, mental illness, dietary inadequacy, and nutrient depletion exacerbating mental illness may be linked in a vicious circle. This model remains to be tested but mirrors and overlaps with what is already known about the course of depressive illness (41).